Sociologist Philip N. Cohen, Part Two

Part One

‘Philip N. Cohen: The End of Men Is Not True’ (Oct. 13, 2012)

Working off the YT transcript and running it through AI.

0:00 Thanks to the organizers; he notes that Hanna Rosin has already left and will have to catch it online.

0:29 Sets up the talk: partly about the book, partly about the same claims circulating outside it, with two general points and a lot of data.

1:03 First point: women are not dominant. He declines to run through the specific errors and distortions, says they are on his blog, and argues instead that the story is wrong at its foundation.

1:41 Why the baseline is worth establishing: the theory of how we supposedly arrived here treats women’s rise as continuous and inevitable, which he calls disempowering.

2:20 The claim that young women out-earn young men, which he says has reached meme status and is false.

2:33 Earnings by age across all full-time workers: close in the late twenties, and no reversal.

2:47 The underlying study compared full-time year-round, never-married, childless, metro-area workers aged 22 to 30.

3:15 His objection: that conditioning set is engineered to find the population with the least measurable labor market inequality.

3:35 Break the same group out by education and men earn more within every level; the pooled group is oddly composed, with many Latino men and many White women.

3:56 The rhetoric of fifty percent, which he distinguishes from a tipping point. A milestone is a round number. A tipping point is where a trend accelerates.

4:20 Managerial and professional occupations are majority female, but the category holds the 89 percent of nurses and 78 percent of elementary and middle school teachers who are women alongside engineering at 12 percent.

4:53 Managers alone are far more male; top executives are 26 percent female, inflated by people running small businesses.

5:07 Degrees. On Rosin’s line that women are beginning to crowd men out of science and engineering, he says beginning to means it has not happened.

5:31 Field composition among 25-to-39-year-olds runs from about 75 percent female in psychology to 20 percent in engineering.

5:56 Wives out-earning husbands sits around 28 percent of married couples, and can be pushed near 40 percent by restricting the base to employed or full-time women.

6:43 The full distribution, 1970 to 2010. Couples where the wife earns nothing fall from 46 percent to 18.7. Couples where she earns everything rise from 0.6 percent to 2.3.

7:32 Newlyweds in the American Community Survey: 36 percent of college-graduate wives out-earn their husbands in the first year, 44 percent among African American women.

8:19 Rosin’s claim that sexual assault rates have fallen so far that criminologists cannot plot them.

8:38 He grants the decline is large and appears in both police reports and the National Crime Victimization Survey.

9:02 New Jersey has the lowest state rate of reported forcible rape at 11.2 per 100,000, which he says is more than twice the median for European countries.

9:36 Rosin’s reversal, that female power will bring more female violence.

9:54 The only rising series is arrests of female juveniles. Victimization data show female offenders down by about half in twelve years, and intimate partner homicide down sharply, with the steeper fall among male victims.

10:46 Alabama. He pauses on the claim that women dominate Alexander City and Auburn, and says he can say they do not without ever having been there.

10:58 The line the talk turns on: if the facts do not support the generalization, the anecdotes are not illustrations of a trend, they are little stories.

11:11 The first female mayor was elected eight years earlier, is a traditional southern Republican from a political family, is stepping down, and five men are running to replace her.

11:45 One woman on a six-member city council; fourteen of fifteen department heads male.

11:55 Against the trailer-park anecdote: 77 percent of men in their twenties in Alexander City are employed, against 53 percent of the women.

12:12 He interrupts himself to say he is not claiming the men are better, only describing a normal pattern.

12:24 Auburn. Council, mayor, and department heads; the within-couple income distribution close to the national one; women at about 70 percent of men’s earnings; men holding about 70 percent of management.

13:18 The velocity claim, that change is too fast to stop and will carry women past parity into dominance.

13:45 Rosin’s list of firsts she calls unthinkable twenty years earlier.

14:06 All of them had already happened, except an Ivy League president, and that was 1993. Marlene Sanders and Jessica Savitch anchoring; Shirley Chisholm and Pat Schroeder running; Roseanne Barr; Penny Marshall on her fourth feature by 1992; Maggie Lena Walker founding a bank in 1902.

14:29 His point: movement toward equality takes politics, social movements, and policy, and does not arrive on its own.

14:47 Women’s labor force participation flat since the late 1990s.

14:57 The pay gap narrowed sharply in the 1980s and early 1990s and has crawled since.

15:04 College-educated women are doing worse relative to college-educated men than they were twenty years earlier.

15:11 Occupational segregation: no progress in twenty years for women without degrees, modest progress for women with them.

15:23 Women entering management: large gains in the 1970s and 1980s, close to nothing since.

15:34 Conclusion. We should think harder about why we are stalled than about how far we have come.

15:56 What he finds wrong in the narrative is the implication that we can sit and watch women collect their gains.

16:03 Google Books mentions of feminism and feminist, both peaking in the 1990s.

16:12 Close: culture, politics, and policy.

The method is complete at age forty-five. This is October 2012, three years before the Goffman comment, four before the archive, nine before he quits the association, and everything the later career runs on is already here: the receipts parked on the blog so the talk can make an argument, the attack on a conditioning set, the correction of a factual detail nobody bothered to check, the willingness to take apart a book whose politics his audience shares, and the flat prosecutorial delivery.

The best thing in the talk is the point about the conditioning set at 2:47. The statistic requires you to hold constant full-time work, year-round work, never having married, having no children, living in a metro area, and being between 22 and 30. His objection is that the controls are the finding. If the gap disappears only after you remove marriage, children, part-time work, and everywhere that is not a city, then marriage, children, part-time work, and geography are the gap. That is the same move he later runs against the success sequence, and it generalizes past gender to any argument that produces equality by subtraction.

The distinction at 3:56. Fifty percent is a round number in base ten and nothing happens at it. A tipping point is a claim about acceleration and has to be shown. Most trend journalism converts the first into the second without noticing, and the conversion is invisible because the arithmetic looks like evidence.

The line at 10:58 is the one to quote. If the facts do not support the generalization, the anecdotes are not illustrations of a trend. It is the clearest statement I have seen of the quarrel between reported texture and aggregate measurement. Rosin was not claiming that Alexander City has a female-majority council. She was reporting a felt reversal of standing inside particular homes among particular men, and Cohen answers with employment rates and department heads. He is right that the aggregates do not show dominance and she is right that aggregates are not what she went to see. The boast at 10:46, that he can say it never having been there, is the whole epistemology in one clause, and it is the same claim that made the ethnographers furious three years later.

The mayor correction at 11:11 is his strongest hit and the model of what he does better than anyone. The claim that a town got its first female mayor that year was checkable in five minutes, was wrong by eight years, and had survived whatever fact-checking the book received because it sounded like the thing being argued. The 14:06 list is the same weapon aimed at chronology: every item presented as unthinkable twenty years earlier had already happened, some of them long before. Trend writing runs on an assertion of novelty and the assertion is the least examined sentence in the genre.

The stall thesis is the part that has aged best. Convergence in pay, segregation, management, and participation happened in the 1970s and 1980s and mostly stopped, and he said so in 2012 and has been broadly right since. It also carries the politics, and he states the link at 14:29: automatic progress requires no movement, and stalled progress requires one.

Two weak spots he would not tolerate in anyone else.

The New Jersey comparison at 9:02 does not work. Cross-national counts of reported rape are shaped by legal definitions and reporting propensity more than by incidence, which is why Sweden looks like a catastrophe and why the countries with the lowest reported rates are often the worst places to be a woman. He built his career on the principle that you check whether the instrument changed before you compare, and he skipped it here for a debating point against a rhetorical flourish. He had already conceded the substantive claim at 8:38.

The Google Books slide at 16:03 is worse. Frequency of the words feminism and feminist in a corpus of published books is a fact about publishing and about corpus composition, and he uses it as a proxy for the vitality of a political movement, with no adjustment and no discussion of what the corpus contains. It is the move he spent the previous fifteen minutes taking apart. It is also the last slide, which is where a speaker reaches for the thing that feels true.

One detail Tom Wolfe would have led with. This is a law review symposium at Boston University, the format is adversarial by design, the subject of the prosecution left before the prosecution began, and Cohen opens by pointing at the empty chair and joking about it. He is the expert witness in a proceeding with no defendant present, which is a fair description of most of his public career, and the joke suggests he knows it.

‘Philip Cohen addresses National Academies on “Against the Generations”‘ (July 30, 2019)

0:06 Opens by thanking the organizers, notes he is the last speaker before lunch and will try to be brief and lively.

0:30 Says the committee’s remit gave him the chance to make a pitch that has been gnawing at him for four years.

0:57 The complaint: generation talk encourages broad generalization, stereotyping, clickbait, character judgment, and echo-chamber thinking. He will illustrate with family demography and the recession.

1:30 The Pew Research Center’s generation scheme, which he calls the leading purveyor of these memes.

1:43 First objection: these are cohorts, the categories are irregular, and the boundaries have no justification. Making the last two spans fifteen years each avoids people sharing a generation with their own children, and families skip generations anyway.

2:13 His own family has no baby boomers; his brother was born in late 1964.

2:32 Once you name them and assign characters, you impose qualities on populations with no basis, and it snowballs into confirmation-bias clickbait that nobody assesses.

3:02 He acknowledges the incentive problem. Abandoning terms that generate millions of clicks would be brave and admirable.

3:24 The 2015 Pew survey that gave people a lineup of generation names and asked which they belonged to. Only 40 percent of Millennials picked Millennial, and respondents could select more than one.

3:53 Uptake was low except among boomers, and the response was not self-reflection. He points out that if racial categories produced this, social science would treat it as a problem for the theory.

4:29 A disclaimer he repeats: he admires Pew, uses their material constantly, including in this talk, and thinks they should get more money. The trouble is an addiction they have to sell.

4:45 Pew’s announcement moving the millennial endpoint back to 1996 and naming Generation Z. On their line that cut points are not an exact science: they are not a science at all, and there is no question they are trying to answer.

5:22 On their claim that the boundaries are not arbitrary: they are completely arbitrary, and there is no agreed formula for the span.

5:42 The baby boom is the only cohort officially designated by the Census Bureau, and it is a demographic event that produced a cohort.

6:06 When Pew shortened the millennial span it generated stories treating the arrival of more Millennials than Gen Xers as news, when the span had simply had more years in it.

6:30 The census work: a birth rate spike in 1946 holding above that level through 1964, the only prolonged rise to a stable high level of fertility in the record, breaking the long-run trend.

7:08 The accident that started everything. The event happened to last roughly the length of a generation, which put the culture on a treadmill of naming them. Had the boom run eleven years instead of nineteen, he does not think the conversation would exist.

7:33 He realizes he never checked the clock.

7:45 The celebrity slide. The Silent Generation runs from Noam Chomsky to Debbie Harry. Donald Trump and Keanu Reeves are in the same generation.

8:09 The line between X and Millennial separates Venus and Serena Williams, sisters one year apart. He says the Gen X is all over Venus, you cannot miss it, this is like astrology, and Serena is a millennial.

8:43 Age-specific fertility rates from the American Community Survey, 2001 to 2017, by birth cohort, with dots marking where each cohort stood in 2010.

9:13 Women born in the early 1970s hit 2010 in their late thirties with their path already written. Late 1970s cohort, same.

9:37 The early 1980s cohort, the first Millennials, were at peak fertility when 2010 arrived, and fertility declined after.

9:52 The late 1980s cohort saw fertility fall from ages 23 to 24 to 25, which he calls extremely unheard of, then rebound into their early thirties, possibly close to catching up on completed fertility.

10:33 The youngest group hit 2010 as teenagers, and their curve rotated clockwise, falling off faster than earlier cohorts, with signs of a post-recession rebound.

11:09 The point: all three of those lines are Millennials, and collapsing them loses the entire story.

11:16 He notes his five-year groupings are arbitrary too. Use one-year or ten-year groups, or groups chosen for a reason.

11:28 Male employment rates by single year of birth from the Current Population Survey.

11:42 Men born in 1982: about 47 percent employed at eighteen, rising to their mid-twenties, then falling as the recession hit them in their late twenties, which he calls ordinary recession behavior.

12:01 Men born in 1987: a scarring event, with employment rates flat from 21 to 23.

12:18 Men born in 1992: 2010 hit at eighteen, and a twenty-point difference in employment at that age against the earlier cohort. All three are Millennials.

12:44 Marriage. Propensity to marry before thirty, controlling for age and sex.

13:00 The decline between people born in 1945 and 1965 is enormous, and both are baby boomers, so the group does not hold together on this measure.

13:20 The question science should be asking: when do we want to group people, and how.

13:29 First positive suggestion: ask whether the question is a cohort question at all.

13:41 On “kids these days,” a thing you could say at any point in history. General Social Survey data showing the under-35 share identifying as liberal is consistently higher.

14:09 If you are using a cross-section, you are describing right now, and labeling those people Millennials adds only the baggage.

14:26 Second suggestion: if you want cohorts, break them into five or ten year groups sized to your sample and eyeball it first.

14:34 Attitudes toward easier divorce, adjusted for age, sex, and education, by the four standard generations, then by decade of birth.

15:01 The decade version shows what the standard grouping hides. Late Millennials are markedly more pro-divorce than early Millennials, and the silent-to-boomer shift happens earlier than expected.

15:19 Naming advice: describing people as born in the eighties or nineties is fine. Calling them the tennis generation or the butterscotch generation assigns a character with no basis.

15:41 Third suggestion: cohort analysis is important where events at a point in the life course have lasting consequences.

15:56 Examples. Children of the Great Depression as the classic. Digital natives, if you study education.

16:14 The story about his daughter texting to ask what her phone number was, and the other daughter’s ruling that if it can send email it is a phone.

16:38 Post-9/11 schooling for security research; leaving school in a recession for labor economics; and, for politics and culture, the high school students shaped by the Parkland moment.

17:07 The close: none of those experiences supports assembling one thing, calling it a generation, and having it make logical sense. He urges the committee not to use the labels at all.

17:39 First question, from John: is there systematic research showing that generation language promotes stereotyping and bad social outcomes?

18:02 His answer: he knows of none. Most sociologists do not use these as real categories in the research, only in the discussion section when making findings seem relevant. He suggests experimental psychology could test what the labels evoke.

18:39 Second question: if Pew hired you, what would you tell them to do other than stop?

19:00 His answer: do not give up cohort analysis. Use decade cohorts and routinize them; the labels exist, the intervals are regular, and they already carry cultural salience. He is not telling Pew to delete their account after a thousand reports.

19:37 His own practice: a script that breaks age into decades and years into decades as a first pass at trends before looking finer.

20:07 A committee member’s comment: cohort and age questions apply universally across the social sciences in a way most topics do not.

20:58 Third question, from Dana, on the great-and-terrible framing: pushing him toward what is redeeming in generational analysis, and noting that many people think astrology has scientific underpinnings.

21:47 His reply that astrology has no scientific value, followed by the crack about a selected expert audience.

22:01 The substantive answer: cohort analysis and the life course perspective are important. Job changes, health insurance at a given age, school crowding, the racial composition of a childhood neighborhood. He wants to see the shadow of events across lives at the cultural level.

22:40 A concession. The millennial idea was that these people came of age in a new era around 2000, which he calls a powerful way of thinking and an opportunity to apply social science.

23:02 He is not discouraging cohort thinking, he wants to improve it, and he acknowledges an institutional juggernaut built around the concepts.

23:34 A return to the Pew survey: if you live in a world where everyone knows what Millennials are, that may indicate you are not interacting with most people.

24:01 The closing self-assessment: this is not his most important work. Understanding demographic change, how it relates to inequality, and communicating it so people can make smarter decisions is his most important work, and he counts this as part of that.

This is the campaign two years before the open letter, delivered to the body most likely to act on it, and it works. He got Pew’s policy change in 2023 and this is where you can watch him earn it: no accusation of bad faith, praise for the target three separate times, a concrete alternative offered under questioning, and a closing line that ranks his own complaint below his real work.

Compare it with the Rosin talk seven years earlier and the difference is the target’s status. Both talks run on the same two moves. First, take a category and show that the things inside it behave differently, which is the argument at 9:13 through 12:44 and the argument about the 22-to-30 conditioning set in 2012. Second, check a claim that nobody checked because it sounded right, which is the Pew self-identification survey here and the Alexander City mayor there. What changed is that in 2012 he was correcting a journalist and in 2019 he is advising a National Academies committee, and the register adjusts without the argument changing at all.

The strongest passage is 6:30 to 7:33 and it is the part nobody quotes. The baby boom is a real demographic event with a documented spike, a documented plateau, and a break in the long-run trend, and the Census Bureau designated it for that reason. Everything downstream is imitation of a thing that happened once. And then the historical accident: the event lasted about as long as a human generation, which licensed the treadmill. Had it run eleven years, he says, there would be no Generation X, no Millennials, no Z. That is a claim about how a measurement artifact became a cultural ontology, and it is the most original thing in either talk.

The Venus and Serena slide is doing more work than the joke suggests. Sisters born a year apart, raised in the same house by the same parents on the same courts, sorted into different generations by a boundary somebody chose. It is a within-family control. He throws it away as comedy at 8:09, which is a habit of his: the sharpest evidence gets delivered as a laugh line.

He concedes the key empirical question at 18:02 and nobody notices. Asked whether there is research showing generation labels cause stereotyping or harm, he says he knows of none. The talk’s opening claim, at 0:57, is that the labels drive stereotyping and character judgment. Under questioning he admits that is an intuition. It is a reasonable intuition and it may be right, and it is unverified, and this is the man who spends his December afternoons counting whether other people’s claims can be checked. He does the honest thing and answers straight, and the honest answer undercuts his own frame.

The astrology answer at 21:47 is the misfire. The questioner was making a subtle point, that a scheme with no scientific basis can still organize how people understand themselves and each other, which is an argument that generation labels might be socially real even if demographically empty. He hears it as a claim that astrology is scientific, corrects her, gets a laugh from the room, and only reaches the substance after the moment has passed. The correction was directed at something she did not say.

And the concession at 22:40 is the largest one he has ever made in public on this subject. The millennial idea, that a cohort coming of age around 2000 encountered something new, is called a powerful way of thinking and a good opportunity for social science. That is close to the thing he spent seventeen minutes urging them to abandon, and it means his position is narrower than the title. He objects to the naming, the character assignment, and the arbitrary cut points, and not to the underlying thought.

The distinction between a demographic event and a generation, with the eleven-years counterfactual, is portable and nobody has credited him with it. And the decade-cohort script he mentions at 19:37, the thing he runs before looking at any trend, is a small biographical detail that shows the habit underneath the arguments.

Transcript flags. At 4:29 “picking up Hewitt” is picking on Pew. At 4:50 “296” is 1996. At 5:12 “your medicines are talking their story ins” is a mangling, probably epidemiologists and historians. At 8:23 “Generation Next” is Generation X, and “Carrie” should be career or possibly aura. At 13:00 “married by age three” is thirty. At 14:43 “general any ‘el” is Generation X and Millennial. At 15:27 “quartzion” is cohorts. At 19:21 “cultural violence” is cultural salience. At 22:53 “melayu” is Millennials. At 24:27 “dr. otter” and “dr. Capelli” are the remote panelists, and Peter Cappelli is the likely spelling.

‘Philip N. Cohen COVID lecture April 3 2020’

0:00 Opens by greeting his social problems class, says he has been missing them, and notes it has been a month since he first opened the class to discuss the then-emerging epidemic.

0:23 Sets the plan: where we are, how we got here, where we are going, then discussion.

0:48 Records this on the afternoon of April 3, 2020.

1:07 Structure: global trajectory, US trajectory, and what comes next.

1:14 The disclaimer he returns to twice more. He is not an expert on pandemics or global public health. He is a social scientist who can understand what he reads, and a teacher.

1:36 Etymology of pandemic, from the Greek for all and people.

1:44 The World Health Organization declared a pandemic on March 11.

2:11 Coronaviruses affect humans and other species, ranging from common colds to serious disease. The novel virus was found in Wuhan late in 2019; the disease is COVID-19.

3:02 Why this one is bad. No human immunity.

3:18 Highly contagious, and he says twice that we do not know exactly how it spreads.

3:34 Asymptomatic transmission, which defeats the strategy of avoiding people who look sick.

3:58 Relatively high mortality. Most people survive and most cases are mild, and the rate is higher than seasonal flu. He says we do not know the rate, that it depends on how much medical care you get, and that it is maybe around one percent.

4:31 Confirmed cases worldwide, probably reaching a million as of the day before, with 80,000 new cases reported yesterday. Exponential increase.

5:11 The global map. Spread followed travel from China to Europe and the United States, with an Iranian outbreak from a cluster of travelers.

5:42 Countries without rigorous testing are unknowns, so low counts in Kenya or Zambia say nothing about incidence.

5:59 Droplet transmission and the origin of the six-foot rule, which comes from how far droplets travel when you cough.

6:16 Asymptomatic spread, plus strengthening evidence of aerosolized droplets that may linger in air, and cases where no liquid droplets appear to have been exchanged.

6:51 Current recommendations: six feet, hand washing. And a forecast, on April 3, that a cloth mask recommendation is coming to protect others from what you may be expelling before you have symptoms.

7:23 The section on what the Trump administration got wrong before the outbreak. First, a general orientation against effective government, expressed in CDC cuts from just under eight billion in 2018 to just over seven in 2019, with a proposed budget near six and a half.

8:11 The pandemic preparedness office in the National Security Council eliminated in a reorganization, and a Center for Strategic and International Studies report from the previous fall noting it remained unclear who would be in charge in a cross-border biological crisis.

9:02 The 2019 federal pandemic exercise, run on a novel influenza scenario, which found the government unprepared and produced recommendations that were ignored.

9:40 Trump’s claim that nobody could have seen this coming. Cohen calls it a total lie, then adds that Trump did not know because he does not know a lot, while people who paid attention did.

10:05 The happy talk section, and Trump’s position that one or two hundred thousand deaths would count as total victory.

10:43 A montage of Trump clips: he always knew it was a pandemic, it is totally under control, one person from China, fifteen cases going down to close to zero, it will disappear like a miracle.

12:46 His verdict on the clips: none of it was true, and the messaging existed to make Trump feel he was doing a good job, which sent the public a signal that this was not serious.

13:09 Operational failures. Testing first. Without it you cannot count cases.

13:25 The standard first-line response: isolate, contact trace, quarantine contacts.

13:57 The CDC’s first test did not work, and the country did not adopt the World Health Organization test that might have permitted faster testing at scale.

14:13 Community spread arrived before anyone could trace it, with cases of unknown origin.

14:22 Containment failures: no restriction on movement between affected and unaffected areas, or within towns, schools, nursing homes, and malls.

14:46 No uniform system for stay-at-home orders. Most places now have rules or strong guidance, not everywhere, and all of it too slow.

15:19 Supply failures. Personal protective equipment, and the fact that it has to be changed constantly.

15:44 Ventilators, and the shortage of both machines and people to operate them. The federal stockpile was insufficient, some units did not work, and distribution was poor.

16:09 His central policy criticism: too much left to the market.

16:42 The log-scale doubling chart, days since a country’s tenth death against deaths, each gridline a doubling, a straight diagonal meaning exponential growth.

17:22 South Korea at about 37 days sits between 128 and 256 deaths. Spain at 24 days is near 10,000. Italy, France, and recent outbreaks in Brazil and Canada at 13 or 14 days follow the same course.

18:03 The United States slope is steeper than the others, which he calls the disturbing part.

18:11 A qualification that improves the analysis. The country is large enough that it is more like Europe, a set of separate outbreaks.

18:28 The same chart by state. New York on Spain’s path, Washington the best at 29 days and under 256 deaths, Michigan, New Jersey, and Louisiana rising dangerously.

19:11 What a straight line means, and what bending looks like. Spain and Italy show early curvature if the numbers are accurate, and he notes the fuzziness.

19:36 China is left off because it would break the graph.

19:53 Flattening the curve. The red distribution shows unmitigated spread piling up cases past the health system’s capacity.

20:23 The logic of closing schools, restaurants, and bars: slow the rate so the sick arrive at a hospital that can take them.

20:54 Even without a vaccine, slowing infection reduces total damage.

21:11 Back to the state chart, with an explanation of the variation. Washington’s first outbreak was largely confined to a nursing home, and the state closed and distanced early, producing a shallower slope. California similar. New York and Michigan did too little early and got explosive outbreaks.

22:19 The window argument: most of the country does not have a bad outbreak yet and can still change its slope.

22:36 What happens next. People arriving at hospitals now were infected ten days to two weeks ago, six to eight days to symptoms and another five to seven to hospitalization, so today’s admissions date to roughly when schools closed.

22:59 Spread continues because nothing stops you driving from New York to Missouri.

23:15 His prediction: many thousands of Americans will die and that is inevitable at this point, with no break in the slope visible.

23:35 How bad it gets over months remains open, and mortality is not a fixed biological rate. It depends on diagnosis, hospital resources, and staffing.

24:10 The Javits Center temporary hospital, and his forecast of more converted convention centers.

24:27 When it ends. Absent a vaccine, nobody is immune, so tamping the numbers down may let people out with waves of local outbreaks and repeated shutdowns.

25:00 The exit condition: few enough cases plus testing and tracing capacity, then normal life with distancing and rapid isolation when someone gets sick.

25:30 Global spread means travel restrictions and quarantine requirements will last much longer than domestic measures.

25:54 Vaccines. A best case around eighteen months, then testing for safety and efficacy, then production and distribution of billions of doses.

26:18 No treatment yet; patients fight it with their own immune systems, and treatment research is underway.

26:34 What this means for students, prefaced by another disclaimer: this is not his expertise, he cares about them, and he is not the person to give personal advice.

26:50 Follow real public health advice. Do not listen to the president or the politicians in his orbit. Listen to your governor, local public health officials, and the CDC.

27:14 It is not only about you but about whom you might affect.

27:21 Take care of yourselves mentally and physically, and do not let physical distance become total social distance.

27:37 Helping others, with time or money or online, can be empowering.

27:52 The close. The world will be different, think about who you want to be in it, and humanity has gotten through bad things before.

This is the only recording in my file so far where Cohen is not prosecuting anybody’s arithmetic.

Watch what he does with uncertainty here. Inside twenty-five minutes he says we do not know exactly how it spreads, we do not know the mortality rate, we do not know how many are infected in countries without testing, we do not know how many will die, we do not know how long it takes, the numbers have fuzziness in them, and China would break the graph so he left it out. That is a man narrating the limits of his evidence in real time to nineteen-year-olds. The audits and the Rosin talk and the Pew talk all run on the opposite posture, certainty about a number somebody else got wrong. The pedagogical setting produces the epistemics his critics say he lacks, which suggests the register is a choice about audience.

The disclaimer at 1:14 is the thesis of Citizen Scholar five years before the book. He tells them he is not an epidemiologist, that he is a social scientist who reads well, and that he is a teacher. He is claiming authority to synthesize while disclaiming authority to know, and then he spends twenty minutes doing exactly what he claimed. That is the role boundary the book argues about, drawn in public, at the moment it was hardest to draw, when a professor with a microphone and frightened students could have taken the larger authority for free.

Then he takes a different one. At 9:40 the lecture turns and stays turned. Ignored warnings, a total lie, the montage, the verdict that the messaging existed to make the president feel good about himself, and at 26:50 a direct instruction to his enrolled students not to listen to the president. He would say the roles are separable and the sourcing supports every claim, and the sourcing does. His own book says the trouble starts when academic authority carries a man from one role into another without the audience noticing the crossing. Here he crosses in front of a captive audience holding his grade book, in the same voice he used for the doubling chart. He announces the crossing at 26:43 by saying this is not his expertise, which is the disclosure his method requires and which does nothing about the power differential. Someone should ask him about it.

He was right about the things his training equipped him to judge and less reliable on the rest. The market-versus-federal-direction argument at 16:09 is his own field and is a live policy dispute he states as settled. The eighteen-month vaccine best case at 25:54 was the consensus and turned out pessimistic, and the reason it was wrong sits outside a demographer’s competence. The forecasts he got right are the ones that follow from arithmetic he could do: exponential slopes do not break on their own, hospital admissions lag infection by two weeks, tamping down produces waves, travel restrictions outlast domestic ones, and thousands would die. The one he got right by a day or two is the mask recommendation at 6:51; the CDC issued it that same day.

Two moves.

At 18:11 he corrects his own chart while presenting it. The American slope looks worse than Europe’s until you notice the country is the size of Europe and holds several separate outbreaks, so the aggregate curve is an artifact of the unit. That is the same objection he makes to generation labels and to the success sequence: the category is doing work the data are not. He applies it to his own slide, unprompted, in front of students who would never have caught it.

And 22:36, the lag argument, is the finest teaching in the video. Today’s hospital is a photograph of two weeks ago. Everything you can see is already history and everything you do today shows up in two weeks. That is a demographer’s habit of mind, the same one behind checking his divorce prediction against data that arrive years later, and it is the single most useful thing an ordinary person could have learned in April 2020.

This belongs in the biography as the missing 2020 entry. Every other document has him fighting somebody. This one has him teaching, and the teaching is patient, careful about what it does not know, and warm at the end in a way the blog never is. The essay currently derives his character from the news desk clock and the audit tables. This is the third setting, and it is the one where the students are, which is where most of a professor’s life happens. Two lines at 27:21 and 27:52, about not letting physical distance become social distance and about deciding who you want to be afterward, are the only places in anything I have read by him where he speaks to a person’s wellbeing.

Transcript flags. “philip gon” at 0:00 is Cohen. “COBIT 19” and “Kovach 19” are COVID-19. “pandemics rent” at 8:37 is pandemic event or similar. “first actor” at 21:37 is first outbreak. “organ bed” at 14:56 is garbled, likely congregating. “an S case scenario” at 26:02 is a best case scenario. “how many people really happen in Kenya” at 5:42 is have it. The chapter titles (“WHAT TRUMP DID WRONG: CUT CDC FUNDING”) are his own.

‘Family violence and abuse in COVID context lecture’ (Apr. 6, 2020)

0:00 Announces a short video lecture on family violence and abuse and the implications of the pandemic.

0:18 Context for why now: he has been revising the violence chapter of his textbook, due out in about half a year.

0:45 The headlines he is responding to, that home is not always safe and that lockdowns are associated with rising domestic violence around the world.

0:59 His first move, and the lecture’s governing posture: we do not have the information to draw those conclusions yet, but we know enough to raise questions and think the issues through.

1:15 Historical backdrop. Violent victimization of people aged 12 to 17 by family members, from the National Crime Victimization Survey, falling from about 14 per thousand to just over two.

2:00 Why it declined, prefaced by a warning that attributing violence trends to particular social changes is speculative.

2:29 Three candidate causes. Shelters and hotlines that provide distance from abusive partners.

2:47 Women’s rising education, employment, and the cultural acceptance of independence and single parenthood, which make leaving easier both structurally and in terms of values.

3:10 The decline in marriage and cohabitation, which reduces exposure to the main risk factor, presence of a spouse or partner.

3:28 Lifetime prevalence. Something over a quarter of both men and women report violence in a relationship, and the gender disparity widens sharply for severe physical violence, contact sexual violence, and stalking.

4:18 Defining the sphere. Family violence means caring or intimate relationships, with partners, children, and elders.

4:48 The structural claim. Secrecy and isolation are made possible by how families are built, literally by the walls and figuratively by the sphere of privacy people value highly, which is the same privacy that makes the problem hard to reach.

5:22 The framing he says he always reinforces: this is not about bad apples, it is a structural feature of family life, which does not mean everyone experiences it and does mean we should ask what our institutions do to enable and reproduce it.

6:04 Child abuse and neglect defined as an act or failure to act by a parent or caretaker resulting in or risking physical or emotional harm, sexual abuse, or exploitation.

6:32 The perpetrator chart, overwhelmingly parents and family members.

6:52 Two glosses. Neglect is definitionally a failure of whoever holds responsibility. The high figure for mothers reflects that many children live with their mothers only.

7:12 Risk factors: parental mental health and impulse control, any history of violence including as a victim, violence between adults in the home.

7:49 Poverty, which raises risk partly through material deprivation and partly through whatever produced the poverty.

8:14 Weak support networks, and the importance of having people around who notice and can intervene.

8:38 A clip from a Cleveland television station on advocates’ concerns, including the governor’s worry about children not seen by anyone outside the family, a center CEO reporting steady calls and heightened fear, and a closing line about checking in on each other.

10:52 The pandemic’s implications, opening with a swipe at the genre of first-world-problems content, the boredom and the same shows and the boring food, which does not tell you the story of families where things are going badly.

11:28 Economic shock. Roughly ten million new unemployment claims in recent weeks as of April 6.

11:45 Isolation from support networks, since family and friends are also sheltering. Who checks in, who notices, who helps, where does somebody escape to.

12:16 Services may be unavailable, government and nonprofit alike, for lack of staff and resources.

12:42 Mandatory reporters. Teachers, clergy, and in some states health care providers are the points where the outside world can reach into a family, and those interactions have been cut.

13:08 His caution again. All of this is theoretical; there is no evidence yet, and reason for concern.

13:23 Intimate partner violence against adults, down from about 16 per thousand women in the mid-1990s to about four, roughly a three-quarters reduction, with a decline for men from a much lower base.

14:04 Homicide, which the survey cannot capture because you have to be alive to answer it. In 2018, a little over a thousand women killed by intimate partners against about 360 men, with male homicide concentrated among associates and strangers.

14:47 The counterintuitive prediction. Lockdown may reduce male homicide more than female homicide, because it removes exposure to acquaintances and rivals while increasing time spent with partners.

15:15 The typology. Situational couple violence, arising from specific disputes without a pattern of escalation, triggered by stress and economic shocks, which is why he expects more of it now.

15:50 Coercive controlling violence, formerly called intimate terrorism, a campaign for control and domination with an escalating pattern.

16:12 Why the pandemic worries him more here. Escaping this pattern usually requires outside support, and isolation removes it.

16:36 The specific route: decisions about who goes out and when, and the negotiation of rules for mobility, feed into controlling behavior.

17:05 Slightly less than half of intimate partner violence is ever reported to police.

17:18 Reasons given for not reporting. Fear of reprisal or of getting the offender in trouble is the most common, covering both punishment for reporting and loss of the family’s economic support. Handling it personally is about 20 percent. Third is lack of faith in police.

17:58 Whether this changes under lockdown he calls speculation.

18:14 Rape and sexual assault. About two thirds by acquaintances and intimate partners, about 24 percent by strangers.

18:59 The two directions the pandemic pushes. More isolation with partners raises one risk while less circulation lowers the other, and the net is unknown.

19:24 Interventions, introduced as things there is reason to believe may help or that many people think are good and are trying.

19:41 Rapid assessment checklists for police at the scene, formalizing judgments about severity and escalation and routing victims toward services and escape planning.

20:13 Civil protection orders, with some success.

20:25 Court-mandated psychological treatment, with a mixed record.

20:35 Specialized family violence courts with judges who develop expertise, again mixed.

20:55 Victim services broadly, including escape planning, assessment, and counseling.

21:16 The forward question. Given a quarter century of decline, what does the present crisis imply.

21:36 The modernity argument. First, the construction of childhood as a protected period of innocence recognized by families, institutions, and the state.

22:19 Second, the rise of state power. Outsourcing dispute resolution to formal authority may reduce interpersonal violence, while raising its own questions about surveillance and oppression.

22:53 Third, the Pinker argument that the rise of reason as an ideal reduces violence because you have to justify what you do.

23:19 The momentum claim: childhood exposure predicts adult violence, so a generation raised with less of it may pass less of it on.

23:42 The unsettling. Two months ago he would have named climate change or Trump; now the pandemic raises the question of whether history has a direction at all, and big crises unsettle any narrative they do not fit.

24:16 The close. He has no answer for most of what he raised, and hopes the context and the background statistics will let students work out what is happening.

On April 3 he played a montage of presidential clips, called a public statement a total lie, and told enrolled students not to listen to the president. On April 6, on a subject where the political temperature is comparable and the policy stakes are immediate, there is nothing. The word Trump appears once, at 23:48, inside a subordinate clause listing the kinds of events that unsettle a story about progress. The rest is data, definitions, a typology, and seven separate admissions that he does not know.

The lecture where he editorialized was the one outside his competence. The lecture where he taught was the one inside it. Whatever governs his register, it is not confidence in the material. My guess, and it is a guess, is that the pandemic lecture had a villain available and this one does not, or has one he is unwilling to name in front of undergraduates whose homes he knows nothing about.

The epistemic discipline here is the best in anything of his I have read. He opens with the headline, tells them the headline cannot yet be confirmed, and then spends twenty minutes explaining what would have to be true for it to be right. He flags speculation at 2:19, at 13:08, at 17:58, at 18:59, at 21:30, and at 24:16. That is a professor teaching students how to hold a live question open, in April 2020, when every incentive ran toward certainty.

Two things in it have aged well and one has not.

The measurement point at 12:42 is the one that mattered most and the one nobody outside the field made in public that spring. Teachers, clergy, and clinicians are the layer through which the state sees inside a family, and closing schools removed it. Which means a fall in reported child maltreatment in 2020 would measure the removal of the observers. Reports did fall. Anyone reading those numbers without his warning would have drawn the opposite conclusion from the correct one, and the subsequent literature on intimate partner violence has been a mess for the same reason, with calls to police and hospital presentations and hotline volume moving in different directions because each measures a different mixture of incidence and access.

The structural claim at 4:48. The privacy that makes a home a refuge is the same privacy that makes it unreachable, and people value it most when things are going well. That is the argument of his textbook and the subtitle of his essay collection compressed into thirty seconds, and it is the reason he can teach this subject without moralizing: nobody has to be a monster for the structure to produce the outcome.

The prediction at 14:47 is the falsifiable one and it looks wrong. He reasoned from exposure structure: male homicide victims are killed by acquaintances and rivals, so lockdown should cut male homicide more than female homicide. American homicide rose sharply in 2020, by roughly a third, the largest single-year increase on record, and the increase was concentrated in exactly the non-intimate gun homicide his exposure logic predicted would fall. He modeled a pandemic as a reduction in interaction, and 2020 turned out to be a reduction in some interactions and a collapse in others, including in policing and in the institutions that suppress street violence. The exposure model was too clean for the year.

Three weaknesses, stated the way he would state them about somebody else.

The mother figure at 6:32 gets a chart and a single sentence of explanation, that many children live only with their mothers. That is a real explanation and an incomplete one. Exposure time, the poverty of single-parent homes, and neglect definitions that by construction fall on the primary caregiver all belong in the answer. He knows all of this. He is presenting a number that gets used badly and giving it a nine-word gloss, which is close to the thing he spends his blog objecting to.

