THE DATA FOR OUR ANALYSES

We use data from a 2009 survey of members of the American Political Science Association (APSA) that was proposed by the APSA Committee on the Status of Women in the Profession, supported by the association president, and enjoyed substantial participation by members of other “status” committees. Following on a systematic survey design and implementation process described in detail in Hesli and Lee (Reference Hesli and Lee2011, 405–407), 1,399 respondents (i.e., 27% of the original sample) completed the survey. Hesli and Lee (Reference Hesli and Lee2011, 405–406) also reported marginals for survey respondents versus APSA members by faculty rank, gender, and type of department, which suggest good representativeness in the survey sample.

The survey included 453 respondents that we verified were employed in PhD-granting departments and who are the scholars analyzed in this study (Hill and Hurley Reference Hill and Hurley2021). Previous research by Hesli and Lee (Reference Hesli and Lee2011; Reference Hesli and Lee2013); Hesli, Lee, and Mitchell (Reference Hesli, Lee and Mitchell2012); and Mitchell and Hesli (Reference Mitchell and Hesli2013) used the full sample of respondents to explore other research questions.

It is possible that some female faculty left PhD departments because of sexist climates before this survey was conducted; therefore, climate perceptions and their effects on productivity could be underreported in our data. However, research (discussed in more detail in the online appendix) indicates that the attrition of women assistant professors is no higher than that for men in PhD social science departments (e.g., Box-Steffensmeier et al. Reference Box-Steffensmeier, Cunha, Verbanov, Hoh, Knisley and Holmes2015). The Box-Steffensmeier et al. data, however, do not track where faculty members who leave PhD departments go. Yet, Hill’s (Reference Hill2021) explication of the career paths of political scientists who entered assistant professor positions in 1988 and 1989 offers two types of evidence on this general concern. Hill demonstrates that more men experienced attrition of any type from these departments (i.e., 26% of men versus 15% of women) and that the majority of moves from PhD departments are to other PhD departments.

This evidence does not mean that cross-sectional datasets like ours are free of survival-bias problems. However, as Box-Steffensmeier et al. (Reference Box-Steffensmeier, Cunha, Verbanov, Hoh, Knisley and Holmes2015) also concluded, we interpret it to mean that such problems are less severe than many fear and that they do not diminish the representation of women more than that of men.

DIMENSIONS OF PERCEIVED DEPARTMENTAL CLIMATE

It is widely agreed that the interpersonal climate in academic departments consists of multiple dimensions (APSA 2005, 6–12; Lee Reference Lee2007). The survey data that we analyze support an especially rich analysis of perceived departmental climate dimensions and how men’s and women’s perceptions differ about them. This survey posed nine semantic differential scales about respondents’ perceptions of their home department. Factor analysis of the set of scales for the PhD faculty uncovered a first dimension (which accounted for 56 of the total variances across the scales) that includes items for how disrespectful–respectful, contentious–collegial, competitive–cooperative, individualistic–collaborative, and hostile–friendly departments were perceived. Hesli and Lee (Reference Hesli and Lee2011, 396) uncovered the same dimension in the full sample, labeled the underlying dimension “collegiality,” included a factor score measure for it in their multivariate models, and found that research productivity was higher in less collegial departments.

Factor analysis also reveals an independent second dimension that includes the items for how sexist–nonsexist, diverse–homogenous, racist–non-racist, and homophobic–non-homophobic departments were perceived (which accounted for 13% of total variances with the PhD faculty).Footnote ¹ Both dimensions could be important for how men and women view and react to the culture of their department, yet the second dimension taps perhaps the most critical concern about climate for women: how sexist or non-sexist it is perceived to be. This second dimension and its components have not been analyzed for how men and women perceive them or how they might affect women’s scholarly productivity.

Two bivariate cross-tabular analyses using some of the data described previously illustrate notable ways that men’s and women’s perceptions of departmental climate converge and diverge. Table 1 presents gender differences in perceptions of how collegial–contentious respondents’ departments are—the “defining element” of the first factor analysis dimension (Hesli and Lee Reference Hesli and Lee2011, 396). Table 2 presents gender-specific perceptions of how non-sexist–sexist departments are perceived to be by male and female respondents.

