Philip N. Cohen’s career can look unusually scattered if you read it by subject. He has written about women’s suffrage, racial inequality, housework, occupational segregation, cohabitation, marriage, divorce, same-sex parenting, genetics and race, fertility, COVID-19, peer review, preprints, open science, public sociology, and the politics of falling birth rates. His own bibliography now divides naturally into two large bodies of work, one on families and inequality and another on scholarly communication and science. But after reading the corpus chronologically, I think that division misses the strongest continuity.
Cohen has spent thirty years doing a sociology of the denominator.
By denominator I mean the rule that decides who has been counted and what has been grouped with what. Cohen’s characteristic move is to encounter an apparently stable social fact and ask what the rule was. Change the universe, the reference period, or the boundary of the category, and the fact may change. Two related habits grow out of that one, and I take them up later: a suspicion of explanations pitched at the level of the individual, and a suspicion of trends converted into directions of history.
Cohen’s first journal article, the 1996 “Nationalism and Suffrage: Gender Struggle in Nation-Building America,” suspects a category that looks more universal than it is. “Women’s suffrage” sounds like the political inclusion of women. Cohen instead reconstructs a movement in which leading white suffragists increasingly made their claim through white citizenship, national strength, racial hierarchy, and an alliance with white men. The category “women” concealed conflicting interests among women themselves. White suffragists could advance their own political status while helping exclude Black women and men from the polity. Cohen ends by insisting that feminist scholars treat those women as agents making purposeful choices.
Three years later, with Suzanne Bianchi, the same intellectual operation becomes statistical. “Marriage, Children, and Women’s Employment: What Do We Know?” begins from an apparently straightforward question: how much do married mothers work? But the Current Population Survey permits different reference periods and different universes. Cohen and Bianchi show that estimates of women’s full-time employment change substantially depending on which is used. Welfare policy was being justified partly by assumptions about how much mothers normally work. If the denominator changes the normal rate, it changes the policy comparison.
The following year, Cohen and Lynne Casper made the same kind of intervention in “How Does POSSLQ Measure Up?” The issue was cohabitation. A government measure that appeared to count unmarried partners was an operational construction whose performance could be tested against alternative ways of identifying couples.
That pattern runs through Cohen’s early inequality research. In “Individuals, Jobs, and Labor Markets: The Devaluation of Women’s Work,” with Matt Huffman, the wage attached to female-dominated work cannot be understood simply by comparing individuals. Jobs are situated within labor markets, and those contexts alter the relationship. In “Racial Wage Inequality: Job Segregation and Devaluation Across U.S. Labor Markets,” segregation and devaluation are analytically separated. In “The Gender Division of Labor: ‘Keeping House’ and Occupational Segregation in the United States,” the usual segregation statistic changes once women doing unpaid housework are restored to the division of labor.
The cumulative lesson is that inequalities attributed to persons are often properties of arrangements around persons. Households, jobs, establishments, occupations, labor markets, and policy regimes keep appearing between the individual and the outcome. Cohen’s 2007 work with Huffman on “Black Underrepresentation in Management across U.S. Labor Markets” asks why establishments operating in areas with larger Black populations are more likely to underrepresent Black workers in management.
Family research remains the largest substantive part of Cohen’s career, but much of it is an argument against treating “the family” as the natural unit at which family outcomes should be explained. Extended households change the employment possibilities of single mothers. National policy changes the relationship between individual bargaining resources and housework. Marriage markets differ by race. Economic shocks affect divorce and fertility. The family is nested inside structures that alter what family means.
Cohen’s second recurring suspicion is directed at trends that are turned too quickly into historical directions.
The titles tell part of the story. “Stalled Progress? Gender Segregation and Wage Inequality Among Managers, 1980–2000.” “Headed Toward Equality? Housework Change in Comparative Perspective.” “The ‘End of Men’ Is Not True.” “The Coming Divorce Decline.” “What’s the Story? Family Demography at the End of Progress.” “Rethinking Marriage Metabolism.”
Cohen encounters a real change and objects to the story attached to it. Women are advancing, therefore men are ending. Housework is becoming more equal, therefore equality is the destination. Marriage is declining, therefore marriage is disappearing. Fertility is falling, therefore population catastrophe is approaching. Cohen distrusts the conversion of movement into destiny.