The interventions section is his weakest four minutes anywhere. Some promise, some success, rather mixed, mixed results, and an introduction that says many people think these are good so they are trying them. No citations, no effect sizes, no discussion of what the evaluations found. Set that beside the ferocity he brings to counting whether a journal article shipped its code and the asymmetry is hard to miss. The standard he applies to other people’s claims about marriage promotion is not the standard he applies to his own summary of what works against family violence.

And he leaves the sharpest tension in the field lying on the floor. At 17:18 he reports that the most common reason women give for not calling police is fear of what calling will cost them, either reprisal or the loss of the household’s income, with distrust of police third. At 19:41 he recommends a set of interventions that mostly route through police and courts. Those two slides are four minutes apart and he never puts them together. A student who noticed would have asked the best question of the semester.

This is the textbook being written in public. He says at 0:18 that he was revising the violence chapter, which makes the lecture a draft of the third edition delivered live, and that is a detail for the biography that otherwise treats the textbook as a finished object.

Transcript flags. Every variant of koban, kovat, Cova, köppen, and Coppa is COVID. At 0:28 “should be halved in half year” is should be out in half a year. At 8:16 “as important networks” is and support networks. At 15:26 “vance” is violence. At 18:11 “not mr. speculation” is that is just speculation. At 21:16 “if we had to speak decline” is we have had this steep decline. At 21:57 “a life of that pain” is a life without pain. At 23:33 “West violence” is less violence. At 23:54 “just history really have a direction” is does history really have a direction. At 6:39 “pare” is parents.

‘Health Disparities and the COVID-19 Pandemic (old version)’ (Apr. 20, 2020)

0:00 Introduces himself, the subject, and the plan: define health disparities, connect them to the pandemic, and let students take it further. Gives his contact information and invites students to get in touch.

0:24 The CDC definition he says he likes: preventable differences in the burden of disease, injury, violence, or opportunities to achieve optimal health, experienced by socially disadvantaged populations.

1:06 Examples from the news, including a nursing home death, junk food, the effect of federal policy on opioid deaths, maternal mortality, and preventable cancers in Alabama.

1:57 The framework, the theory of fundamental causes, which he says sounds larger than it is. It concerns the relationship between socioeconomic status and health outcomes, and asks why general inequality stays so persistently attached to inequality in illness and death.

2:24 The improving baseline: water and sewer systems, nutrition, refrigeration and canning, medical advances, and rising education that lets people absorb health information.

3:00 Life expectancy in the United States rising from a little under 50 to almost 79 across about 120 years, with the caveat that the figure assumes nothing changes.

3:36 The theory’s answer. Income and education bring money, knowledge, prestige, power, and beneficial social connections, and the key insight is that as the health system evolves, higher status keeps converting into better outcomes. He calls it a moving target.

4:31 A study of people born in 1960, asking the odds of surviving from 50 to 85 by income quintile. Roughly 77 percent for women and 66 for men in the top fifth, against about 30 percent in the poorest fifth.

5:33 Smoking by education. A steep gradient from graduate degrees at the low end to GED holders at 36 percent, which he attributes to what education teaches and to the circles and resources it brings.

6:28 Infant mortality by race. In 1850, 340 per thousand Black infants and 217 White. By 2017, 11 and 5. Enormous progress for both, with the gap persisting.

7:22 What the theory predicts, and the routes: living conditions, nutrition, health care, occupations, and discrimination running from housing to schooling to the health system.

7:53 A qualification. Persistence so far does not make persistence inevitable, and it does mean nobody should be surprised to keep seeing it.

8:02 The opioid epidemic, and a map showing overdose deaths concentrated in Appalachia, the Upper Midwest, and other places where the industrial economy collapsed.

8:38 The route he describes: lost jobs, fewer opportunities, and opioids introduced through workplace injury, snowballing into mass addiction across prescription drugs, heroin, and fentanyl.

9:11 Background on the pandemic, the coronavirus family, and the Wuhan discovery in late 2019.

9:58 Transmission. Droplets, surfaces, hands to mouth, with two hedges: good evidence of asymptomatic spread and evidence of aerosolized particles that float.

11:14 Why this one is bad. Novel, so nobody is immune. Highly contagious. Asymptomatic transmission, which means the infected keep moving instead of lying still, which he notes is much better for containment. Relatively high mortality, rate unknown.

12:03 The counts as he speaks: over 160,000 deaths worldwide, about 40,000 in the United States, which he says has the worst epidemic in the world.

12:26 The inversion. This epidemic began in richer parts of the country, in coastal metros, and spread from there, which is not what the previous discussion would lead you to expect.

12:54 About 80 percent of counties have cases, with a visible pattern of a dark county surrounded by lighter ones as the disease moves from a city to its suburbs.

13:30 Concentrated outbreaks. Nursing homes, where people cannot move away, infection goes unnoticed until late, and residents are the most vulnerable. Several thousand deaths already in New York State facilities.

14:04 Prisons and jails, with the largest outbreak in the country at Marion in Ohio at 1,800 infected, and another at Cook County Jail.

14:37 Factories, with meatpacking singled out. About 800 cases at a Smithfield plant in South Dakota, another at a plant in Greeley, Colorado.

15:20 The sources of disparity. Physical concentration first.

15:31 Exposed workers. Health care staff, and service workers in restaurants, groceries, and delivery, and then the general category of anyone who cannot work from home, which he contrasts with his own position recording a lecture in his studio.

16:21 Severity once infected: asthma and breathing problems, high blood pressure, compromised immunity, and age, which accumulates all of them.

17:03 A clip from Elise Gould of the Economic Policy Institute. The CDC says stay home when sick; 27 percent of private sector workers have no paid sick leave; over 30 million workers; the United States is one of two OECD countries without a federal mandate; taking time off can mean missed bills and calling in sick can mean dismissal; the workers without it are concentrated in restaurants, hotels, transportation, and tourism, with under half of leisure and hospitality workers covered and 30 percent of the bottom tenth of earners.

18:38 His gloss: the law does not require paid sick leave, which puts workers and everyone around them at risk.

18:55 Bureau of Labor Statistics data on who can work at home, by occupation, education, and race. Managers, professionals, sales, and office workers far more than transportation, production, service, and construction. Over half of college graduates against about 5 percent of those without a high school degree. Asians and Whites most able.

19:51 The consequence: the instruction to stay home opens an immediate inequality between who complies and who goes out and takes the risk.

20:16 The Black-White difference in COVID deaths, with the caution that these are very early days for the data.

20:42 Five jurisdictions reporting deaths by race, Washington, Chicago, Louisiana, New York State, and New York City, all showing Black residents dying at higher rates.

21:14 The explanation through the theory. Prior living conditions had already produced worse baseline health, and he notes these are deaths, so most of the people counted had an existing condition.

21:38 Education and discrimination in access to jobs with benefits, flexibility, and the option to work from home, all unequally distributed.

21:52 Employer policies, including sick leave.

22:05 The health system, both the care people did or did not get before, since many have no insurance, and the hospital and doctor they reach once infected.

22:42 The close. This is a taste of one disparity, and the pattern shows the kind of thing to look for as the epidemic unfolds.

Thoughts

Three lectures in eighteen days, and the register changes each time.

April 3, on epidemiology, where he twice said he lacked expertise: a montage of presidential clips, the phrase total lie, and an instruction to enrolled students about which officials to believe. April 6, on family violence, squarely his field: no politics at all. April 20, on health disparities, also squarely his field: the most structurally radical content of the three, and not one word of partisan heat.

The lecture argues for federal paid sick leave mandate. Universal health coverage. That racial disparity in employment and benefits is discrimination. Those are contested political positions with serious opposition, and he advances all three inside a framework any medical sociologist would recognize as standard. A student leaves this lecture with a more consequential set of policy commitments than he leaves the April 3 lecture with, and without ever being told what to think about anybody.

Except once. At 17:03 he stops teaching and plays three minutes of advocacy video from the Economic Policy Institute, a labor-aligned think tank, and thanks the presenter afterward. He does not tell students what EPI is or who funds it. That is the same move as the April 3 montage with a fig leaf: a professor who selects an advocacy clip has made the argument, and the fig leaf is that somebody else’s mouth delivered it.

Cohen might say the paid sick leave numbers are correct, checkable, and drawn from BLS, and that this is what distinguishes EPI here from the Institute for Family Studies work he attacks. That defense holds for the statistics. It does not cover the framing, the policy conclusion, or the decision to let an outside organization narrate three minutes of a course. Anybody who thinks he is right about IFS should still want him to say out loud that he applies a different procedure to a think tank whose conclusions he shares.

Now the part that would matter to him more, which is that he undersold his own evidence twice by skipping the measurement work he demands from everyone else.

The infant mortality slide at 6:28 is presented as a gap that persisted. Run the numbers he put on the screen. In 1850, 340 against 217, a ratio of about 1.6. In 2017, 11 against 5, a ratio of about 2.2. The absolute gap collapsed by a factor of thirty and the relative gap widened by about forty percent. That is a stronger result than the one he stated and it is the result his theory predicts, since fundamental causes is a claim about relative advantage surviving absolute improvement. He had the better fact on the slide and told the students the weaker one.

The race and mortality slide at 20:42 has the same problem in the other direction. Those spring 2020 jurisdictional reports were crude counts and shares. The Black population in the United States is younger than the White population, and COVID mortality rose steeply with age, so age-adjusting those figures makes the disparity larger, not smaller. He shows unadjusted numbers to introductory students without saying they are unadjusted. This is the man who spent a decade arguing that a claim arrives with its instrument attached, who three weeks earlier corrected his own slide on live video when he noticed the American outbreak curve was an artifact of treating a continent-sized country as one unit. Here the caveat would have strengthened his case and he skipped it.

There is a defensible explanation, which is that this is twenty-three minutes for an introductory course and compression is the job. It is also the pattern from the family violence lecture, where the interventions section ran on some promise and mixed results with no effect sizes. His rigor is a function of whether he is prosecuting or teaching, and it runs in the direction you would not predict.

Two things here have aged well.

The prediction at 12:26 is his best public call of the pandemic and nobody has credited him with it. He notes that the epidemic inverted the usual gradient by starting in rich coastal metros, and then argues from the theory that the gradient will reassert through exposure, comorbidity, and treatment access. That is what happened. By the later waves the mortality burden had moved down the education and income distribution and stayed there. Set it beside the homicide prediction from April 6, which was wrong, and the difference is instructive: this one applies a well-tested theory to structural position, and that one was a bespoke exposure model built for the occasion. He is reliable when he reasons from the frame and less so when he improvises.

And the outbreak trio at 13:30 is the analytic core of the lecture, delivered so fast that most students would have missed it. Nursing homes, prisons, and meatpacking plants are three institutions with nothing in common except that they hold people in place who cannot leave. He names confinement as a variable, and the mechanism operates identically whether the confinement is medical, penal, or economic. That is a better sentence about the American pandemic than most of what was published that spring.

Two small things for my file.

At 16:08 he interrupts a list of workers who cannot work from home to note that he is delivering the lecture from his own studio. That is one clause, unforced, in front of students who mostly could not have caught it. It is the disclosure ethic operating without an audience to perform for, and it is the best evidence in this file that the ethic is a habit.

And the opioid map at 8:02 is the seed of published research. His 2024 work on marital status and deaths from alcohol, overdose, and suicide grows from exactly this material, and here he is teaching it to undergraduates four years earlier. The initial biography treats his research program and his teaching as separate strands. This is the place they are visibly the same strand.

YT transcript flags. Every instance of köppen, koban, covin, cope at, and Cova is COVID. At 0:00 “Philip Conn” is Cohen. At 5:52 “37% of them smoke” for graduate degrees is 3.7 percent; the gradient is roughly tenfold and the transcript flattens it to nothing, so this one will wreck the sentence if you quote it. At 4:48 “income Phipps” is income fifths. At 6:39 “1853 forty” is 1850, 340. At 6:49 “217 of the white babies” needs the per-thousand restored. At 8:08 “rotavirus epidemic” is coronavirus. At 10:39 “spread among people our age” is probably aerosols, given what follows. At 11:31 “de jure disease” is the disease. At 13:22 “the poor Daria’s first” is the poor areas first. At 19:41 “they say they couldn’t work from home” is could. At 21:22 “were already half conditioned” is have already conditioned. At 21:29 “that Charlie just showed” is that chart. The named facilities are correct: Marion Correctional Institution in Ohio, Cook County Jail, Smithfield in South Dakota, and the Greeley plant in Colorado. Elise Gould of the Economic Policy Institute is correctly identified. The 160,000 global and 40,000 American death figures place the recording on or about April 19 or 20, 2020, which matches the upload date.

‘COVID-19, preprints, and the information ecosystem’ (Jun. 8, 2020)

0:00 Introduces the talk on COVID-19, preprints, and the information ecosystem, gives his contact information, and invites people to get in touch.

0:13 The cartoon he opens with: you judge how bad things are by which specialist appears on the news. A volcanologist means trouble, an archaeologist means relax, and for this crisis a biologist sits close to the volcanologist end.

0:42 The ecosystem model, borrowed from ecology. The foundational layer holds government agencies, news organizations, and scientific publishers.

1:08 The access layer. Devices, subscriptions, paywalls, and language.

1:21 The marketplace layer, where demand is generated: people’s information needs, the production and movement of information, and its use.

2:01 The influencer network, running on individual decisions about whom to trust, both for one’s own needs and for the people relying on you.

2:23 The impact layer, and his summary that the whole thing is large, complicated, and hard to grasp.

2:45 Peer review has not always worked the way we assume.

2:53 Einstein and the Physical Review in 1936. He was offended to learn the editor had sent his paper out for review, the paper turned out to need correction, and he withdrew it.

3:28 The general case before the mid-twentieth century: editors decided what to publish and consulted experts at their own discretion. Gatekeeping and authority existed; editors held them.

3:51 The origin story he offers. After the Second World War, facing constricted research funding, academics saw that they could lose control of the process to government agencies, and the mystique of anonymous peer review arose from the insistence that only academics could decide what was trustworthy.

4:31 Watson and Crick on the double helix was not peer reviewed. The editor decided to publish it.

4:38 The cycle of scientific work: search and discovery, developing ideas, designing projects, gathering data, analyzing, publishing, and the results returning to the pool.

5:22 A qualification. Nothing published is right forever, things get superseded and added to, and the process expands.

5:42 Where journal peer review sits, between analysis and publication.

5:58 Publishers coordinate it using their own status criteria, who the researchers are and where they work, alongside the reviewers’ judgments, and the process is a black box conducted in secret.

6:17 Other evaluation points in the system: awards, grant funding decisions, and hiring.

6:51 Where open science advocates have intervened. Preregistration, which lets designs be reviewed before results exist.

7:23 Sharing materials and methods in repositories.

7:34 Publication beyond journals, including preprints, news media, and social media.

7:50 Two definitions of a preprint. The finished draft sent to a journal before review, and the accepted manuscript before typesetting.

8:25 The main servers. arXiv from math and physics in the 1990s, now over a million papers; bioRxiv and medRxiv; and SocArXiv in the social sciences, which he directs.

9:11 The pandemic explosion. A paper by Fraser and colleagues counts 16,000 COVID papers from January through April, of which 6,000 were preprints.

9:36 The follow-up finding that many were published in journals quickly with few substantive changes, which he reads as evidence they were in decent shape when posted.

9:46 Some were shared tens of thousands of times and shaped the discourse. Some turned out to be wrong, and the controversy raised the question of whether people should be reading and sharing them.

10:08 The medical sensitivity problem, and the caution notice bioRxiv and medRxiv now carry: preliminary reports, not certified, not to be relied on for clinical practice, not to be reported as established.

10:38 How journalists handle them anyway, using their own reliability indicators: who the researchers are, where they work, what outside experts say, their own reading, and newsworthiness against the cost of waiting weeks or months.

11:21 Preprints that became public knowledge. Excess deaths exceeding the official COVID count, indicating undercounting. The estimate that an earlier lockdown would have saved tens of thousands of lives, from medRxiv, on the front pages of the Washington Post and the New York Times.

12:01 A third case, the men-doing-more-housework finding, published by the Council on Contemporary Families, which does its own vetting outside both journals and preprint servers.

12:19 Retractions from the New England Journal of Medicine, The Lancet, and Annals of Internal Medicine, alongside a withdrawn bioRxiv preprint that had claimed a human-made origin for the virus on faulty genetic analysis.

12:49 His summary. Errors appear in preprints and in published papers, different systems of trust and authority are at work, and everything is moving fast.

13:00 The system runs on trust, normally formalized into predetermined authorities, prestigious journals and big universities, so that status, reputation, and legitimacy do the work.

13:27 A constant churn of deciding whom to believe, of which peer review is one part.

13:33 His verdict on the crisis. Preprints succeeded at accelerating research and getting knowledge out, and they taxed the system of trust, making explicit what is usually implicit.

13:54 Transparency as the emerging criterion. Whether researchers share data and code is now one of the ways we decide whether to believe them.

14:12 What a researcher can do. Publicly endorse work and share only what you think is reliable.

14:29 The Plaudit tool, a browser extension that lets anyone with an ORCID endorse a piece of scholarship. He has endorsed the Watson and Crick article, and he endorses a paper on pandemic dog adoptions when sharing it on Twitter. If people already trust him, his endorsement carries something.

15:14 Understand the ecosystem you work in and allocate your labor toward organizations that support open science: post preprints when ready, and decide carefully which journals to edit, review, and write for.

15:53 Closing reading list. The National Academies’ Open Science by Design, a resource for journalists on covering preprints, and his own report on scholarly communication in sociology.

This is the most theoretically interesting thing of his I have read, and he does not seem to notice.

The historical section from 2:45 to 4:38 is an argument that peer review is roughly seventy years old, that it arose from a professional turf defense, and that the two most consequential papers he names, Einstein’s 1936 submission on the gravitational wave and the Watson and Crick’s double-helix paper, went through editors, he says. The claim is broadly correct and it is more radical than anything in his blog. Anonymous peer review is a mid-century institution built when academics realized that if they did not control quality certification, somebody with appropriations authority would.

Einstein’s 1936 submission to Physical Review was sent to an anonymous referee, whose criticism led Einstein to withdraw it before a revised version appeared as “On Gravitational Waves” in 1937.

Notice what that does to his own program. If the mystique of peer review was a professional guild move dressed as epistemology, then the transparency criterion he names at 13:54 has to answer the same question about itself. Is checkability a discovery about how knowledge works, or is it the certification standard preferred by the faction that finds it easy to meet? Quantitative researchers with code and data can satisfy it at the cost of a day’s file management. Ethnographers, historians, and theorists cannot satisfy it at any cost. A standard that redistributes authority toward the people who already hold the tools is exactly what his own history predicts would emerge from a crisis in an existing certification regime. He tells that story about the generation before him and does not turn it on himself. That is the essay somebody should write, and he handed over the materials.

The Plaudit passage at 14:29 is the strangest four minutes in this file. Follow what he proposes. Peer review is a black box run by publishers on status criteria. The replacement he demonstrates is a browser extension where he clicks endorse and then says that if you already trust him for other reasons, his tweet might mean something to you.

That is status criteria with the intermediary removed. The anonymous referee at least did not benefit personally from the paper’s reception. A named endorser trades on his reputation, accrues from being seen to endorse, and is embedded in a network of people he will need favors from. He replaced a black box with a visible reputation market and described the visibility as the improvement, which it partly is, and never addressed the thing that makes reputation markets go wrong, which is that they reward endorsing the people who can endorse you back. A man who spent his career pointing out that prestige networks distort what gets believed just proposed prestige networks as the correction.

He also endorsed Watson and Crick with a browser extension in 2020 and offered it as a demonstration. The joke is good. It also quietly concedes that the tool has no way to distinguish an endorsement that costs the endorser something from one that costs nothing.

Two things here are sharper than his usual work.

The Fraser finding at 9:36 is the best empirical defense of preprints anybody made that year. If work posted before review appears in journals shortly after with few substantive changes, then review added little to those papers, and the months of delay bought the reader nothing. That is a testable claim about the value added by an institution, of the kind he demands from everyone else, and it comes with a citation.

And the 12:19 slide is the honest move. He puts preprint failures next to retractions from the New England Journal of Medicine, The Lancet, and Annals of Internal Medicine, which is the right comparison and the one an advocate would have skipped. The Surgisphere retractions from The Lancet and the NEJM had happened days before this recording, and both journals had cleared the same peer review he was being asked to defend as the safeguard preprints lack.

Two weaknesses.

The ecosystem model from 0:42 to 2:23 does no work. Five layers, no argument, no prediction, and nothing later in the talk uses it. Strike it and the talk loses nothing, which is the test he would apply to somebody else’s framework.

And the accuracy question goes unanswered. He says at 9:46 that some preprints turned out to be wrong and at 12:49 that some published papers did too, and he never tells you the rates. Whether preprints are wrong more often than reviewed papers is answerable, at least roughly, and it is the question anyone watching would want settled. He has the data literacy to know that the presence of errors on both sides is compatible with any ratio between them. This is the counting he does not do when the answer might be inconvenient, and it is the same pattern as the interventions section in the family violence lecture.

For my file, this is where the archive stops being an administrative project and becomes an argument about what certification is. The biography covers SocArXiv through its founding, the prize, and the AI crisis, and the crisis reads as an accident that befell a working institution. This talk shows the position was always there: certification is a historically contingent arrangement that some professionals built to keep control, and preprints tax the trust system by making the implicit explicit. Six years later he wrote a policy conceding that his moderators cannot state the rule and will use human judgment. The taxing of the trust system turned out to have no floor, and he described the direction of travel in June 2020 while assuming it stopped somewhere convenient.

One line to carry forward: at 13:33 he says the crisis made explicit what is usually implicit. That is the whole of his career in eight words.

Transcript flags. Kobe, kovat, COBIT, koban, Cobin, and club in nineteen are all COVID-19. At 8:48 “social kind” and “sai archive” are SocArXiv. At 8:26 “the major print print” is preprint. At 12:09 “bedding” is vetting. At 12:53 “systems of crust” is trust. At 11:52 “a pre current published” is a preprint. At 14:32 “the plaudit tool” is Plaudit, correctly named. “Orchid ID” is ORCID. At 15:44 “review right for” is review and write for. The Fraser paper is Fraser et al. on the preprint literature during the pandemic; the version circulating in mid-2020 was a preprint.

A smoothed transcript: 2:45, on peer review as a historical artifact. “Now, it’s important to realize, when we talk about peer-reviewed science and the way we assume science works, that it has not always been this way. There’s a famous interaction between Albert Einstein and the journal Physical Review, where he was extremely distressed to discover that the editor of the journal had sent his paper to somebody else to review it before deciding to publish it. It turned out that this particular paper was wrong and had to be corrected. But Einstein was offended, incensed: on the basis of this incident, I prefer to take my paper elsewhere. He hadn’t authorized it to be shown to anybody else.

“So that’s 1936. Peer review, and especially anonymous peer review, was completely offensive to Einstein. Of course, he’s Einstein. But it’s the general case that before the middle of the twentieth century, journal editors were the ones who decided what to publish, and they consulted with experts at their own discretion. Which doesn’t mean there was no gatekeeping, and it doesn’t mean there was no authority and no expert review. It was editors who were in charge of it. They decided who would review what, and they were the gatekeepers.

“Then, after World War II, as there was the threat of the constriction of research funding, academics realized that they could lose control of this process if it was handled instead by government agencies, without their direct influence. That’s really when we see the mystique of peer review come up, where academics insisted that only academics could decide what was valuable. That’s when we got this idea of the anonymous peer review system, and especially the idea that this was essential for science to be considered trustworthy.

“There’s a good example. The Watson and Crick paper on the double helix structure of DNA. Not peer reviewed. The editor just decided to publish it.”

5:42, on where review sits and what else does the same work

“The peer review process as we know it in journal publishing happens at that point between the analysis of the data and the publication of the results. That’s when we bring in these expert peers, sometimes anonymous. The effort is usually coordinated by publishers, especially journal publishers, and they’re going by their own status criteria: who are these researchers, are they reputable people, where do they work. And then they’re also consulting their peer reviewers. The process is very black box. We don’t know exactly how it works. It all happens in secret.

“Now, there are other places where evaluation goes on in this system that we don’t think of as peer review in that specific sense. After results are published they can be elevated; they can be given awards. We have grant funding agencies that evaluate research ideas and whether or not to fund them. We have people who decide who’s going to get jobs, and which are going to be the important jobs. So there’s a lot of peer evaluation that goes on in the system besides that moment of publication we focus on.”

7:50, what a preprint is

“There are two main definitions of preprints that are relevant. One is the finished draft of scholarly work, not yet peer reviewed. Think of the version of the work that scientists are sending off to a journal to be reviewed. Another definition is the version after it’s been reviewed and accepted, but before it’s been typeset and published by the journal. People sometimes share that, and that can also be called a preprint. So preprints are essentially before the typeset version that comes from a journal, at some stage of development.”

“The major servers: arXiv, the first big one, out of math and physics, started in the 1990s, now well over a million papers. bioRxiv and then medRxiv in the life sciences and medicine. In social science, SocArXiv, which I direct. And a number of other discipline-specific archives working on similar principles, archiving and disseminating preprints more rapidly and openly than traditional journals.”

9:11, the pandemic volume, and the Fraser finding

“In the current pandemic environment there’s been an explosion of research, and much of it in preprints. This paper by Fraser et al. finds that in the first four months, January through April, of sixteen thousand papers published having to do with COVID-19, six thousand were preprints. So a huge volume of work, pre-peer-review, coming out available to the public.

“They also did some follow-up and found that a lot of these studies are published quite quickly in journals with few substantive changes, indicating the preprints were in pretty good shape when they were posted. Some of these preprints have been shared thousands of times, or tens of thousands of times, and really helped shape the discourse on COVID-19. Some of them have turned out to be wrong, and there’s been controversy with some of them, and that’s part of what’s raised the issue of whether we should be reading preprints and talking about them and sharing them.”

10:08, the warning label, and what journalists do instead

“Because we’re dealing with medicine, and people are very sensitive about it, you’re going to give clinical advice, you can talk about specific medications, doctors might be looking at these. medRxiv and bioRxiv, which are run by the same organization, are putting this caution right on their homepage: preprints are preliminary reports that are not certified, they should not be relied on to guide clinical practice, and they should not be reported as established information.

“But preprints are all over the news. So how is that happening? Journalists and news organizations are using their own indicators of reliability. Who are these researchers, where do they work. They might talk to other experts who are not involved in the research and say, what do you think, is this reasonable, should we write a story about this. They have their own assessment. The science journalists and reporters and editors have their own expertise, and they can read and decide what to trust. And then there’s the newsworthiness. How important is it? Do we need to get this out there right now, instead of waiting weeks or months for it to be published in a journal?”

11:21, preprints that became public knowledge

“Some of the things you’ve probably heard that are now established knowledge in the public sphere around COVID-19 have come from preprints. For example, the estimates that the number of deaths going on in the country far exceeds the number counted by the official statistics, indicating we were undercounting infection-related deaths. Page one of the Washington Post. Page one of the New York Times: if we had done the lockdown sooner, tens of thousands of lives would have been saved. That’s also from a preprint, published on medRxiv. So these things have been tremendously influential, and lots of people know about them.

“Here’s another one. Men are doing more housework, say researchers. This one is not on a preprint server. It’s published by the Council on Contemporary Families, which does its own vetting of unpublished research. They publish it themselves, but not in a journal, and not with a formal peer review process.”

12:19, errors on both sides

“Now, some of these preprints have been wrong. But then again, some of the published articles have been wrong. Here are some retractions from major journals, the New England Journal of Medicine, The Lancet, Annals of Internal Medicine, which turned out to be wrong and had to be retracted after getting a lot of attention. I include here one bioRxiv preprint that was withdrawn after suggesting that the coronavirus was made by humans, because of a faulty analysis of the genetics of the virus.

“So there have been errors in preprints and among published papers, and we have different systems of trust and authority, and the whole thing is moving very rapidly.”

13:00, the passage the essay turns on

“So what do we trust? What do we believe we know? This whole system runs on trust. Normally it’s formalized, and we have authorities that are predetermined to be things we should trust: Nature and other important journals, and big universities, and so on. We use the status and the reputation and the legitimacy of the people who did the study, and the people who published the study, and the people who report on the study. There’s a constant churning process of deciding who we’re going to believe. Peer review is just one part of that.

“This crisis has really shown that preprints have been tremendously successful at accelerating the pace of research and the efficiency of getting that knowledge out there. And it has really taxed the system of trust, and made more explicit part of what’s usually implicit. One of the things that we now look at is transparency. One of the ways that we evaluate whether or not we should believe a piece of research is: are they sharing their data and code and so on. Have they made their process transparent. If they have, that indicates that they’re more trustworthy.”

14:12, the replacement he proposes

“I want to talk a little bit at the end here about what you can do if you are a researcher, who works not just on this but on anything. One thing you can do is take steps to publicly endorse work, and make sure you only share work in your networks that you think is reliable. For example, the Plaudit tool. I really like it. It’s a browser extension you can install, I use Chrome, where if you have an ORCID ID you can endorse any piece of scholarship. You can see here I have endorsed the Watson and Crick article. And also, if I’m going to share a paper on Twitter, here’s a paper I shared about dog adoptions during the pandemic, I’m giving it my endorsement. If you trust me already for other reasons, then my tweet might mean something to you.

“So if you’re a researcher, I just urge you to spend some of your time doing the work of publicly endorsing things that you think are good.

“I also think it’s important for us to try to understand this ecosystem that we work in, and then to make decisions about how we allocate our own labor to support those organizations and institutions that support open science and support our values. To practice open science ourselves. To publish preprints when they’re ready, to help accelerate and transmit knowledge out to other researchers and the public. And to decide who we’re going to edit journals for, review for, write for. To think of ourselves as part of this system, and try to push the system in a positive direction while we do our work.”

Notes on the editing

Removed “sort of,” “you know,” “um,” and repeated false starts. Repaired the speech-recognition damage: Kobe, kovat, COBIT, koban, Cobin, and club in nineteen all became COVID-19; “social kind” and “sai archive” became SocArXiv; “the major print print” became preprints; “bedding” became vetting; “systems of crust” became trust; “a pre current published” became a preprint; “review right for” became review for, write for; “orchid ID” became ORCID. Joined sentences he left hanging. Nothing was added, no order was changed, and where a phrase was ambiguous I kept his looser version.

At 3:11 he delivers Einstein’s line as though quoting: on the basis of this incident, I prefer to take my paper elsewhere. The documented letter of July 27, 1936, reads that on the basis of this incident he prefers to publish the paper elsewhere. His rendering is close and not exact. If you are quoting Cohen, use his words as smoothed above. If you are quoting Einstein, use the letter, which is reproduced in the gravitational-wave histories I sent with the peer review essay. Do not let the two blur, since the whole point of that passage is what counts as a faithful record.

One flag on the Council on Contemporary Families passage at 12:01. He presents CCF as a third model, an organization that vets and publishes outside both journals and preprint servers, without saying whether he has any relationship to it. CCF has published his work and he has been associated with the organization.

What I cut and why. The ecosystem model (0:42 to 2:38) is five labeled layers with no argument attached, and nothing later in the talk uses it. The virus background (9:11 area, in the other lecture’s language) duplicates material he covered better on April 3. The closing reading list is three citations, all recoverable from the slides at https://osf.io/5mue7/. The retraction slide carries numbers that never made it into the audio.

If you want a verbatim passage rather than a smoothed one, the two places where the exact wording is likely to matter are 3:51, the after-World-War-II turf-defense sentence, and 13:33, the taxing of the system of trust.

In the early summer of 1936, Albert Einstein (1879–1955) and his assistant Nathan Rosen (1909–1995) sent a paper to Physical Review arguing that gravitational waves do not exist. The editor, John Torrence Tate (1889–1950), had published Einstein before without sending the work to anybody. This one struck him as suspicious. He mailed it to Howard Percy Robertson (1903–1961), a relativist at Princeton, who read it and sent back ten typed pages explaining that the authors had made a mathematical error.

Tate forwarded the report and asked for a response. Einstein wrote back that he and Rosen had submitted the manuscript for publication and had not authorized anyone to show it to specialists beforehand, that he saw no reason to address the comments of Tate’s anonymous expert, and that he preferred to publish elsewhere. Tate answered that he could not print a paper the author would not let his editorial board see. Einstein took it to the Journal of the Franklin Institute and never submitted to Physical Review again.

What happened next is the part that gets left off.

Rosen went to the Soviet Union. A young Polish physicist named Leopold Infeld (1898–1968) arrived at Princeton to replace him. Robertson, back from a sabbatical in Pasadena and by all accounts sociable, befriended Infeld, got him talking about what Einstein was working on, and expressed doubt. The two of them went through the manuscript together. Infeld took the objection to Einstein. Einstein revised the paper before returning his galleys, reversed its conclusion, changed the title, and thanked Robertson in a footnote for friendly assistance in clarifying the original error. He never learned that Robertson had been the referee.

The correction arrived. It traveled through a man who happened to be in Princeton that year and happened to like talking to graduate students. The institution built to carry it had been refused at the door.

Philip Cohen tells the first half of that story in a lecture he recorded in June 2020 on preprints and the information ecosystem, and he uses it to make a claim that is larger than his listeners probably noticed. He follows it with Watson and Crick on the double helix, which no referee read; the editor decided to publish. Then the general point. Before the middle of the twentieth century, journal editors decided what to publish and consulted experts when they felt like it. Gatekeeping existed. Authority existed. Editors held both. And then, he says, after the Second World War, facing the prospect that research funding decisions might be made by government agencies without academic control, academics insisted that only academics could judge what was true, and the mystique of anonymous peer review was born.

That is an argument that the central credentialing procedure of modern science is a professional turf defense that later acquired an epistemology. He states it in ninety seconds, in a lecture for undergraduates, and moves on.

The history is right in structure and wrong in date, and the correct date makes his case stronger.

Melinda Baldwin, a historian of science who teaches at his own university, published the definitive account in Isis in 2018, “Scientific Autonomy, Public Accountability, and the Rise of ‘Peer Review’ in the Cold War United States.” External referee reports were an optional feature of journal editing and grant making through the nineteenth century and most of the twentieth. The idea that refereeing is a requirement of scientific legitimacy arose in the Cold War United States, and the decisive moment came in the 1970s. American science was under sustained congressional attack over how it spent public money. Scientists needed to appear accountable to the people funding them while retaining control of funding decisions. The term peer review, which had been migrating out of grant administration, was pressed into service as the thing that made science credible as such. Nature did not require formal review of everything it published until 1973. The Lancet came later still.

So the sequence is: an existing certification regime comes under external pressure, the profession selects a standard that answers the pressure while keeping the decision rights inside the profession, and the standard is then described as constitutive of knowledge.

Cohen tells that story about the generation before him. At the end of the same lecture he proposes the successor standard and does not turn the story on it.

The passage runs about twenty seconds. The crisis has taxed the system of trust, he says, and made explicit what is usually implicit, and one of the things we now look at is transparency. Are they sharing their data and code. Have they made their process inspectable.

Ask of that standard the question Baldwin asks of the last one. Who can meet it, at what cost, and who decides whether it has been met.

A quantitative researcher working with public survey data can meet it for the price of a day’s work. That figure comes from Cohen’s own comment section, where a sociologist who had published in a journal with a mandatory policy reported that assembling his first replication package took a full day despite tidy file management, and that most people default to available upon request because collation for a stranger is labor. One day, once, and then the habit is cheap. Every subsequent paper is a matter of directory hygiene.

An ethnographer cannot meet it at any price. The field notes are the instrument, the consent forms usually forbid release, and even redacted material carries the risk of deductive disclosure in a neighborhood where everyone knows who the researcher followed for two years. An archival historian cannot meet it either. The archive is the data, and it sits in boxes in a building, often behind permissions granted to individuals, and citation to box and folder is the most a reader will ever get. An interview researcher who promised confidentiality is barred by the promise. A demographer working inside a Census research data center is barred by federal law. A theorist has nothing to deposit.

That is a standard whose compliance costs run from one day to infinity depending on what kind of evidence a field is built out of, and it arrived in a discipline whose quantitative wing had spent thirty years gaining ground on its qualitative wing.

Political science ran the experiment first and the record is public.

The American Political Science Association wrote data access and research transparency into its ethics guide in 2012. In October 2014, twenty-seven journal editors signed a joint statement committing to require, by January 15, 2016, that cited data be available in a trusted digital repository at the time of publication, with exceptions for restricted material at editorial discretion. The following autumn a petition circulated asking for a delay. It gathered more than seven hundred fifty signatures within days and close to a thousand within a week, including senior figures and sitting editorial board members.

The signatories did not say they wished to be believed without evidence. Their complaint was about discretion. The rules were clear enough for people with datasets and unspecifiable for everyone else, which meant that ten editors could produce ten interpretations, and a scholar planning a project could not know whether a given kind of fieldwork would close off a set of journals. The standard did not tell them what compliance was. It told them who would decide.

The association’s past, current, and future presidents answered with a joint statement supporting implementation, and the ground they chose is the one Baldwin would predict. Their legitimacy as scholars speaking to the public, they wrote, rests partly on maintaining common standards for evaluating evidence-based knowledge claims. External legitimacy, purchased with an internal certification rule, in the middle of a decade when the social sciences were being asked what their findings were worth. That is 1975 with different equipment.

The deadline held at some journals and slipped at others, and the association’s qualitative and multi-method section opened a multi-year deliberation to work out what transparency could mean for research that cannot deposit anything. The deliberation produced sensible documents. The distributional question it was convened to settle has not been settled anywhere.

Now the counter-case.

Showing that a standard has winners does not show that it is false. Peer review’s origins in a guild’s anxiety about congressional appropriations tell you nothing about whether referees catch errors, and Robertson caught a real one. In the same way, the fact that checkability favors people with spreadsheets does not make checkability arbitrary. There is a difference between a claim a stranger can test and a claim he must take on the author’s word, the difference is not a matter of taste, and every discipline that has looked hard at its own literature in the past fifteen years has found things that a stranger’s inspection would have caught earlier.

The second point. If your evidence cannot be inspected, that is a fact about your evidence. It may be an unavoidable fact, forced by ethics and by law, and it still bears on how much weight a reader should place on your conclusion. A discipline that answers the transparency demand by saying its methods are different has said something true and has also said trust me, which is the position everyone occupied before 1970 and which produced a literature nobody can now audit.

And the third point runs against Cohen’s own side. Compliance costs fall hardest on people with the least slack. A day of file management is a day for a tenured professor with course relief. It is a week that does not exist for an adjunct teaching five sections, and the standard has no way to see the difference.

Peer review answered an accountability crisis with a rule that kept the deciding inside the profession. The test of any successor is whether it moves the deciding outward, and this is where Cohen’s program deserves better than the easy debunking, because on its face it does exactly that. A replication package is not an opinion delivered by a credentialed stranger to an editor in confidence. It is a public artifact. Anyone can download it. The authority to check has been handed to whoever wants it. That is a real answer to the guild problem, and nobody else in his discipline proposed one.