Table 1

Perceptions of Contentiousness in Respondents’ Home Departmental Climate ^(a)

Notes: ^(a)Cell entries are the count and the column percentage. Kendall’s tau-c=0.05, n.s.

Table 2

Perceptions of Sexism in Respondents’ Home Departmental Climate ^(a)

Notes: ^(a)Cell entries are the count and the column percentage. Kendall’s tau-c=0.35, p<0.05.

Table 1 demonstrates that there are no gender differences in perceptions of the level of contentiousness within departments. At the most extreme ends of the scale, men are slightly more likely to rate their department as uncontentious and women are slightly more likely to rate it as very contentious—but the differences are not statistically significant. In contrast, table 2 demonstrates notable gender differences in perceptions of whether a respondent’s department is sexist. Men are far more likely than women to perceive their department as non-sexist: nearly 52% of men placed it at the least-sexist scale point whereas only 21% of women selected that position. In contrast, 7.5% of women rated their department at the most-sexist end of the scale whereas fewer than 1% of men chose that position. The percentages at the most-sexist end of the scale are small, but the same pattern persists at position 4 of the 5-point scale (where 5 indicates most sexist): 15.8% of women chose this position versus only 5.6% of men. Gender differences in perceptions of sexism are pronounced and statistically significant. Viewed together, the patterns in tables 1 and 2 also indicate that perceptions of sexism are not a byproduct of perceptions of contentiousness.

MORE EVIDENCE ABOUT HOW WOMEN AND MEN PERCEIVE DEPARTMENTAL CLIMATES

Three types of evidence enrich the information in tables 1 and 2. First, women’s evaluations of how respectful, cooperative, collaborative, and friendly (i.e., other components of the first underlying dimension) their department is do not differ significantly from those of men in cross-tabular analyses. Yet, women’s perceptions of how homogenous, racist, and homophobic (i.e., other components of the second dimension) their department is differ identically from men’s perceptions on their views of sexism. Thus, women distinctively perceive what we call the “cultural climate.”

Second, men’s and women’s evaluations of departmental sexism demonstrate notable within-gender similarity across academic ranks, as demonstrated in the cross-tabulations in tables 3 and 4. Men at the rank of professor (in table 3) as well as men who are assistant professors (in table 4) are heavily clustered at the two least-sexist points on the scale (i.e., 88% of senior men versus 77% of junior men). Women at the senior rank do not perceive high sexism as commonly as assistant professor women, yet they both do so far more than their generational male peers. Of the senior women, 19% perceive sexism at one of the top two levels compared to only 3% of similarly ranked senior men. Further, 24% of assistant professor women have this perception compared to only 7% of men of the same rank.

Table 3

Perceptions of Sexism in Respondents’ Home Departmental Climate—Faculty at the Rank of Professor ^(a)

Notes: ^(a)Cell entries are the count and the column percentage. Kendall’s tau-c=0.29, p<0.05.

Table 4

Perceptions of Sexism in Respondents’ Home Departmental Climate—Faculty at the Assistant Professor Rank ^(a)

Notes: ^(a)Cell entries are the count and the column percentage. Kendall’s tau-c=0.37, p<0.05.

Third, we could question whether some women’s views of sexism are distinctive perhaps because of distinctive experiences in their department, but other evidence suggests that this is rarely the case. We received responses from women in 70 departments; for 33 of those departments, we had two or three female respondents. In seven of those 33 departments, women respondents agreed on the same scale score of sexism; in 14 departments, women respondents chose adjacent scale score ratings. Thus, in two thirds of the departments with multiple women respondents, their climate assessments were very similar.

THE EFFECT OF PERCEIVED CLIMATES ON RESEARCH PRODUCTIVITY GENERALLY AND ON WOMEN’S PRODUCTIVITY SPECIFICALLY

A major concern about this survey for multivariate analyses is the substantial amount of missing data, some of which arises because of the common behavior of respondents skipping questions. Also, many respondents evidently became “fatigued” and stopped answering the questions near the end of the survey, most of which were not germane to our interests. There also were odd cases of miscoding in some variables.

