Democrat-to-Republican ratios among political scientists have been estimated at 5.6:1 for reported voting records (Klein and Stern Reference Klein and Stern2005, 264) and at 10:1 for party affiliation (Gross and Simmons Reference Gross, Simmons, Gross and Simmons2014, 30). This imbalance might produce negative consequences for political science, but practices can be adopted to reduce these negative consequences.

Inadequate political balance in a discipline can limit the range of research questions (Maranto, Redding, and Hess Reference Maranto, Redding, Hess, Maranto, Redding and Hess2009, 5) and embed into methods assumptions of the majority group (Duarte et al. Reference Duarte, Crawford, Stern, Haidt, Jussim and Tetlock2015, 4ff). Consider the Nteta, Schaffner, and MacWilliams (Reference Nteta, Brian Schaffner and MacWilliams2017) post on attitudes about the right of professional athletes to kneel during the national anthem to protest the treatment of African Americans. The post reported that whites’ strong opposition to this right associated with more negative stereotyping of African Americans relative to whites, controlling for factors such as participant patriotism. However, the post did not report that the data also indicated that strong opposition to this right associated with higher levels of participant patriotism (p<0.001) (personal communication with Brian Schaffner). This omission produced a less-balanced understanding than if the post had reported results for participant patriotism.

Or consider research on gender attitudes and support for Donald Trump in Bracic, Israel-Trummel, and Shortle (Reference Bracic, Israel-Trummel and Shortle2018); Cassese and Barnes (Reference Cassese and Barnes2018); Cassese and Holman (2019); Frasure-Yokley (Reference Frasure-Yokley and Lorrie2018); Knuckey (Reference Knuckey2018); Schaffner, MacWilliams, and Nteta (Reference Schaffner, MacWilliams and Nteta2018); Setzler and Yanus (Reference Setzler and Yanus2018); and Valentino, Wayne, and Oceno (Reference Valentino, Wayne and Oceno2018). Each article had “sexism” in the title or abstract, but the gender-attitudes items did not measure attitudes disfavoring men. For example, “to capture gender attitudes,” Bracic, Israel-Trummel, and Shortle (2018, 9) measured responses only to statements that “Most men are better suited emotionally for politics than are most women” and “A man should be in control of his wife”; these items permitted responses ranging from disfavoring women (“agree” responses) to not disfavoring women (“disagree” responses) but did not permit responses disfavoring men, which could have been captured by, for instance, measuring perceptions of whether—compared to men—women are better, worse, or equally suited for politics (Taylor et al. Reference Taylor, Morin, Cohn, Clark and Wang2008, 14).

A similar asymmetry occurred in the questionnaires for the 2016 American National Election Studies Time Series Study (ANES 2017), which by my count in the supplementary materials contained 10 gender-attitudes items about only women and one gender-attitudes item about only men. This item about only men paralleled an item about only women: “How much discrimination is there in the United States today against each of the following groups?” Results for these items are reported in figure 1. If “working from the perspective that denial of discrimination is a manifestation of modern sexism” (Cassese and Barnes Reference Cassese and Barnes2018, 14), then this denial-of-discrimination manifestation of modern sexism appeared among Donald Trump voters more frequently directed at men than at women and appeared more frequently among Hillary Clinton voters directed at men than among Trump voters directed at women. About half of Clinton voters denied that there is discrimination among men in the United States today, suggesting that many of these voters harbored anti-men attitudes.

Figure 1

Ratings of No Discrimination against Women and Men

Note: The figure reports weighted percentages and 95% confidence intervals for Hillary Clinton voters and Donald Trump voters who selected “None at all” for targets of women or men when asked: “How much discrimination is there in the United States today against each of the following groups?”; other response options were “A great deal,” “A lot,” “A moderate amount,” and “A little.” Analyses were conducted in Stata 15 (StataCorp 2017). The figure was produced in R (R Core Team 2018) with ggplot2 (Wickham Reference Wickham2016). Data: American National Election Studies, University of Michigan, and Stanford University (2017).