Except that the thing that got adopted is not checking.

What his audits count is whether a package exists. He said so in 2020, that he had assessed them at a glance and had not tested anybody’s code, and he has never claimed otherwise. The proportion of quantitative articles in the discipline’s flagship journal arriving with packages went from four of fifteen to twelve of eighteen across five years, and what that number measures is documentation. Nobody has run the code. Nobody has recomputed the tables. The public who received the authority to check has not exercised it and shows no sign of intending to.

Which leaves a standard that certifies compliance with a documentation regime, adjudicated by journal editors and by a professor clicking links on a December afternoon. The outward transfer of authority is real in principle and dormant in practice, and while it sleeps, the rule functions the way Baldwin’s model says these rules function: as a gate, operated by people whose evidence happens to fit through it, carrying the language of universal accessibility.

Then documentation became free to produce.

In the winter of 2025 Cohen’s archive began receiving papers with everything in place. Abstracts, real citations to real work, competent models, code in public repositories, and disclosure statements listing every tool used to generate the analysis and draft the text. One of them he described in detail because it would have been accepted a few years earlier. Dull, unoriginal, thin in its literature review, empty where it claimed a contribution, and immaculate on every checkable dimension.

A standard that certifies the existence of an audit trail becomes worthless at the moment audit trails cost nothing. That is what happens to any proxy once the thing it stood for stops being expensive.

The policy his committee published in March 2026 refuses automated detection, declines to state a rule that could be written down completely, commits the moderators to judgment made by people using perception and experience, and invites appeals. A small group of trusted individuals, embedded in a community, deciding what to publish on criteria they cannot fully specify, sometimes wrong, answerable to nobody but their own standing.

That is John Tate at his desk in 1936.

The narrative Cohen told his students in June 2020 ran from editorial discretion, through a guild’s invention of peer review under political pressure, to transparency as the successor standard. He was describing a sequence he thought he was at the end of. Six years later the third term collapsed and the first term returned, and the whole circuit closed inside a single career.

The counting moved a journal that three years of committee work had failed to move, and a literature with code in it is better than a literature without. What the history says is narrower and harder to shrug off. Every certification standard is a settlement about whom to believe, arrived at by people with something at stake, under pressure from outside, and every one of them gets described afterward as a discovery about the nature of knowledge. He can see this clearly when the subject is the generation that trained his teachers. The materials for seeing it about his own generation are on his own slides.

Robertson’s ten pages never persuaded Einstein of anything. What persuaded him was a man he liked, at his own institution, talking to his assistant over the course of an autumn. Every certification regime since has been an attempt to make that transmission work without depending on who happens to be in Princeton that year. None has managed it. Each has had to decide, in the attempt, whose evidence counts as evidence.

Notes

The Einstein episode. Daniel Kennefick’s research, first published in Physics Today in September 2005 and developed in Traveling at the Speed of Thought (Princeton, 2007), is the source for Robertson’s identification as the referee. The submission was “Do Gravitational Waves Exist?”, sent to Physical Review in June 1936; Tate wrote to Einstein on July 23 with the referee report; Einstein replied on July 27. Verbatim quotation used: “comments of your anonymous expert,” from that letter, which is reproduced in full at https://arxiv.org/pdf/1609.09400 and https://arxiv.org/pdf/1610.08803. Tate’s reply of July 30, that he could not publish a paper the author was unwilling to have shown to the editorial board, is in the same sources. The Infeld and Robertson sequence, the reversal of the conclusion, the retitling to “On Gravitational Waves,” and the acknowledgment thanking Robertson are documented in both arXiv histories and in https://www.science.org/doi/10.1126/science.347.6226.1085. The detail that Tate had published Einstein, Podolsky and Rosen in 1935 without refereeing is from https://www.discovermagazine.com/the-sciences/einstein-vs-physical-review.

The history of peer review. Melinda Baldwin, “Scientific Autonomy, Public Accountability, and the Rise of ‘Peer Review’ in the Cold War United States,” Isis 109, no. 3 (September 2018): 538–558, https://www.journals.uchicago.edu/doi/abs/10.1086/700070, with a copy at https://gwern.net/doc/statistics/peer-review/2018-baldwin.pdf. Her earlier “Credibility, Peer Review, and Nature, 1945–1990,” Notes and Records 69 (2015), and John Burnham’s 1990 JAMA piece on the evolution of editorial peer review, are the standard companions. Nature dated at 1973 and The Lancet later; the term appearing in JAMA in 1970; Colin Norman’s “Peer-Review under Review” in Nature in 1975. She is on the faculty of the University of Maryland’s history department.

DA-RT. The joint statement of twenty-seven political science journal editors, October 2014, with the January 15, 2016 deadline and the restricted-data exception. The APSA ethics guide change dates to 2012 and the initiative to an ad hoc committee formed around 2010. Petition coverage at https://www.insidehighered.com/news/2015/11/16/political-scientists-seek-delay-transparency-standards-publications, with signature counts at https://dialogueondartdotorg.wordpress.com/faqs/ (more than 750 as of November 8, 2015) and https://politicalbehavior.wordpress.com/2015/11/11/statement-on-da-rt-and-the-jets/ (almost 1,000 by November 11; the organizers were Hirschmann, Htun, Mansbridge, Thelen, Wedeen, and Wood). Verbatim quotation used: “common standards for evaluating evidence-based knowledge claims,” from the joint statement of the APSA past, current, and future presidents, November 24, 2015, linked from https://www.dartstatement.org/2014-journal-editors-statement-jets. The Qualitative Transparency Deliberations are at https://qualtd.net/page/about/, and the discretion objection I paraphrase is from a contributor post at https://www.qualtd.net/viewtopic.php?t=67.

Cohen material. The June 2020 lecture, “COVID-19, preprints, and the information ecosystem,” slides at https://osf.io/5mue7/, for the Einstein and Watson-Crick examples, the postwar turf-defense account at 3:51, and the transparency criterion at 13:54. The 2020 audit with the glance caveat, the December 2025 audit with the counts, and the commenter’s day of file management are all linked in my earlier notes. The AI submission and the March 2026 policy are at socopen.org.

The concession structure, which is the essay’s exposed flank. Three counter-arguments run before the payoff: that genealogy is not refutation, that a field’s inability to share is a fact about its evidence, and that compliance costs fall hardest on the least resourced. A hostile reader will say the concessions are placed to be survived. The honest answer is that the first one is decisive and the essay says so, and that the argument turns on the narrower claim about who adjudicates. If you want the essay less exposed, strengthen the second concession, since it is the one most likely to be true.

The claim that is mine. That what got adopted is documentation, that the public who received the authority to inspect has not used it, and that the standard therefore operates as a gate while carrying the rhetoric of universal access. Cohen concedes the first part of that (he has never claimed to run the code) and would dispute the conclusion. The strongest reply available to him is that a dormant right to check disciplines behavior even when unexercised. I have not found anyone who has tested it.

The audit measures a regime whose effects do not depend on the regime being used. Tax compliance runs far ahead of the audit rate. Agencies write differently under freedom of information statutes whose requests mostly never come. A firm’s filings are shaped by a short-seller who never appears. If that holds for replication packages, then counting deposits is counting something real, and my complaint about dormancy is a complaint about the wrong variable.

The trouble with the reply is that deterrence is doing four different jobs in it, and they have different requirements and different failure conditions.

The first is a forcing function that needs no inspection at all. To assemble a package you have to run the pipeline from raw data to published table in one pass, and that pass finds errors whether or not anybody else ever looks. Cohen’s commenter described this without naming it: a full day of work for a first package, despite tidy file management. Much of what takes a day is discovering that your own directory does not do what you thought it did. This version of the reply is the strongest, because it operates at a zero inspection rate, and it is the one Cohen’s audits are aimed at, since what they count is whether the package was built.

The second is deterrence proper, which needs a nonzero perceived chance of being caught, and which targets choices made before any file exists: the outlier you would have dropped, the specification you settled on after seeing which one worked.

The third is conditional inspection, and it is the strongest structural point in Cohen’s favor. Nobody checks at random. People check when a finding is surprising, contested, or load-bearing for something. So the unconditional rate at which packages get downloaded badly understates the discipline, because the probability of inspection is high exactly where a wrong result would do damage and near zero where it would not. Michael LaCour’s fabricated study was exposed because David Broockman obtained the data and found that it matched an existing survey too closely. The right was dormant across the whole literature and got exercised at the one point that mattered.

The fourth is norm signaling, which does not require inspection either, and which is probably how most compliance with most rules works.

Now the evidence, which exists, and which separates these cleanly.

In March 2015 the American Journal of Political Science adopted third-party verification. Every accepted quantitative manuscript goes to the Odum Institute at Chapel Hill, where staff run the code and compare the output to the tables in the paper before the article is published. Authors know this. They know it with certainty, they know it will happen within weeks, and they know publication waits on it.

By February 1, 2018, Odum had processed 149 manuscripts and completed 127. Of those 127, eight reproduced all the analyses on the first attempt. Six percent. The mean number of resubmissions was about 1.8 and the maximum was five. A later account gives a pass rate around ten percent, and the process adds fifty to sixty-five days to publication.

That number kills the forcing-function version of the reply. These are authors operating under near-certain inspection by a named institute on a known timetable, which is the most favorable condition the argument could ask for, and ninety-four percent of them handed over packages that did not run. Whatever anticipatory self-auditing the deposit requirement induces, it does not survive contact with a stranger’s machine. If it fails under certainty, it is not doing much under dormancy.

The economics numbers point the same way from a different angle. When Kingi, Vilhuber, Herbert, and Stanchi went through 303 papers in American Economic Journal: Applied Economics, ninety-eight of them, about a third, relied on confidential or proprietary data and could not be reproduced by an outsider at all. Of the eligible attempts, roughly forty-three percent fully reproduced the analysis and another forty-two percent partly did. That corpus was published under a data availability policy the American Economic Association had held since 2005. Fifteen years of a deposit requirement, and under half of the checkable papers checked out.

So the deposit is a weak proxy for the thing Cohen cares about, and the gap is measurable, and it has been measured, and the measurement was not done by him.

What the AJPS record also shows, and this is the part that cuts against me, is that the requirement does enormous work anyway. The errors got found. They got fixed before publication, without a retraction, without a correction notice, without anybody’s reputation entering the transaction. That is close to the ideal case for research correction and it is invisible in every count of retractions, including Cohen’s count of one this century.

But look at what produced it. Not the right to check. Somebody checking. Funded staff, fifty to sixty-five days per article, a five-hundred-thousand-dollar Sloan grant in 2018 to automate part of the workload. The improvement came from labor, and the labor is the thing that does not happen when a package sits in a repository with a right attached to it.

Two things survive for the deterrence reply, and I want to state them because they are not weak.

The AJPS numbers refute the forcing function and say nothing about deterrence proper. A package that fails to run is evidence about file hygiene. It is not evidence about whether the author dropped an inconvenient case, and it could not be, because those decisions leave no residue in the deposited files. The archived script documents the path taken, never the search. Whatever restraint the possibility of inspection exerts on specification choices is not something a verification failure rate can measure, in either direction.

And the conditional-inspection point stands untouched. A regime under which most packages are never opened and the consequential ones reliably are is a good regime, possibly better than random auditing, and its performance cannot be assessed from the average.

What would settle it is three studies, none of them expensive.

Compare AJPS packages deposited before March 2015 with packages deposited after, re-run today by a third party. If the post-policy packages verify at higher rates years later, the regime changed author behavior durably. If the improvement disappears once you account for the Odum round-trips, then verification worked and the requirement by itself did nothing. Somebody proposed this in 2019 and I cannot find that it was completed.

Compare journals that require deposit without verification against journals that verify. Sociology, political science, and economics now sit on opposite sides of that line, and the comparison is available.

And look at the download logs. Dataverse and openICPSR know how often each package has been retrieved and by whom. The dormancy rate is not a matter for speculation. It is sitting in server logs at institutions that would hand it over to a researcher who asked.

Which brings the whole thing back around on Cohen. He counts a proxy. The relationship between his proxy and the outcome he cares about is measurable. It has been measured in two neighboring disciplines and never in his own. And the answer his own standard would demand is one sociology journal paying somebody to run the code, which costs money and two months per article, and which nobody in sociology has proposed.

As of February 2018, three years into the policy, the journal reported that it had accepted no manuscripts with qualitative analyses since the policy took effect. There are other possible explanations, including submission mix and editorial preference, and it is a single journal. It is also the cleanest available datum on the distributional question, and it points where the petitioners said it would.

Sources

AJPS verification figures from the journal’s own policy document, “The American Journal of Political Science Replication and Verification Policy,” dated February 27, 2018: https://ajps.org/wp-content/uploads/2019/03/ajps-replic-and-verif-policy-2-27-18.pdf. Source for the March 26, 2015 announcement, the 149 processed and 127 completed manuscripts, the eight first-attempt successes, the mean of 1.8 resubmissions and maximum of five, and the sentence about no qualitative manuscripts having been accepted. The roughly ten percent pass rate and the division of labor between Odum and the Qualitative Data Repository at Syracuse are at https://wholetale.readthedocs.io/en/stable/project/cases/political-science.html, which also notes the 2018 Sloan grant of five hundred thousand dollars. The fifty to sixty-five day publication delay is from a conference report at https://www.csescienceeditor.org/article/science-editor-symposium-reproducibility-reporting-guidelines/. The workflow case study is Thu-Mai Christian et al., “Operationalizing the Replication Standard,” International Journal of Digital Curation 13, no. 1 (2018), https://ijdc.net/article/view/555.

Economics figures from Kingi, Vilhuber, Herbert, and Stanchi, “The Reproducibility of Economics Research: A Case Study,” working paper at https://sylverieherbert.github.io/files/Replication_aejae.pdf, published as Herbert, Kingi, Stanchi, and Vilhuber, “Reproduce to Validate,” Canadian Journal of Economics 57, no. 3 (2024): 961–988. The eligible-attempt percentages in the working paper text are ambiguously worded and should be checked against the published version before you print them. The AEA’s data availability policy dates to 2005, and prepublication verification under the Data Editor began in 2019: https://aeadataeditor.github.io/.

The LaCour case is the standard reference for consequential inspection of an available dataset. Broockman, Kalla, and Aronow’s May 2015 report is the primary document.

The proposed pre- and post-policy comparison at AJPS is described at https://wholetale.org/internship_projects/2019-analyzing-ajps.html. I could not find a published result and would not assume there is none without a proper search.

I have not checked whether anybody has published dormancy rates from Dataverse or openICPSR logs. If nobody has, that is a small original paper sitting unclaimed, and it is the direct empirical answer to the question you asked. It is also the sort of thing Cohen could do in an afternoon and has not.

‘Families, Inequality, and Sociology in Pandemic Times’ (Aug. 1, 2020)

0:04 Opens with the discipline’s premise: sociology looks for where macro and micro meet, where biography meets history, and much of that happens in families. Families are where social structure gets internalized into personality early in life.

0:24 The example that carries the talk. People marry when they have confidence in the future, and many have children when they feel able to shape their own destiny. If people are being laid off and housing, health care, and education are out of reach, they will not make long-term commitments.

1:07 The Gini index and half a century of rising American inequality.

1:19 Runaway wealth at the top. He grants that Bezos doubling his fortune does not affect him personally, and argues the aggregate effect corrupts other parts of the economy, since a few firms set wages and conditions for millions.

1:46 The formulation: the richest people make the rules for themselves and everyone else works around it. We want Gates to solve malaria because he has more money than many governments, which would be good, and it is unfortunate that this is where we ended up.

2:04 The weak social floor. How far people are allowed to fall, and the toll even on those whose necessities are currently met, since losing them easily wears on health, anxiety, and fear for children.

2:25 The parenting consequence. Do you let your children be whoever they want, or do you fear they will land on the wrong side of the gap.

2:43 Self-efficacy again, the sense of controlling your own destiny, which he says modern society prizes and which shapes every decision and relationship when absent.

2:55 The decline in marriage. When women’s earnings and education accelerated from the 1970s, some feared feminism would end the family, which he says now seems quaint.

3:16 The arithmetic under the fear. In the 1950s more than 80 percent of Americans married before 25, so marriage was near universal and there was nowhere for the rate to go but down once opting out became possible.

3:38 What nobody anticipated: richer women marrying more. He attributes it partly to better economic prospects among their available partners, and to men increasingly judging partners by earnings.

3:54 Why poorer people and African Americans marry less. He calls it an unbalanced gamble, a long-term commitment in a universe offering no reason to think long-term decisions will be reciprocated by reality.

4:09 Middle-class anxiety, driven by college costs and fear that children will end up broke, which he says contaminates opportunity with fear and produces cynicism, materialism, and a duty to keep children on track that limits their range.

4:31 The other side. People look for a realm they can control, and families are part of that. The haven in a heartless world, which he attributes to an old book. Emotional connections over material principles, pulling together, trusting each other. That is the ideal and some get near it.

4:57 How it is achieved lower down: extended family, especially among immigrants, fictive kin, informal care networks, people watching each other’s children.

5:24 The correction he stops to make. He does not want race collapsed into class. Take two college graduates with mid-level incomes, one White and one Black. Because of history and context the Black graduate is likelier to have poor relatives and friends, so people come knocking for help, and likelier to be less secure and to face racism from work relationships to policing to exclusion in public. Track only class and you miss it.

6:06 Impression management, part of the human condition, and heightened on social media.

6:12 The connection back. When 80 percent married before 25 you did not have to justify the arrangement. More choices mean justifying them to others and to yourself, which feeds the presentation of lifestyle.

6:35 His refusal to call it all bad. The humble brag about the messy office, the real brag about the cooking or the clothes, and the point that presenting a controllable image gives people some real control.

6:53 The Zoom green screen as the pandemic case: creating the visual context for relationships that exist only online.

7:05 A shallow cynical dimension and a positive one, since making your own image can be empowering.

7:13 Teenage anxiety. Adolescents may know rationally and cannot fully grasp that everyone is presenting a positive image, so they conclude they are the only sad one when everyone is posting one happy minute from yesterday.

7:31 The unanswerable question: how much is people having fun and how much is producing the image of fun.

7:38 Large shocks push things further in the direction they were already going, or that is an illusion and a confirmation bias, and everywhere he looks he sees inequality growing.

7:51 Education. Children with computers, devices, wi-fi, and parents pulling away from the rest.

7:58 Home. Some are trapped and getting more family time, which is basically good even when tiresome, and some are trapped in an abusive hellscape.

8:13 The measurement problem, stated as sharply as he has ever put it. Many tools for measuring family violence are activated only when people leave home, so tracking is limited.

8:29 School as the eight-hour daily respite from home, and how much that matters for children.

8:42 The magnification argument. Those with enough rooms for everyone to work and their own computers are riding it out, and the existing advantage is grossly magnified.

8:56 The saving that comes from not traveling, not buying coffee, and not eating out, being readied for a tutor or a private pod teacher next year.

9:13 Permanent cuts. Workplaces doing without certain workers may never restore them, and an office without a receptionist, or without an office, takes down building staff, parking attendants, and coffee shop workers.

9:38 A one-line aside on the advice to try something new, and that it is not voluntary.

9:42 Long-term medical effects, including possible lifetime consequences of infection in cases that look mild now, which given disparities in infection rates becomes future health inequality.

10:02 Fertility. No data yet, and certainly many are postponing children in a country where birth rates were already falling. Some delay becomes births forgone through breakups, restarting, and conception difficulty at older ages, with ripples like fewer kindergarteners in six years.

10:35 Postponed and cancelled weddings, and the open question of how many are temporary.

10:41 Children not going to camp, not traveling, not making friends or seeing new things, who will be different adults.

10:50 Single people, and whether they are meeting anyone.

10:56 Some of what is happening is the things that are not happening, and that absence will have effects nobody can predict.

11:03 The millions infected or bereaved, and how prolonged risk might change personalities or culture permanently.

11:16 The close. If the political system can function at all, he expects the crisis to produce universal health care, child care, and paid family leave, and says those are for a different talk.

Five recordings across four months and this is the one where the theory shows. The other four were a professor explaining things he had checked. This one is a man saying what he thinks families are, and the position turns out to be more conservative than his reputation allows and more radical than his critics would grant.

Start with what he says at 0:24 and repeats at 2:43 and again at 3:54. People marry when they are confident about the future. People have children when they believe they can shape their own destiny. Poor people do not marry because the gamble is unbalanced, a long-term commitment made in a universe that gives no reason to expect long-term decisions to be honored.

That is a claim that marriage requires conditions, and that the conditions have been withdrawn from most of the people being lectured about it. Hand that argument to a reader who thinks family breakdown is the central American problem and he will find that Cohen agrees with him about what has been lost and disagrees about the address to send the complaint to. The disagreement with Wilcox is about whether you can produce confidence in the future by exhorting people who have no reason for it. This talk is the clearest statement of his side of that argument I have seen from him, and it contains no accusation against anybody.

The haven in a heartless world, at 4:31, is quoted from Christopher Lasch’s 1977 book, and Cohen names it only as an old book. It is a strange citation to leave anonymous, since Lasch’s argument was that the haven had been hollowed out by the therapeutic professions and the market, and Cohen is using the phrase as an ideal that some families reach. Whether he is endorsing Lasch, correcting him, or reaching for a familiar phrase is impossible to tell from the recording, and it is the one place where a reader would want a footnote.

Two passages carry the analysis.

At 5:24 he stops the flow to say he will not collapse race into class, and gives the argument about two college graduates with the same income, one of whom fields requests for help from poor relatives and is exposed to racism from the workplace to the police. That is the wealth-versus-income point and the kin-network point, delivered in ninety seconds without a chart, and it is the single most portable thing in the talk. It also answers, in advance, the objection that his structural story is a class story wearing a racial costume.

At 8:13 he names the thing that broke the data. Many measures of family violence activate only when people leave home. Teachers, clinicians, and clerks are the observers, and the lockdown removed them, so the numbers that came out of 2020 measure the observers as much as the violence. He said the same thing in April and it is sharper here. Anyone reading the pandemic literature on family violence without that sentence in hand will draw the wrong conclusion from a falling number.

And 10:56 is the line of the talk. Some of what is happening is the things that are not happening. That is a demographer’s habit turned into a sentence a general reader can carry: absences have consequences, births forgone, friendships never made, meetings that did not occur, and none of it registers in any count of events.

Now the weaknesses.

The confirmation-bias moment at 7:38 is admirable and unfinished. He says large shocks push things further in the direction they were already going, then says that might be an illusion, then says everywhere he looks he sees inequality growing, then proceeds for three minutes on the assumption that it is growing. Naming the bias and then not acting on it is worse than not naming it, because the acknowledgment buys credibility that the subsequent argument does not earn. On the substance he was partly right and partly wrong, which is what you would expect: pandemic-era learning loss and childcare disruption fell hardest on poorer children, and household income at the bottom rose sharply through the transfer programs, and wage compression at the low end over the following two years ran against the story he was telling. A man who spends his Decembers counting other people’s evidence made a directional forecast here on the strength of what he could see from his window, which he says at 8:35, and never went back to it in public that I can find.

The Bezos passage at 1:19 concedes and then abandons. He grants that one man’s wealth doubling does not affect him, then says the aggregate corrupts the economy, and the connective tissue is one sentence about a few firms dictating wages. That is the weakest link in a chain the whole talk hangs from, and he would demolish it in anyone else’s argument. The concentration story about employers is empirically serious, monopsony research is real, and it has nothing to do with the personal fortunes he opens with.

And the ending is a wish presented as a forecast. There is every reason to believe, he says, that the crisis will produce universal health care, child care, and paid family leave. In August 2020 that was a hope with a political condition attached, and he attaches the condition and then discards it in the same breath. None of the three arrived. The one that came closest, the expanded child tax credit, ran for six months and lapsed. It is the only sentence in five recordings where he says something will happen because he wants it to, and he closes on it.

For my file, this is the piece the biography most needed and did not have. Everything else in the record shows him checking somebody. This shows him theorizing, and the theory is coherent: structure enters the person through the family by way of confidence about the future, inequality operates by removing that confidence, and the removal shows up as decisions not mad. That is a thesis, it is the same one underneath the divorce paper and the marriage-market work and the pronatalism arguments, and he has never stated it this clearly in print as far as I can see.

Two smaller things. The view count is the highest of the five, at nearly five thousand, and it is the least data-dense of them, which is either a fact about audiences or a fact about him. And the description, a short talk about what is on my mind as a family and inequality sociologist today, is the only one of the five that promises opinion, which suggests he knew the difference.

Transcript flags. At 1:09 “genie index” is Gini. At 4:42 “an old book” is Christopher Lasch, Haven in a Heartless World: The Family Besieged (1977), and you should name it if you quote the passage. At 8:02 “some people are trapped at home are but” is garbled. At 9:38 the Ivanka Trump reference is to the “Find Something New” campaign launched in July 2020, which dates the recording. At 11:24 “true universal health care” may be real universal health care. The talk appears to be scripted, given the absence of the verbal stumbles that fill his classroom recordings.

‘Diversity, Inequality, and Social Change: A Framework for Family Sociology’ (Aug. 4, 2020)

Julia Hall of W. W. Norton makes the introduction; the lecture is Cohen presenting the framework of his textbook, Family: Diversity, Inequality, and Social Change, then taking questions.

1:42 “The book is called Family: Diversity, Inequality, and Social Change. That is my framework for teaching family sociology.”

2:19 “The framework I use takes diversity as an array of choices and options and differences, some voluntary, some not. It’s the proliferation of difference when it comes to family life. We’ve concentrated on four types: cultural, legal, political and technological things in the social environment that facilitate a wider array of family structures and family experiences.”

3:25 “Diversity is all about choice and freedom and look at all the different things we can do. But of course they’re not really freely chosen. The amount of constraints and the type of constraints vary a lot, and that’s a way of framing the issue of inequalities. What choices do you have?”

4:15 “My basic theoretical framework is modernity theory and demographic perspectives. Modernity theory, the institutionalization of the individual.”

4:46 “Some people, maybe in your advanced courses, are going to call this something like the neoliberal world order, when we talk about the elevation of choice as something that holds everybody accountable. So yes, you have freedom, but on the other hand, if you’re poor it’s your fault. That’s the dilemma of modernity.”

5:18 “More people living more different kinds of lives than before. That’s a startling thing to say, in the sense of, how could you quantify that?”

5:44 On the census: the 1960 form carried six relationship categories, including inmate. By 1980 there were nine, adding brother, sister, father, mother, partner and roommate. “Some of those were romantic and some were not. They weren’t really differentiating. They were afraid it was rude to ask about unmarried couples at the time.” By 1990, eleven. By 2000, fifteen. In 2010, unmarried partners sit on the unrelated side of the form. “When they calculate the poverty rate they don’t include the income of the unmarried partners as of 2010.” By 2020, sixteen categories, now with the sex of partners.

8:42 “There’s both more diversity in terms of family structures and more recognition and incorporation of those different types.”

9:08 “You can see the rise of conformity through the first half of the 20th century, 1960 being the census where there was the most dominance of the married couple household, and after that the peacock’s tail of diversity. It’s not that one new kind of household replaces the old dominant household, but rather the proliferation of categories, especially people living alone, extended families, more single parents… A dominant, highly conformist society being replaced by more difference.”

10:09 “You can calculate a diversity score, the probability that two people selected at random are in different categories. Diversity in terms of race and ethnicity, at least with these six categories, increased from 0.29 to 0.57 over that fifty-year period.”

11:07 The YouTube wedding exercise. “There are new wedding videos uploaded every second on YouTube. You can search by ethnicity, by gender, by location, religion.” Students record the parameters. “How big, how formal, what are the family configurations, how expensive, who’s paying. It’s the things that are unspoken about just the way a wedding should be, or they’re deliberately countering traditions.”

13:29 On age at first marriage: “It’s not that everybody’s getting married at older ages. What was happening was everybody was getting married young, and now we have more choices. Some people are getting married older, which pulls up the average, but you miss the diversity story.” In 1960 nearly half of people married between twenty and twenty-four and about eighty percent married before twenty-five. “The mean age at marriage has risen, but the curve is also flattened out.”

14:44 The four types of inequality. Between families, the poverty rate, about twelve percent of the population in 2018, with single mother families historically at higher risk. Within families, “gender is the big kind of inequality within families,” shown through housework: women’s declining since the 1960s, men’s rising until the 1980s and flat after, men’s child care still climbing.

16:20 “The other kind of inequality that’s also about families is who gets a family, and who gets the family that they want. We don’t want to assume that everybody wants the same kind of family.”

16:45 On his paper with Joanna Pepin, plotting every metropolitan area: “Most single white women are living in metropolitan areas where there are more single men than single women. Almost all the black women are on the other side, where there’s a shortage of men per woman. For both groups there’s a positive slope, so the more men there are the more likely they are to get married.” The shortage traces to “incarceration, mortality, intermarriage.”

18:05 “In polygamous societies where rich men have lots of wives, obviously there’s a lot of poor men who get no wives. Historically this has always been one of the ways that inequality works out.”

18:42 The quote he uses on marriage equality: “I’m a forty-five-year-old woman. I’ve been in love with a woman for ten years. I don’t have a word to tell anybody about that.” Cohen: “The formal institutional designation of marriage, even in the most intimate way, gives a couple language to speak to each other.”

19:30 “Somewhere in 2009 or 2010 was the first time I had no debate in my class about same-sex marriage, and I realized I had to change my script. I couldn’t get any students to express the opinion that same-sex marriage should be illegal. Possibly it was because there were already students in the room who had gay parents, and so all of a sudden it became rude interpersonally. It’s fascinating to watch that happen.”

20:14 On the social distance exercise: “It is difficult to get people to talk about their prejudice and biases. If you ask, do you disapprove of whites and blacks getting married, you’re going to be hard-pressed to find somebody who says they do. But if you ask, would your parents mind, would your grandparents mind, it’s a good way to get people talking about how they were raised.”

21:29 “People don’t want to be racist and don’t feel that they are racist. They may feel empowered, and they feel good about talking about how their grandparents are racist.”

22:24 “Life expectancy rising from forty-five to eighty over a century and a quarter, and the birth rate falling from thirty per thousand to ten. People live longer with fewer children.” He adds: “I never met my great-grandparents. A lot of our kids in school today have. With fewer siblings they have more vertical relationships.”

23:32 “The great thing about demography is that it’s basically arithmetic. If we have a huge drop in births next year, as it appears very likely that we will, that’s something that will be with us forever. The tomorrow centenarians are in our classes now. You can’t reduce the number of people who were born last year no matter what you do. They could die, but they still were born.”

24:36 The work-family life history calendar. Students interview one person and fill a grid from left to right, top to bottom. “You start with the major events. When did you get married, when did you get divorced, when did you have kids, when did people die in the family. People will remember those. Then you work your way across to the things that are harder to remember, and the respondent synchronizes them with the things they do remember.” His worked example: she married Tom at twenty-four, lived with him at twenty-three, had Jen, her father died and her mother moved in when she was twenty-nine, she divorced two years later and started nursing school the same year. “You can ask, is there any meaning to this sequence?”

27:31 “You interview one person but you get data for every row, so it’s a way of making an unrealistically large sample.” Pooled across semesters he has 341 calendars, enough to plot women’s employment rates by marital status and age.

29:12 The Pew item he wishes had been repeated: “There seems to be a growing variety in the types of family arrangements that people live in. Overall, do you think this is a good thing, a bad thing, or don’t you think it makes a difference?” Cohen: “Good thing, bad thing, no difference corresponds to liberal, conservative and radical views of the family. Liberals take the middle view, whatever people want is fine. Conservatives are upset about the decline of the so-called traditional family. Radicals say good riddance.”

31:01 On the conservative-liberal-critical mapping exercise: “Sometimes it’s easier to describe other people’s political views, especially on culture war issues, than their own, especially if they’re feeling like they might be in a minority in the class. I don’t like to put people on the spot for what’s your opinion. Maybe some of you are better at doing that than I am.”

32:03 On names: “Almost fourteen percent of children born in the United States in 1840 who were girls were named Mary. We never had that much dominance before or since. Mary was number one until 1961,” by which point the share had fallen to about two percent.

33:08 “People don’t coordinate the names they give to each other over time, but somehow the name Mary has a trend. If you knew the number of girls named Mary in the last five years you can predict amazingly accurately how many girls are going to be named Mary next year. It’s bizarre. All those very personal individual decisions fit in a trend like this.”

33:56 “In the forties about two-thirds of girls were given names in the top hundred, and now it’s down to about thirty percent.”

34:32 “In the old days you had three kids named John, partly because you weren’t sure they would survive and you wanted to make sure one of them ended up named John, and partly because people just didn’t think every child was so different from every other child.”

36:27 On predictions: “I don’t like to make predictions. A much better way is to flag issues of, let’s talk about where this might be going.”

37:06 On declining marriage rates: “The conservative response is, hey, people got married during the Depression, so obviously the culture has changed. The culture has changed. Not being married is more of an option than it used to be. It’s not that people don’t appreciate marriage. I’m in the category of people who think we put marriage on more of a pedestal than we used to, and that’s one of the reasons why it’s declined. People want it to be perfect. They want to have an Instagram wedding. They want to have a marriage they can be proud of.”

38:05 “We’ve been living together for years, but I don’t want to get married until we can afford a house. This kind of stuff drives conservatives bananas. Why don’t you just get married now, like we did? But people have a high standard for it.”

38:45 On the title: “It was a very deliberate choice to call it Family. I’m old enough to remember the Journal of Marriage and the Family. Feminists took that out because we didn’t want to have one kind of family anymore, and I’m totally appreciative of that. But the way I think of the family is as an institutional arena. It’s like the workplace or the economy. The family includes good families and bad families, big families and little families, married and unmarried families, gay couples and straight couples.”

39:40 “Inequity I think of as unfairness, and inequality as objectively unequal quantities of something. If you have two people and they have different amounts of money, that’s inequality even if it’s totally fair.”

40:12 On global content: “Norton did a big survey when I started this. There was not much demand for global stuff in this book among US teachers, and it’s partly a problem with US sociology. There’s a tradition in family sociology that I don’t love so much, which is that there are a few common comparisons that we make, like arranged marriage and polygamy and the caste system. It’s kind of superficial. I have not found a great way to do global stuff in a way that doesn’t lose more teachers than I gain.”

42:13 On teaching race and family: “When I talk about the lower marriage rate among blacks I talk about incarceration. To some people that sounds a lot like I’m saying blacks are criminals. When I say, wow, look at these incarceration rates, they’re really high, to me that’s horrific. We have mass incarceration, we have to do something about that. It’s just really difficult and important to have that conversation.”

43:19 On a pandemic baby boom: “There was an idea that because of the quarantine we’d have more babies, but that’s not really the way fertility works. It’s not that when you stick couples together in their house they just have more babies, because we have modern birth control. People are meeting less. A lot of our births are unplanned. Isolation and staying home is just less sex, less people meeting, less new relationships.”

45:22 “We work on jargon. Less jargon. I don’t talk about the neoliberal world order. I try to just be clear on the writing and the figures.”

46:05 On whether married two-parent families produce better child outcomes: “All the research on that is extremely complicated and difficult because of selection. People who are better off choose that kind of family arrangement more often, and then they do better. Different kinds of privilege go together. The third thing is that stability is good for kids, more than what kind of parents you have. Transitions tend to go with other stuff like violence and abuse. Stability is better for kids, and that to me makes more intuitive sense than traditional families are better, not just politically because it rubs me the wrong way, but I think it logically makes sense.”

The framework does one thing, and the whole lecture is built to make that one thing feel obvious. Cohen converts a normative fight about family decline into a descriptive claim about variance. Marriage did not fall, the distribution spread. Age at marriage did not rise, the curve flattened. Married-couple households did not lose to some rival household type, they lost their monopoly. Once you accept variance as the object of study, the conservative argument about decline has nowhere to stand, because decline requires a baseline and variance measurement supplies none. This is a strong move and he never announces it as a move.

The evidence he brings for it varies in quality. The census-category series is the weakest, and he half concedes as much. Counting the relationship codes on a federal form measures what the state decided to record. He says he thinks it captures both real change and recognition, and offers nothing to separate them. The household composition series from 1900 does better because the categories stay fixed while the shares move. The names data does best of all, and it does something the family data cannot. Nobody administers first names. No agency recognizes them. The collapse of Mary from fourteen percent to nothing, and the fall of the top hundred from two-thirds of girls to thirty percent, gives him a measure of declining conformity that owes nothing to state categories and nothing to the family-values argument. He lays it next to the family trends and lets the reader draw the line. He never asserts the causal claim, which is the right restraint, though it also means the strongest support for his framework arrives as an aside about a hobby.

The most revealing thirty seconds come at 19:30. Around 2009 the debate over same-sex marriage vanished from his classroom and he had to rewrite his lecture. He reads this as social change and finds it fascinating. Ten minutes earlier he explains why he built the social distance exercise: ask people about their own prejudices and they will not tell you, so ask about their grandparents instead. He designed a tool on the premise that students conceal what they think, then read the silence of students on a live culture-war question as the absence of the view. Preference falsification and attitude change produce identical classroom transcripts. He mentions that students with gay parents may have been present and that it became rude, which is the preference-falsification account, and he still calls the result change. At 31:01 he closes the circle without noticing, saying students find it easier to describe other people’s politics than their own, especially when they suspect they are in a minority in the room. He has stated the objection to his own inference twice and drawn the conclusion anyway.

He is candid about constraints on his teaching, and the candor is the most useful thing here for anyone studying how academic knowledge gets packaged. Three admissions stand out. Norton’s market research found no appetite among American instructors for comparative material, so the book stays domestic, and he says he cannot add global content without losing more adopters than he gains. He drops the term neoliberal world order because undergraduates do not need the label, while keeping its content, which means students receive a critique of individual responsibility with no name attached to argue with. And he tells a room of sociologists that he avoids putting students on the spot about their own views. Textbook content shaped by adoption numbers, delivered by an instructor who has decided not to ask students what they think.

His best unrehearsed moment is the marriage-pedestal answer at 37:06. He grants the conservative premise, that people married through the Depression, so material hardship cannot be the whole story, and then flips the valence. Marriage declined because it rose in status. It became a capstone with an entry price, a house and a wedding worth photographing, and people who cannot pay the price wait. That argument is testable, it explains the class gradient in marriage better than the deprivation story alone, and it cedes real ground to the other side while keeping the field’s commitments intact. He has not developed it in print with the force he gives it here.

The child-outcomes answer at 46:05 is where the argument thins. Selection is real and he is right to lead with it. Privilege clustering is real. Then he lands on stability as the thing that matters, and stability is the escape hatch. It concedes almost all the empirical content of the two-parent finding while removing the word marriage from the sentence, and it inherits the exact selection problem he opened with, since the households that stay stable are not randomly assigned either. He then names his own prior aloud: the traditional-family conclusion rubs him the wrong way, and he also thinks it makes logical sense. Naming a prior is not controlling for one. It is the same move as the “such as me” aside in the COVID lecture, the analyst placing himself in the distribution, and it is more attractive than the alternative. It is not a substitute for the decomposition.