When Hanna Rosin claimed that “young women” were outearning young men, Cohen unpacked “young women” and showed that the statistic described single, childless women under thirty who lived in metropolitan areas and worked full-time and year-round. Rosin said women had become a majority of managers. Cohen noticed that the underlying category was “managerial and professional,” which combined managers with female-dominated professions such as nursing and elementary education. The category created the headline.
Cohen begins that essay with a sentence that could serve as a manifesto for his career: statistics are not “mere technical details.” They are attempts to represent reality. That is why the boundaries matter.
The anti-teleological position becomes explicit in “What’s the Story? Family Demography at the End of Progress.” Cohen argues against imagining normal modernity as a track along which societies travel while pandemics, climate change, inequality, identity conflict, and political disorder interrupt the journey. Those things may be constitutive features of the world whose demography we are trying to explain. Even the second demographic transition begins to look suspicious if it is treated as another version of history moving in an incontrovertible direction.
This helps explain Cohen’s more recent interest in falling fertility. His 2025 article, “Negative Views of Falling Birth Rates in the United States Come Mostly from the Right Wing,” and his continuing work on population panic fit the older pattern. The subject has changed. The argumentative opponent has not. A demographic trend is being recruited into a large story about social decline. Cohen’s instinct is to separate the measured change from the narrative laid over it.
Return now to the first habit, because it is the one that carries him into a new field.
“Women.” “Managers.” “The Black middle class.” “Cohabitors.” “Married people.” “Young women.” “Science.” “Peer review.” Each can operate as a useful category. Each can also conceal enough heterogeneity to make a claim true only because unlike things have been pooled.
That is what makes the later open-science work a logical extension of his interests.
The 2022 collaborative paper “PReF: Describing Key Preprint Review Features” begins from a classification problem. “Preprint review” sounds like a recognizable practice, but the activities collected under that term differ in who initiates review, how reviewers are selected, whether identities are known, whether competing interests are checked, whether authors can respond, and what outcome the review produces. The solution is to decompose the label into standardized features so different systems can be compared on common dimensions.
That is Cohen’s old move in a new field.
In 1999, “women’s employment” had to be decomposed by universe and reference period before a trend could be interpreted. In 2013, “young women” had to be decomposed before Rosin’s claim could be evaluated. In PReF, “peer review” has to be decomposed before anyone can compare review systems or measure their evolution. The substantive distance between those projects is enormous. The methodological distance is small.
The same continuity appears at the level of causal explanation. Around 2018, Cohen’s bibliography begins to develop a second program alongside family demography: open scholarship, SocArXiv, peer review, scholarly infrastructure, and the production of knowledge. He never stops doing demography. His recent work includes marriage expectations, marriage events, fertility, mortality, and birth rates at the same time that he is publishing about scholarly communication.
The new program again moves the explanation away from individual behavior and toward context. In “The Scholarly Knowledge Ecosystem: Challenges and Opportunities for the Field of Information,” Micah Altman and Cohen define the scholarly ecosystem broadly enough to include stakeholders, markets, policies, organizations, norms, infrastructure, and educational systems. Scientific reliability is an outcome produced inside an institutional ecology.
The analogy with the earlier labor-market work fits. An employee’s outcome cannot be understood without the job, establishment, occupation, and labor market. A scientific paper cannot be understood without journals, repositories, incentives, review systems, publishers, universities, funders, and technical infrastructure. In both cases Cohen resists explanations that stop at the individual because the relevant machinery is operating one or more levels above.
There is a political continuity too.
Cohen’s 1996 suffrage paper is a critical account of white feminism, nationalism, and racial exclusion. His 2007 review essay “Confronting Economic Gender Inequality” is hardly value-neutral. What changes is that Cohen becomes increasingly explicit about what political commitments mean for epistemology.
The pivotal document is probably his 2015 “How Troubling Is Our Inheritance? A Review of Genetics and Race in the Social Sciences.” Cohen argues against Nicholas Wade’s attempt to connect population genetics to racial differences in social behavior. He says that he sets the evidentiary bar high for racial-genetic explanations because the consequences of a false positive are potentially severe. He calls this a political judgment and compares it to the asymmetric treatment of error in criminal justice and medical research.