Hesli and Lee (Reference Hesli and Lee2011) recognized the general missing-data problem and adopted a two-part hypothesis-testing strategy. They analyzed their large set of predictor variables with the conventional use of list-wise case deletion, losing 54% of the cases in their most comprehensive analysis. Then they replicated the preceding analyses with a missing-data replacement strategy using Amelia. The results of these two estimation strategies were almost identical—implying that even the limited set of cases with no missing values represented well the structure in the data.

Our estimation strategy was to rely as much as possible on true-score data on the respondents and on a parsimonious set of rival hypotheses for which there is notable evidence in prior research—both of which minimized the loss of cases in our analyses. Thus, we first used information in the original dataset to replace as much missing data as possible (see the online appendix for more details). Most notably, 19% of the values for gender for respondents in PhD departments was missing because this variable was near the end of the survey, where many respondents had dropped out. We reduced that figure to 2% by using identifying information on individual respondents in the original data.

Hesli and Lee (Reference Hesli and Lee2011) also found a significant negative relationship between the year that respondents earned their PhD and the measures of research productivity. However, a striking pattern of numerous missing values on the former variable suggests that many respondents did not navigate correctly a drop-down year-counter variable in the survey—a problem with such survey items that is documented in online marketing research (e.g., Baymard Institute 2021). Thus, we did not include this variable in our analyses. However, the variable for academic rank was correlated with the year of degree for the valid cases at r=-0.70 and likely captures the same general concept of time in the profession. We also reduced missing values or replaced miscoded values in other variables (explained in the online appendix).

Second, systematic studies of research productivity of political scientists in diverse institutions (Djupe et al. Reference Djupe, Hill, Smith and Sokhey2020; Hesli and Lee Reference Hesli and Lee2011) demonstrated support for only a small number of predictor variables in our data: faculty rank, gender, teaching load, and frequency of conference attendance. The latter variable, however, was never a significant predictor in our more homogenous subset of faculty in PhD programs; therefore, we omitted it from our analyses. Djupe et al. (Reference Djupe, Hill, Smith and Sokhey2020)—and Hesli and Lee (Reference Hesli and Lee2011) only in their imputed data analyses—also found that minority scholars have fewer publications. However, exceptional missing data for this variable that we cannot reduce with true-score estimates precludes us from using it. Hesli and Lee (Reference Hesli and Lee2011) also demonstrated that more research resources in a department and a climate perceived as highly contentious were associated with higher productivity—and we included those variables in our analyses. Our unique research question required us to add the measure of perceived departmental sexism as a predictor.

Table 5 presents four OLS regression models that demonstrate the potency of our set of direct predictors, as well as interaction tests for the influence of perceived sexism on the productivity of women. The two dependent variables are those used by Hesli and Lee (Reference Hesli and Lee2011): a logged measure of “total publications” that sums self-reported numbers of published refereed articles, books, book chapters, and edited books; and a logged measure of self-reported refereed journal articles. Columns 1 and 3 in table 5 present the “restricted models” for each dependent variable that exclude the interaction term.

Table 5

The Effects of Perceived Climate Contentiousness and Sexism on Research Productivity

Note: *p<0.05 in a one-tailed test.

In both restricted models, the measures of faculty rank, gender, count of overall research resources provided to respondents by their department, and their teaching load are all significant and signed as in the results for the entire sample reported by Hesli and Lee (Reference Hesli and Lee2011). Neither climate variable, however, has a significant direct effect on either of these two models, which is contrary to the negative effect of a collegial climate reported by Hesli and Lee (Reference Hesli and Lee2011).

In our sample of PhD-program faculty, neither of these discrete climate variables nor the factor scores for the two climate dimensions has a significant bivariate or multivariate relationship with either dependent variable (even with more or fewer predictors in the regression models). We conclude that Hesli and Lee’s (Reference Hesli and Lee2011) negative finding for the collegial climate variable is likely a product of greater heterogeneity in the full sample, in which 57% of the cases were from non-PhD departments and perhaps because departmental climates in those institutions are notably different from those in PhD departments.

Third, we observed that in the two unrestricted models in columns 2 and 4 of table 5 neither of the F tests for the presence of an interaction is significant—and neither of the interactions between the female dummy variable and the degree of perceived departmental sexism is significant. Thus, the research productivity of women is not generally reduced by working in a department that they perceive as relatively highly sexist.Footnote ²