One absence runs through the whole hour. The framework has an account of constrained choice for poor people, and none for people whose choices went wrong on their own terms. Desired fertility exceeds achieved fertility across the developed world by a wide margin. People who wanted three children and had one are a large population, and their frustration has nothing to do with conservative nostalgia. In this framework the birth rate falling from thirty per thousand to ten arrives as a demographic fact and a source of longer grandparent relationships. Nobody in the lecture regrets anything. Diversity accumulates and constraint operates only on the poor. A version of this course that took disappointment as seriously as it takes inequality would be a harder course, and a better one.

‘Joanna Pepin interviews Philip Cohen’ (Aug. 23, 2020)

0:00 Joanna Pepin introduces herself as being at the University at Buffalo and introduces Cohen and the textbook.

0:25 First question: what led him to become a family sociologist.

0:33 His answer, that inequality questions kept routing him back to families.

“I was interested in inequality, and a lot of the inequality questions I was thinking about kept leading me back to families, especially things having to do with the intersections of race, class, and gender. If you ask why women do more housework and child care, part of that is because they’re single parents, and that’s a family question. Why are more parents single now than they used to be? Who decides who’s married, and why are people married, and what are the consequences of that for health and well-being and income? I just kept coming back to family topics.”

1:20 The policy version of the same point.

“What is the government going to do? Is it going to give people money if they’re poor? Is that going to be based on their family structure? Is it going to penalize people for their family choices? Is it going to let people get married if they’re gay? All these things that have to do with people’s well-being and social inequality go through family. So that’s how I got led into it.”

1:52 Pepin notes the textbook’s life course approach and asks for the biggest transformation of American families.

2:06 His answer: two changes that arrived together over a couple of centuries, fewer children and women working outside the home. Fewer children means less care work at home, partly because women had other opportunities and partly the reverse, since women doing other things had fewer children.

2:41 Why he calls it foundational.

“Having fewer children changes the relationship between parents and children. It means we’re investing more in each child than we used to, so kids spend more years in school with more resources. It also means our relationships are longer. People are living longer, so you have fewer children and now you have a seventy- or eighty-year relationship with your children, and there’s only one or two of them, instead of having a forty-year relationship with eight or ten kids. It’s a whole different relationship.”

3:27 Pepin asks about a common misperception he finds himself correcting.

3:40 The decline narrative first, with a hedge.

“There’s a common sense that families are going down the drain in general, that in the good old days things used to be better. Some things used to be better for some people, some things used to be worse for some people. It’s hard to generalize that way.”

3:58 Then divorce, which is the substantive answer.

“Divorce is more common than it used to be a long time ago, like a hundred years ago or fifty years ago. But in the last ten or twenty years, divorce is becoming less common. That’s surprising to people, and it clues us into some important changes. People are getting married later, people are getting married after they have more education, there’s more selectivity in marriage. Some people are choosing not to get married, or they don’t have anybody they want to marry.”

4:39 What the decline is made of, stated without comfort.

“You have fewer people getting married just because that’s what you do, fewer people getting married because they’re stuck with somebody, fewer people stuck in marriages because they can’t support themselves on their own and so they can’t leave an abusive marriage. When you put them together, the fewer marriages and the marrying later have turned into a decline in divorce, even though divorce is more available than it used to be, which is great, and it’s happening less. But partly because marriage is available to fewer people. So it’s a mixed bag.”

5:31 Pepin asks what he has had to update for the third edition.

5:48 Same-sex marriage, and the measurement problem underneath it.

“In the second edition, same-sex marriage was just coming out, and we didn’t really have a lot of good data about what was going on with it. That’s improved. I worked really hard to try to get better information. It’s still pretty hard. A lot of states don’t record the gender of the people getting married, which seems odd, but they just have spouse one and spouse two on their marriage certificates. So the data is still blotchy. But we’re starting to learn more, and in a lot of ways we’re learning that same-sex married couples look pretty much the same as different-sex married couples.”

6:42 The faster-moving revision.

“At the same time, the rapid change that’s happening is on sex and gender and language and culture and attitudes. Transgender identities, non-binary identities. The language had to get revamped even though the last version of the book was only a couple of years ago. I used to have a whole section on androgyny, and now I don’t even talk about androgyny anymore. Now we have non-binary identities.”

7:24 Pepin asks him to speculate on the pandemic’s impact, noting that data collection is hard right now.

7:43 His framing principle, delivered as an aside.

“Things that are happening now will tend to continue happening. That’s just a way that society works. So working at home is going to be more common now forever.”

8:11 The disrupted side. Cancelled weddings, postponed childbearing, and the observation that postponement becomes permanent for some people, so the drop in marriage and childbearing will probably be large this year and next.

8:42 Inequality, prefaced by self-awareness.

“In terms of the subjects I pay most attention to, I think inequality is going to be exacerbated. I sound a little like a broken record on inequality, because I think everything’s related to inequality.”

8:56 The question underneath his whole program. “I’m interested in the things that we don’t see about families, like who gets to have a family. We talk a lot as if everybody’s in a family, but not everybody’s in a family, and some people aren’t in families because they don’t have access to the resources or the people. When gay marriage was illegal, it was, okay, there’s a bunch of people who can’t have the kind of family they want. But now we have people who are stuck with their families that they don’t want to be stuck with. We have trans young people who are living with families that don’t accept them. We have people living in abusive situations.”

9:39 School at home, and the magnification argument: children whose parents have education, resources, spare rooms, and computers against children whose parents have none of that.

9:57 The summary.

“All the gaps and inequalities we have, whether class and race or between happy and abusive, I just feel like the gaps are opening up right now, and it’s going to get more unequal before we get out of this.”

10:32 The close, on what follows for the discipline: a lot of work to do, not only to study these things but to fix them, and a hope that a semester of teaching and learning will turn up ways to address some of them.

Thoughts

Eleven minutes, a friendly interviewer who is his own former doctoral student, and no adversary anywhere in the frame. It is the softest recording in the file and it contains the two clearest statements of his position that exist on tape.

The first is at 4:39. He is asked about a misperception and gives the divorce decline, which is his best-known finding, and then does something almost nobody does with their own famous result: he itemizes what the good news is made of. Fewer marriages entered out of default. Fewer people trapped with someone. Fewer people unable to leave an abusive marriage because they cannot support themselves. And then marriage available to fewer people, which is the part that does not fit the story, and he says it in the same breath and calls the whole thing a mixed bag. A researcher protecting a headline finding leaves that last clause out. He puts it in, to a sympathetic interviewer, with no one pressing him.

The second is at 8:56, and it is the sentence I would build any essay on him around. Who gets to have a family. He has spent a career being read as the sociologist who says the traditional family is not in decline, and what he says here is that his real subject is the people outside the frame entirely: not counted in the marriage rate, not counted in the divorce rate, absent from the data because the thing they would have been counted for never happened. That is the same insight as some of what is happening is the things that are not happening, from three weeks earlier, and it is the same insight as the marriage-market work with the interviewer sitting across from him. The pattern of his mind is to ask what the denominator excludes.

And notice how he uses same-sex marriage in that passage. He reaches for it as the historical example of people barred from the family they wanted, and then pivots to people trapped in families they cannot leave, including trans teenagers living with parents who do not accept them and people in abusive homes. The two cases are structurally identical in his account: the constraint runs on access to resources and to other people, and it operates whether it locks you out or locks you in.

Three small things.

The androgyny detail at 7:02 is the best single piece of evidence in the file for how a textbook works as an artifact. He had a whole section on androgyny in an edition published two years earlier, and by 2020 the concept had been superseded in his own book by non-binary identity. That is a discipline’s vocabulary turning over inside one author’s revision cycle, reported without defensiveness and without celebration. Whatever a reader thinks about the substance, the mechanic is the same one that produced his complaint about generation labels: the category changes, and the trend line moves with it, and somebody has to decide whether the world moved or the instrument did.

The marriage certificate detail at 6:07 is a measurement anecdote he throws away. States record spouse one and spouse two, so the administrative record of who married whom is unusable for the question everyone wants answered. He calls the data blotchy and moves on. That is the same category as the mandatory reporters vanishing from the family violence data and the observers vanishing from the child maltreatment counts. His most durable contribution may be that he keeps noticing when the instrument produced the number.

And at 7:43 he states a forecasting rule so casually that it sounds like nothing: things happening now tend to keep happening. He then applies it to remote work, where it held, and in the same minute applies the opposite reasoning to weddings and births, where he expects the disruption to produce permanent losses through postponement. Both turned out roughly right, and the two claims run on different logics that he never reconciles. It is the closest thing he has to a stated method for talking about the future, and it is four seconds long.

Two more things.

He is a broken record on inequality and says so at 8:47, and the self-awareness does not change the fact. Asked what the pandemic will do to families, he answers that inequality will grow, which was the answer he gave to a different question three weeks earlier and the answer he gives at every level of the analysis from Bezos to the spare bedroom. A frame that explains everything explains less than it appears to, and his own joke is the acknowledgment that no correction follows.

And the interview format flatters him. Pepin was his doctoral student, published with him on the marriage-market work, and asks five questions with no follow-up. Every recording in this file has him talking to people who either agree with him or have no standing to push. There is no tape anywhere of Cohen being interrogated by someone who disagrees, which is a gap in the record on a man whose reputation rests on interrogating others.

For my file, this is the interview to mine for direct quotation, because it is the only one where he speaks in his own voice about what he is trying to do. The biography currently derives his position from his writing. Three of these passages, the origin story at 0:33, the mixed bag at 4:39, and who gets to have a family at 8:56, would let him state it himself, and quoted speech does something in a profile that paraphrase of published work cannot.

Transcript flags. At 0:03 “university of buffalo” is the University at Buffalo, which is how the institution styles itself. At 0:11 the textbook title is garbled as “diversity in equality and social change” and is The Family: Diversity, Inequality, and Social Change. At 7:28 kovid is COVID. At 8:18 and 8:23 “that we’re gonna” is “that were going to,” which reverses the meaning of both sentences and matters: people who were going to get married and were going to have children. At 10:32 “upgrade news” is garbled beyond recovery and appears to be Pepin’s transitional remark; leave it out. The third edition he describes working on published in 2021, and the fourth in 2024, so this catches him mid-revision on the third.

‘COVID-19 Pandemic Overview’ (Aug. 25, 2020)

0:00 Opens the lecture, and lays out the plan: history and terms, the global situation, where the United States fits, the American failure, the overlapping crises, and the future.

0:37 Pandemic from the Greek for all people, meaning the worldwide spread of a new disease. The World Health Organization declared this one on March 11, 2020. The American epidemic is part of the global pandemic.

1:14 Precedents. The 1918 influenza pandemic, which he notes probably did not come from Spain, infected about a third of the world and killed around 50 million people including more than half a million Americans, against a much smaller world population. The 1957 and 1968 influenza pandemics killed about a million each. SARS in 2003 killed about 800 and was extremely frightening before it was controlled. H1N1 in 2009 killed about 200,000.

2:53 The virus. Discovered in 2019, probably from a bat to a person, possibly through another species, in or near Wuhan. SARS-CoV-2 causes COVID-19.

3:38 Why it is bad, in four parts.

“It’s a novel disease, which means we are not immune to it. As far as we know, nobody was immune when this started. And it’s very highly contagious, so it spreads very easily between people. All they have to do is interact or be close to each other, and if they’re not taking any precautions the virus transmits and nobody is immune. To make matters worse, it transmits asymptomatically. There are people who show no signs that they are sick who can transmit it, so it’s hard to avoid people who are transmitting the disease. It transmits through aerosols, which are tiny airborne particles you can get just from breathing, so you don’t have to spit or cough or sneeze on somebody to transmit the virus.”

4:27 Mortality. Seasonal influenza kills somewhere between 0.3 and 0.7 of every thousand infected. He puts COVID at roughly ten times that, perhaps six to ten per thousand, with the rate uncertain, and notes the seriously ill who survive.

5:12 The counts in late August. The world above 20 million confirmed cases and approaching 25 million, with more infections than confirmed cases. The United States above five million. Global deaths in the hundreds of thousands and American deaths approaching 200,000, which he calls about a quarter of the global total.

6:09 The world distribution. Started in China, spread to Europe, then the United States, which now has the most confirmed cases. India approaching three million, Brazil above three million, Russia approaching a million, South Africa at half a million.

6:58 Comparison with peer countries, defined as those above thirty thousand dollars of annual income and five million population. The United States has more cases per thousand than any of them by a wide margin, around 17, with Israel next and then Singapore. On deaths per million, five large rich countries were then higher: Sweden, Italy, Spain, the United Kingdom, and Belgium. He immediately says the cumulative comparison is misleading without the trajectory.

7:59 The trajectory chart, and the central finding of the lecture.

“In late April, when we had our peak, there were those five countries with higher peaks than us, including much higher, like Italy and Spain, many more deaths at that time per person in their populations. We all started to decline after that. But the key thing is that those countries had a faster decline and a more continuous decline. By June our daily deaths had stopped declining, and started rising again in July and August, which is virtually unique among the rich countries. Now we’re at about ten times or more deaths per day, relative to our population, than the other rich countries of the world.”

9:39 Racial and ethnic disparities. Black Americans with the highest death rate at 85 per hundred thousand, White, Asian and Pacific Islander Americans the lowest, Hispanic and American Indian populations between. The causes he lists: discrimination, inequitable access to good health care, work in front-line and essential occupations involving contact with the public, inequalities in income, education, and wealth, and housing quality and crowding.

11:02 CDC funding, cut from above seven and a half billion two years earlier by about a billion for the 2020 budget.

11:35 The National Security Council pandemic office, closed, and the warning from the Center for Strategic and International Studies that it was unclear who would be in charge at the White House in the event of a pandemic. He adds that this turned out to be true.

12:04 The ignored exercise. His administration ran a functional scenario gaming out a global pandemic originating in China, which produced recommendations that went unused, against the president’s claim in March that nobody could have foreseen it.

12:44 Published warnings. Ed Yong’s 2018 piece asking whether America was ready, on underfunded public health programs and stretched hospitals. Biden in the fall of 2019 saying the country was not prepared.

13:19 His summary of the foreknowledge question.

“This was not that much of a surprise. The details are always a surprise, but the fact that this was inevitable at some point, and that we were not prepared, was known.”

13:25 The messaging failure, framed as motive: the president more concerned with how it made him look than with dealing with the epidemic.

13:53 A montage of presidential clips running roughly ninety seconds, longer than the version he used in April. Totally under control. One person from China. It goes away in April with the heat. Eleven or twelve cases and many of them in the shade. Fifteen going down to close to zero. It will disappear like a miracle.

15:08 The compounding failures. Not enough testing from the beginning, so nobody knew who was infected. Without that, no contact tracing. Without containment, spread, with nothing stopping travel from New York to New Orleans for Mardi Gras. Numbers too large for tracing to work. Slow and weak policy once the federal government caught up, with no strong lockdown and no internal travel restrictions. Mismanagement of equipment and staff when New York was overwhelmed, so more people died than needed to.

16:53 Institutional outbreaks in nursing homes, prisons, and factories, largely unchecked at first, which then seeded the surrounding communities, which he notes are largely low income and minority.

17:17 The cultural failures: happy talk, mixed messages, a cynical attitude toward science and expertise from the top, conspiracy theories, and what he calls useless hero talk.

“This sort of useless hero talk. Oh, we’re warriors, we’re tough, we’re strong. That’s not how you fight a virus. Basically you fight a virus by acting weak and staying home, and we didn’t do that.”

17:56 The Italy case study, which is the analytical heart of the lecture. A steeper and higher spike than the American one, arriving about a month earlier, followed by a much steeper decline, so that by late August the United States had more than ten times Italy’s ongoing death rate while Italy prepared to reopen schools.

19:01 The Italian sequence, step by step. Two deaths at the start of February. First, a few northern towns locked down with checkpoints and barricades, which failed because the virus had already left. By early March, the whole northern region. Within about a week, all schools and sporting events nationwide. By the second week of March, a national lockdown.

19:50 What that lockdown contained, and the comparison he draws.

“Their lockdown was much stronger than anything we had in this country. You could not leave the house unless you were going to the doctor, to get necessary supplies, or to do an essential job, and the police would stop you and you had to explain what you were doing. We had nothing like that in this country. You could not travel between regions of the country. By and large you couldn’t go from one state to another. We never had that restriction.”

20:17 All non-essential workplaces closed, including factories, for five or six weeks. Then phased reopening as cases fell ahead of the death curve: travel within regions, some factories, most businesses with distancing and masks, then pools and gyms in late May, then theaters by June 15.

21:26 The endpoint he wants students to see.

“Now they have a low enough number of cases that they can do testing and tracing, isolate people who get infected, and contain the epidemic. We never reached that level of containment in this country. We never had that level of closure, and we never had a small enough number of cases that we could jump on them, isolate them, and stop outbreaks as they occur. So our epidemic is relatively uncontrolled still at this point. It’s only controlled to the extent that people are wearing masks and staying home as much as possible.”

22:05 Why the failure. Beyond the presidential decisions, a partisan attitude toward the science.

22:23 Four survey questions he says have correct answers, with party gaps. That social distancing slows the spread, agreed by 76 percent of liberal Democrats and 47 percent of conservative Republicans. That the virus spreads more easily than most other infectious diseases, 78 against 58. That there is not enough testing, 82 against 31.

23:20 His diagnosis.

“Science can answer these questions, but because of the way our politics corrupts our science, or our science fails to penetrate into our political bubbles, we are impaired from being able to do smart things.”

23:36 Misinformation. The claim that the pandemic was engineered for profit, the 5G theory, the claim that masks harm health, and fake cures, including a hydroxychloroquine post the president shared. He notes how hard this makes it for anyone who is not a full-time debunker.

24:42 Racism and xenophobia. White Americans treating it as a disease of non-White people, the same way people in Texas and Arizona treated it as a New York problem. Presidential language about the China virus, vandalism against Asian businesses, attacks on Asian people, and a clip of a rally crowd cheering the phrase kung flu.

25:47 The southern border, where he notes an inversion.

“At this point it’s probably better for Mexicans than for Americans to have the border closed.”

26:05 A clip of the president claiming that public health authority at the southern border prevented a catastrophe, followed by his correction that the American side, and the Rio Grande Valley in particular, has been extremely hard hit.

26:44 The economic crisis. Non-farm employment rising from about 130 million in the early 2000s, dropping in 2008, recovering past 150 million, then falling by 20 million in a single month in April.

27:56 Food lines in Queens set beside New York City lines from 1931.

28:10 Education. Online math participation dropped about 20 percent in higher-income zip codes in March and more than 60 percent in low-income ones, and the richer areas recovered by May while the poorer ones never did. Estimates of about seven months of lost learning on average without in-class instruction, more for Black and Hispanic students and much more for low-income students.

29:28 The other side of the choice, which he refuses to resolve. Hospitalization rates five to eight times higher for Black and Hispanic children than White children. And the exposure of the household: 4 percent of White seniors live with a school-aged child against 11 percent of Black seniors and 13 to 19 percent in other groups.

30:35 His conclusion on schools: there are no good choices.

30:42 Family violence, and the measurement problem he has now stated three times across the recordings.

“We don’t have great data on this, because the way we collect data on this is from people leaving their houses and doing things like going to work and going to school and going to the doctor, and a lot of that has not been happening. One study found that there were fewer visits to the emergency room with injuries related to domestic violence, but the injuries that were showing up were more serious. So it looks like a lot of people are having abuse happen in their homes but are not making it to the emergency room unless their injuries are very serious.”

31:21 The gender impact. Women with school-aged children were the group most likely to leave the labor force between February and April, with some evidence of men taking on more of the domestic work when home.

32:12 The mental health section: fear, isolation, separation from friends and from services, and despair at the accumulation of bad news.

32:44 A pause to address the students directly, which is the only moment of its kind in any of these recordings.

“I just want to pause to say here, if you are anxious and depressed, if you are thinking about suicide, you are obviously not alone. A lot of people are, and a lot of people are not talking about it. There are the numbers there for you to call and seek help at any time, and you should not feel like you’re the only one, or that you are burdening anybody by asking for that help. So please do.”

33:08 Survey figures on young adults reporting serious thoughts of suicide, rising through the 2010s and then a much higher figure in a June 2020 survey of adults under 25. He says he was astounded by it.

33:52 The worst-case arithmetic, flagged as speculation. About 177,000 deaths and 5.7 million confirmed cases, a case fatality rate near three percent, an infection fatality rate perhaps ten times lower if infections run to 50 million. Herd immunity possibly around 70 percent, which for 330 million people would mean roughly 230 million infected and perhaps 700,000 deaths, unless treatment improves or a vaccine slows transmission.

35:58 When it ends. Waves for months and hopefully not years. Tracing only works once numbers are low, since nobody can call a hundred thousand contacts a day. Limited world travel for an unknown period. Vaccines under test with unknown efficacy and safety, unlikely to affect students personally inside eighteen months, which he calls speculation. And a caveat on his own arithmetic: herd immunity is theory, and it may not arrive at 70 percent.

37:39 The advice. Follow real public health advice from state and local government, and then: your sociology professor is probably a good source too.

38:03 Stay home if you can, granting that it is boring, frustrating, and lonely, and may be impossible to pull off.

38:27 Take care of yourself, since nobody alive has experienced this. Watch your friends, neighbors, and relatives, work out what they need, and provide it.

38:45 If distressed, find ways to help others, online, with money, through volunteering or activism, which gives a sense of purpose.

39:16 The close. The world, politics, society, and science will be different, career plans may have to change radically, and it is a time to think big about what should be done afterward.

This is the longest and most political of the recordings so far, and it is the opening lecture of a required introductory course. Not a guest talk, not an elective seminar, not a video for people who sought him out. The first thing a nineteen-year-old sees in Social Problems, fall semester, includes ninety seconds of presidential clips, a rally crowd chanting a racial epithet, a partisan knowledge slide, and at 37:57 the instruction that their sociology professor is a good source of pandemic advice.

Every factual claim in it is sourced and checkable. The clips are real, the CDC appropriations are real, the exercise happened, the Yong article exists. The question this recording raises is is that a man who wrote a book about the boundary between the scholar’s role and the citizen’s role crossed it in the first week of a course, in front of students who cannot leave, and the crossing came packaged with a recommendation that they treat him as an authority.

His defense is that there was no neutral version of that lecture in August 2020, and a professor who withheld the trajectory chart to appear balanced would have failed at the only job Fish thinks he has. That defense covers the charts. It does not cover the montage, the rally clip, or the clause at 37:57.

Now the strongest thing in the lecture, which is not the politics.

The Italy sequence from 17:56 to 21:26 is the best twenty minutes of teaching in this entire file. It is a counterfactual done properly. Same disease, worse start, higher peak, earlier arrival, and a better position five months later. Then the specific timeline of what Italy did, escalating from a few towns to a region to a national shutdown, with the detail that police stopped people and required an explanation. Then the reopening sequence, with the observation that cases were falling faster than deaths because deaths lag. A student comes out of that able to answer the question that dominated American argument all year, which is whether the outcome was fixed by circumstance. Cohen shows it was not, and he shows it with a case.

The line at 17:43 is the one to keep. You fight a virus by acting weak and staying home. That is a sociologist noticing that a culture’s heroic register was the wrong tool, and it says in one sentence what a great deal of pandemic commentary took paragraphs to miss.

Three failures of his own standard.

The partisan knowledge slide at 22:23 does the thing he attacks in other people. He introduces four items as questions that have correct answers, and one of them is whether there is enough testing. He calls it uncontrovertibly true, just a true fact. Sufficiency is a judgment about a threshold, resource allocation, and what testing is for. There were serious people arguing at the time about testing strategy, targeting, and turnaround. He may be right about the policy and he has converted a policy position into a factual item and scored a partisan gap on it, which is the operation he objects to when the Institute for Family Studies presents a family ideal as an empirical conclusion. The other three items are closer to factual, and the causal story he attaches at 22:29, that Republicans believe their political leaders instead of the science, is an assertion about why the correlation exists.

The suicide comparison at 33:08 is a measurement error of the kind he built his reputation on catching. The earlier figures he cites are twelve-month prevalence. The June 2020 figure is thirty-day prevalence, from an online panel with a low response rate. A thirty-day number that exceeds a prior twelve-month number is partly an artifact of the windows, and any comparison of the two overstates the change by an unknown amount. He says he was astounded, and part of what astounded him was the instrument. The underlying concern was real, the direction was probably right, and the number as presented does not support the size of the jump.

To his credit, the passage at 32:44 is the most careful thing he does in any of these recordings. He stops the lecture, speaks to the students directly, tells them they are not alone, that many people are not talking about it, that seeking help is not a burden, and points them to the numbers on the slide. That is right, and it is the only moment across six hours of tape where he steps out of the analytic role to address a person.

The worst-case arithmetic at 33:52 is his most honest passage and his least accurate forecast. He shows every step, labels it speculation, states the herd immunity assumption, and then at 37:13 comes back and says the assumption itself is unknown and may fail. American deaths passed his worst case of 700,000 in the autumn of 2021 and have since gone well beyond it. His worst case was optimistic, and the reason is the assumption he flagged: immunity at 70 percent did not stop transmission once variants and waning protection arrived. A forecast that fails on the premise its author identified as the weak point is a better performance than most, and it is still a failure.

For my file, two things.

This lecture and the April 3 one are the same lecture five months apart, which makes them the only place in the record where you can check him against himself on a short horizon. In April he said thousands would die, that nothing would break the slope, that waves and repeated shutdowns were coming, and that travel restrictions would outlast domestic ones. By August, 177,000 dead, the slope unbroken, waves arriving, and he repeats the same forecast with more evidence behind it. The stable predictions are the ones that come from arithmetic. The unstable ones come from assumptions about biology he had no standing to make and said so.

And the family violence measurement point now appears in three separate recordings across five months, in April, in June, and here. He said it before the data came in, he kept saying it, and it is the thing about the pandemic he was most reliably right about and least credited for.

Transcript flags. At 5:49 the world death figure reads as over 80,000, which is wrong by a factor of ten and should be over 800,000; check the slide before quoting. Kobit, covet, corvette, cobin, and chronovirus are all COVID or coronavirus. At 3:10 “somebody ate a bad or a bad bit somebody” is bat. At 3:28 “the viruses stars” is SARS-CoV-2. At 24:17 “a bad high druxy chloroquine” is hydroxychloroquine. At 13:03 “even joe biden running for president last fall” places that quote in fall 2019. At 14:35 “we’re a bridge” is garbled. At 33:36 “26 of adults under age 25” should be checked against the slide, since the underlying survey reported 25.5 percent for ages 18 to 24. Slides are at https://osf.io/9vw2a/ and will settle the death figure, the disparity rates, and the suicide chart.

‘Introduction to Social Problems’ (Aug. 29, 2020)

0:00 Lays out the lecture: what sociology and social problems are, the main theoretical perspectives, and the research methods.

0:24 Sociology as the science of society, the study of individuals, groups, and the social structures that emerge from their interaction, with attention to the role of those structures in creating problems and solutions.

1:00 A social problem defined as a condition or pattern of behavior with negative consequences for individuals, for the social world of families, education, and the economy, or for the physical world of housing and environment.

1:27 Gallup’s most-important-problem question in August 2020, where coronavirus dominates, followed by government and the economy, with everything that would normally rank high pushed down.

2:14 The objective and subjective duality.

“Objectively, does the particular condition exist? Do we have a pandemic? Yes, we have a pandemic, we have a virus which is spreading, which has killed a lot of people around the world. So that’s objective reality. But do we think it’s a problem, and how do we shape or define that problem in our collective minds? Do we think the problem is the pandemic? Do we think the problem is the way the government is dealing with it? Do we think the problem is the government telling us what to do? What is the problem?”

3:10 Social constructionism, with the illustration he uses to mark its limit.

“Social constructionism is the view that social problems only become problems when they’re perceived to be problems. If a meteor hit us and we were all obliterated, that wouldn’t really be a social problem. That would be the end of all problems. It’s called constructionism because it’s acknowledging that social reality is the meaning we attach to it. There is a reality out there beyond our consciousness, but in terms of society, what matters is how we construct it.”

4:06 The back and forth is iterative. Some problems emerge fast, like a pandemic, and some persist a long time, like racial inequality, and the defining, blaming, and fighting over solutions goes on continuously.

4:45 Stages of the process, with the caveat that they are categories for thinking, and that some problems do not move through them in order.

5:09 Stage one, claims making. Somebody has to come forward and say this is a problem. With a pandemic there may be no single actor, but the disparities and consequences require someone to stand up.

5:53 Stage two, getting weight behind the claim, especially from government agencies willing to accept the definition and take on the role.

6:23 Stage three, negotiation and power struggle over whether the response is adequate, effective, and moving in the right direction.

7:01 Stage four, when invested groups conclude the established system will not produce the results they want and have to work outside it or change it radically. He notes that not every social problem reaches this.

7:43 Why courses on social problems focus on protest: it is the most explicit version of the process, where people gather and literally state what they think the problem is.

8:03 A photograph he took at the Women’s March the day after the 2017 inauguration, and the observation that people bring signs, friends, and relatives.

9:04 What protest does beyond making claims.

“In this process of making claims and having protests, people are also building identities around their definition of problems, and who else thinks it’s a problem, and who defines it the same way we do. The formation of a social group comes out of that process of making the demand that something be considered a problem.”

9:23 A photograph he took at the March for Life, where teenagers from religious schools came from around the country, and the same analysis applied.

“It’s not just people saying, hey, we think you should stop women from having abortions. It’s this process of we are all in this together, we all believe this, we reinforce it. I’m looking at my friend, my friend is really into it, I’m really into it, we’re really into it together. And it’s building solidarity and identity.”

10:12 C. Wright Mills and The Sociological Imagination, and the link between personal experience and the social world, described as the intersection of biography and history.

11:00 Personal troubles against public issues.

“If I’m hungry right now, that’s a personal trouble. I’m hungry, maybe I just need to go in the other room and have some food, or maybe I don’t have enough food. But it’s still one person who doesn’t have enough food. The point at which it becomes a public issue is when it affects a lot of people and we start to acknowledge that.”

11:29 The illustrations: cars lined up for free food in San Antonio, and bags of food at a drive in Des Moines.

12:07 Theory defined as a set of assumptions or propositions used to explain, predict, or understand, stated explicitly and systematically.

13:37 Why the theories were already in place.

“We may even be able to make predictions. We’ve seen pandemics before, we understand that they’re likely to take a certain course. If we could already in January, February say some intelligent things about what was likely to happen, it’s because we had experience and we already had theories about how pandemics work.”

14:04 Macro and micro as a continuum. Macro asks why this happens in one society. Micro asks why an individual gets in line at a food drive.

15:05 Functionalism, traced to Durkheim, and the body metaphor: organs with distinct functions maintaining a balance, and institutions like family, politics, and religion working in some harmony to meet social needs.

16:18 Change as the problem, from that perspective, producing anomie.

“When society changes rapidly, when the social order changes rapidly, it disrupts that balance, and we end up not knowing what the rules are, what the norms are, how we should behave, how our institutions should work together, where individuals fit into the changing social order. This produces a state that Durkheim described as anomie, or normlessness.”

18:04 The functionalist question: where did the problem come from, and is it serving some function, the way a fever serves the body against an infection.

18:41 Merton, and the argument that crime may help mark the boundary between good and bad behavior, and poverty may motivate people to avoid destitution. He calls the poverty version cold-hearted and grants there is a logic to it.

19:29 “There’s a potentially faulty logic to this, which is that if you see something happen in society you assume there must be a good reason for it. Every society seems to have poverty and inequality, so it must be good, right? Like human beings have lungs, everybody’s got lungs, let’s see why we have lungs. You can see the potential logical problem in this metaphor is you tend to define things as good and then try to rationalize them after the fact, working from the assumption that they are good. So this tends to be politically a conservative theory, because it tends to be describing the status quo as somehow good, and it’s suspicious of rapid change.”

20:12 Conflict theory as the opposite, with a teaching disclosure: when we teach sociology we build up this opposition because it is easier to learn that way.

20:31 Society defined by conflict, held together through power and coercion, with social problems emerging from conflict between groups.

21:53 The attitude toward change reverses. Conflict theories are positive about social change, especially where the tension can be harnessed, and especially if the less powerful side wins.

22:33 Marx introduced with a qualification.

“Karl Marx was not an academic sociologist. He was a writer and philosopher, and really a political activist in the nineteenth century. He wanted to change society in what he saw as a progressive direction. But his framing, and the way he identified how society works, is really the basis for what we’re now simplifying into conflict theory.”

23:01 Workers who own nothing but their labor and owners who own capital, the proletariat and the bourgeoisie, brought together non-consensually and unequally.

23:56 Alienation, set beside anomie as a parallel condition arriving by a different route: workers do not control the process because the owner makes the rules, and do not own the product.

24:46 The contemporary illustration: iPhone factory workers in China who cannot afford iPhones, do not set how the factory runs, and produce for people in other societies.

25:24 Where the two theories part on the same fact.

“When the subordinate group, the weaker group, gets an awareness of their position, that is when they are ready to make change. This is what concerns or frightens functionalists, and what excites Marx and the conflict theorists.”

26:10 Feminist perspectives, introduced with a caveat that they are not all conflict theories, though they lean that way because of the focus on inequality between the genders.

27:14 Patriarchy defined as a male-dominated society or subsystem, family, workplace, or school, with the term originating in the family and extended outward.

27:55 Intersectional feminism, its origin in the position of Black women, and its extension to class and sexuality. Patricia Hill Collins, who he notes retired from his own department, on Black feminist thought. Angela Davis, from activism into academia. Audre Lorde, writer, activist, and poet.

30:00 An aside on the migration of the term.

“One of the interesting things about this is that for something that kind of came from academia, intersectional feminism has really emerged into politics in the last few years in a really interesting way, where you see people actually holding up signs at rallies that say intersectional feminism. Which if you’re old like me strikes you as a big change in the world.”

30:33 Interactionism as the third of the big three, a micro-level theory about how people create and maintain social reality through language, words, and symbols.

31:34 George Herbert Mead, and society as organized and patterned interaction among individuals, so that society is real at the macro level and built from interaction at the micro level.

32:12 The self as both mental and social.

“You can’t have a self if you never interact with anybody else. Our ability to see others in relation to ourselves is reflective. Once we have social interactions and they shape our identity and our personality, we know that when we act a certain way, when we say or do something, we can use our imaginations to predict how people will react, or to understand how they react when we’re seeing their reaction. So this reflective ability is what really creates the self.”

34:07 The summary table in speech: functionalists on order held together by institutions, conflict and feminist theories on tension and coercion and the engines of change, interactionists on definition, anticipation, and labeling.

35:20 The line that governs the whole lecture.

“The key thing is that most people are not one or another of these theories. We use different kinds of ways of thinking to ask different kinds of questions.”

36:33 Research methods. Survey research first, as probably the most common, with the tradeoff stated in one breath: you can ask many people efficiently and get answers to questions like whether older people support Trump more than younger people, and it is not a rich question and will not tell you about their hopes and dreams.

37:23 Secondary data analysis, and his own practice.

“Most sociologists are not actually doing those surveys. We’re using the surveys that other people have done. This is what we call existing data analysis or secondary data analysis. Someone has done a survey, or we’ll pool all of our resources, through the government, and say let’s do a big survey, and then we’ll all analyze it. I do a lot of research using surveys. I don’t conduct a lot of surveys. I conduct a few, but mostly I’m getting the data that someone else has collected and analyzing it.”

38:02 Experiments, less common in sociology because of ethics and construction difficulty, and not necessarily reflective of the world, and useful for specific questions about response and learning under different conditions.

38:55 Qualitative research: in-depth interviews, observation, participation. And its cost.

“Qualitative research is much more time intensive, in the sense that you’re spending a lot of time studying a relatively small number of people, and there’s a lot of interpretation that has to go into making these results meaningful.”

39:47 Historical and comparative methods, comparing countries or one country across time, on questions like why some societies became rich, what causes civil wars, and what happens when democracy breaks down.

Twenty-four thousand views, twenty-five times the audience of anything else in this file, and it is the recording with the least of him in it. Somebody’s syllabus is assigning it. The largest audience Cohen has ever reached with a lecture that I have seen so far came from doing the ordinary work of his job well, and it dwarfs the reach of the fights.

The lecture is also the answer to a question the rest of the file raises. Two days earlier he opened Social Problems with presidential clips and a rally chanting a slur. This is the next lecture in the same course, and there is no politics in it at all. When he teaches conservative theory he teaches it, and when he criticizes it he criticizes the logic.

That passage at 19:29 is the best thing in the lecture. He gives functionalism its strongest form, the fever fighting the infection, then Merton on crime, then poverty, which he calls cold-hearted while granting the logic. Then he takes it apart on the ground it stands on: the body metaphor licenses you to assume that whatever exists has a purpose, and then you reason backwards to the purpose from the assumption. Everybody has lungs, let us find out why we have lungs, therefore every society has poverty, let us find out why poverty is good. He identifies the conservatism of the theory as a consequence of the metaphor.

And then he does something that his blog almost never does. He gives conflict theory the identical treatment structurally, notes at 20:22 that the opposition between the two is a teaching device, and closes at 35:20 by telling students that nobody holds one of these positions and that they are question-generating instruments. A man who is often accused of prosecuting from a fixed position spends forty minutes teaching that theoretical positions are tools you pick up and put down.

Three things carry beyond the classroom.

The construction passage at 2:14 is the most useful thing in the lecture for a general reader. The pandemic exists. What the problem is remains open: the virus, the government’s handling, or the government telling people what to do. Anyone who lived through 2020 watched that fight and mostly did not have the vocabulary for it, and he hands it over in ninety seconds without indicating which answer he prefers. Set that beside the previous lecture, where he had made his own answer unmistakable, and you have the two halves of the citizen scholar problem in a single week of one course.

The March for Life passage at 9:23 is the one that would surprise readers who know him only as a combatant. He photographed the march, and he analyzes the teenagers from religious schools with the identical apparatus he applies to the Women’s March, and the analysis is respectful. The solidarity is real, the identity work is real, and the mechanism is the same on both sides.

And at 37:23 he tells his students something about himself that the rest of the record only implies. He does not collect data. He analyzes what other people collected. That sentence explains an enormous amount: why measurement is his obsession, why he audits instruments rather than gathering fresh evidence, why his reflex when a number moves is to ask whether the survey changed, and why his disputes with ethnographers run the way they do. A researcher who lives entirely inside other people’s instruments will be exquisitely sensitive to what those instruments do and unmoved by the claim that presence in the field is its own warrant. His method and his quarrels have the same source, and he states it in one aside while explaining research designs to freshmen.