That passage makes it difficult to describe Cohen’s later open-science work as an attempt to construct a value-free machine for deciding truth. He does not believe such a machine exists. Altman and Cohen’s scholarly-ecosystem work argues for information systems aligned with human values. The point of transparency is to make judgment inspectable.
That position is developed in Cohen’s 2025 book, Citizen Scholar: Public Engagement for Social Scientists. Its chapter sequence moves from description through open scholarship and peer review to social media, activism, and citizenship. The book’s argument is that scholarship and citizenship can be integrated, provided the scholar remains answerable to evidence.
Seen chronologically, there is therefore a critic-to-builder trajectory in Cohen’s career. By the early 2010s he is increasingly visible as an auditor of public and scholarly claims. Rosin’s statistics. Mark Regnerus’s same-sex parenting research. Nicholas Wade’s racial genetics. Alice Goffman’s methods. Claims about poverty, marriage, and family decline. The scholar stands outside the claim and asks whether it survives inspection.
Then Cohen becomes involved in building the infrastructure through which claims circulate. SocArXiv launches in 2016. He works on scholarly communication, preprints, review taxonomies, openness, editorial boards, and research infrastructure. The person who spent years inspecting products coming out of the scientific machine increasingly starts working on the machine.
His own scholarly-communication research states the methodological principle. In “Interventions in Scholarly Communication: Design Lessons from Public Health,” Altman, Cohen, and Jessica Polka argue that scholarly reform suffers from an absence of standard measurement, systematic data collection, and comparable evaluation. They mock the common intervention strategy as “Do it now, check sometime.” Their proposed remedy is to identify common measures and outcomes, build infrastructure for tracking them, incorporate assessment into interventions from the beginning, and create consistent longitudinal evidence.
PReF makes the same point in another form. If heterogeneous review practices are going to be compared across time, first define stable features that distinguish them. Only then measure the trend.
Public argument rewards compression. “ASR went from one-third to two-thirds” is intelligible in a way that a five-category taxonomy of data accessibility is not. A scholar trying to participate in public life must turn a complicated empirical world into categories, numbers, sentences, and stories. Yet Cohen’s scientific work repeatedly demonstrates that the compression is where trouble enters.
This tension sits near the center of Citizen Scholar. Cohen wants scholars to leave the seminar room and enter public argument without surrendering the standards that make scholarship worth hearing. But his own career suggests that the two activities place different pressures on description. The public intellectual needs a story. The empirical sociologist keeps discovering reasons the story is too clean.
His three informal audits of replication materials at the American Sociological Review are where that pressure becomes visible and countable, and I have taken them up separately.
That may be the most interesting trajectory in the entire corpus — a progression from studying how social arrangements produce observable outcomes to studying how knowledge arrangements produce observable facts, followed finally by an attempt to live inside both systems at once as a citizen scholar.
The objects keep getting larger. Women within families. Workers within jobs. Jobs within establishments. Establishments within labor markets. Families within policy regimes. Papers within journals. Scholars within an information ecosystem. But Cohen keeps asking variants of the same question: what structure surrounding the observation made it look the way it does?
And alongside that question is another: what disappeared when we named it?
“Women” can hide racial conflict. “Young women” can hide a highly selected demographic subgroup. “Managers” can become managers plus professionals. “Peer review” can hide radically different institutional processes.
Cohen’s sociology is strongest when he refuses to let the noun settle the empirical question.
That is why denominator skepticism seems a better description of his method. Before accepting a social fact, ask who counts. Ask what the comparison class is. Ask whether the category means the same thing in every observation. Ask whether a measured difference has been converted into a causal story, or a trend into a theory of history.
Those habits connect work that otherwise looks unrelated.
Cohen argues that statistics matter enough that their construction has to be exposed. Categories can mislead because reality exists outside them. A better denominator can produce a better description.
That faith in better description survives into the metascience work. The solution to unreliable scholarly communication is to expose provenance, make evidence inspectable, improve measurement, clarify categories, build institutions that permit criticism, and give outsiders enough information to decide whom to trust.