Even if these imbalances in gender-attitudes items were due to the reasonable perception that bias against women is or has been larger in politics or in society than bias against men, the dearth of gender-attitudes items about men nonetheless undercuts our understanding about gender attitudes and the 2016 US presidential election, because, even though the previously mentioned articles provided evidence that Trump supporters, on average, reported less positive or sympathetic responses about women than Clinton supporters did, without parallel measures of attitudes about men, the articles cannot support inferences about whether Clinton was on net disadvantaged because of gender attitudes among the electorate.

Researchers interested in how patriotism associates with anthem-protest attitudes or how attitudes about men associate with 2016 US presidential-candidate preference could produce publications that parallel these articles. However, such balancing is implausible as an effective mechanism for countering politically asymmetric research, given the political imbalance in political science membership. Nonetheless, more balanced research might be produced through a stronger peer-review process, which, in the previous examples, suggested or required (1) reporting of results for participant patriotism, or (2) a battery on the 2016 ANES that measured attitudes about men.

Inadequate political balance in an academic discipline might undercut the vetting of research findings (Lukianoff and Haidt Reference Lukianoff and Haidt2018, 109; Tetlock Reference Tetlock1994, 510). The political imbalance in political science produces a nontrivial chance that no peers on the political right would be invited to review a manuscript with an implication favoring the political left, which plausibly reduces the rigor of the peer-review process, given evidence that “liberals were better at identifying flawed arguments supporting conservative beliefs and conservatives were better at identifying flawed arguments supporting liberal beliefs” (Gampa et al. Reference Gampa, Wojcik, Motyl, Nosek and Ditto2019, 29) and meta-analysis evidence that “both liberals and conservatives show a consistent tendency to be less skeptical consumers of information that supports rather than challenges their political beliefs” (Ditto et al. Reference Ditto, Liu, Clark, Wojcik, Chen, Grady, Celniker and Zinger2019). A plausible method to strengthen peer review is for journals to open manuscript review to volunteers, thereby increasing the number of reviews from skeptics of a manuscript’s inferences.

Commenting systems such as Hypothesis and PubPeer are available, respectively, for Open Science Framework preprints (Staines Reference Staines2018) and post-publication review; however, other than to correct published articles, there does not appear to be a mechanism to require comments in these systems to be addressed. Nonetheless, journals could post online conditionally accepted manuscripts, permit public comments on these manuscripts similar to public comments for the ANES Online Commons, and then require that a manuscript address particular conditional-acceptance comments as a condition of publication (see Meta-Psychology for a peer-review process that incorporates public comments, available at https://web.archive.org/web/20180315124052/https://open.lnu.se/index.php/metapsychology/about).

Conditional-acceptance comments would not be needed for reject-or-accept recommendations but instead could identify potential manuscript flaws that nonpublic peer review might have missed. Consider Mitchell and Martin (2018, 648), which reported a comparison of official university course evaluations and Rate My Professors comments for one woman instructor to corresponding evaluations and comments for one man instructor, in which “Findings suggest that the relationship between gender and teaching evaluations may indicate that the use of evaluations in employment decisions is discriminatory against women.” Conditional-acceptance comments about Mitchell and Martin (Reference Mitchell and Martin2018) could have (1) noted that, in table 2, the use of one asterisk to indicate a p-value under 0.1 for an N=54 difference between 0% and 6.6% is apparently inconsistent with the use of two asterisks to indicate a p-value under 0.05 for an N=54 difference between 0% and 5.5%; (2) requested that the manuscript indicate whether the table 1 N=68 for the official university course evaluations and the table 2 N=54 for the Rate My Professors comments are per-instructor sample sizes (in the unlikely event that the two instructors had identical numbers of comments in each table) or are pooled-across-instructor sample sizes and, if the latter, requested the instructor-level Ns that could have produced the p-values indicated by the asterisks; and (3) inquired whether table 4 inferential statistics were from analyses that treated multiple ratings from a student in a category as independent ratings, as suggested by reported sample sizes (see the supplementary materials). A Mitchell and Martin (Reference Mitchell and Martin2018) addressing these comments might be less likely to be accepted as evidence that “student course evaluations...have been shown to be biased against female instructors” (Sen Reference Sen2018, 336).