Two weaknesses, both small.

The qualitative passage at 38:55 is thin next to the survey passage. Interviews, observation, participation, time intensive, lots of interpretation required. It is accurate and it does not convey why anyone would do it or what it can reach that a survey cannot. The survey section gets its tradeoff stated with real specificity. The qualitative section gets a cost with no corresponding account of the benefit, and a student would come away unsure why that method exists.

And intersectionality gets nine minutes of names and a promise to return, which is fine for an intro lecture, and it is the one section where he tells students to go look people up. He names Collins, Davis, and Lorde and never states what the concept claims, beyond that different inequalities interact. Given how contested the term is in public argument, a definition would have served students better than a reading list.

Transcript flags. At 15:24 “email durkheim” is Émile Durkheim (1858–1917). At 17:14 and 25:14 “enemy” is anomie. At 18:46 Robert Merton (1910–2003) is credited with “dysfunctionalism,” and the term he actually coined is dysfunction, alongside manifest and latent functions; check the slide, since the concept of latent function is what the crime and poverty examples illustrate, and the poverty argument is usually credited to Herbert Gans (1927–2023) in his 1971 essay on the uses of poverty. At 19:50 “let’s see why we have loves” is lungs. At 22:37 Karl Marx (1818–1883) is dated correctly in the chapter title. At 23:03 “the economic director of capitalism” is probably the economic structure. At 31:35 George Herbert Mead (1863–1931). C. Wright Mills (1916–1962) at 10:19, The Sociological Imagination published 1959. Patricia Hill Collins (b. 1948), Angela Davis (b. 1944), Audre Lorde (1934–1992). At 5:12 “emergency claims making” is emergence and claims making, two stages run together.

‘Measuring Inequality with the Gini Index’ (Sep. 3, 2020)

0:00 States the question: is there a way to put a number on how much inequality there is, so it can be studied.

0:17 Why a summary number is needed before the normative question can be asked.

“We know there are poor people, rich people, and people in between. But can we say anything more? To answer how much is too much, we need to be able to answer how much is there.”

0:46 The zero-inequality case. Four people, Betty, Jack, Jane, and Bob, one stack of cash each.

1:09 Building the grid. Each person is 25 percent of the people with 25 percent of the money, so the cumulative shares track one to one and produce a diagonal, which he labels equality.

2:14 A more realistic case. Betty now has five stacks and the others still have one. The first three columns no longer reach 75 percent of the money, so the curve sags below the diagonal and then shoots up at the end.

3:06 The Lorenz curve, named for an economist he places in the early twentieth century, with the rule that the further the curve sits from the diagonal, the more inequality there is.

3:29 The Gini index derived by counting squares, which is the pedagogical move that carries the lecture.

“There’s sixteen squares overall in our grid, eight squares in that lower right triangle. That curved line takes about three of those eight squares. I drew it by hand, but it’s about three. We’re going to call that area Gini. That number is the percentage of that lower right triangle that is above the Lorenz curve. In this case it’s about three of the eight squares, or about 37.5 percent, so 0.375 is our Gini index.”

4:16 The two endpoints. If the curve matches the diagonal it is zero. If one person has everything and everyone before her has nothing, the whole triangle sits above the curve and it is one. So the index runs from zero to one, or zero to a hundred depending on the scale.

4:49 The United States. Census asks a large random sample what their family earns, and the answers get lined up from poorest to richest to see how far the curve bends.

5:13 Points on the American curve. The bottom 25 percent of families hold about 6 percent of income, below $42,500. The bottom 49 percent, below $77,500, hold 19 percent of income together. At the other end, 93 percent of families are under $250,000, so 7 percent are above it, and they hold 25 percent of all income.

6:03 The American family Gini at 0.45.

6:13 What having the number lets you do, which is the point of the lecture.

“Now that we have put a number on it, we can ask questions like: is it going up or down, is it more or less than France, what makes it go up or down, is it higher in New York than in California. Being able to take all the different numbers, everybody’s income is a different number, and describe that whole distribution with one number that quantifies the amount of inequality, turns out to be super useful for analyzing society.”

6:42 The American series from 1947. Falling until about the end of the 1960s, rising since, from about 35 to about 45 between the early 1970s and 2018.

7:16 Causes, offered as a few main reasons rather than a lecture. The rise of the super rich, visible in famous names and also in ordinary top executives, bankers, and lawyers.

7:39 The decline of manufacturing and unions, which had supplied middle-income jobs without much schooling.

7:58 The service economy, which he describes as bifurcated: airplane pilots and cashiers, doctors and janitors, with fewer of the middle-income jobs that had been holding inequality down.

8:16 The minimum wage falling in real terms because increases have not kept up with inflation.

8:23 Family structure, since this is family income. More dual-earner families at the top as women’s employment rose, and more low-income single-earner families at the bottom.

8:47 The Economist chart comparing Gini before and after taxes and transfers.

9:09 What the earlier numbers left out.

“The numbers I was showing you before were just total income, how much income did you have this year. It didn’t take into account how much you ended up having to pay in taxes, and what else the government did for you. Did it give you food stamps, rent subsidies, education subsidies, free health care. Once you take all that into account, you can see how much the amount of inequality you start with in income is changed by the government moving money around.”

9:44 The American position: a red dot out past 45 before taxes and transfers, higher than the other countries shown.

10:04 Ireland, which starts with more inequality and knocks about twenty points off its Gini through taxes and transfers, in cash and also in education, health care, and housing. France, Canada, and Sweden also doing more than the United States.

10:38 The conclusion of that section: the United States has a lot of income inequality and does less about it than a lot of other rich countries.

10:48 Global inequality, and a first pass at it.

10:58 The between-country method: treat every person in a country as though they had that country’s average income. India near the bottom with about 20 percent of world population and 6 percent of income. China near the world average, about 22 percent of population and 21 percent of income. The United States at the top with fewer people and a large share of income.

12:04 The result: treating each country like a family gives a world Gini of 0.46, close to the American family figure.

12:21 The refinement. Instead of 116 countries, each country is split into seven slices, bottom 10 percent, 10 to 20, 20 to 40, 40 to 60, 60 to 80, 80 to 90, top 10, with everyone inside a slice assumed to have the same income.

13:00 The single most striking fact in the lecture.

“Down at the red circle at the bottom there is the bottom 10 percent of the US. The bottom 10 percent of the US is above half the world’s population in terms of income.”

13:27 The result of the refinement, and what it implies.

“This gets us to a Gini of 0.58. You can see more inequality, but much bigger is simply the inequality between countries in the first place. So that’s one lesson you can take from that. There’s a lot of inequality between countries, rich countries and poor countries.”

13:45 The summary. Many numbers turned into one number, which then permits comparison across time and place, with a note that the Gini is only one of several ways to describe a distribution and is chosen here for its intuitive reading.

This is the best-constructed thing of his in the file so far, and it is fourteen minutes on a topic that any textbook covers in two paragraphs.

He builds the grid from four named people, moves one person’s money, watches the line bend, and only then counts squares. By the time the word Gini arrives at 3:50 the student has already seen the quantity it measures. And the counting is done by eye, by hand, on sixteen squares, so a student who cannot integrate can still see that three of eight is 0.375. He never writes an integral.

Two things in the lecture matter beyond the pedagogy.

The pre-tax and post-tax comparison at 8:47 is the whole American inequality argument in one chart, and he lets the chart make it. The United States starts with high market inequality and then does less redistribution than its peers, so the gap at the end is wider than the gap at the beginning. Ireland is the example that does the work, since it begins worse and finishes considerably better. That is an argument for a policy position, and he presents it as a comparison rather than a conclusion, which is the register he uses when he is teaching rather than prosecuting.

The global section is the more interesting half and it contains the sentence that would unsettle his own students most. The bottom tenth of Americans sits above half the world’s population in income. He states it flatly, does not soften it, does not moralize it, and moves on. A course that has spent the semester on American inequality now has a fact in it that reframes the whole subject, and it comes from applying the same tool at a different scale. That is what a summary statistic is for.

And the finding at 13:27 is the substantive payoff. Splitting every country into seven slices raises the world Gini from 0.46 to 0.58, which means within-country inequality adds something and between-country inequality dominates. Where a person is born explains more of world income inequality than where they land inside their own country. This is Branko Milanović’s territory and Cohen does not cite him.

Now the weaknesses, and this lecture has one that would draw his own fire.

He teaches the Gini as though it were transparent. Two distributions with different shapes can produce the same index. The measure is most sensitive to changes around the middle of the distribution and least sensitive at the tails, which is a strange property for a country whose inequality story is overwhelmingly a story about the top one percent and the top tenth of one percent. He lists the rise of the super rich as cause number one at 7:23, using a statistic that is close to indifferent to what the super rich do. Top-share measures, the 90-50 and 50-10 ratios, and the Theil index all exist because the Gini hides things. He gestures at this at 14:04, saying the Gini is one kind of number and we could get into other math, and then stops. A man who built a career on the argument that the instrument shapes the finding taught a fourteen-minute lecture on an instrument and gave its limitations eight seconds.

The family-income Gini he shows and the tax-and-transfer chart from the Economist are not the same measure on the same units, so the 0.45 and the red dot are not strictly comparable, and a student would reasonably think they were. And the causes list at 7:16 mixes settled findings, like the decline of manufacturing employment, with contested ones, like the family-structure contribution, without marking the difference.

This is the recording that shows what he is when nobody is watching him fight. The Ithaca news desk is in it: fixed length, deadline discipline, an explanation that has to land in one pass with no chance to take a question. So is the Census Bureau apprenticeship, in the reflex of asking what the number was built from before asking what it means. And so is the thing he told Pepin about not collecting his own data, because this is a lecture about how to read somebody else’s.

Transcript flags. Every instance of genie is Gini, after Corrado Gini (1884–1965), the Italian statistician who introduced it in 1912. The Lorenz curve is after Max O. Lorenz (1876–1959), who published it in 1905, so his early twentieth century is right by a whisker. At 12:37 “812 quantiles” is garbled and should be 812 country-quantiles, which is 116 countries times seven slices; the arithmetic works, so the figure is recoverable. At 1:38 “20 00 five seven five hundred percent” is 25, 50, 75, 100 percent. At 8:34 “low single learner” is single-earner. At 13:19 “the averaging” is the average income. At 5:29 the $42,500 and $77,500 thresholds and the 6 and 19 percent shares should be checked against the slide before you print them, since the audio gives no year for the underlying Census file.

‘COVID-19 Economic Crisis’ (Sep. 20, 2020)

0:00 Sets the plan: what led to the crisis, its contours and extent, and some proposed solutions.

0:15 The pandemic curve from March to late September, with the April spike in deaths, the summer rise in cases, and a smaller rise in deaths.

0:37 He rewinds to March 15, when the epidemic looked small and the economic response was already enormous.

0:45 The New York Times front page of March 16: soaring toll, economy shutting down, business grinding to a halt, schools and bars and restaurants closing.

0:58 The distinction that opens the analysis.

“It wasn’t all because of the policy of locking things down. It was also that people stopped traveling, people became afraid, people stayed home. So you had a big contraction in the economy of all the things that involve people going places and doing things with other people.”

1:22 A teaching note: every slide carries the take-home message written in red, and students taking notes should look there first.

1:36 Google mobility data across all fifty states, showing movement falling by close to 40 percent in New York by April, with varying declines elsewhere, then partial rebounds. He points out the Sturgis motorcycle rally visible in the South Dakota line in August.

2:33 Global perspective. The ten largest economies from 2009, with the 2008 crisis producing a decline he calls what used to be considered very big, then a decade of growth, then the first two quarters of 2020, with the United States approaching a 10 percent drop in the second quarter and not the worst performer.

3:39 American employment from 2000. The 2007 to 2009 decline to about 129 million, the long rise through the Obama and Trump years past 150 million, then 22 million jobs lost in March and April.

4:32 Employment rates for prime-age workers, 25 to 54, down 11 percent from February to April, proportionally more for women, and below where it stood during the 2008 crisis.

5:03 The shape question, which is the most useful forecast in the lecture.

“That purple line is the 2020 recession. The drop is unprecedented, but the other thing to see is that the shape of what’s happening now may be different. We had a very rapid rebound in the first three months after April. We’re still down eight percent in employment, and you can see the contrast with previous recessions, which happened slower and then took longer to recover. We don’t know if we’re going to be rocketing back up or if we’re going to be spending years and years coming back to where we were.”

5:48 Business closures tracked through Yelp listings, approaching 100,000, with the temporary category increasingly converting to permanent.

6:24 The uneven distribution across sectors. Restaurants and retail hardest hit, breakfast and brunch places, burger and sandwich shops, gift shops, clothing stores. Least affected: lawyers, architects, accountants, and outdoor trades like tree work and roofing.

7:07 Occupations with the largest employment losses: waiters, cooks, cashiers, retail sales, maids and housekeepers, and some construction and laborer work.

7:36 Census data on households losing employment income, which he reads as a catalog of American inequality.

“This basically is a catalog of US inequality. Hispanic and Black workers are the most likely to have lost employment income, over fifty percent of them. It’s also single people, divorced and separated people. Then look at the income gradient: the people with the lowest incomes are by far the most likely to have seen a loss of income during this period.”

8:17 Health status in the same chart, with the interpretation.

“People who already had more health problems are more likely to be suffering. That’s because of where they were socially located in the first place. The people with these service jobs, and other jobs that were hard hit in places that were hard hit, were people who were likely to already have health problems.”

9:08 A composition argument, worked out aloud, for why the gender difference looks small. Most households contain a man and a woman experiencing the same shock, more single parents losing jobs are women, and more retired people on fixed incomes are women, so the two run against each other.

9:43 Unequal receipt of unemployment benefits.

“It’s a common myth that everybody who’s unemployed gets the benefits. In fact it requires paperwork, it requires somebody to approve the paperwork, it requires the time and energy and know-how of how to get it, and you have to be in a place where the bureaucracy is supportive of that. For all of these reasons, whites are much more likely than blacks who are unemployed to be getting unemployment benefits.”

10:36 Food insecurity. Twenty-two million American adults in households that sometimes or often did not have enough to eat in the previous seven days, as of the end of August. Higher among households that lost a job, among Black and Hispanic respondents, among those under $25,000 where more than a quarter reported it, among the less educated, and among families with children.

11:28 Housing background. Rent burdens already high and rising through the previous decade for households under $75,000, with most of those under $15,000 paying between 30 and 50 percent of income on rent.

12:15 Matthew Desmond’s formulation, that the rent eats first.

12:24 Census figures on households not caught up on rent, at 15 percent, highest among Black households, with the same education and income gradients.

12:55 A moment of stated ignorance worth marking.

“I don’t know what’s going on with the people over two hundred thousand dollars, why twelve percent of them are having a hard time paying their rent right now. That’s something to look into.”

13:02 Eviction, and its place in American poverty.

13:20 A news segment from Harris County, Texas, running about three and a half minutes. A constable’s deputy executing eviction orders, a father who lost his job and fell thousands behind, a family whose possessions fit in a stroller, the deputy noting he has six children of his own, 200 eviction orders through the county courts in one week against a normal month’s hundred, and an elderly tenant given a one-day reprieve because it is too hot to put her out.

16:57 The CDC eviction moratorium explained. A health-based order rather than a housing one, with five conditions: seeking government assistance, income under $100,000 or $200,000 for a couple, an affirmed substantial loss of income or extraordinary medical expense, partial payments where possible, and a showing that eviction would produce homelessness or crowded doubling up.

17:52 Why it may not work, which is the sharpest policy analysis in the lecture.

“It’s unclear, because the guidelines are vague and subjective, who’s going to actually be protected under this moratorium. It turns out it really depends what state you live in, even what county, even what judge you get when your eviction case comes to court. People still may have to go to court to fight these evictions. The other problem is it’s not paying rent for people, it’s just saying you can’t evict them right now. So there’s no relief for landlords, and a lot of landlords can’t make their payments while they have tenants who are not able to make their rent payments. And the rent is accruing. When the moratorium is up, the rent you owed will now be a debt. So we’ll have to resolve that later, or else we’re going to evict millions and millions of people when the moratorium ends.”

18:43 He introduces a set of proposals from an economist and former labor secretary, and plays the video.

19:04 The embedded argument, in five parts: contain the virus first because there is no recovery until it is contained, citing Germany, South Korea, and Italy on testing, tracing, and isolation; extend the supplemental unemployment benefits due to expire for roughly 25 million people; prevent a wave of evictions and foreclosures, citing 32 percent of households missing a July rent or mortgage payment; shore up state and local budgets against cuts to schools, child care, nutrition, mental health, housing, and health care; and stop letting national debt concerns limit the bill, since the relevant figure is the ratio of debt to the size of the economy.

20:26 Inside that video, the contested claim.

“Now some say the extra unemployment benefits have discouraged recipients from seeking jobs. That’s rubbish. Given the size of the economic collapse, few jobs are available anyway, and normal unemployment benefits typically pay a small fraction of the wages of jobs that were lost.”

23:16 His own summary in his own voice. Not everyone agrees, and there are common elements across the debate.

“One key point that economists tell us is that it’s good, it’s necessary, and it’s fine for the government to be borrowing money, a lot of money right now, to help mitigate this economic crisis. Interest rates are very low. Getting the economy going is better for the national debt in the long run anyway. So spending money is good. The question is how to spend the money.”

23:50 His list: cash stimulus, continued emergency unemployment insurance, rental assistance rather than only a moratorium so that renters and landlords both survive, expanded food support including meals for children whose schools are closed, and aid to state and local governments facing layoffs of firefighters, medical workers, and municipal staff.

Seven recordings now, and a pattern has closed that none of them shows alone.

Where Cohen teaches his own research field, family demography, he does not editorialize. The family violence lecture has no politics in it. The Gini lecture has none. The introduction to social problems teaches functionalism fairly and takes it apart on logic rather than on politics.

Where he teaches an adjacent field, he imports an advocate and hands over the microphone. In April it was three minutes from the Economic Policy Institute on paid sick leave. Here it is four minutes from Robert Reich (b. 1946), who served in a Democratic cabinet, on what should be in the stimulus bill. Both times the numbers in the clip are accurate. Both times the framing and the policy conclusion arrive in somebody else’s voice, and both times the students are not told what the organization is or where it sits.

He is careful in the domain where he could be caught. Outside it, he outsources.

Look at what the outsourcing carries here. At 20:26 the video dismisses as rubbish the claim that supplemental unemployment benefits reduced job search. That question was live and had real evidence on both sides. The $600 weekly supplement produced replacement rates above 100 percent of prior earnings for roughly two-thirds of eligible workers, which is arithmetic rather than ideology. What the research mostly found, then and since, is that the disincentive effect on employment was small despite those replacement rates, which is a striking result worth teaching. So Reich landed close to the right answer and called the question rubbish, and a class watching that clip learned the conclusion without learning that anyone had checked. Cohen, who spends his Decembers counting whether other people’s claims can be inspected, played it without comment.

The strongest analysis in the lecture is his own and it is at 17:52. The eviction moratorium came from the CDC on public health authority, its criteria were vague enough that protection depended on the state, the county, and the judge, it left landlords with no revenue, and it converted unpaid rent into accumulating debt that would come due all at once. Every one of those turned out right. The moratorium’s legal foundation collapsed at the Supreme Court in August 2021, enforcement did vary enormously by jurisdiction, and rental debt was the central problem the emergency rental assistance program was built to address. He worked that out in three minutes from the text of the order.

The employment-shape question at 5:03 is the other thing he got right by refusing to answer it. He put up the comparison, noted the rebound was unlike anything in the prior recessions, and said we do not know whether this rockets back or takes years. It rocketed back. A more confident lecturer would have picked one.

Two problems with the instruments, and they run in his blind spot.

Google mobility, Yelp closures, and the Census Household Pulse Survey are the backbone of this lecture, and he treats all three as though they were the American Community Survey. The Pulse was explicitly experimental, fielded weekly online with a response rate around ten percent, and Census attached warnings to it that he does not relay. The twenty-two million food insecure adults and the fifteen percent behind on rent both come from it. Yelp’s closure counts describe businesses that maintain Yelp listings. Google’s mobility index describes people carrying phones with location history enabled. Each is a reasonable emergency instrument and none of them is a probability sample of the country, and the man who built a career on asking what an instrument was made of skipped the question three times in one lecture.

The other is inside the news segment. At 13:47 the reporter says forty million Americans face eviction. That figure came from an Aspen Institute projection and was an upper bound built on the number of renters in households experiencing income loss. Actual eviction filings in 2020 ran well below normal levels, partly because of the moratoria and partly because of the rental assistance. He plays it without correction. That is the kind of number he would take apart in a Wilcox post inside a paragraph.

The moment I would keep for a profile is at 12:55. He puts up a chart, notices that twelve percent of households above two hundred thousand dollars report trouble making rent, says he does not know why, and says it is something to look into. No hypothesis offered, no gesture at one, no embarrassment. That is a small thing and it is the honest professional reflex the rest of the file keeps circling: the number does not fit, say so, move on, come back to it.

Transcript flags. At 19:01 “robert wright” is Robert Reich. Kobit and covent and copenhagen are COVID. At 1:01 “the potamic” is the pandemic. At 12:22 “the rent deeds first” is the rent eats first, from Matthew Desmond’s Evicted (2016). At 6:30 “all ass all sectors” is all sectors. At 8:18 “fair health poor health good health” is a chart category list. At 9:23 “old women” is older women. At 14:07 “idiom russia’s but doing that get everything we need” is unrecoverable. At 15:25 “before copenhagen” is before COVID.

‘Covid-19 and health disparities’ (Oct. 5, 2020)

0:14 He opens with the CDC definition. Health disparities are the preventable differences in the burden of disease, injury, violence, or opportunity to achieve good health, borne by socially disadvantaged populations. He stresses the word preventable. Something in the social system and the health system produces the gap.

1:07 The theoretical spine is the theory of fundamental causes, which comes from Bruce Link and Jo Phelan. It exists to explain a paradox. Health improved enormously over the last century and a half through water, sewers, nutrition, medicine, education, and changes in behavior. Life expectancy rose for everyone. Yet socioeconomic status still predicts health as strongly as ever. The answer is that income and education deliver money, knowledge, prestige, power, and social connections, and those resources protect against whatever the current threat happens to be. The threat changes. The resources keep working.

3:00 Life expectancy and income. For the 1960 birth cohort, among men in the poorest fifth, roughly 30 percent survived from age 50 to 85. In the richest fifth the figure runs about two thirds to three quarters higher, and higher for women than men.

3:51 Smoking and education. Among people with graduate degrees, 3.7 percent smoke. Among high school graduates the figure approaches 20 percent. Among people whose credential is a GED, more than a third.

4:48 Infant mortality and race, back to 1850. Infant mortality has fallen enormously. The Black and White gap persists and now exceeds two to one, a wider ratio than in 1850.

6:14 The education gradient in infant mortality holds inside every racial group, and race persists inside every education level. Black women with college degrees have higher infant mortality than White women who finished only high school.

7:39 The opioid epidemic as a second illustration. Overdose death rates cluster in Appalachia, parts of the Northeast, poor counties in the West, and reservation areas. These are places hit by deindustrialization and by shrinking state revenue for schools and health care. Geography of socioeconomic status producing health disparities.

8:50 Why the pandemic is severe. A novel coronavirus, so no prior immunity. High contagiousness. Asymptomatic transmission, which defeats isolation. And the disease is bad, worse than seasonal flu. As of the recording, of roughly one million deaths worldwide, more than 200,000 were in the United States, far above the American share of world population.

10:01 Transmission. Droplets from coughing, sneezing, loud speech, yelling, singing, plus aerosols that linger. Risk rises with proximity, duration, and poor ventilation. People who can separate themselves from others get sick less.

11:07 Institutional outbreaks. Nursing homes account for something like 40 percent of all deaths, with more than 77,000 residents and staff dead. In many rural counties the nursing home was the local epicenter. Prisons show more than 200,000 infections and 1,265 deaths by a New York Times count, with 3,000 positives at Avenal State Prison and 2,500 at San Quentin. Prisons churn, so the virus travels in and out with prisoners and staff. Factories follow the same pattern, with more than a thousand cases each at the Smithfield plant in South Dakota and the Tyson plant in Iowa.

13:59 The telework gradient. Bureau of Labor Statistics data on who reports working from home because of the pandemic. In computer and math occupations, over 60 percent. Lawyers, architects, engineers, managers, and scientists cluster high, all above $1,500 a week. Food service, health care support, personal care, and building and grounds workers sit near $500 to $600 a week, and under 10 percent of them work from home. Truck drivers, bus drivers, sales and office support keep going out.

15:38 The same gradient by schooling. Half of workers with advanced degrees, and Cohen adds “such as me,” work at home at least part of the time. Among workers without a high school diploma, four percent. He draws the analogy directly to smoking: a behavior that raises health risk, concentrated among the less educated, and in this case nobody chose it.

16:37 Paid sick leave. He runs a produced video segment. The CDC says stay home when sick. Twenty-seven percent of private sector workers get no paid sick leave, over 30 million people. The United States is one of two OECD countries without a federal mandate. Those workers concentrate in restaurants, hotels, transportation, travel, and tourism, exactly the contact-heavy jobs. Fewer than half of leisure and hospitality workers have paid sick leave. Among the bottom tenth of wage earners, 30 percent do.

18:27 Disparate impact. COVID-19 death rates per 100,000 by age and race. Deaths rise with age in every group. The Black and Hispanic gap against White rates is largest among working-age adults, 18 to 65, where Black and Hispanic death rates run five to eight times higher. For Black Americans over 85 the rate reaches 1,600 per 100,000, meaning 1.6 percent of that entire population died.

20:16 The wrap. Resources shape living conditions. Education and discrimination shape which jobs people hold. Workplace policy shapes exposure. Health care shapes outcomes after infection. He points at the president and his circle as the illustration: reckless behavior, infection, and then medical care that no ordinary patient could obtain.

Cohen teaches one theory and then climbs a ladder of cases toward the pandemic, so that by the time COVID arrives the students already know what they are supposed to see. Smoking, then opioids, then telework. Each rung is a health-relevant behavior that tracks education. The third rung is the strongest thing in the twenty-two minutes, because a warehouse worker cannot choose to telecommute, and once you accept that, the choice language around smoking starts to look shakier too. He touches this and moves on. He should have stopped and hammered it. It is the argument the whole ladder was built to deliver.

The single most portable fact he gives is the crossing pattern in infant mortality: Black college graduates faring worse than White high school graduates. He delivers it in one sentence. That finding is the strongest available evidence that socioeconomic status does not absorb race, and it is also the finding most contested on interpretation, since SES measured by credential is a poor proxy for wealth, and maternal age structure and immigrant selection among Black mothers both complicate the comparison. A stronger lecture names the objections and answers them. This one states the number and keeps walking.

The weak point is mediation. He shows a five to eightfold race gap among working-age adults and then says he thinks we can trace it back to preexisting conditions and to the inability to work at home. That is the theory predicting the answer rather than the data producing it. Household size, multigenerational living, transit dependence, and the accident that the virus seeded first in New York City were all carrying weight in the fall of 2020. None appear. The lecture never names a rival explanation, which means students learn a framework and simultaneously learn that nothing else was ever on the table.

The nursing home number sits unreconciled with the occupational story. If 40 percent of deaths occurred in and around long-term care, the causal chain cannot run mainly through who commutes to work, because residents do not commute. It has to run through who ends up in which facility, how those facilities are staffed and reimbursed, and how thin the wage floor is for the aides who move between them. Fundamental cause theory handles this well. Cohen has the pieces on the same slide and leaves them apart.

If resources protect against whatever threat is current, then the affluent should capture tests, then treatments, then vaccines, in that order and ahead of everyone else. In October 2020 that prediction was two months from its natural experiment. He gets close with the president’s care and treats it as an anecdote about recklessness instead of the test case.

Two cautions. At the oldest ages, differential survival to 85 complicates any simple race comparison, since the Black population reaching 85 has already passed through heavier attrition. And the rate he reads as 1.6 percent of all Black Americans over 85 is a cumulative figure over about seven months, which students will hear as an annual risk unless told.

The sick leave segment changes genre. Music comes up, a produced advocacy video plays, and a named policy is called essential. Whatever one thinks of the policy, and the public health case for it is strong, the students watching are graded by the man who queued the clip. Same with the closing turn on the president, delivered a month before a presidential election. This is the pattern from his corpus generally. The arithmetic travels anywhere. The verdicts travel only among people who already share them, and they give anyone looking for a reason to discard the arithmetic a free one.

When he reaches the education gradient in telework, he says half of workers with advanced degrees, “such as me,” are working from home. He puts himself inside the distribution he is describing, on the protected side, while explaining why the protection exists.

‘The Pandemic and the Family’ (Oct. 27, 2020)

0:07 Introduces the talk as the family demography edition, added because the number of family topics touched by the pandemic is effectively infinite.

0:23 Sets aside the direct effects, sickness and death inside families, because the mortality and infection data cannot yet be located inside households.

0:57 The frame he offers instead: social metabolism, his term for the general slowdown, which he says is inaccurately called a lockdown.

1:11 Then unemployment and economic insecurity, which he describes with a rare personal hedge.

“What has really turned into a seismic inequality event. I don’t want to say we’ve never seen anything like this, but I’ve never seen anything like this that I can think of, that has been not only this big but this unequally experienced.”

1:43 The demographic agenda: births, then marriage, then divorce, with the acknowledgment that in October the births could not yet show pandemic effects.

2:23 The thesis, stated as an empirical observation rather than a theory.

“The hypothesis is basically there’s less of everything demographic happening. And so what does that mean for families and the future of families and life experience? I wish I had more wisdom on that, but that’s my empirical observation.”

2:46 The sources, with the caveat attached before the evidence: vital statistics from various states, some miscellaneous items, and a lot of Google search data, which he corrects himself on mid-sentence. “It’s at best speculative. Or I should say suggestive.”

3:09 Google mobility data, aggregated from pings at groceries and pharmacies, transit stations, retail and recreation, residential, and workplaces. He notes he averaged counties up to states without weighting: “Don’t hold me to this methodology exactly.”

3:56 Mobility crashing 15 to 40 percent across states by late March, the reopening rebound, and renewed declines.

4:21 Face-validity checks he points to himself: Hawaii not recovering because of tourism, and the Sturgis motorcycle rally visible as a bump in the South Dakota line.

5:01 Google search terms as a proxy for social life, prefaced by the disclosure that matters most in the talk.

“This is for suggestive purposes. I didn’t keep a record of all the things I tried that didn’t look good. And Google isn’t very transparent with how they release this data.”

5:26 The searches that fell: breathalyzers, party invitations, weekend, and happy hour, which he says was decimated, all dropping sharply after March 9.

6:12 Young adults living at home. Normally about 42 percent of people aged 18 to 29 live with parents, grandparents, or in-laws, and the figure spiked to almost 50 percent in June, roughly three million more.

6:54 The measurement caveat he supplies unprompted.

“We can’t say from this if they were moving back, or this could also have been achieved just by people who normally move out not moving out.”

7:24 Employment change by occupation since January, top nine occupations for each sex.

7:47 The class gradient, delivered with the one joke in the talk.

“Chief executives, that includes small business owners, managers, software developers actually increased. Thank you, Zoom. They were much less impacted than the working-class jobs: janitors, retail sales, construction workers.”

8:22 Retail sales as the rare occupation with enough of both sexes to compare: almost 60 percent of women in retail sales were out of work in April against about 40 percent of men, and by October still 20 percent against about 5 percent.

8:44 The reversal from the previous recession. Women lost more jobs than men, unlike the 2009 downturn.

8:58 The same chart for Black men and women, with larger shocks, and personal care aides where nearly 60 percent of Black women were out of work at the trough.

9:30 Labor force participation for parents with resident children, down about four percent overall, recovering faster for fathers, with the autumn decline steeper for mothers, which he links to school closure without asserting it.

10:32 The Census Household Pulse Survey, which he calls “pretty good quality,” on households losing income.

10:57 The result.

“Overall it’s about 45 percent of people, them or a household member, lost income, which is insane. Whites the least, Hispanics and Blacks the most. Never married or divorced people the most, poor people the most, people in bad health the most, people with less education, people with children, and younger people.”

11:28 Gender as the exception, and his own working explanation for it, offered as unverified.

“Gender is a wash, and I think that’s because you have old women on fixed incomes who haven’t lost income and young women, single mothers, who have lost income, and so it balances. I haven’t looked at the micro data, so I’d have to sort this out.”

11:51 The theory of planned behavior as the framework for family decisions: background factors, demographic characteristics, social context, beliefs about consequences, expected social support, perceived control, intention, birth.

12:56 What the pandemic does to that diagram, walked through item by item: emotions shot, income devastated, social context wiped out, beliefs about consequences unknown, social support out the window, interfering factors everywhere. “A giant red splotch flies onto this graph.”

13:51 An aside noting that the photographs in the slides are his own, taken when he came out of his basement once.

14:07 The pandemic baby boom prediction, dismissed.

“I don’t know if you remember that charming era six months ago when people were saying things like, oh, there’ll be a pandemic baby boom because people are all stuck at home. That was a hypothesis. It was not well grounded, and I don’t think it’s going to bear out.”

14:23 Condom sales as an indirect measure: Durex slumping, double-digit consumption declines at the company that owns Trojan in the second quarter, most brands recovering in July and Trojan not.

15:03 The inference, and the teaching aside inside it.

“This is a great methods assignment. Condoms prevent births. If there are fewer condoms, will there be more births? No. So I think people are not having as much sex. That’s what I’m getting at.”

15:18 The Guttmacher survey fielded in late April and early May: about a third of women aged 18 to 49 said the pandemic had made them want to delay childbearing or have fewer children, with Black, Hispanic, queer, and poor women more likely to say so.

16:24 The comparison to a version of the same question asked in 2008, which shows a larger effect now.

16:46 The care gap alongside it: 39 percent more careful than previously about contraceptive use, 25 percent more worried about affording it, and delayed or cancelled reproductive health care. “This is not a recipe for happy family development.”

17:17 Google Trends methodology, explained in full: each term scored zero to a hundred by week, five-year window, terms chosen for contemporaneous correlation, combined into indices with a reported alpha.

17:50 The contraceptive index, built from birth control, condoms, and depo shot, alpha 0.7.

18:31 The comparison that saves him from a bad inference: three grey lines for 2016 through 2018, 2019 in black, 2020 in red, showing that a decline was already underway in 2019.

“I do that just to show you that we already had a decline in 2019, and this is a theme we have, which is declining fertility in the US. So that’s going to complicate our interpretation of all of these.”

19:27 Sex-related searches as a proxy, and the self-correction two sentences later.

“What would people Google if they were sexually active? Premature ejaculation, pain during sex, pain after sex, orgasm during sex. These all show declines to varying degrees this year compared with last year. Orgasm during sex, I realize later, maybe that’s actually porn searches. I’m not sure.”

20:39 The pregnancy index: am I pregnant, pregnancy test, morning sickness, missed period, alpha 0.74, with a large drop in March and April and no normal spring bounce.

21:44 The alternative reading he supplies against his own conclusion.

“You never know, maybe people are just Googling other things and not bothering to think about whether or not they’re pregnant. But with the sex and the contraception and the condom sales and all this, I think there’s less sex and less pregnancy. That’s where I’m going with that.”

22:17 Births, which he says are too early to read, except for the possibility of miscarriages and abortions, which he flags as outside his expertise and poorly measured, with the stress hypothesis called “purely speculative.”

22:59 Monthly rolling birth data from Florida and California, the only two states he could find publishing it.

23:29 The comparison line on each chart is the average annual decline over the prior three years, about 0.7 percent for Florida and closer to 3 percent for California, so the pandemic drop is measured against an existing trend.

23:53 California’s September figures, posted the previous Sunday, showing a roughly 10 percent drop against September 2019, with four possible explanations offered and none chosen: premature births, miscarriages, immigrants going home, or something else.

24:18 The forecast, and the disposability of his own talk.

“This is going to be updated when we see probably a catastrophically giant crash in births in December and January and February. If that happens, we won’t remember this talk.”

24:44 An apology to the strangers photographed on the National Mall who ended up illustrating his marriage slide.

25:07 The wedding index: wedding invitations, wedding shower, bridal shower, wedding dresses, alpha 0.94, showing a catastrophic crash in 2020.

25:54 The four readings he will not choose between: people sneaking to the courthouse, people completing licenses without acting on them, relationships falling apart, or minds changed about marriage.

26:41 The forgone-versus-delayed distinction, carried over from fertility.

“For every hundred births delayed, some are forgone, some never occur. Relationships break up, people’s bodies get older. The recovery of delayed births is less than a hundred percent. I don’t really have the basis for guessing what the recovery on delayed marriages will be, and that’ll be a super interesting question.”

27:13 Marriage license searches, excluded from the index because they did not correlate, showing a first-quarter drop and substantial recovery.

27:52 The paper with Brandon Wagner and Kate Choi, “Decline in Marriage Associated with the Pandemic,” and the results from three jurisdictions.

28:08 Florida: cumulative recorded marriages down roughly 25 percent, with the monthly shortfall tapering by June and July.

28:41 Hawaii: close to a 50 percent drop, most of it non-residents not coming to marry there.

29:04 Seattle marriage applications: down about 10 percent, and tapering.

29:28 What the gap between the three means.

“Some evidence that the decline in weddings is going to be greater than the decline in marriage. Maybe some people are managing or deciding to execute their marriage without having the wedding they dreamed of. That gives a hint that some of these lost marriages may be recovered, in the sense that a lost wedding may still become a marriage.”

30:08 Divorce, and the news story he takes apart.

“A big news story that broke a while ago: divorces are rocketing up during the pandemic. I don’t believe that. That story was based on a company press release, a company that sells an online divorce kit, and their downloads went up. It seems pretty likely that people who were considering divorce would be shifting from in-person to online divorce processing.”

30:56 Why the timing makes an increase implausible: divorces take processing time and some states impose waiting periods.

31:03 His expectation instead.

“Even if, as we have reason to believe, this pandemic is catastrophic for marriages and families, and more people want to get divorced, I wouldn’t expect to see more divorces yet. It may very well be not unlike the Great Recession: decline in divorces, possibly followed by some rebound as the people who had to put off their divorce because of housing and the expense and the moving and the kids, all the reasons that people who kind of want to get divorced don’t get divorced, that may rebound when the logistics become more practical.”

31:43 The divorce ideation index: divorce lawyer, divorce attorney, get a divorce, how do I get a divorce. A decline followed by a rebound that pushes above earlier years.

32:22 And then the sentence that is the best thing in the talk.

“This is very speculative, but it does fit my presumption, which is probably why I’m showing it to you.”