There is a revealing passage in Cohen’s 2023 chapter “How Do We Tell What’s True?” in which he refuses to put conventional peer review on a pedestal. Expert credentials, institutional reputation, disclosed political commitments, linked evidence, open criticism, and transparent sourcing all provide imperfect signals. Knowledge becomes more trustworthy when claims can be traced and interrogated.
That is also a description of Cohen at his best.
His enduring subject is the distance between a claim and the machinery that produced it.
For thirty years Cohen has been opening that machinery.
Andrew Gelman is the nearest single analogue. Substantive researcher, then methods critic, then a daily blog auditing other people’s published claims, then a participant in reforming what the field accepts.
The three acts match. Substantive researcher, then critic of method, then a public auditing operation, then a hand in what the field will accept. Cohen’s blog has run since 2008 and Gelman’s since 2004, and both use the same instrument: take a published claim, go to the numbers behind it, show the reader the arithmetic, name the paper and the author. Both write for a mixed audience of colleagues and journalists. Both get called a bully by people they have corrected.
Two differences. Gelman reruns. When he takes apart a paper he pulls the data where he can and does the analysis, and his commenters do it too, and the corrections that come out of that blog are corrections about what the numbers show. Cohen inspects. His audits ask whether a package exists and what it contains, and he says he does not run the code. That is the difference between an auditor and an inventory clerk, and it explains why his findings are about availability while Gelman’s are about results.
Gelman also audits his own side. His targets are chosen by whether a claim looks statistically implausible, and the record includes his own papers, his friends’ papers, and work whose conclusions he likes. That indifference is his central asset.
Kieran Healy holds Cohen’s epistemology. His work argues that the measure and the rendering are constitutive of the finding, and his 2017 essay against nuance argues that sociology overvalues theoretical elaboration and undervalues clear description. That last point is Cohen’s descriptive turn stated as a general position about the discipline, published before Cohen said it at Wisconsin, by a sociologist Cohen surely reads.
Healy built no audit. No annual counts, no named lists, no preprint server, no correction demands. He wrote a book about how to draw graphs and taught a generation of sociologists to think about rendering.
So the epistemology does not produce the enforcement. Two men hold roughly the same view of what a measure is and only one of them audits. Whatever explains Cohen’s project, it is not the belief that categories manufacture findings, because Healy holds that belief and does something else with it.
Jeremy Freese occupies the position in the discipline that Cohen might have taken. He is sociology’s designated metascience thinker, coauthor of the field’s review article on replication, a writer on preregistration and research transparency. He works through the literature rather than through counts and lists.
Freese and Peterson cite Nicholas Eubank. Cohen does not. The sociologist doing metascience as scholarship engaged the one journal that ran the experiment; the sociologist doing metascience as enforcement did not.
Otis Dudley Duncan (1921-2004) built the tools. The dissimilarity index he developed with Beverly Duncan became the standard measure of residential segregation, and he brought path analysis into sociology. Then he spent his later career arguing that sociology’s measurement was pre-scientific, that its quantities frequently did not measure what their names claimed, and that the discipline had adopted numerical machinery without asking what the numbers meant. His late book on social measurement is a sustained warning against exactly the pooling Cohen keeps undoing.
Duncan questioned the segregation index he had built and the causal modeling he had introduced. Cohen has done that with PReF, where he decomposed a category to build a measurement tool. He did not do it with the audit categories. Duncan is the standard, and the standard is set inside Cohen’s own lineage: this is a demographer working on segregation, which is Cohen’s own subject, who made auditing his own instruments the last act of his career.
Barry Wimpfheimer’s argument is that a category which arrived later, legal-analytic reading, has governed how earlier material gets read, and that the material looks different once you stop letting the later category set the terms. That is the same move on texts rather than on variables, which tells you the operation is a habit of asking what the naming did.
What the set adds up to is a way of locating Cohen.
The category operation is not his: Healy has it, Duncan had it, Wimpfheimer has it in another material. The three-act trajectory is not his: Gelman ran it a decade earlier in an adjacent field. The metascience is not his alone: Freese has been writing it in sociology throughout.
What is distinctive is the combination plus the enforcement. Cohen alone among these men publishes lists of named papers with a verdict beside each, repeatedly, over years. That is the residue after the comparisons subtract everything shared.