33:00 Florida divorce records in three-year bins back to 2006, with the pre-existing downward trend visible, and a large 2020 drop. He canvasses whether clerks were closed and reports that his coauthors found most were not.

33:48 The rebound in the percentage difference by September, with the arithmetic point that a rebound only recoups the lost divorces if the cumulative gap closes to zero.

34:28 Conclusions, illustrated with a photograph of a basketball court near his house where the hoops were removed and people now play soccer.

34:36 The open question: what recovers, against what gets worse.

“The fact that the economic shock is not as bad now as it was a few months ago doesn’t mean that the shock to the decision-making system of families deciding whether or not to have children is reduced. Maybe they’re catching their breath and thinking it over and thinking, definitely not having children.”

35:23 The forecast he stakes out.

“The big demographic story of 2021 is likely to be the very large decline in births. Everything points that way, and I think that’s going to be quite shocking when we see how big that is.”

35:46 What he is not covering: violence and abuse, “with reason to believe there’s more violence and abuse being detected less”; educational disparities; the gender division of labor, where more men are doing housework and child care while more work has moved into a home that “is generally not the place where you find the most gender equality”; a mental health catastrophe; and the opioid epidemic merging with the rest.

37:00 The close: “It’s a disaster, it’s a catastrophe. If you’re studying anything in this space, you’re going to be studying this for a long time. I expect to be studying this for the rest of my career, however long that is.”

This is the best single document in the file so far for understanding how he works, because it is the only one where he shows the method operating under conditions where the method barely holds.

The subject is a set of demographic events that have not happened yet. He wants to know about births in October 2020, and the first pandemic conceptions could not have been born before December. So he assembles proxies: condom sales from earnings calls, Google search indices for contraception and pregnancy and weddings and divorce lawyers, mobility pings, two states’ rolling vital statistics, and a survey question fielded in April. Every one is weak. He says so about every one, before showing it.

Count the disclaimers. At 2:46 the data is suggestive rather than speculative, corrected in the same breath. At 3:40 don’t hold me to this methodology. At 5:10 he did not keep a record of the searches he tried that did not look good, which is a confession of the file-drawer problem in his own slide deck, volunteered to a room of demographers. At 6:54 the living-at-home spike could be people returning or people never leaving, and the data cannot distinguish them. At 11:35 he has not looked at the micro data behind his own explanation of the gender wash. At 18:31 he puts up three prior years to show that fertility was already falling, which complicates every reading that follows. At 20:07 he wonders aloud whether one of his own search terms is measuring pornography. At 21:44 he supplies the alternative explanation for his pregnancy index. At 22:33 he says nobody measures miscarriages well.

And then at 32:22, the line: this is very speculative, but it does fit my presumption, which is probably why I’m showing it to you.

I have not seen a working scientist say that about his own slide. It is the file-drawer problem, confirmation bias, and motivated selection admitted in one clause, at the moment of committing them, without any apparent discomfort. Anyone writing about his epistemics has to reckon with that sentence, and it complicates the picture the audits give. The man who counts other people’s replication packages is capable of narrating his own bias while it operates.

The forecast record here is his best. He predicted a large decline in 2021 births, and American births fell in 2020 and the first months of 2021 before recovering, with the birth-dearth story dominating demographic coverage that year exactly as he said. He predicted that divorces would fall rather than rise and that the viral story to the contrary rested on a vendor press release, and he was right on both. He predicted that the wedding decline would exceed the marriage decline, and he had the Seattle-versus-Hawaii spread to support it before anyone else was looking. He predicted a partial rather than full recovery of delayed events, which is the standard demographic result and which held.

The September California drop of 10 percent, offered with four candidate explanations, is the sort of single-month figure in a monthly series that usually turns out to be reporting lag, and he lists that possibility only obliquely.

Two things for the biography.

The condom passage at 14:23 is what a demographer does when the data he needs does not exist for eighteen months. He reads earnings calls. He then converts it into a teaching problem, out loud, in the middle of the talk: fewer condoms should mean more births, the sign is wrong, therefore the mediating variable is sex. That is the whole discipline in ninety seconds, and it is funnier and more instructive than anything in the lecture recordings.

And the divorce-kit takedown at 30:08 is the Rosin method applied at conference speed. A number went up, the number came from a company selling the product, the vendor is new, the shift from in-person to online is the obvious confound, and divorces have statutory waiting periods so the timing is impossible. Four objections in forty seconds, none requiring data he did not have. That is what he was doing to Rosin in 2012 and to Goffman in 2015, and here he does it in passing to a press release.

He builds indices from Google Trends, reports Cronbach’s alpha for each, and posts none of the search terms as a file, no code, no downloaded series. The audits he runs every December ask other people to ship exactly that. This talk is a slide deck, and slide decks are not held to the standard, and the asymmetry sits there unremarked.

Transcript flags. At 15:19 “the good mocker institute” is the Guttmacher Institute. At 27:52 “kate troy” and “k troy” are Kate Choi, and the coauthor is Brandon Wagner; the paper is “Decline in Marriage Associated with the COVID-19 Pandemic in the United States,” which appeared in Socius in 2021, so link the published version rather than the preprint. At 8:51 “the heek man session” is the he-cession, and the 2009 coinage was mancession. At 15:11 and 24:11 “birds” is births. At 16:40 “the 2008 procession” is the 2008 recession. At 33:17 “rotating clockwise” is his own borrowed phrase, credited to David in the August 2020 lecture. At 36:44 the suicide figure is given as 11 percent of adults and about a quarter of young adults, which is the June 2020 CDC survey and carries the same thirty-day versus twelve-month problem I flagged in the August lecture.

‘Pandemic and after interview’ (Dec. 4, 2020)

0:04 Opens on what will last: online interaction, and specifically the regularly scheduled contact people now keep with distant friends and relatives to compensate for lost travel and local socializing, which he expects them to want to continue.

0:43 The first pass at winners and losers.

“The people who are doing really well in this situation are probably the people who are doing really well already. I don’t think there are a lot of people who are having troubled or difficult or terrible relationships and life at home who are now finding themselves suddenly doing better.”

0:58 A methodological warning inside a claim about survey response.

“If you interview people and ask them, has the pandemic improved your relationships and your family life, the people who say yes are probably already headed in that direction.”

1:10 Things on hold, split into two kinds. Some are mechanically blocked: weddings cancelled, divorces hard to process.

1:33 The other kind, which is his actual subject.

“Other things are not happening because the drivers of family change are just being impaired. People are meeting new people less. Casual interaction is not happening as much. People are putting a stop to plans that they would have had before.”

1:48 The forecast, made in early December 2020.

“I think it’s very likely that we’re going to see a substantial drop in birth rates. People are holding off, at least temporarily, from their decisions to have children.”

2:01 Young adults living with parents in the millions, who are not marrying, probably not meeting people, and not having children, through two routes he keeps separate: opportunities impaired by isolation, and changed decision-making under uncertainty.

2:29 The summary, with the hedge attached.

“It means everything has ground to a halt in family life. It’s a slight exaggeration, but a lot of things have ground to a halt or slowed down substantially.”

2:37 What makes families work, stated as plainly as anywhere in the record.

“Successful families are a lot about security and comfortable interdependence and being able to rely on each other. That’s especially the case for children, but it’s also the case for people in intimate relationships. So the people who were secure, in the emotional as well as material sense, are probably the people who are doing better now.”

3:03 The other side, and the measurement point he had been making since April.

“The downside is, we can’t measure this, but the terrible and abusive situations that are being experienced in many homes and families are going undetected. Both our formal ability to detect and root those out, like school social workers and nurses and hospitals, but also just the informal interactions and connections that people have that are buffers against trouble at home.”

3:48 The gender division of labor, worked through as a genuine two-sided problem rather than a talking point.

“People are spending more time at home, which is a gender-unequal space, and so women are doing more parenting, homeschooling and all that. On the other hand, because men are home more, they’re also doing more. So they’re maybe not doing so much a greater share, but they’re doing a greater quantity of housework and child care and care work generally.”

4:22 Who handles that well, and the condition he attaches.

“The people who already had the relationship infrastructure in place to have those conversations successfully are probably also doing better at handling that flexibility. People who don’t talk about those things, or couldn’t talk about those things, or couldn’t come to agreement because of conflict, I’m not sure how they are able to work out the changes that they have to make now.”

5:01 Low-stakes unplanned interaction, and why its loss is hard to see.

“One thing about low-stakes, informal, unplanned interactions we count on so much is that we don’t realize we counted on them so much, and we often don’t do a lot to make them intentional. And that’s part of the point.”

5:20 The strength of weak ties, explained without naming Granovetter.

“In sociology we’re very aware of what we call the strength of weak ties, the people that you know a little. Those are the people you really need when it’s time to get a job. The people who you know enough to ask for help, but they’re distant enough from you that they know a lot of people that you don’t. Your best friend isn’t the one who’s going to help you get a job, because you know all the same people.”

5:54 The design problem that follows, and why he thinks the technological fix is hard.

“The rules of that kind of interaction are not explicit, so remaking it is complicated, and the norms are sort of up for grabs in a space we haven’t developed before. In my case, I have students who might drop into my office to ask me a question, but they might be reluctant to schedule an appointment for an online chat, to get themselves on the calendar.”

6:43 The task he sets for his own discipline: sit down, take apart, and try to rebuild the things about relationships nobody realized were important.

6:56 On the phrase everyone was using that autumn.

“Going back to normal is a myth. For one thing, normal is how we got here. So we don’t really want to go back to normal, being unprepared, and having all the things that have made this worse. The seeds for that were already there.”

7:16 The rebound question, framed as genuinely open, with the historical precedent named.

“Will there be a baby boom of people who are so excited they just start having children again? Will we have a golden age, a roaring twenties, which is what we had after the last great pandemic, of exuberance and outward-looking excitement? That’s possible. It probably depends on whether the victory over the virus in the end is real and sudden and observable, or gradual and a demoralizing war of attrition.”

8:14 The last ten percent, which is his best economic observation in any of these recordings.

“The things that don’t come back could be very big. If we don’t have cruises and movies anymore, well, those are giant industries. What about Broadway, if that doesn’t come back? That last ten percent of getting back to normal makes a very big difference.”

8:36 The divergence thesis, and the part he flags as fear rather than finding.

“I’m afraid it’s the people who are less social or less pro-social that are going to remain turned inward, and keep those barriers up around them and their families and their homes. And I’m afraid that that will have a corrosive impact on society and social life.”

9:13 The generalization: a story of divergence, where those better prepared psychologically, socially, and economically integrate their new selves into the new reality better.

9:31 The link from national narrative to private decision, which is the argument underneath his whole body of work.

“One of the things that is dislocating for people is the overall narrative of history: where is our society going? That has a lot to do with social mobility. Do we think our kids are going to do better than ourselves? If we’re safe with that assumption, then we feel ourselves the masters of our universe, and we’re comfortable making long-term plans.”

9:56 The taxonomy that follows, which is the best single sentence in the interview.

“Making banana bread is a short-term response. Redesigning your home, planning to have more children, making long-term career investments that you don’t expect to have short-term rewards, these are things that people do when they’re confident about the future. For family life, that means the lack of confidence in the future means fewer children, less marriage.”

10:21 His confidence level on that: pretty confident those trends, already underway, continue as the data comes in.

10:30 Masks, and a claim about motivation he offers as observation.

“We were told that we should wear masks to protect others. I don’t think that’s the way most people are built psychologically. If you look at the way people are wearing their masks, I think they’re thinking about protecting themselves.”

10:53 The in-group finding, stated as something he sees rather than something he measured.

“People are much less careful when they’re around people who they see as like them. So around their own neighborhood they’re less likely to wear a mask than if they leave the neighborhood, or interact with people who are more different from themselves.”

11:12 The cost he draws from it.

“I fear that the downside of that is building our walls higher and our distrust of others increasing. So that is the downside of reshaping our home lives in ways that seem positive. We’re going to do great things at home, but it’s partly because we’re afraid of and we distrust the outside world. That’s a real challenge for social cohesion in the future.”

11:36 The close, which is the same close he gave his students in August.

“Who are we going to be after this? That’s an important question to ask ourselves, if we can make that intentional. Who do we want to be after this? Some of that could be altruistic and outward thinking: I can’t wait to get out there again and help people and meet new people and be friendly toward them.”

Twelve minutes, one side of an interview, no slides, no data, no antagonist. It is the only recording in the file where he is asked to think rather than to show, and the result is that the theory underneath thirty years of demographic work comes out uncovered.

Whether a society believes its children will do better determines whether individuals feel like masters of their own lives, which determines whether they make commitments whose payoff is distant, and the commitments whose payoff is most distant are marriage and children. So low fertility and late marriage are, in his account, a readout on collective confidence rather than a change in values. The banana bread line at 9:56 is the clearest thing he has ever said, because it gives the reader a way to sort his own behavior into short-horizon and long-horizon without any technical apparatus.

Set that beside the family-restoration argument and the disagreement gets smaller. Wilcox and the people around him would accept nearly every sentence from 2:37 through 10:21. Security, comfortable interdependence, reliance on each other, and the observation that people without confidence in the future stop making long-term commitments. The dispute is about whether the conditions can be produced by exhortation. Anyone writing about that fight should quote this passage rather than the blog posts, because the blog posts are where the argument gets narrower and louder.

Two passages here are his best work outside his field.

The weak ties explanation at 5:20 is Granovetter delivered without jargon and without the citation, which is the standard way a discipline’s finding enters general speech. And what he does with it at 5:54 is the interesting part. The reason the office drop-in cannot be replaced by a scheduled Zoom is that unplanned interaction has no explicit rules, so it cannot be reconstructed by writing rules for it. The student who would put his head round a door will not put himself on a calendar. That is a design constraint, stated in 2020, and it turned out to be the central problem of hybrid work for the following five years. He was ahead of almost everyone on it.

And the last-ten-percent point at 8:14 is an economic observation a demographer had no obligation to make. Aggregate recovery statistics hide the sectors that do not return, and cruises, cinema, and Broadway are not marginal. Broadway grosses recovered slowly and cinema admissions never returned to 2019 levels. He named the right industries.

The weaknesses are of a kind that recur across the file.

The mask claims at 10:30 and 10:53 are presented as observations and are not observations of anything he measured. He says he thinks people are protecting themselves rather than others, and that people mask less around neighbors who resemble them. The second is plausible, has since been studied, and at the time was his impression from a window. This is the same man who spent 2015 demonstrating that a celebrated ethnographer’s household counts did not match census figures, offering behavioral generalizations about a nation from what he could see near his house. He does not flag it, which he did flag repeatedly in the October talk three weeks earlier.

The divergence thesis at 8:36 has a moral valence buried in a descriptive claim. The people who remain turned inward are described as less social or less pro-social, which makes withdrawal a trait rather than a circumstance. His own account elsewhere says exactly the opposite: withdrawal follows from insecurity, from having no buffer, from an unsafe home. He does say “I’m afraid” twice, which marks it as prediction rather than finding, and the framing still sorts people into better and worse types in a way the rest of his work resists.

And the second sentence of the interview, at 0:43, is a strong empirical claim about who benefited, offered with no data and immediately followed by an explanation of why survey evidence on the question would be unreliable. That is a man who has ruled out his own means of checking and then answered anyway.

For the biography, this is where the December 2020 forecast belongs, and it is the version to quote, because he made it in plain speech before the data existed and it held: births did fall, and the birth-dearth story dominated 2021. It is also the recording that shows the theory rather than the fights, and the only one in the file where the last word is about what people might want to become.

Transcript flags. At 2:29 and 2:34 “grown to a halt” is ground to a halt. At 3:22 “route those out” is root those out. At 3:59 “a gender and equal space” is a gender-unequal space, and getting this wrong reverses the sentence, so check it against the audio before quoting. At 4:14 “a greater shorter share” is a greater share. At 10:58 “in their social people are much less careful” is a stutter and the sentence resumes cleanly. The weak-ties concept is Mark Granovetter’s, “The Strength of Weak Ties,” American Journal of Sociology 78:6 (1973).

‘Abortion and Abortion Rights lecture for Social Problems sociology class’ (Dec. 7, 2021)

0:07 Frames the topic as two things at once, a personal matter many people have their own experience of, and a public policy, political, and legal question.

0:58 The naming problem, handled fairly to both sides.

“Neither side in this debate wants to be described as anti-something. They both want to be described as pro-something. So that’s why we have pro-choice and pro-life. If you ask the pro-choice people, they would say we are pro-choice and the other side is anti-abortion. But the pro-life people don’t want to be anti-anything. If you ask the pro-life people what to call the pro-choice people, they don’t see it as choice, they see it as abortion. They think the other side should be called pro-abortion.”

2:12 The country is roughly split on the identity question, a little under half calling themselves pro-life, with the distinction he insists on.

“The country is about split on this, which does not mean that the country is about split on the issue. The country is about split on the question of how people identify themselves.”

2:41 Roe polling: about a third say they want it overturned, roughly 60 percent do not, which is more than the share who call themselves pro-choice. His inference: some self-identified pro-life respondents do not want states allowed to ban abortion.

3:45 Legality polling with a middle option, and the effect of offering one.

“This allows people, when you give people a more sort of middle-ground position, they often take the middle-ground position. So half the people say it should be legal in certain circumstances.”

4:20 The 19 percent, and the logical argument he builds on it.

“You really only get 19 percent as of right now who think abortion should just be illegal, right, like murder. The pro-life politics often speaks of abortion as if it is murder, or a holocaust, but it’s a small minority of people who want to treat it that way in terms of the law. If abortion is murder, then it does not matter why or how the person became pregnant. In the case of rape or incest, which is a common exception, that doesn’t matter. The unborn child is still innocent, if you think abortion is murder.”

5:17 What the middle-ground respondents are picturing: very young mothers, rape or incest, early pregnancy. And the conclusion he draws, that anyone taking those exceptions is not treating abortion simply as the taking of a human life.

5:48 The three questions he wants kept apart.

“There’s a question of what’s your identity, I’m pro-life, I’m pro-choice. There’s a question of what do you want to do about the law. And there’s the question, do you want to criminalize abortion, in the sense of put the doctor in jail, put the pregnant person in jail.”

6:18 Group breakdowns: 62 percent of women and 56 percent of men favoring legality in all or most cases, small gender gap; Black and Asian respondents somewhat more favorable than White and Hispanic; younger more than older, more educated more than less, religiously unaffiliated the most favorable, White evangelicals the least.

8:00 Catholics, and the aside he does not walk back.

“The majority of Catholics think abortion should be legal. The majority of Catholics use contraception. The majority of Catholics think divorce should be legal. American Catholics are not good Catholics. They disagree with the Pope. What can you say. It’s not my opinion, good or bad. I’m just saying they disagree with the Pope.”

8:34 Measurement, which is the strongest section in the lecture.

“One thing to say about abortion is, we don’t measure it that well. If you just do a survey and you ask people did you have an abortion in the last year, some people don’t answer you, some people don’t give you the true answer. There’s shame and stigma. There’s also no federal registry. We have a registry of births in the country, so we measure every birth quite well, but there’s no registry of abortions. So any number on abortions is based on an estimate.”

9:24 The Guttmacher figures for 2017 as the best available: a little under a million abortions, about 13.5 per thousand women of childbearing age, so about 1.4 percent of such women in a given year, and almost 40 percent of them medication abortions.

10:22 About 18 percent of pregnancies excluding miscarriages ending in abortion, with the rate falling for forty years.

11:21 His explanation for the decline, offered as the one thing he can say clearly.

“Teenagers and very young women are getting pregnant a lot less than they used to, and that’s one big reason that the abortion rate has fallen. The age of marriage has gone up, and people are allowed to be single mothers now in the culture much more than they used to be. There’s less unintended pregnancy among young women, who are the group most likely to have abortions.”

11:59 Roe as a 1973 decision recognizing a right to decide, with states permitted to limit it around viability, then about 28 weeks.

12:54 The basis, and why the argument never ended.

“The basis for Roe v. Wade was privacy. Privacy is not in the Constitution. So the Supreme Court had to do some interpretation here, and that’s what allows them to still be arguing about it today. They did not say everybody has a right to have an abortion. They said women have a right to make the decision about abortion. That’s why pro-choice people are called pro-choice, because that’s the way they won Roe v. Wade.”

13:37 The line of privacy cases behind it: contraception first, and what he calls the same principle later applied to gay sex, with the morality laws that once restricted birth control to married couples.

14:43 The Hyde Amendment and the distinction it exploited.

“They didn’t say you have a right to have an abortion like you have a right to vote. You have a right to vote, and that means the government has to make it equally possible for everybody to vote. In the case of abortion, you have a right to decide about an abortion, and the government has to let everybody have equal access to that decision, but they don’t have to pay for it. So Congress passed the Hyde Amendment: fine, you want legal abortion, it could be legal, but the federal government’s not going to pay for it. So this right is not going to apply to poor people.”

16:07 Casey in 1992, with medicine having pushed viability earlier, permitting states to ban after viability where a fetus could survive outside the body.

17:32 Dobbs, argued the previous week, and the origin of the Mississippi statute.

“That is the case in Mississippi, which was passed in Mississippi just for the purpose of getting to the Supreme Court, so that the new Trump-appointed judges could have an opportunity to overturn Roe v. Wade.”

18:09 The precedent problem, stated in a way that gives the overturning argument its due.

“Each of these decisions reaffirmed Roe v. Wade. It said we’re not overturning Roe, we’re just clarifying it. But the abortion right is still not in the Constitution, so it’s really just a matter of Supreme Court precedent. And so the Supreme Court can change its mind, can change precedent if they want to. Usually they’re pretty cautious about doing that, and that’s what they’re arguing about now.”

18:44 The Mississippi law at 15 weeks, before viability, and his forecast.

“If the Supreme Court says this law is okay, to one degree or another they are overturning Roe v. Wade, and it will depend on the wording of what they say how severe that is. But to some degree that’s going to happen.”

19:37 On Casey, answering a student who is evidently unconvinced: it affirmed Roe explicitly, and after it you still could not ban abortion outright. “It seems semantic, but it turns out to be kind of important.”

20:19 The Texas Heartbeat Act at six weeks, and the terminology.

“Six weeks, most women don’t know they’re pregnant. They call it a Heartbeat Act because there’s an electrical signal discernible in the fetus at that point. It’s not a heartbeat. There’s no heart pumping blood, but there are the electrical firings, the organ is beginning to form. That’s very semantic, but that’s how they like to describe it.”

21:07 The enforcement design: private suits for ten thousand dollars rather than criminal prosecution, which he says would end provision because no provider could absorb the litigation.

21:50 Why the Court’s handling of that case was the signal.

“Prior to this court, with six Republican judges, the Supreme Court would have obviously thrown that out and just said this obviously violates Roe, we have a precedent, so we’re throwing it out. Instead the Supreme Court said, ah, interesting, we’ll discuss it. That was a very big moment, because that’s when it became clear where we’re probably headed.”

22:50 Declining access, which he treats as a separate track from the law: funding restrictions making practice hard to sustain, plus protesters and the risk of violence narrowing the pool of providers.

23:38 The county maps: in Mississippi, Arkansas, and Missouri more than 75 percent of women live in counties with no abortion provider.

24:36 The catalogue of state restrictions: physician-only requirements, hospital or admitting-privileges rules, gestational limits in 43 states, so-called partial-birth bans in 21 states, prohibitions on state funding and in some places on private insurance coverage, waiting periods in 25 states, mandatory counseling, and parental consent or notification for minors.

27:09 The analogy he lands on for what kind of right it is.

“It’s not like we have a right to abortion in some universal sense. There are lots of restrictions on the right. It’s more like you have a right to drive a car, but they can do all kinds of things to limit that. Not like voting. Voting you have a right to do, period.”

27:38 Bans on abortion for sex selection, with the obvious objection supplied: “Of course, how could they know what your reason truly is.”

28:10 Gestational timing: 79 or 80 percent of abortions before 10 weeks, the largest group under six weeks, very few in the later weeks.

28:42 What the late ones are.

“All of the abortions that take place after the age of viability, if hypothetically a fetus is 24 weeks along and usually a 24-week-old can survive, then if you’re going to do an abortion at 24 weeks it is only going to be if that baby cannot survive. It’s when they discover an abnormality which is so extreme it can’t survive, or something that would not allow the mother to survive or would greatly threaten her.”

29:22 Why that distribution mattered to the movement.

“You can see why feminists were not so distressed at the idea that states would only allow abortions before 24 weeks. That did not seem like the end of abortion rights, because that’s not what we’re talking about in the vast majority of cases.”

29:55 A campaign video for Plan C, played in full, on medication abortion, the FDA restrictions, Aid Access, and the three demands it makes of viewers.

31:30 His comment on it, which includes the legal risk.

“Right now in Texas, if you want to use this method, it’s pretty easy to get the pills and do it, but somebody could sue whoever provides the person getting an abortion with those pills under the Texas Heartbeat Act. So it’s risky.”

32:08 The trigger-law forecast: a number of states already have statutes written to ban abortion the moment Roe falls.

32:40 Relative risk, introduced through the counseling requirements.

“It has always been the case that abortion bans don’t stop rich people from getting abortions. The number of deaths per birth is 20 out of every 100,000, and the number of deaths per abortion is less than one half out of every 100,000. It should be obvious that having a baby is more dangerous than having an abortion.”

34:02 The argument that reversed, which he flags as a genuine turn.

“What goes around comes around. Feminists said we need reproductive rights because we need to be able to control our lives, if you make us have babies then we won’t be able to have careers. One of the things that happens in the legal cases now is the anti-abortion people say, tons of women have children and successful careers. You don’t need to not have a baby in order to have a successful career anymore, so why do we still need this protection?”

34:42 He introduces and plays Justice Amy Coney Barrett’s safe-haven exchange from the Dobbs argument, pausing first to explain what safe haven laws are, and telling students to listen to what she is actually saying.

35:44 Barrett’s question in full, on whether relinquishment after birth answers the burden-of-parenting argument, and her acknowledgment of “an infringement on bodily autonomy which we have in other contexts, like vaccines.”

37:04 His restatement of her argument before the answer.

“What she’s saying is, if the law now says 23 weeks, and we overturn Roe, we’re making women basically only be pregnant for 15 more weeks, then they can put the baby in the fire station, and so it’s really not that big of a burden.”

37:22 The lawyer’s answer, that adoption was available in 1973 too, and that pregnancy imposes unique physical demands beyond the burdens of parenting.

38:03 His reading of Barrett, which gives her the argument on its own terms.

“We understand that if we say an abortion is a human life, and you cannot just terminate a human life at will, that we are imposing on women, on their bodily autonomy, where the law is intervening and saying you don’t get to do exactly what you want with your body. Just like with the vaccine rule. She was just using an analogy to another case where that’s obviously considered acceptable nowadays. So they are saying, you don’t get to do whatever you want with your body if there’s a public interest.”

39:09 What he thinks the lawyer left out: that carrying a pregnancy and relinquishing a child is a process people often do not want to go through, rather than thirty weeks of lost work time.

40:00 The concession he makes to the other side, at length.

“There is a sense that you cannot avoid the fact that in this policy we are in fact balancing the well-being of fetuses with the well-being of the people who carry them. You can’t pretend that there is no issue there.”

40:27 The viability standard as a drawn line, and what it concedes on both ends.

41:23 The limits of science in the argument.

“Some people would like to not have a debate about abortion that recognizes that there’s a continuum of life here. If you have a very religious view, you can say every fertilized egg is a human life and they’re all completely equal, but science is never going to say that. On the other hand, science can’t draw a line and say at this moment it becomes a human being. That’s why this viability standard is a ballpark attempt, and in a reasonable world something like a viability standard would maybe not be unreasonable.”

42:08 Catharine MacKinnon’s position, offered as the argument most pro-choice people will not make.

“Most pro-choice people don’t want to say that, or they don’t agree with it. But even if they did agree with it, they don’t want to say it, because nobody wants to say you’re allowing infanticide.”

43:05 A discussion of an evidently dystopian novel from the 1980s, on how quickly rights can be withdrawn: women’s bank cards stopping, an emergency rule, a rapid crackdown, soldiers enforcing it.

44:13 The closing, which is the most careful thing he says on any contested subject in the file.

“I just want to say to you all, younger people, in some ways these are disturbing times. Even if you are unsure about, or against, abortion rights, there are aspects of this politics which are disturbing to most people. It’s only nineteen percent of Americans who think abortion should always be against the law. I would never tell religious people what to believe. But there are absolutely trade-offs between some people’s rights and other people’s rights, and the way we manage those things is with politics. So there’s no easy answer. This is just my way of saying I understand this is difficult, so I’m sorry.”

This is the hardest test in the file and he mostly passes it.

The structure of the lecture is the same structure he used on functionalism in August 2020. State the position, give it its strongest form, then examine the logic. He does this to the pro-life argument at 4:20: if abortion is the taking of an innocent human life, the circumstances of conception are irrelevant, so a rape exception is a concession rather than a principle. That is the argument the most serious pro-life thinkers make against their own movement’s political compromises, and a student who came in pro-life would recognize it as the hard question rather than a caricature.

Then he does the same thing in the other direction. At 34:02 he reports that the workplace-access argument feminists won on has been turned around and used against them. At 40:00 he says the policy balances the well-being of fetuses against the well-being of the people carrying them and that you cannot pretend otherwise. At 41:23 he tells the class that science will not settle it in either direction, which forecloses the most common move on his own side. And at 42:08 he puts up Catharine MacKinnon (b. 1946) saying abortion should be legal even granting that a fetus is a human life, and notes that most people who agree with him will not say it out loud.

The Barrett passage is the strongest evidence in the whole file against reading him as a man who only prosecutes. He plays the audio, stops it, explains safe-haven laws neutrally, restates her question in its strongest form, plays the answer, and then explains her bodily-autonomy analogy sympathetically before saying where he thinks the response fell short. That is what teaching a contested question looks like, and it is a different man from the one who plays montages of a president lying.

Three places where a pro-life student would have a fair complaint.

The Catholics passage at 8:00 is the worst moment. He says American Catholics are not good Catholics, then says it is not his opinion, then, in the part the transcript catches mid-stumble, appears to say that it does make them bad Catholics. Whatever he meant, he has made a judgment about the authenticity of somebody’s religious practice from a lectern, on a subject where a good number of his students are the people being described. It costs him nothing to have said instead that American Catholic opinion diverges from Church teaching, which is the same fact without the verdict.

The Plan C video at 29:55 is the third time in this file he has handed the microphone to an advocacy organization. Paid sick leave in April 2020, the stimulus bill in September 2020, and now abortion pills. All three clips are accurate on their numbers. All three arrive with framing and a call to action in somebody else’s voice, and none is identified for students as advocacy rather than reporting. The pattern is consistent enough to be a finding: he is careful in his own voice and outsources the argument at the edges of his expertise.

And the coda at 43:05 is the loosest thing in the lecture. He uses a 1985 dystopian novel as a warning about the direction of American politics, describing soldiers stoning women. He flags it as fiction from the 1980s. He does not flag that the distance between a Mississippi 15-week statute and that scenario is enormous, and a student inclined toward the other side would hear the comparison and stop listening. The closing three minutes are so careful that the coda is the only part working against them.

On the record, his forecasting is his best in the file. On December 7, 2021 he said the Court would overturn Roe to some degree, that the wording would determine severity, and that trigger laws in a number of states would fire immediately. Dobbs came down on June 24, 2022. Every part of that held. He also said the medication route would go underground and become the practical center of the fight, which is what happened, with the shield-law and mail-order disputes that followed. A demographer making a confident prediction about a constitutional case is outside his expertise, and he read it correctly from the Court’s handling of the Texas procedural posture rather than from the merits, which is the right signal to have watched.

One methodological complaint of the kind he would make about someone else. The 19 percent figure carries a lot of weight in his argument, and it comes from a poll option reading illegal in all circumstances. Someone who believes abortion is homicide but accepts an exception where the mother’s life is at risk will not choose that option, and there is no contradiction in that position. So the gap between what people believe and how they answer the question is smaller than his inference implies. The inference is still substantially right, and he presents an estimate as though it measured the thing exactly, which is the move he audits others for.

For the biography, this lecture belongs in the same paragraph as the March for Life photograph. Together they make the case that on subjects inside sociology he can hold a contested question open in front of students who disagree with him, and that the register changes when the subject moves to a named political actor. The difference is whether there is somebody to accuse.

Transcript flags. At 17:32 “doubts what we’re calling doubles” is Dobbs, and every variant of “roe versus wage,” “robert’s weight,” and “brokers his ways” is Roe v. Wade. At 9:24 “goodmaker institute” is the Guttmacher Institute. At 15:29 and 16:02 “hype amendment” and “the harris decision” are the Hyde Amendment and Harris v. McRae (1980). At 19:37 “planned parenthood” is Planned Parenthood v. Casey (1992). At 2:24 “protrusions” is pro-choice or. At 4:07 “milligram position” is middle-ground position. At 20:36 “most women don’t know their credit” is don’t know they’re pregnant, and the sentence is unusable as transcribed. At 24:42 “licensed position” is licensed physician. At 34:47 “amy coney barrett” is correct (b. 1972); the lawyer she is questioning, transcribed as “miss rickleman,” is Julie Rikelman, then of the Center for Reproductive Rights. At 36:12 “amiki” is amici. At 42:15 “kathryn mckinnon” is Catharine MacKinnon, and you should pull her sentence from the original rather than the transcript, since the version here is garbled around “I can’t follow that.” At 31:24 “plancypills.org” is plancpills.org. At 43:05 the novel is Margaret Atwood’s The Handmaid’s Tale (1985), which he never names. The 13.5 per thousand, 39 percent medication, and 18 percent of pregnancies figures all match Guttmacher’s 2017 estimates, and the 79 percent before ten weeks matches CDC surveillance, so those are safe.

‘Open Social Science and Public Engagement’ (Feb. 18, 2022)

0:03 Sophia of the Scholars Strategy Network NYC chapter opens, describing SSN as connecting scholars to policymakers, civic leaders, media, and the public, with roughly 37 or 38 chapters nationally.

1:32 She introduces him, including the Trump blocking lawsuit, which she says she wanted to make sure to mention.

2:11 He notes the session is scheduled for 90 minutes and asks for questions in the chat, warning he will not see them while sharing slides.

3:03 SocArXiv’s funding position, stated more bluntly than in print.

“It’s now run out of the University of Maryland libraries, which means I serve at the pleasure of the dean of the libraries, basically. And that means they pay for it. Not that much, but not to diminish their contribution. It’s a low-budget operation.”

3:34 The audience he assumes: social scientists who want influence on policy, politics, other scholars, or the public, in the context of an enrollment crisis, budget problems, and a crisis of democracy.

4:14 His formulation of the goal.

“The way for us as scholars to address these goals is to combine being public in our intellectual life and intellectual in our public life. I’m looking for ways to integrate these two aspects in ways that strengthen them both.”

4:44 The principle underneath the whole talk.

“We have to have a level of accountability in all of the communication that we do, and the key principle for accountability in science and social science is openness. It doesn’t mean everything is always open to everybody, but openness is a key aspect of our accountability and the trust that we ask people to put in us.”

5:25 Publication lag at the association’s journals, which he says he ballparked from first-round review plus one revision plus production, excluding the author’s own revision time: about twenty months at Sociological Theory, around seven at the fastest, an average near nine months.

6:18 The paywall, framed as friction rather than only cost: shared PDFs, temporary logins, paying extra for an open version.

6:50 An interruption from the host asking him to fix his screen share, and his response that he had thought he already had.

7:56 Market concentration: of the five thousand most recent sociology articles he checked, 80 percent came from five publishers, with Sage the largest.

8:12 On his former association, in the sharpest phrasing he uses anywhere.

“Our association, the American Sociological Association, is pretty well captured by Sage, and organizationally not interested in changing this beyond what might be necessary for tinkering, daily functions, and PR.”

8:44 Preprints as a workaround rather than a solution: a way to address the problems without waiting for the system to be restructured.

9:15 The servers, with a pronunciation note that the X in bioRxiv is a Greek chi, and the 2016 arrival of SocArXiv and PsyArXiv.

9:46 What a preprint is, and the qualification that not everything on his archive fits the definition, since some items are accepted versions of peer-reviewed papers.

10:10 The framing he prefers for a social science audience.

“A preprint is like what we would call in social science historically a working paper. It’s a formal scholarly output, it’s permanent and citable, and it’s a research event when it’s published. It happens, like publication in a journal, and becomes part of the record.”

10:37 Three reasons: efficiency, engagement, inclusivity.

10:44 The waiting problem, described from the author’s side.

“You’ve all probably experienced the frustration of doing work that you’re excited by, and then it enters into this doldrums period where you’re just waiting. And the frustrating thing is you’re waiting for people who are having their arms twisted to read it, while you suspect, or you hope, that there are people actually out there who want to read it, and they have to wait.”

11:22 Access as principle rather than convenience: the public supports the work through grants and salaries, and “we don’t aspire to have our work hidden from the people it most affects.”

11:56 The third reason, which he says is underrated: the same friction that blocks the public also slows connection between scholars, and engagement means building collaborative spaces rather than broadcasting.

12:28 Volume: approaching 400,000 preprints a year across the major servers, about 14 percent of the annual article count in Web of Science, with the caveat that the denominator is imperfect.

13:49 The Fraser et al. finding on COVID preprints, and what he takes from it.

“They followed some of these papers to see what they looked like when they were published in journals, and very often they were published with hardly any substantive change, or no change in the main findings. That was in a way reassuring. It implied the preprint system was accelerating the distribution of research without producing tons of junk. However, of course some of it is wrong.”

15:07 The concession he makes in the same breath, which he has not made this plainly elsewhere.

“Some peer-reviewed work is wrong, some preprints are wrong. Probably more preprints than peer-reviewed papers are wrong.”

15:35 The medRxiv caution notice, and whether the press ignoring it was bad.

“Mostly not, because there was also great journalism happening. The good journalists were using other indicators of reliability. Who are the researchers, they would interview other experts. You would see an article in the New York Times or the Washington Post about some preprint with three top experts interviewed, and that’s essentially what we get out of peer review, but in the space of a few hours.”

16:49 Preprints that carried the pandemic: excess mortality, which introduced the concept to a general audience, and the finding that delayed lockdowns cost lives, published in a journal six months later with the main finding intact.

18:28 The legacy workflow, laid out as a sequence of submissions and rejections, and then the three functions journals perform: establishing the record and precedence, promotion and distribution, and the signal that experts reviewed it.

20:37 The preprint workflow, and what it does to those functions.

“It’s separating the functions of what journals now do, and taking part of what journals now do and moving it earlier in the process.”

22:06 On whether this sacrifices trust, answered by widening the frame to the whole research life cycle.

23:03 His argument that evaluation happens at many points, and the one that is least often named.

“If you have a job, if you have tenure, if you’re a graduate student, your status has already been adjudicated to some degree. You don’t want to assume that someone from one school’s work is better than another, but we do carry a reputation which contains within it some assessment of our veracity as a scholar.”

24:35 The broader open-science agenda: preregistration, sharing materials and code, and a note that NIH had announced a data management and sharing requirement the previous day.

25:18 SocArXiv’s scale: ten thousand papers since 2016, which he says is small against 400,000 preprints a year, and which he says does not matter for the individual decision.

26:01 A walkthrough of one paper on the platform, using his own divorce decline paper: download count, hypothes.is annotation, Plaudit endorsement, unlimited supplemental materials, two DOIs linked so citations aggregate, license options, taxonomy from ASA section names, and versioning.

29:12 Why versioning matters, and the detail that the paper shown is on version five with all earlier versions retrievable, since the American Community Survey data updates annually.

30:08 A second example, the Stone Center working paper series, and a Branko Milanović paper posted in July 2020 and published in Review of Income and Wealth ten months later, downloaded about 800 times in between.

32:39 Objections, taken one at a time. First, the fear that a paper is not good enough.

“A lot of people think it’s not good enough, or they think they need to have peer review before they feel permitted to share their work. If you look at the difference between peer-reviewed work and initial drafts, usually this concern is not borne out. I think more people err on the side of being too cautious. That’s my personal view.”

33:55 Second, that posting will block later journal publication, which he calls a vanishing problem and directs people to check on Sherpa Romeo.

34:41 Third, that ideas will be stolen, where he inverts the objection and then adds the part that is the most valuable sentence in the talk.

“I think the problem of stealing ideas probably happens more in people’s inner circles, especially their mentors and advisers and committee members and journal peer reviewers. The people who are stealing ideas are usually not getting them from the public square. They have some private access to your ideas, and they’re abusing that position of trust. So in a case like that, posting it publicly can help you.”

35:52 When to post, presented as four cut points by nervousness: whenever you want feedback, at conference submission, at journal submission, or at acceptance.

37:06 On the last option, and the practice it should replace.

“You see this all the time on social media. I’m so happy I got my paper accepted, in a few months I’ll share it with you, or send me an email and I’ll share it with you. People think that’s a friendly, nice thing to say, but it actually gets you a lot fewer readers than if you just put up a link.”

38:37 The disclaimer that governs the whole open-science pitch.

“I do not mean to imply here that everything must be shared with everyone at all times. There are times you do need to keep things private. Sometimes you have preliminary results that you’re really not sure about and you don’t want to share outside of a circle. That’s totally fine. It’s just a question about being purposive in how you do this.”

39:58 The modular workflow he actually uses: shared and unshared folders, working out of the unshared ones and dropping files across when ready.

41:17 On tools, and the line that got the biggest reaction.

“We don’t build the tools that we’re using, by and large. People love Google, Google Drive, Google Scholar. Google is not really your friend. It’s okay to use Google tools because they’re awesome, but they won’t love you back in the end.”

42:43 The old convention that no longer holds: mailing a draft marked do not quote or cite without permission.

43:25 What replaced the audience.

“The audience is less the people that you’re speaking to, and now much more a network. It’s the people that you are reaching right now and the people that they’re reaching right now, and you don’t know who they’re going to be.”

43:57 Who is now reading: journalists who do their own data analysis, and journalists who marshal peer review by calling three experts before filing.

“That is not violating the principle of peer review, that’s actually honoring the principle of peer review, if they do it well.”

44:32 The chaotic disciplinary mashup, where an economist shares a sociologist’s paper and an unexpected audience arrives.

44:56 The argument against trying to control it.

“We do not get to choose whether we want to share things just with our friends and not with our enemies. It’s very hard to share something and have it be completely insulated. So we want to get on top of it instead of trying to restrict it, because that’s a losing battle, and if we succeed we don’t really want to live in that world where so few people read our work.”

46:31 The claim he flags as unproven and believes anyway.

“It’s hard to empirically prove this, but I really believe it. Even if people do not read your data and code and validate everything you do, the fact that it’s available says a lot. It says that you are open to being held accountable, that you are open to collaboration, that you’re not trying to slow other people’s work down, that you’re not trying to own the work more than necessary. I think openness is part of the accountability that we need to communicate in order to earn trust, and I think it’s important even if nobody ever replicates your work.”

48:32 The correction he offers to his own field’s language about dissemination.

“When we say we want to use social media to get our work out there, we also want to get other people’s work in here, and we want to hear from other people. Social media, if used well, is just as much about listening as it is about speaking.”

49:19 A story from Twitter about a reviewer plagiarizing a paper, and what he draws from it.

“If the incentives were different and the communication structure were different, that person who stole your idea, if you had connected with them earlier in a more transparent and collegial environment, you might have become collaborators. It’s a shame when the only opportunity to communicate you have with people is an anonymous peer review.”

51:03 First question, from Heath, on journalists and editors trained to cite only peer-reviewed work.

52:24 His answer, that the premise is out of date, with the list: think tank reports, in-house data journalism at FiveThirtyEight, Vox, and the New York Times, Pew.

53:47 The questioner pushes back, arguing the elite outlets are the exception and that reporters and editors under thirty at smaller outlets remain risk-averse and would benefit from education. Cohen concedes the point and adds one of his own.

“There’s a business interest in maintaining this distinction, because the publishers want to reinforce this distinction all the time. If you want their PR help, or your university’s PR operation, they might insist that things are peer reviewed.”

56:34 Second question, from Alicia, on whether journals object to preprints and on the cost barrier to open access. His answer walks through author agreements, the point at which publishers start to feel they have added value, and the Sage and Wiley default permitting the accepted manuscript on a nonprofit repository.

59:26 Where a bad agreement bites, with a consequence most people miss.

“Some will say only 12 months later, and that would be a worse agreement, because in the worst case the revisions that you did for the journal corrected some important errors, and you don’t want to keep circulating the wrong one.”

1:00:02 An administrative fact he had not published anywhere.

“We have never received a request from a journal to take down a preprint. We have to take it down if we get an appropriate request under the Digital Millennium Copyright Act, and we’ve never received one.”

1:00:27 On what moderation does and does not do: authors click a box saying they have permission, and the archive does not check author agreements, but rejects a submission that looks like a publisher’s PDF with a copyright line.

1:01:04 Why journals are coming around, with the bioRxiv direct-submission button and the citation arithmetic behind it.

1:03:05 Third question, from a demographer at Columbia teaching dissemination, asking what to advise students. His answer opens with a policy he had not mentioned.

“When you post a paper on SocArXiv you can’t take it down later. We don’t let you say I decided this was bad and I want to take it down, because we have a commitment to the scholarly record, and it happened, and you can’t unring the bell.”

1:04:20 The cut point he recommends for students, and the standard he attaches to it.

“If you’re standing up and giving a research presentation for fifteen minutes at PAA, I think there should be a paper behind you that you’re willing to share. That’s just important accountability. I also think that paper should have data and code behind it, but even if it doesn’t, you should be prepared to say, you know that graph you put up with the hundred coefficients that no one can read, they should be able to read that on their own time.”

1:05:14 The parallel test for journal submission: if you think it might be embarrassing to share, it is probably not ready for the reviewers either.

1:05:53 A follow-up on journals whose author guidelines forbid posting at submission, and his answer that guidelines and agreements should not conflict, that the agreement governs, and that he would not recommend ignoring it.

1:07:15 The blinding problem, and the delicate position it creates.

“An editorial assistant may Google the title, find the preprint, and say you have to take this down because it’ll mess up the blinding. Then you’re in that awkward position where you don’t want to annoy them and make them your enemy, but they’re wrong. It may mess up the blindness, but they probably don’t have a policy that you actually have to take it down. It just makes them uncomfortable, or they’re old.”

1:08:12 On blind review as a value rather than a rule.

“Blind peer review is not a sacred principle. It’s a good idea, it has some benefits, it has some downsides. Anonymity protects evildoers among the reviewers, and if they were being held accountable, it might be a fair trade-off.”

1:09:56 A question relayed from a public health co-chair on participation rates from public health and medical fields, which he cannot answer numerically and works through aloud, ending with an admission: “I guess I don’t know exactly what the difference is between medical and public health.”

1:11:36 A question from Sarah on the annotation tool, and a candid answer about a feature he does not promote.

“The author is not notified when you put a comment on their paper. I really wish they were. I don’t promote this feature that heavily, because I’m not sure people want to post their papers and have comments appear that they’re not notified of. If somebody posts a whole thing about why your paper is wrong, I’d like to know right away.”

1:13:23 The overlay journal idea, spelled out: solicit reviews through the annotation layer, publish a list of papers that pass, and you have a journal sitting on top of the archive.

1:14:12 A question on licensing, and his walkthrough of CC0, CC BY, and the non-commercial, no-derivatives, and share-alike variants, with the ethical point that CC0 does not license plagiarism.

1:15:56 What he uses and why, including a publisher that repackages open-access articles into books.

“It’s offensive, it violates our norms, but it doesn’t really hurt you. The open access version is still open. It’s a shame if they’re fooling people into paying for something they shouldn’t be paying for, but it’s not really your job to stop that.”

1:18:13 The last question, from the host, on qualitative research and the ethics of sharing data from vulnerable populations.

1:18:47 His answer, which distinguishes papers from materials and then works through options: naming participants where they consent, sharing interview guides and coding guides and recruitment materials, and blinding transcripts.

1:20:27 The proposal he arrives at in real time.

“I think we should have a way for reviewers to have access to everything but not the public. An intermediate category where ten people can read all your notes and transcripts, but under an agreement that they won’t share it beyond that. We do that with confidential quantitative data to some degree. I think we could do that with transcripts and qualitative materials also.”

1:21:40 And the concession to the ethnographers that closes the session.

“There’s the other issue with ethnographers, which is that just because you read my transcripts doesn’t mean you can analyze my data. The data is me. I was there, it’s the experience. There’s no such thing as reproducibility, because you couldn’t have the interaction and the experience that I had as me at that moment. So reproducibility is not really the issue. But maybe something like accountability is. You still might have lied in what you reported, or been wrong.”

Eighty-two minutes, 109 views, and it is the single most useful recording in the file for anyone trying to describe what he believes, because the question period breaks him out of the prepared pitch four separate times.

The prepared half is the June 2020 talk with better slides. The unprepared half is where things happen.

Start with 1:21:40, because it is the largest concession he has ever made in public on the question he is most associated with. An ethnographer’s data is the ethnographer. Reproducibility is not the issue. That is the ethnographic position, stated by the man whose census arithmetic against Alice Goffman is the thing his critics in that camp will never forgive. He does not abandon anything: he replaces reproducibility with accountability, and holds that a reader still needs some way to check whether the account is honest. But he grants the premise, unprompted, at the end of a long session, to a friendly questioner. Anyone writing about the Goffman affair should have this passage.

And the proposal at 1:20:27 is the constructive version nobody has built. A restricted-access tier where a small number of vetted readers can see transcripts under agreement, on the model of Census research data centers. That is the obvious answer to the qualitative transparency fight that consumed political science after 2015, and he arrives at it live, says it is not his area, and says it would be neat. Six years on, the Qualitative Data Repository at Syracuse does roughly this and he cannot remember its name until a listener supplies it, which tells you how far the idea had traveled inside his own field of vision.

Three other places where the questions produce better material than the slides.

At 15:07 he says, flatly, that probably more preprints than peer-reviewed papers are wrong. That is the sentence missing from the June 2020 version of this talk, where he listed errors on both sides and declined to say which side had more. Here he says it. It costs his own project something and it is almost certainly true.

At 34:41 he identifies where idea theft actually happens: advisers, committee members, and anonymous reviewers, people with private access abusing a position of trust. That is the strongest argument for preprints anybody makes. The usual pitch is speed and access. This one is protection for the powerless against the people who supervise them, and it converts an open-science talking point into a labor argument about graduate students. It is also, quietly, an argument against the anonymity he goes on to question at 1:08:12.

And the Heath exchange from 51:03 to 55:05 is the only moment across every recording in this file where somebody pushes back on him in real time and he loses the point. He says the peer-reviewed-only rule is not reality anymore and reels off the Times, Vox, FiveThirtyEight, Pew. The questioner answers that those are the elite outlets with data staff, and that a twenty-five-year-old reporter with a thirty-year-old editor at a local paper is risk-averse and untrained, and that the education opportunity lies there. Cohen concedes. He is right about the national press and generalized from it, which is a sampling error of exactly the kind he catches in others, and the correction came from a listener rather than from him.

The self-implicating material is thick here. At 3:03 he says the archive is a low-budget operation and that he serves at the pleasure of a library dean, which is the honest description of the infrastructure position he was later given a prize for. At 46:31 he says the value of open data is hard to prove empirically and that he believes it anyway, and asks the room what they think, which is the same admission I flagged in his December 2025 audit: the connection between availability and correctness is asserted rather than demonstrated. At 1:11:36 he explains that he does not promote his own annotation feature because it notifies nobody, which is a product manager saying his own product is broken. And at 1:10:15 he simply does not know the answer to a question about his own platform’s disciplinary composition and says so at length rather than guessing.

Two weaknesses.

The association line at 8:12, that it is captured by Sage and interested in change only for PR, is delivered to a professional audience as a settled fact three months after he resigned. It is a characterization of motive, unsourced, about people who cannot answer. It is also the register he uses for institutions and not for individuals in this talk, which is a distinction worth noticing: nobody is named.

And the workflow advice from 39:58 is thin. Shared folders and unshared folders, drop a file across when ready. That is a personal habit rather than a system, offered to a room asking how to build one, and it sits oddly beside eighty minutes on infrastructure.

For the biography, this recording supplies three things the current draft lacks: the low-budget line, which grounds the SocArXiv paragraph in what the operation is; the private-access argument about idea theft, which is his best original contribution to the open-science case; and the ethnography concession, which belongs in the Goffman paragraph as the thing he came to grant later.

Transcript flags. Every instance of “social archive,” “search archive,” “site archive,” “cyrica,” and “sai archive” is SocArXiv or PsyArXiv; “bio archive” and “med archive” are bioRxiv and medRxiv. At 1:14 “a long and brilliant cd” is CV. At 3:20 “not that much” is his own aside about the budget and should be quoted carefully. At 14:08 “fraser at all” is Fraser et al. At 26:43 “plotted endorsement” is Plaudit. At 31:08 “milanovich” is Branko Milanović. At 33:44 “not worn out” is not borne out. At 44:06 “marshall peer review” is marshal. At 48:24 “pentagonation” appears to be his own coinage and he says he does not know why; treat it as a joke rather than a term. At 59:52 “sherpa romeo” is correct. At 1:10:32 “copic 19” is COVID-19. At 1:21:36 “quality of data repository syracuse” is the Qualitative Data Repository. The host is identified only as Sophia and the questioners as Heath, Alicia, Gilly or Jilly, Sarah, and a relayed question from Diana Hatto; none of the surnames are recoverable from the audio, so describe them by role rather than naming them.

‘Race and Racism in America’ (Sep. 27, 2020)

There’s no YT transcript.

‘Family Sociology Lecture: Race, Ethnicity, and Immigration, part 2’ (Feb. 11, 2025)

0:03 Opens the second half of a lecture on chapter three of his textbook, noting that people who found it on the internet are welcome to follow along.

0:42 Whites as a group with internal ethnicities, the most common ancestries reported being German, English, and Irish.

1:28 Why those boundaries sit differently.

“White groups do not have the same kind of boundaries around them, especially with regard to families and intermarriage, that some other groups do. It’s not such a big deal if someone with German great-grandparents marries someone with English great-grandparents. Not nearly as much of a big deal as it is to cross the more recent boundaries.”

2:00 Census composition from 1970 to 2020: the White share losing numerical dominance, the Latino population the fastest growing and above 60 million, and the non-Hispanic White population peaking and declining slightly between 2010 and 2020, which he calls largely symbolic.

2:46 The arithmetic behind it, worked out for a single day in 2023: about 6,300 White deaths against about 4,900 White births, so roughly 1,400 fewer Whites a day before immigration, against a natural increase of a few thousand among everyone else.

3:40 The correction to his own framing, which is the substantive point of the section.

“Natural increase, or the difference in birth rates and death rates, is not really what’s driving our ethnic composition in the US. It is mostly immigration.”

3:53 The long series from 1900: a previous peak near 15 percent foreign-born around 1910, overwhelmingly European and Canadian, that cohort dying out across the century, then a new influx after 1965 from Latin America, the Caribbean, and Asia, bringing the share back above 14 percent.

4:49 American Indians. Something like seven million before European arrival, reduced by roughly two thirds by 1900 through disease, war, and what he calls the genocide of conquest.

5:23 His explanation for the recent growth in the count, which is a measurement point rather than a demographic one.

“That rapid growth over the last century is due largely to the greater desire of those with mixed ancestry to express that identity. We have a lot of people who have some Indian and some non-Indian heritage, and a lot more of them are now coming forward, especially on the census.”

5:51 About one in five living on reservations, mostly rural, far from centers of economic opportunity, with poverty, isolation, diabetes, alcoholism, obesity, and family violence, which he describes as a continuation of the conquest.

6:28 Pandemic mortality, and the measure he chooses to illustrate it: about 600 per 100,000 American Indian children who lost a caregiver.

7:00 African Americans, and W. E. B. Du Bois (1868–1963), the first African American to take a Harvard doctorate, and his Philadelphia study.

7:40 What Du Bois concluded about the marriage rates he found.

“He tried to understand how this was happening. He definitely saw this as problematic, and his conclusion was it was the legacy of the oppression of slavery, and then the discrimination that they experienced later in urban ghettos, that had devastating consequences for Black family life.”

8:24 The competing account, and the reversal in it.

“An analyst like Daniel Patrick Moynihan produced reports that essentially reinforced racist stereotypes about these problems by attributing racial inequality to family structure rather than the other way around. Rather than saying the family structure resulted from inequality, as Du Bois had said, he tended to look at it the other way.”

9:29 The family resilience perspective, presented as the Du Bois tradition: adaptive, fluid family structure as a response to persistent oppression rather than a cause of poverty, and the historical point that nuclear and separate-spheres arrangements were limited for Black families in the twentieth century by the same poverty, immobility, and mortality that had limited them for Whites in the nineteenth.

10:28 Segregation, illustrated with three maps he made of the same area of Milwaukee in 2021, one for each population.

11:01 The Great Migration from about 1910 to 1970, Black workers leaving southern agriculture for northern factory work, and the segregation that met them.

11:38 Deindustrialization, and the reversal it produced.

“The fact that they had moved historically into these cities, good opportunity at the time, boomeranged and became a bad situation when they were stuck in de-industrialized cities.”

12:31 The family profile: lower college completion, about 70 percent of Black children born to a mother not married at the time against 40 percent overall, higher poverty.

13:14 The trend line since 1950, and the inference he draws from it, which is the analytical core of the lecture.

“34 percent of Black and 51 percent of White adults are married as of 2021. But if you go back to 1950, these rates were much more similar. So this is not a direct consequence of slavery, in the sense that we always had this big gap in marriage rates. The gap in marriage rates has in fact opened up, and the de-industrialization story that I told on the last slide is a big part of it.”

13:59 The marriage market argument, his own research area, delivered to freshmen.

14:24 The four-city comparison, and how the measure is built.

“It’s the number of employed unmarried Black men for every hundred unmarried Black women. So the idea is Black women are thinking about who to marry, as are White women, and they’re thinking about marrying a man with a job. It’s kind of a traditional, slightly sexist assumption, but not crazy. And so the combination of health problems, incarceration, poverty, and unemployment is what creates that relative shortage of men.”

15:23 Latinos as the largest minority group, with the aside on Puerto Rico.

“Puerto Rican people are US citizens but are still considered immigrants from Latin America. It’s just a colony that we have, that America happens to have, in Latin America.”

16:07 Recent arrivals driven by economic crisis, political violence, and natural disasters, and the argument he builds on the continuous flow.

“That continued flow of immigration in general helps maintain cultural continuity. If you drop a population of people from one place to another and don’t have a flow of new migrants, you can imagine their culture kind of dissipating over time. If you have a steady stream of new people arriving bringing with them their cultural ways, then you end up with more cultural continuity.”

17:05 Familism defined as family obligation high in the priority order, self-sacrifice for family well-being, intergenerational ties, and extended household living.

17:31 Profiles for Mexican, Central American, and Puerto Rican populations: lower college completion, similar nonmarital birth shares, high foreign-born shares among Central Americans, higher poverty among Puerto Ricans.

18:03 Asian Americans and educational attainment, and the two explanations offered in order.

“One, you’ll hear this a lot, has to do with the Confucian background, the tradition of striving for educational excellence, the strong parental support for education. On the other hand, that can be overplayed in the analysis, and it’s also the case that many Asian Americans have come from immigrant families who had higher education, or came here to continue professional careers they had already started. So they already had advantages when they came, in many cases.”

19:04 The selection point behind it: immigration law privileges high-skilled arrivals, so the flow itself is selected.

19:30 Profiles for Chinese, Filipino, and Asian Indian populations, with the exception noted for Chinese Americans, a longer-settled population with higher poverty.

20:08 Racism, prefaced by a warning against overgeneralizing, since some Southeast Asian groups have higher poverty and blue-collar employment.

20:42 The Chinese Exclusion Act and the fear of labor competition, then the other edge of it.

“The model minority stereotype, the idea that Asian Americans are inherently or automatically successful, good students and so on, which is a form of stereotyping which seems positive but also comes back to be not so positive for people who don’t achieve that standard, and also for people who just want to be treated as individuals.”

21:28 The pandemic rise in anti-Asian racism, which he attributes in part to presidential language about the Chinese flu.

21:59 Immigration proper: almost 15 percent foreign-born, about a quarter of the population in an immigrant family counting the children of immigrants.

22:33 A consequence he names that is rarely mentioned: immigrants now live in places that had few before, so “we’ve had a lot of people meeting immigrants in large numbers for the first time in the last few decades.”

22:51 The 1965 law, family reunification, and the end of the quota system.

23:26 The two-sided effect, given without a thumb on the scale.

“This has allowed immigrant communities to thrive and be rejuvenated by the influx of recent arrivals. But that has the possible consequence of increasing social distance. You have more people arriving who don’t speak English, you have more communities building up around the influx of new groups, instead of more dispersed, more integrated groups integrating into the rest of the society.”

24:00 The legal history, framed as a seesaw of crackdowns and welcomes, beginning with the 1882 Chinese Exclusion Act.

25:03 United States v. Wong Kim Ark (1898), the San Francisco-born son of Chinese immigrants stopped on return from China.

25:30 The Fourteenth Amendment standard, and the live argument he sees coming.

“Anybody born here is an American citizen, as long as they are subject to the jurisdiction of the United States at the time they are born, which is basically everybody born here unless your parents are diplomats or you’re part of an invading army. That’s what Trump is going to say when his case comes to the Supreme Court about ending birthright citizenship. He’s going to say that today’s undocumented immigrants are part of an invading army, so their children should not be considered US citizens when they’re born.”

26:16 The 1924 restrictions, the Bracero program and its temporary permits for millions of Mexican workers alongside deportations, and the 1965 reopening.

27:00 The border-policing irony, which is the most portable thing in the lecture.

“One ironic consequence of more strict policing of the border is that more migrants decide to stay here. When it’s easier to cross the border, people kind of come and go for jobs and work. When you make the border hard to cross, people are afraid to go back, so they end up staying.”

27:22 DACA, described as administration policy rather than statute and therefore vulnerable, with the dreamers’ position unclear.

28:05 The wall period, then the Biden-era asylum surge, and the people now facing possible mass deportation.

28:47 Acculturation defined as immigrants learning the ways of the new country.

29:16 Assimilation defined as something else, which is the distinction he most wants students to hold.

“Assimilation is the reduction of ethnic distinction between immigrants and the mainstream or host society. Not just the immigrants changing, but the two populations moving toward each other. Assimilation can only work when the host society is welcoming and accepting of the new group. If you eat Mexican food now and you are not of Mexican ancestry, that’s part of assimilation. It’s not just Mexican immigrants and their descendants changing, it’s the mainstream society changing as well.”

30:10 His own grandmother, photographed in front of the bar and liquor store his grandparents ran, and an earlier photograph of her with her siblings and mother around the time they came from Poland.

30:32 The generational taxonomy: the 0.5 generation arriving at retirement age, the first generation as working-age adults, the 1.5 generation arriving as children, the second generation born here, and the third.

32:01 On the 1.5 generation and what it does to a family.

“If you’re 40, or 35, and you come here with your six- or ten-year-old child, that child’s going to learn English faster than you do and is going to end up being more American than you. This produces some generational conflict, but also opportunities for those children.”

32:43 The third generation, and himself in it.

“The grandchildren of immigrants, like me. We may retain our identity as part of an immigrant family, but we come to see that ethnicity as family history rather than our experience. I don’t have an immigration experience. I had grandparents who had a Yiddish accent.”

33:08 Intermarriage as the measure, and why the course spends time on it.

“You could say intermarriage is a litmus test of racial and ethnic difference. If people don’t marry each other, then they really are more separate. If they do marry each other, then they start to grow closer together.”

33:57 Loving v. Virginia (1967), with the note that most African Americans lived under anti-miscegenation law through the middle of the twentieth century.

34:28 The denominator problem, explained carefully.

“Through 2017, 8 percent of Black marriages involved Black people marrying White people. Is that a lot? I don’t know. But the percentage of Whites marrying Blacks is less than 1 percent. That’s because the White population is bigger, so the same number of marriages is 8 percent of Black marriages and less than 1 percent of White marriages.”

35:16 Attitudes, from the General Social Survey: in 1990, close to 70 percent of Whites over 35 somewhat or very opposed to a close relative marrying a Black person, half of younger White adults; by the recent data, a little under one in five over 35 and less than one in ten under 35.

36:04 Social distance defined as the level of acceptance one group holds toward another, with both the attitude decline and the intermarriage rise read as declining social distance.

This is the textbook being taught by its author in 2025, and it is the recording that shows what he does with the material his critics find hardest.

The Du Bois and Moynihan passage from 7:00 to 9:29 is the whole family-structure argument in ninety seconds, and it is more honest than the way the same argument runs on his blog. He gives Du Bois the founding position, that family patterns followed from oppression. He gives Moynihan the reversal, that family patterns became the cause. Then he gives the resilience perspective as the Du Bois tradition continued. What is missing is any acknowledgment that a serious version of the second position exists among people currently working, which is the position he fights Brad Wilcox over every year. A student leaves this lecture knowing there are two accounts and knowing which one the professor holds, without knowing the second is live.

But then he does something at 13:14 that repairs a good deal of that, and it is the best analytic moment in the lecture. He puts up the marriage rate series back to 1950 and shows the Black and White rates were close then, which means the gap opened after 1950 and cannot be a direct inheritance from slavery. That is an argument against a lazy version of his own side’s story, made with a chart, in front of students who would never have caught it. And it sets up deindustrialization as the explanation, which is Wilson’s argument and is checkable.

The marriage market passage at 14:24 is where his own research enters the syllabus, and the way he handles the assumption inside it is worth noting. He says women are looking for a man with a job, calls the assumption traditional and slightly sexist, says it is not crazy, and moves on. He neither hides the normative content of his own measure nor pretends it is neutral. That is a small thing done correctly, and it is the thing he demands from other researchers.

Two places where an unsympathetic student would have a fair objection.

At 21:28 the anti-Asian racism section attributes the rise in part to the president’s language, and the slide apparently carries his picture. That claim is defensible and contested, and it arrives without the treatment he gives contested claims elsewhere in the same lecture. Compare it with 23:26, where the effects of the 1965 law are given as thriving communities on one hand and increased social distance on the other, with no verdict. He can hold a live question open. He does it three times in this lecture. He does not do it for this one.

And 25:30 is a prediction about a legal argument that had not been made yet, stated as what Trump is going to say. He turned out to be broadly right about the shape of the birthright citizenship fight, and in February 2025 it was a forecast delivered as a certainty in a course where students are being graded.

The lecture also contains one of the better lines of political demography anyone has offered undergraduates, and he tosses it away. At 27:00 he explains that harder borders increase the resident undocumented population, because circular migration stops and people who cross stay. That is a finding with a large literature behind it, it runs against the intuition of everyone on both sides of the argument, and he gives it fifteen seconds without a citation.

Two notes.

At 30:10 he puts his own grandmother in the textbook and in the slide deck, in front of her Polish family photograph and in front of the liquor store she and her husband ran, and at 32:43 he places himself in the third generation, with grandparents who had a Yiddish accent. That is the only time in any of these recordings where he uses his own family as course material rather than as blog material, and it belongs beside the passage about his mother in any profile.

And the measurement reflex is everywhere, unprompted. The American Indian population growth is explained as identification change rather than demography at 5:23. The intermarriage percentages get a denominator lecture at 34:28. The Asian education finding gets a selection explanation at 19:04 that undercuts the cultural one offered first. He cannot teach a number without teaching where it came from, and that is the thing his students will retain longest.

Transcript flags. At 0:36 “New Mina middleast North African” is the new MENA category. At 7:24 and 9:14 “web Deo,” “W EB de Bo,” and “Deo” are W. E. B. Du Bois. At 8:53 “Daniel Patrick Mahan” is Daniel Patrick Moynihan (1927–2003), and the report is the 1965 The Negro Family: The Case for National Action. At 11:07 “Doug rate migration” is the Great Migration. At 25:12 and 26:15 “Wong Kim Arc” and “won Mark” are Wong Kim Ark, United States v. Wong Kim Ark (1898). At 26:32 “bercero” is Bracero. At 24:36 “the 1980 the 1882 Chinese Exclusion Act” is 1882, self-corrected. At 27:28 “daaka” is DACA. At 34:14 “it did not become illegal to ban intermarriage until 1967” is what he means and the sentence as delivered is tangled; Loving v. Virginia was decided June 12, 1967. At 35:46 “in 1900” is 1990. At 35:20 “litness test” is litmus test. At 12:59 “at the time they get married” is at the time of the birth. At 3:19 the daily figures around “40 42 new ones who die” are garbled and should be read off the slide.

‘Philip N. Cohen, Citizen Scholar: Public Engagement for Social Scientists’ (Mar. 1, 2025)

0:05 Opens by dating the book against the political calendar: proposal written just after Biden’s election, draft finished as Trump was winning the 2024 primaries.

“It’s really an interregnum product, and now all of a sudden the new world has arrived, and we can stress test the book and see how we think it works.”

0:48 A self-deprecating framing of what follows: you reach an age where you tell the same story over and over, and he is going to formalize that.

1:07 SocArXiv, described modestly.

“We were trying to make sort of modest increments on the system of scholarly communication. Like a preprint server. We’re not replacing journals, not eliminating peer review, we’re just trying to make the system better.”

1:33 The shift in his own research, stated more plainly than anywhere else in the record.

“In my research I’ve taken a pretty descriptive turn. I’m less about trying to discover new theories and mechanisms, and more about trying to figure out what’s actually happening, and find ways to communicate about that to other scholars and various publics.”

1:54 Fertility as the current example, and the answer he keeps having to give journalists.

“I keep getting journalists and other people who ask me, what works to raise fertility? And I’m trying to communicate, nothing that we know of works. The dot on each one of these lines is the lowest point on the line, and they’re mostly at the end. So for most countries, the lowest low is now.”

2:24 Pandemic work he says he was not prepared to do, including the correlation between Trump vote share and vaccination rate, which he calls grotesque and says cost thousands of Republican lives for no good reason.

3:00 A description of his public register that nobody else in the file would give him.

“A lot of my public work is sort of in the category of debunking, or being a crank and complaining about things. Trying to be more constructive than just that, but that’s a recurring theme.”

3:35 On the lawsuit and what it raised.

“I call this activism. I guess it’s activism if you’re expressing your political opinions in public. It’s pretty lightweight activism, me complaining about Trump on Twitter. But it did lead to me being part of a lawsuit. And that raised the question, is this discrediting, is this crediting, is this good or bad for our integrity, our reputation, our discipline.”

4:14 The book’s origin with editor Eric Schwartz at Columbia, and the title.

“I hit on this concept of citizen scholar, which I’m aware is quite pretentious sounding. Who calls himself a scholar nowadays? But I kind of want to reclaim that a little and make it less pretentious and more just descriptive. The citizen scholar is somebody who accepts the duties of a scholar in society. It’s in the practice of citizenship that we generate knowledge.”

5:03 The political situation that arrived after the book was finished, and what he says is new about it.

“The idea is quite foreign to those of us who’ve been around for a while, that we live in a modern democratic capitalist society where, at the highest levels of power, their perception is that the main problem with our society is too much higher education, too many universities, universities where they don’t pursue knowledge and truth, they pursue deceit and lies.”

6:01 The concept he says he underweighted in the book: common sense as the populist substitute for evidence, in the second inaugural address and in the response to the Washington plane crash.

“That is how the authoritarian fills the vacuum that results from the attack on scholarship and university and science.”

7:03 A preamble on trust, built up in three stages. Fish follow the nearest fish because they have no memory of which fish is reliable. Humans add reputation through gossip, which he calls a collective memory. And then institutionalization.

8:07 The double edge of the institutional stage.

“You walk out there with a sign on you that says I work at a major university, and some people will believe you because of that. On the other hand, some people won’t.”

8:36 Do your own research as a meme, and what he says it means.

“Doing your own research is not trusting experts and science and the accumulated wisdom of that institutional framework. And we now know that people who do their own research know less than people who don’t do their own research. You keep using that phrase, and I don’t think it means what you think it means.”

9:16 The elevator, which is the governing image of the talk. Nobody knows how one works, and everyone trusts them.

9:36 The distinction he wants to draw, which is the most useful idea in the lecture.

“More than a fact check, which is a particular fact that something is true, we have veracity, which requires systemic trust. The fact may or may not individually be true, but it comes in a package which has some trustworthiness.”

10:06 Anthony Giddens on disembedded systems, applied to the elevator certificate.

“The person who inspected the elevator might have their name on the certificate, but nobody’s telling us how it works, and that they checked it that day, and that they know it’s still going to work. There’s no statistics on elevators. The person that we’re trusting is not there. They’re disembedded. They’re part of an institutional framework, not an individual person.”

10:38 Why that matters: modern economies and democracy run on trusting people who are not there, and the absence of a person who can be trusted is a fragility in the system.

11:26 The book’s central proposal, which follows from that.

“One thing we can do is be reembedding agents. We can show up at that moment, sort of in the elevator, and say, you don’t have to worry about this, because we’ve figured this out and here’s how we did it. And I’m a person who lives in your community, and I have some firsthand knowledge, and you can trust me for some reason. So a moment of reembedding.”

12:15 A two-by-two of democratic against scientific, with China in one corner and the current United States in another, and his own preference for secular democracy stated as such.

14:01 Riding the wave, and the lag problem.

“It takes time to do projects. Grants, data collection, analysis, peer review takes years. If you’re making decisions about what you’re going to publish in three years based on something that somebody started working on five years earlier, academics can really lag.”

14:52 The exact moment he understood the pandemic: March 2, 2020, reading about Washington state cases with no known origin, and looking at a map of planes in the air.

15:22 The citation spike: on a log scale, social science articles keyed to pandemics rising from about a hundred a year to ten thousand in one year, alongside earlier spikes for hurricanes after Katrina and terrorism after 2001.

16:10 What the life sciences did: shorter papers faster, first-time preprint use, media sourcing them, hundreds of special issues, publishers dropping paywalls, and real rather than nominal interdisciplinarity.

17:25 Description as the thing he wants elevated, illustrated with the New York City Health Department’s 1913 parade of statistics, where people stood in the rain to watch statistics go by on floats, and with Du Bois’s charts of marital status around 1900.

18:40 The research life cycle with description overlaid, and the loop he wants it to run in: listen to what people care about, gather what exists, do descriptive work, push it back out, find out what people think now.

20:17 The professional cost of that, and his explanation for why academic prose sounds the way it does.

“One of the things about the tendency of academic work to be dismissed as jargon-written and elitist is partly that we don’t credit descriptive work. If you have a descriptive finding, it’s considered to be less than, in a lot of academic circles, and so we end up doing more puffery just to make our work sound like it’s not just descriptive. Very off-putting to the public, in my opinion.”

20:52 The county-spike example: sudden rural case clusters he traced to prisons, meatpacking, and nursing homes, written up fast in a minor open access journal.

21:32 Why he did it, said without embarrassment.

“I wanted to plant a flag that says here’s this data, you can get it from me, here’s a quick analysis of something I think we should be paying more attention to. To try to get myself on the radar as somebody that people might want to talk to about this.”

21:55 Openness, introduced through the fraud cases, and the argument he builds on them.

“We found out they were frauds because other people checked their work. We were able to check their work because of the norms of transparency that were required in their discipline. And more important than that, practically, is that people know that they can. We’re working in an environment where somebody can check our work, and it’s one of our secret powers. The willingness to have our work scrutinized separates us from a lot of other professions.”

24:17 The collective action problem: openness benefits the community and is not in any individual’s immediate interest, which is why it needs institutions. Brian Nosek’s pyramid of infrastructure, then community, then institutionalization.

25:02 The cost he refuses to minimize.

“It does also make us vulnerable. Anytime anybody can read our work, it makes us vulnerable to whatever they might react to. Particular bad actors and people who really don’t like us can create misinformation or harass us. I don’t minimize that at all.”

25:57 Peer review, where he says he is being deliberately provocative.

“Peer review is an accepted language of trustworthiness. Even Robert F. Kennedy Jr. is going to say he’s got peer-reviewed research that shows whatever nonsense he’s talking about.”

26:47 The scale of it: millions of articles, tens of thousands of full-time equivalents of labor.

27:10 The claim about ownership.

“We, the cogs in that machine, don’t have to let the owners of that machine define and control that process to the extent that they do. We can recognize the principles involved in peer review and say, actually, what we are trying to do is peer assessment. Assessment by experts who know about particular topics, making public their evaluation of the work. Some of it happens in journals, but it doesn’t all have to happen in journals.”

27:45 What has changed, which is the strongest passage in the talk.

“This is especially important in the new world where publishing is essentially free, where peer review can no longer stop bad research from appearing in the world. It used to be that printing was expensive and difficult, and you could use peer review to decide what was going to be printed and shipped and bound and stuck on a shelf. That is not the case anymore. You cannot stop bad research by just saying it’s bad. That’s not how it works, and that’s not the goal of peer review anymore. What peer review is doing in real life today is filtering or curating and labeling research as good or bad. We’re not gatekeeping in the sense of preventing things from appearing. We’re helping people evaluate and decide what to read. It’s a very different job, but it’s the same principle.”

29:20 Generation labels as the example of something peer review never touched, since the terms came from marketing.

30:05 The Pew campaign, told with the credit given.

“I like the Pew Research Center a lot. They do really good work. They just got addicted to generation clicks, which has happened to a lot of people. Whatever your research is, if you put millennial in the title it gets more clicks. And then eventually they changed their process. Minor victory. Didn’t stop the train of generations, unfortunately.”

30:52 What he is trying to do with his own public complaining.

“I’m trying to elevate what looks like us just complaining about stuff in public to a higher level of value, and say, no, actually what we’re doing is important. We’re discussing other people’s research.”

31:16 His rules for it: find a role, behave well, be constructive, practice accountability, listen as well as speak, and be open to admitting your own errors.

31:38 Social media, and the metaphor he rejects.

“The megaphone is a very common metaphor for this, and it’s terrible. It’s a terrible way to communicate, except in very specific situations, like when you need to make an announcement. At one point I reviewed three books on public sociology and they all had some version of the megaphone on their cover.”

32:39 Rogers Brubaker on hyperconnectivity, and the four problems he takes from it: abundance, miniaturization and context collapse, convenience that serves bad actors equally, and a quantification treadmill.

34:24 A discipline-and-quality problem, where scholars become trapped in the engine of their own engagement and define themselves on its terms.

35:43 The advice on promotion, which cuts against academic manners.

“Many academics feel like it’s improper, or that it makes them look bad, if they’re always running around saying, hey, this reminds me of my own work, pay attention to me. That’s exactly what you should do. When you write an article, that’s what you’re doing. The article is not the research. The article is the description of the research. The article is telling people that you did the research. Same as a tweet.”

36:44 On punching down, and on sharing paywalled material.

“Sharing paywalled stuff should be considered taboo more than it is. You can make your reputation much more positive just by sharing open links.”

37:34 On practice, with a disclosure about the talk itself.

“You get to practice and develop your voice and find out what works. I tried explaining it this way and nobody cared, I tried explaining it this way and it worked better. You can delete your old tweets that nobody liked. This is the fourth time I’m giving a version of this talk, and I’m trying to pay attention to what works and what doesn’t.”

38:49 Activism, where he places himself.

“I don’t think activism is bad for your scholarship or for your reputation as a scholar. Life is too short not to speak out as a citizen. It’s our job in life to be active citizens, whether that means literally protesting or not. Imposing silence on ourselves is not appropriate or good.”

39:47 The duality he wants to keep, offered to people who reject it: as a citizen you take a position, as a scholar you are pressured to maintain distance, and both can go too far.

40:57 What exposure buys.

“If you go around speaking your opinion all the time you’re taking a risk, and that’s part of the value of what you do. When you stand up in a meeting and say this is what I think, the fact that you’re exposing yourself is part of what makes it valuable.”

41:39 The asymmetry between the roles, which is the distinction the book turns on.

“If you’re an activist, it’s not your job to list all the ways that you’ve been wrong before and all the ways you might be wrong right now. It might make you better in your activism, but it might also slow you down. Different from our role as scholars: being wrong is good and important in our scholarship.”

42:14 Positionality, where he takes a middle position.

“I don’t think I need to state my political opinions in every demography article that I write. On the other hand, I think that should be available to people, and I want people to be able to say, wait a minute, I think your work is not trustworthy because I happen to know something about your political beliefs. I think we can work with that tension instead of trying to avoid it.”

43:00 The challenge he says he must accept.

“If somebody wants to say, you’re biased because you think poverty is bad and we should spend the money to eliminate poverty, so I don’t trust your research, that’s a challenge I should be willing to accept.”

43:33 The policy implications section, taken apart.

“You finally publish an article and you get two or three paragraphs to state your policy implications at the end. This feels like a great privilege, and it is, it’s sort of a courtesy. You’ve done all this work and they’re going to let you attach a few paragraphs of opinion to it. But it’s not really very effectual. If you look at actual policy implications that I have, they’re often so broad that anybody could have written them about almost anything. I find that kids in poverty do worse in school, therefore we should eliminate inequality. You could have said that already. It’s not really a result of your research.”

44:44 The case that proves the point.

“The terrible recent example is the expansion of the child tax credit in 2021, which lifted millions of children out of poverty. We had research to back that policy, the policy worked, and then they cancelled it. So just getting the research out, even with the policy implication section, is not enough. We have to actually do politics. The reason I invoked Du Bois is because he looked at that problem over many years and concluded that sometimes you just can’t wait for the last word of science. You have to actually do the politics.”

45:36 His own Twitter campaign, described as strategy rather than impulse: a theory of what line would be effective, a large platform, a favorable algorithm.

46:14 Why the identities did not separate.

“I was doing it not separately from my scholarly identity, but as a sociologist. It doesn’t mean that I went out there every day and said, I’m a sociologist and my research shows that Trump is corrupt and incompetent and authoritarian. But I was putting my reputation on it. Part of the lesson of the social media chapter is you don’t get to have different identities in different contexts unless you really work at it, with anonymity.”

46:42 The line the room laughs at.

“So I sued the president. Of course I did. Sociologist sues the president. I can’t tell you how many times people said something like, of course you did.”

46:55 The defense, with the risk named.

“I don’t think it undermined my research that had nothing to do with those topics, although it would not be hard for somebody to find one of my articles on divorce and post a thing about it on Twitter with some antisemitic comment.”

47:19 Why he was selected as a plaintiff, described without vanity.

“They chose me through the reality show process they use to choose plaintiffs for these things: reasonable person, professional career, peer-reviewed research, responsible citizen, doesn’t shout profanity or throw ketchup on the Mona Lisa. For better or worse, that’s the role I played in that group.”

47:55 The theoretical question underneath it.

“We show our good acts in one area to help generate legitimacy in another. Sort of the Bourdieu question of the exchange rate between different types of capital. Do I get to write an op-ed about Trump because I have written articles about divorce rates? I could be wrong about this and I’m happy to argue about it either way.”

48:26 Why he thinks hiding will not work.

“When Vance says the universities disseminate deceit and lies, it’s really the social sciences and humanities, and maybe the climate people and public health, that they’re going after, because that’s the work that’s intelligible to their followers. But what they’re really doing is undermining the whole system. If we cowered in the corner, if we stopped mentioning diversity or structural racism in our work, it would not help. They would find another way, because we’re not actually the target. We’re the target they use, but the actual target is the whole system of knowledge. I don’t think we can win the battle by hiding.”

49:40 The conclusion on roles: interdependent rather than opposed, since one person holds all of them and a career and a life do not divide.

50:24 The last point, on credit for public work, and the position that separates him from most people making his argument.

“I care more about having a university that protects my right to speak than one that agrees with me. I’m not trying to get my university to put out a statement about Palestine. I’m trying to get my university to put out a statement protecting people who make statements about Palestine. I’m not trying to get them to agree with me. That’s a losing battle.”

51:24 And the reason he does not want the credit.

“There is a neoliberal trap, in the sense that once you start trying to get credit, now you’re negotiating and you’re going to end up playing a game you might not want to play. I don’t think we should just get paid for being a good citizen. If you’re out there talking to people and learning new things and having interactions with real people, your work will become better, your writing will be better, you’ll be more successful without getting credit for your public work.”

This is the book’s argument in fifty minutes, delivered by a man who knows he is about to be tested on it, and it is the most self-aware document in the file.

Three things here are new.

The first is the elevator. The distinction at 9:36 between a fact check and veracity, and the Giddens point at 10:06 about disembedded systems, is the theoretical spine of everything he does, and it explains why he is so preoccupied with replication packages. A fact check certifies one claim. A working elevator certifies a system, and nobody involved in certifying it is present when you step in. What Cohen wants from open science is what an elevator inspection certificate is meant to be: a promise that a stranger could check, from a system that has to keep working after everyone who built it has gone. And the reembedding proposal at 11:26 is the answer to the question I have been raising across these notes, which is what the dormant right to check is for. His answer is that it is a signal rather than a service. Availability tells you the system expects to be inspected. That is the argument, and here he states it explicitly rather than assuming it.

The second is the peer review passage at 27:45. Print scarcity was what made gatekeeping possible, print scarcity is gone, and therefore peer review can no longer prevent bad work from existing. What it can do is label. That reframes the whole open-access argument from access to curation, and it is a more honest description of what journals now do than journals give of themselves. It also, quietly, concedes the objection to his own archive: if peer review is labeling, then a repository with no labels has given up something real, and the March 2026 AI policy is him discovering exactly that a year later.

The third is 41:39. An activist who catalogues his own past errors is a worse activist. A scholar who does not is a worse scholar. The two roles have opposite relationships to being wrong. That is the actual tension, it cannot be dissolved by disclosure, and he does not pretend otherwise here. Anyone writing about the April 2020 lecture, where he crossed from one role to the other in front of a captive class, should put this passage beside it, because he has named the problem himself with more precision than his critics have.

Now the places where the talk works against him.

At 3:00 he calls his own public work debunking and being a crank. At 21:32 he says he wrote up the prison and meatpacking finding partly to get himself on the radar. At 45:36 he says the Twitter campaign was strategic, that the algorithm favored him, and he says congratulations to me. At 37:34 he tells the room this is the fourth version of this talk and that he is watching which lines land. That is an unusual quantity of visible self-interest, all volunteered, and it will read to a hostile audience as either disarming honesty or as a man who has worked out that admitting the thing inoculates him against the charge. I lean to the first reading because none of it is flattering, and the second is available.

Two arguments are weaker than they sound. The claim at 8:36 that people who do their own research know less rests on a body of survey work with real measurement problems, and he states it as settled. And the two-by-two at 12:15 is a set of assertions with a joke attached; he apologizes for it in advance, which is the right instinct, and then uses it anyway.

The most consequential passage for my purposes is 50:24. He says he wants a university that protects his speech rather than one that agrees with him, and he uses the Palestine example to make it concrete. That is a position with costs on both flanks in February 2025, and it is the clearest evidence in this file that the citizen-scholar argument is a procedural commitment rather than a partisan one. It also sits oddly with the man who told an introductory class not to listen to the president, and he does not reconcile them.

For the biography, this recording supplies the theoretical vocabulary the current draft lacks. The essay describes what he does. This is where he says why, in terms borrowed from Giddens, and where he names the descriptive turn as a deliberate choice rather than a drift.

Transcript flags. At 1:09 “soci archive” is SocArXiv. At 11:29 “in utions” is institutions. At 18:20 “WB de boys” and at 45:19 and 45:23 “a boy” and “the boys” are W. E. B. Du Bois. At 4:22 “Eric Schwarz” is Eric Schwartz. At 22:08 “Franchesco Gino Michael Lor dri staple” is Francesca Gino, Michael LaCour, and Diederik Stapel; get all three names right if you use the passage. At 24:39 “Brian noek” is Brian Nosek. At 32:43 “Rogers Brewbaker” is Rogers Brubaker, and the book is Hyperconnectivity and Its Discontents (2022). At 34:19 “the ordinal Society” is Fourcade and Healy’s The Ordinal Society (2024). At 48:00 “the Bordan question” is Bourdieu. At 10:07 “Anthony giddens” is correct. At 39:47 to 40:57 there is a gap of roughly thirty seconds in the transcript where the audio drops mid-sentence; if you quote around it, note the break. At 46:24 “corrupt and comp competent” is corrupt and incompetent, echoing his 2017 tweet.

‘Philip Cohen: The Family, Chapter 7: Love and Romantic Relationships’ (Mar. 13, 2025)

0:00 Opens a recorded lecture on chapter seven of his textbook, love and romantic relationships, and tells viewers to follow along.

0:11 Social scripts, defined as a commonly understood pattern of interaction that serves as a model of behavior in familiar situations.

0:32 What a script does, and why it is shared rather than private.

“The key thing about a social script is it helps you enter into a new social interaction without having to reinvent it from scratch. You don’t have to figure out what kind of situation you’re in, decide how to act, decide what your role is in relation to other people. A script is not something you have secretly. A script is something that is shared socially.”

1:46 The store example: you are a customer, this is a store, the relationship will be superficial, and all of that is settled before anyone speaks.

2:16 Where the scripts have thinned.

“When it comes to relationship formation, we still have scripts, but they are becoming less clear as starting and building relationships is becoming more diverse. There’s more different kinds of situations and different kinds of relationships, so communicating about them has become more complicated.”

2:47 Four reasons: the rigid mid-century dating system is gone, cohabitation is now a stage that has to be negotiated rather than assumed, more divorced and older single people are in the market and some have children, and the technology keeps changing.

4:05 Love, and the evolutionary account he gives for it.

“We made an evolutionary trade-off where we produce babies with giant heads. In order to make a baby that has a giant brain, its body is very underdeveloped, so babies need tons of love and support or they will die. And the only way to get adults to reliably provide that, from the point of view of evolution, is a set of emotions that we could call love.”

5:20 The inference he draws from that, which is the frame’s payoff.

“Loving a child is probably the original love.”

5:27 Love defined as deep affection and concern for another person with whom one feels a strong emotional bond, and romantic love as passionate devotion and attraction.

6:05 Ann Swidler on the modern romantic ideal as mythical, with its traits: unambiguous, unique, permanent.

6:38 The role of obstacles.

“People prove and demonstrate their love at the same time by overcoming obstacles in the quest for love. When somebody has to overcome obstacles, has to outcompete somebody else, has to fly around the world, has to do something dramatic to demonstrate this, they are both proving and demonstrating their love in a way that reinforces it for both people involved.”

7:04 How the marriage structure used to underwrite the ideal: one person, chosen young, faithfully, forever.

7:32 The historical caveat.

“We should acknowledge, historically speaking, that the idea that spouses would really love each other is a modern ideal. This is not the way humans always structured their families, certainly not in premodern societies where marriage was much more practical or economic or religious, and was not all about passionate love.”

7:56 Utilitarian love, defined as the practical, rational dedication of one person to another based on shared understanding and emotional commitment, and described as less cosmic and more down to earth.

8:42 The parallel he draws, which is the most ambitious claim in the lecture.

“It parallels the growth of rational science replacing irrational religion as a way of explaining the world, and also the spread of democracy instead of monarchy. We choose our leaders rationally instead of having them just be born leaders. Science is a rational system, a list of pros and cons of all the evidence that we have, rather than a religious dogma or doctrine that is just handed down to us.”

9:20 Romantic relationships defined generically as mutually acknowledged ongoing interactions with heightened affection and intensity, including married and unmarried, not necessarily sexual.

9:52 Why the definition has to stay loose.

“In modern relationships there’s not necessarily a fixed beginning or end. There’s not a formal status, and they might not even be called the same thing. One person says this is my boyfriend, the other person says they’re not sure what they are, or they say different things to different people.”

10:19 Mate selection defined, with the note that the term applies even where no literal mating occurs.

10:45 Why the concept carries so much weight in the course.

“Mate selection is the building block of families, and the elements of mate selection will determine other social things like the level of inequality, inclusion or exclusion of certain groups. If poor people marry rich people, if certain religious groups exclude other religious groups, if you have a process that leads you to have older men with younger women, that will change the way the relationships evolve.”

11:29 The mid-century dating system, and the figure he uses to convey how routinized it was: a survey showing college women going on an average of twelve dates a month with five different men.

12:13 What replaced it: a date as a single event, dating as a more stable relationship, and language that is less formal and more open to negotiation.

12:36 Online dating, with about 20 percent of young adults having met a current partner through an app.

13:05 Who benefits most.

“It’s especially important for people who are members of small groups or marginalized groups, where you can’t just walk into a bar or walk into a crowded room and find somebody who is like you or meets your criteria.”

13:24 What the technology does to the model of love.

“It does promote utilitarian models. Once we have the metrics, the app can list and rank people for you, and that’s really utilitarian thinking. You’re not looking for just a spark, you’re sorting and searching on a whole set of criteria: age, sex, race, weight, occupation, whatever else.”

13:52 The commitment worry, flagged as unresolved.

“There is a concern, and this is hard to pin down, whether or not this is a legitimate concern, that it reduces commitment. If it’s so easy to meet new people all the time, then maybe people will be reluctant to commit to the person right in front of them, because in the back of their mind they’re always thinking, yeah, but there’s another person just a click away.”

14:23 Hookup culture, drawing on Lisa Wade, and the framing point that it is an environment people live in whether or not they participate.

14:57 The distribution, which cuts against the usual picture: more than a third of college students in a survey about ten years old reported no hookups, and only a small number reported ten or more.

15:16 Hookups defined as casual sexual or romantic contact without implied commitment or exclusivity, with the definition acknowledged as subjective and locally variable.

15:38 The finding that most surprises students.

“College students who do this, whatever this is, hookup culture, they’re just as likely as those who don’t to think of themselves as in a committed relationship, or to think of themselves as heading toward committed relationships. It is a way of meeting people. The difference between dating and hooking up is the expectation of sexual contact on the very first encounter. So what you’re dealing with is people who have a sexual encounter before they get to know each other, rather than the other way around.”

16:11 Wade’s interactional work, and what it consists of.

“There’s this interactional work that often goes on afterwards where people try to establish that it didn’t mean anything. So you first have the hookup, then you communicate to each other what you mean to each other, and often that’s about not meaning a lot to each other, to sort of extricate themselves from the situation.”

16:33 The risks: sexual contact between people who do not know each other well, accentuated by alcohol and drugs.

16:54 Beauty as a sociological object, and why it belongs in a mate selection chapter: norms, expectations, culture, and power.

17:29 The argument against reading attraction as biology.

“A beauty characteristic may be common, but that doesn’t mean its attraction is biological. Many of the traits associated with attractiveness in modern Western society have nothing to do with health or reproductive fitness. It’s not as simple as people find beautiful the bodies of people who will produce healthy babies. In some cases what has been considered attractive by mainstream society has been so unhealthy that it’s really just implausible that evolution would have produced it.”

18:48 The gender asymmetry, and the account he gives of it.

“Historically, and in the scholarship on this, the focus has often been on what defines beauty for women, partly because men, as the more powerful people in society, were choosing their spouses. The more powerful the man, the more choice he had. So it was a situation where powerful men were choosing women and women were being chosen, and the beauty standard was the way that women worked in that sexist milieu.”

19:53 High heels as the worked example, which he notes is his own research area: they change how a woman walks, limit mobility, and that limitation was itself found attractive.

20:30 The mechanics: altered posture, accentuated hip and pelvis movement, attention to the waist-to-hip ratio, and long-term physical damage.

21:03 The function he emphasizes.

“One of the things high heels do is help establish the difference between men and women, taking whatever differences there are naturally and accentuating them, so men and women seem more different. In terms of the theory of symbolic interaction, high heels are one way that women perform femaleness and that men perform looking at femaleness.”

22:00 Foot binding in China as the comparison case, with the class signal it carried.

“What it communicated from the point of view of the family was, in our family the girls don’t have to work in the fields, we are richer than that. And so that was the desirable quality in the marriage market.”

22:54 The disanalogy he supplies himself.

“Obviously it’s very different. Wearing high heels is voluntary, for one thing. It’s not something which young girls are locked into before they’re old enough to have any say. But there’s a socialization element here, and I think it’s a fair comparison, at least to discuss.”

23:15 His own research on the injuries: on the order of 40 or 50 women a day in American emergency rooms before the pandemic, with foot, leg, knee, and hip injuries from falling and twisting.

23:45 The 2020 drop, and the check he ran on it.

“Either a lot fewer people were getting injured because they weren’t going out to parties and bars and work and wearing heels, or it’s also possible they weren’t going to the hospital when they got injured because they were afraid because of the pandemic. One of the things I checked was, are the injuries more severe during the pandemic? They were not really more severe. So it’s not that people with minor injuries were staying home.”

24:24 The line he closes the section on: “One good thing about the pandemic is fewer women were breaking their ankles from wearing high-heeled shoes.”

24:33 Homophily, defined as the principle that similar people have more contact with each other, visible in friendship, collaboration, and romantic relationships.

25:25 Endogamy and exogamy defined, with the etymological note that gamy is marriage.

25:59 Why the second matters.

“Exogamy is good for the gene pool, it’s good for biological diversity, it’s good for cultural diversity. So you need some exogamy, even if people are more comfortable with endogamy.”

26:19 The closing observation, which loops the two halves of the lecture together.

“It’s interesting that nowadays, when we have this concept of utilitarian love, it often reinforces endogamy. If you’re making that checklist of things that you want, it often ends up producing people who are kind of like you. Same religion, same area, maybe same social class background. Unless you’re trying to get away from all that, then you could be making a list thinking, I don’t want someone from my religion, I don’t want someone from my town. It still can feed into the utilitarian view, but more often the similar person shows up on the checklist.”

Thoughts

Twenty-seven minutes, 272 views, and it contains the single best argument in the file, which he builds across the lecture without announcing it and lands in the last thirty seconds.

The structure runs like this. Romantic love is a mythical ideal: unambiguous, unique, permanent, proved by obstacles. Utilitarian love is its modern counterpart: a checklist, a weighing of pros and cons, parallel to science replacing religion and election replacing inheritance. The dating apps push people toward the second, because sorting on age, income, and education is what the interface is built for. And then at 26:19: utilitarian love reinforces endogamy. The checklist returns people to their own kind.

That is an argument with teeth. Assortative mating on education is one of the better-documented contributors to household income inequality over the past forty years, and here is a route to it that nobody would have predicted. The rationalization of intimacy, which reads as liberation, produces more marrying within one’s own group than the mythical version did, because a spark can cross a line that a filter cannot. He gets there in one clause and does not stop to admire it.

It is also the same argument as his work on marriage markets and the divorce decline. Marriage becomes rarer, more deliberate, and more concentrated among people who already have resources. The sorting mechanism he describes in this lecture is how that happens at the individual level. The lecture and the research program are the same claim at two scales, and he does not say so.

The evolutionary passage at 4:05 is done correctly, which is rare. He gives the adaptive story for why humans love, grounds it in the obstetric trade-off, and then extends it outward: the original love is love of a child, and pair-bonding love is derivative. Then at 17:29 he uses the same evolutionary frame against itself, pointing out that the beauty standards actually in circulation are frequently incompatible with reproductive fitness, so the just-so story does not work as an explanation of attraction. That is using biology where it explains and refusing it where it does not, in the same lecture, without hedging in either direction.

And the hookup section at 14:57 is the corrective a nineteen-year-old most needs. More than a third of students in the survey had none. The heavy participants were a small group. And the students who do participate are just as likely to want committed relationships. Every part of that runs against what a freshman believes about his own campus, and it comes from Lisa Wade’s own data rather than from reassurance.

Three weaknesses.

The high heels material runs about five minutes in a twenty-seven minute lecture, and he says himself at 21:03 that he uses the example because he has spent a lot of time on it. That is a researcher’s hobbyhorse taking a fifth of a compressed lecture. He also introduces his own injury paper into the class, which is fine and which he does not flag as his own until midway through.

The foot binding comparison at 22:00 is where a student would push back hardest, and he anticipates it well: voluntary, adult, not irreversible, and he says the comparison is fair to discuss rather than fair full stop. That is close to the right handling. What is missing is any account of why women wear them, offered from women who do. The entire passage runs on what heels communicate to men and what they do to bodies. A student who wears them would recognize nothing of her own reasoning in it.

And the parallel at 8:42 between utilitarian love, science replacing religion, and democracy replacing monarchy is a large modernization claim dressed as a definition. It carries the assumption that rational selection is the later and more developed form, which is precisely what his own closing point about endogamy calls into question. He has the material to complicate his own frame and does not use it.

For the biography, this is the recording that shows him teaching theory rather than data, and it shows the same reflex that runs through everything else: define the concept, give its strongest form, then find the case where it fails. Swidler’s romantic ideal, then the observation that premodern marriage was never about it. The evolutionary account of love, then the beauty standards it cannot explain. Utilitarian love as modern rationality, then the finding that it sorts people back into their own groups.

Transcript flags before you quote. At 6:09 “Anne swidler” is Ann Swidler (b. 1944), and the relevant work is “Love and Adulthood in American Culture” (1980) and Talk of Love (2001). At 7:38 “spors” is spouses. At 14:35 Lisa Wade’s book is American Hookup: The New Culture of Sex on Campus (2017), which he does not name. At 24:33 and 24:39 “homop” is homophily, and the classic reference is McPherson, Smith-Lovin and Cook, “Birds of a Feather” (2001). At 25:04 “common Mercers” is unrecoverable and appears to describe a photograph on the slide. At 19:57 and 22:58 “ey heel” and “high deals” are high heels. At 20:59 “Dage” is damage. At 25:38 “gami is marriage” is his own etymological aside, correctly. The twelve-dates-a-month figure comes from mid-century survey data cited in his textbook and should be checked against the book before you print it, since the number sounds implausible without its source. His high heels paper is on medRxiv, which he mentioned in the February 2022 talk, and the emergency room data come from the National Electronic Injury Surveillance System.

‘Philip Cohen: The Family, Chapter 8: Marriage and Cohabitation’ Mar. 27, 2025

0:00 Opens a recorded lecture on the first half of chapter eight, marriage and cohabitation.

0:13 Kristin Perry and Sandra Stier, the couple who sued for the right to marry in California, used to open the chapter and to make the point that marriage carries both symbolic and legal weight.

“It matters to people that their marriage counts, that other people consider them to be married, and it also affects their rights and obligations, reciprocal duties between them and society.”

0:56 The organizing question: how cohabitation has changed the meaning of marriage.

1:21 Diversity, in three items: about two-thirds of people who marry have lived together first, more same-sex couples marrying and cohabiting, and a large rise in remarriage, so more married couples arrive carrying a prior marital history.

2:30 Inequality, opened with the cost of a full-service 2023 wedding, itemized down to the videographer and the stylist, arriving at about $47,000 for 117 guests.

3:10 The connection he draws from it.

“Weddings are expensive, and the people who get married are increasingly richer. Couples are increasingly pairing up by education level, both high and low, but more often high. Marriage is now more common for people with college degrees, and marriage is less likely among people who are poor.”

3:44 A second form of inequality: the lack of social recognition for same-sex couples.

4:07 Social change, framed as the growth of modern individualism.

“The growth of individualism in our culture and the goals of individual self-fulfillment associated with it have weakened the ties of marriage. Fewer people getting married, fewer people staying married than in the old days. Not necessarily more and more divorces every year, but the ties of marriage are weaker than they used to be.”

4:36 The legal side of it: the law increasingly treats people as individuals rather than as families or couples, married or not.

5:26 Three key trends: falling marriage rates, sharper decline among African Americans, and marriage concentrating among the richer and better educated.

6:19 The 1970 to 2021 comparison, by age and sex. Among men aged 20 to 24, about 45 percent had ever married in 1970 against about 8 percent now. Among women, about 64 percent against about 12 percent.

7:02 The correction at the other end of the age range, which is the analytic point.

“If you go all the way out to age 45 to 54, the decline has been much more modest. It was 93 percent of men ever married by the time they were about 50, now 83 percent. Among women it was 94 percent, now about 86 percent. So it’s not so much that people are not getting married ever, but they’re getting married later than they used to, and somewhat less altogether, but definitely much less in their 20s and 30s.”

7:38 The annual rate: close to 100 marriages per thousand unmarried people in 1950, down to about 30 now.

8:16 The figure that does the most work in the lecture.

“In 1960 the average American could expect to be married for 29 of the years between age 18 and 55. Now it’s only 18 of those years. So a big drop in the amount of time people spend married in that core adulthood age range.”

9:00 Why marriage declined, beginning with the cultural explanation and then turning it over.

“Is it just individualism and self-centeredness, people don’t want to be married anymore? Maybe they don’t care as much about marriage. But on the other hand, maybe they care more about marriage. Maybe they idealize marriage more than they used to, so they are less eager to get into it because they have put it on a pedestal. They don’t think their life or their relationship is worthy of a marriage. Maybe they delay marriage not because they don’t want to, but because they want a better marriage.”

9:53 The version of it that applies further down the income distribution.

“Among poorer people, working-class people, the long-term commitment may be postponed by a sense of insecurity or uncertainty. They’re not sure what their income or their living situation or their job or their education is going to be in the future. They’re not sure what their partner’s is going to be. And so they’re reluctant to make a long-term commitment that would lock them into something they’re not sure about.”

10:26 Economic independence, especially women’s, reducing the perceived need for marriage, and the corresponding change for men who can no longer assume a wife devoted to the home.

11:09 The puzzle that follows, and his answer to it.

“If it was just about independence, then you might think rich people wouldn’t get married. So why are rich people getting married more than poorer people? Well, they don’t need to get married as much, but more people want to marry them. They’re more sought after in what we’re going to discuss as the marriage market.”

11:51 The calculation for people without resources, delivered with an unusual amount of self-awareness about how it sounds.

“Marriage for them may seem like a commitment that brings more burden than benefit. If you’re relatively poor, marrying somebody who’s relatively poor, what do you gain from it? That’s a perception anyway. Most poor people want to get married, but they’re either not persuaded by the prospects they have, or they can’t find someone who they can persuade to marry them. It’s a harsh way of putting it.”

12:51 A demographic case: the baby boom producing a shortage of slightly older men for women coming of marriage age about twenty-five years later, which squeezed marriage rates for reasons that had nothing to do with attitudes.

13:46 First marriage rates across the five major groups, with about 49 per thousand for never-married Asian women at the top and Black Americans at the bottom, and the observation that women’s rates exceed men’s in every group except Black Americans.

14:46 The marriage squeeze explained through deindustrialization.

“The loss of manufacturing jobs meant that men without college degrees had a harder and harder time finding steady work. One of the things that makes people willing to take the plunge in marriage is feeling secure that they or their partner has a good financial future. Losing all those manufacturing jobs undermined that, especially in urban Black communities. So for those poor urban women, their chances of finding and marrying men with decent jobs were reduced.”

16:02 Incarceration compounding it, with the scarring point.

“Even if they’re only incarcerated for a few years in their early adulthood, that affects them negatively, and affects their chances of getting married and getting a good job ever.”

16:40 The trend line, and what it rules out.

“In 1950, Black and white Americans had very similar rates of ever being married, and the gap really opened up after that. There’s that period in the 70s and 80s when the Black marriage rate was falling much faster than the white marriage rate. There’s now a very big gap that really was not present in the 1950s, sort of in the heyday of our manufacturing industrial economy.”

17:15 His own research with Joanna Pepin, now at Toronto, plotted as a scatter of metropolitan areas.

17:35 How to read it.

“Each dot represents one metropolitan area for either Black or white women. The vertical line shows that in areas on the right there are more unmarried men than unmarried women, ages 20 to 45. Most white young adults live in metropolitan areas with more unmarried men than women, and the opposite holds for most Black women. The clouds drift up and to the right, which means that in places where there are more unmarried men, women are more likely to get married.”

18:41 Why people still marry, starting with what they say. Pew in 2019: love first, then companionship and the desire to make a commitment, then children, with financial sense, convenience, and pregnancy well down the list.

19:24 Then the structural reading of the same question: state and federal rights and benefits, employer incentives like health insurance, and informal pressure from parents and others.

20:22 The last route, offered tentatively.

“Especially in a world with so many legitimate or acceptable choices, people sometimes do make major decisions just out of imitation. The obvious example is a celebrity endorsement for a product. You’re a very good football player, so maybe I’ll trust your judgment on what kind of car to buy. That ends up also happening with major life choices. People don’t really literally think that way, but sociologically we can track these traces through the culture.”

21:23 Marriage markets, illustrated with the Obamas and the Trumps.

21:47 What makes it a market, in the economists’ terms he attributes to Gary Becker: it must be voluntary, and there must be competition.

22:50 The two couples read against the model. The Trumps as an exchange across dimensions. And the Obamas as the opposite case.

“In the case of the Obamas it’s a very different situation, much more equal in the sense that they were both already professional, Ivy League educated, employed lawyers in the same city. So their marriage market, if you will, was much smaller, in the sense that if they wanted to marry another successful Black lawyer in Chicago, there just weren’t that many to choose from.”

23:50 The criticisms of the market frame, which he gives at length. Desires conflict rather than reducing to one currency.

24:15 And the harder objection.

“Maybe more fundamentally, if you look at all the bad decisions people make, it really is hard to describe their behavior as something like rational, the way you might think of it in terms of car shopping, if indeed that is rational either. It’s not reasonable to assume that spouses reflect a conscious choice from many alternatives, because they have fewer choices than you might think.”

25:05 His verdict on the frame, which is the model of how he handles a theory he half accepts.

“In some ways this marriage market framing doesn’t work for everything, but we teach it because it helps understand certain kinds of decisions that people make, and it may be one of several factors in play.”

25:19 Endogamy by race and ethnicity: 2 percent of new marriages crossing major group lines in 1960, about 20 percent now, a tenfold increase.

26:25 The qualification he attaches to that.

“That rate of 20 percent is still quite low compared to the potential for intermarriage that we have as our society has become more and more diverse. If people married without regard to race and ethnicity altogether, we would see a number quite a bit higher than 20 percent. So there’s still selection into marriage going on within these major groups.”

26:49 His own tabulation from people who reported marrying in the previous year, 2021, read along the diagonal: 86 percent of white men married a white non-Hispanic spouse, 69 percent of Black men a Black spouse, 28 percent of American Indian men an American Indian spouse, 68 percent among Asian and Pacific Islanders, 71 percent among Hispanics.

27:49 The size explanation for the outlier.

“You would expect that smaller groups would have higher rates of intermarriage. The American Indian population is the smallest group here. If they walk out the door and marry somebody, they’re less likely to marry somebody from their same group.”

28:14 The sex comparison across the table, ending on the case that reverses.

“Black women have higher rates of marrying within their own racial ethnic group than men do, and they’re the only group on the chart that has that. So that partly explains the lower marriage rates among Black women, that they’re less likely to marry somebody outside their group. And there’s racism and all kinds of other factors that contribute to that.”

29:35 Educational endogamy, with the social observation attached.

“Choosing a spouse of the same education level is not only common but expected at this point, especially for people with higher levels of education. People with a college degree who marry somebody without a college degree raise some eyebrows. It’s not normal or commonly expected as much as it used to be.”

29:59 The figure: in 2022, about 77 percent of married couples were either both college graduates or both not, three quarters of American couples on the same side of the college divide, and higher than it was historically.

30:25 Why that follows from the ideal.

“This is consistent with the modern ideal of marriage as a companionship of equals. The idea that people have compatible tastes, and in the pursuit of individual goals that they will have somebody they have a lot in common with. It’s not surprising that education is one of the key factors.”

30:52 Same-sex couples as the exception to like marrying like, with the measurement problem he has raised elsewhere.

“Most states do not even report data on the sex or gender of people getting married. They may list two people, but they don’t tell you their genders, or report statistics on how many couples were different gender versus same gender, including big states like California and New York. So we don’t have official statistics, we have to go by surveys.”

31:45 The Census estimate: under one percent of married couples, fewer than a million same-sex married couple households, and a knowledge base still being built.

This is the chapter where his own research is the syllabus, and the interesting thing is how much of the lecture is spent undermining frames he needs.

Watch the individualism section from 9:00. He states the standard cultural explanation, that people are more self-centered and care less about marriage, and then inverts it in the same breath: maybe they care more, maybe they have put it on a pedestal so high that their own relationship does not qualify. That is the Edin and Kefalas finding delivered without the citation, and it does something politically unusual. It hands the decline-narrative side its strongest evidence, that Americans revere marriage more than ever, and turns it into support for the structural account. People who idealize marriage and lack the security to enter it will postpone it.

Then he does the same thing to the marriage market. He teaches Becker, applies it to two famous couples, and then spends ninety seconds at 23:50 taking it apart: desires conflict, people make bad decisions, nobody has the range of alternatives the model assumes, and shopping for a spouse is not shopping for a car, “if indeed that is rational either.” And then the verdict at 25:05, that they teach it because it explains some decisions and may be one factor among several. That is how a professional uses a model he does not believe, and it is the intellectual habit that most distinguishes this lecture from a partisan one.

The strongest empirical passage is 14:46 through 17:15. The Black and white marriage rates were close in 1950 and diverged afterward, which rules out any direct inheritance from slavery and points at the postwar collapse of urban manufacturing plus mass incarceration. Then the Pepin scatter arrives as the individual-level route: most white women live in metropolitan areas with a surplus of unmarried men, most Black women in areas with a deficit, and marriage rates track the ratio. That is the same argument he makes against the success sequence, delivered to freshmen as a chart rather than as an accusation, and it is far more persuasive in this form.

The number I would steal is at 8:16. Twenty-nine of the years between eighteen and fifty-five spent married in 1960, eighteen now. That converts an abstraction about declining marriage into a quantity a general reader can hold, and it does something the marriage rate cannot: it captures later entry, lower entry, and divorce in one figure. He gives it fifteen seconds.

Three weaknesses.

The Trump paragraph at 22:50 is beneath the lecture. He describes a rich man who may or may not be the most attractive man in the world paired with a poorer woman who is above average attractive. That is a real illustration of Becker’s exchange model and it is also a professor making a joke about the appearance of two named living people in a course recording. The Obama half of the comparison is analytically much better and needs no dig.

The imitation passage at 20:22 is the weakest reasoning in the lecture. Celebrity endorsement of a car is a documented advertising phenomenon; celebrity marriage prompting ordinary marriage is not something he cites, and he says outright that people do not think that way, and then keeps it anyway. It is a mechanism proposed and disclaimed in the same breath.

And the whole lecture takes for granted that cohabitation is a stage in a sequence, without asking whether cohabitation is doing what marriage used to do for the people who never reach it. He tells students the chapter’s organizing question at 0:56 is how cohabitation changed the meaning of marriage, and then spends thirty-two minutes on marriage. Presumably the second half of the chapter answers it.

Two items for your file. At 30:52 the same-sex marriage measurement complaint appears for the third time across these recordings, in 2020 to Pepin and again here in 2025, unchanged: states record two spouses and not their sexes, so the administrative record cannot answer the question. Five years and it has not moved, which is itself a finding about how administrative data changes.

And the educational sorting figure at 29:59, three quarters of couples on the same side of the college line, is the marriage half of the argument he made in the chapter seven lecture about utilitarian love reinforcing endogamy. He builds the same case twice, two weeks apart, from different starting points, and does not tell students they are the same case.

Transcript flags. At 0:13 “Christian Styer and Sandra Perry” is Kristin Perry and Sandra Stier, and the case is Hollingsworth v. Perry; the names are transposed in the transcript. At 17:19 “Joanna pepen” is Joanna Pepin. At 22:38 Gary Becker (1930–2014) is correct. At 26:13 “by 1921” is 2021. At 28:59 “133% of them marry a white man” is 13 percent. At 1:33 “cooptation” is cohabitation. At 3:31 “marriag” is marriage. At 25:58 “meic and Venezuelan” is Mexican and Venezuelan. At 30:03 “34s or 77%” is three quarters. At 30:57 “most people marry somebody of the same gender” is a slip and should be a different gender; the sentence as transcribed reverses his meaning and is unusable. At 16:33 “reduce the shortage” should be produce or create the shortage. The $47,000 wedding figure and the 117 guests come from an industry survey reproduced in his textbook and should be sourced to the book rather than stated flatly.