In addition to opening peer review to volunteers, journal editors can reduce the potential for political imbalance in research reporting by adopting policies to further reduce reporting flexibility, given that reported estimates and inferences can depend on how data are analyzed and on whether data are omitted from reported analyses (Simmons, Nelson, and Simonsohn Reference Simmons, Nelson and Simonsohn2011; Zigerell Reference Zigerell2017). Journal policies to further reduce reporting flexibility can, for instance, require that articles reporting on nonpublic surveys include full questionnaires in the article or appendices, with reasonable exceptions.

Political imbalance in political science also might bias which perceived problems are addressed in the discipline. For example, in 2018, the American Political Science Association (APSA) had status committees on Asian Pacific Americans; blacks; community colleges; contingent faculty; first-generation scholars; graduate students; Latinos and Latinas; women; and lesbians, gays, bisexuals, and transgender individuals (APSA 2018c). The charge for APSA’s Committee on the Status of Women in the Profession discussed discrimination and underrepresentation of women (APSA 2018a), and these concerns also could also explain establishment of other status committees. However, APSA has no status committee on political minorities, despite evidence that some academics are willing to discriminate based on ideology (Honeycutt and Freberg Reference Honeycutt and Freberg2017) and despite the previously mentioned evidence of Republican underrepresentation in political science relative to the US population.

Political imbalance in political science might influence the choice of policies to address perceived problems. General Social Survey data (Smith et al. Reference Smith, Davern, Freese, Hout, Smith, Davern, Freese and Morgan2019) indicated that, in 2018 and between 2000 and 2018 inclusive, strong Democrats and extreme liberals were more likely than strong Republicans and extreme conservatives to favor preferential hiring and promotion of women, with and without controls for sex, race, age, and education. Similar preferential treatment occurs in political science, with women explicitly favored in the provision of publicity by the Women Also Know Stuff project and in the provision of educational opportunities at Visions in Methodology conferences and Journeys in World Politics workshops (see the supplementary materials). To the extent that such discrimination is in tension with principles from the American Association of University Professors (2009) “Statement on Professional Ethics” endorsed in APSA’s “A Guide to Professional Ethics in Political Science,” that professors “avoid any exploitation, harassment, or discriminatory treatment of students” and “do not discriminate against or harass colleagues” (APSA 2012, 5; emphasis added), the APSA Committee on Professional Ethics, Rights, and Freedoms should issue an advisory opinion regarding ethical limits on protected-class discrimination in the discipline. Given the importance of peer evaluations to professional advancement in political science, the committee also could indicate whether political scientists involved in open protected-class discrimination against students or colleagues should recuse themselves from nonblinded decision making in domains such as hiring—even if only to avoid risking the appearance of bias in such decisions.

Moreover, APSA can provide nondiscriminatory alternatives. For example, as an alternative to the exclusionary lists of experts maintained by Women Also Know Stuff and People of Color Also Know Stuff, APSA could maintain a list of experts open to everyone in the discipline that permits filtering by expert characteristics such as gender identity, race, and volunteered ideology, which could permit selection of a diverse set of experts and permit more intersectionality in searches than lists that exclude experts because of their gender identity or color.

The optimal policy for political science might not be nondiscrimination but instead be modeled on US equal-protection jurisprudence to help cabin discrimination to discrimination that, for instance, is narrowly tailored to further a compelling discipline interest, such as remediating effects of unfair discrimination in the discipline’s past or present, producing a diverse set of experts for an event, or reducing political asymmetry in questionnaires for publicly funded surveys. In these cases, for discrimination justified on empirical claims, discipline discriminators could announce the criteria that could be met to stop or reverse their discrimination. For example, APSA webpages list Asian Pacific Americans as one of the “underrepresented racial or ethnic minority groups” for which Ralph Bunche Summer Institute Program eligibility criteria are relaxed (APSA 2019) and as possessing one of the “underrepresented backgrounds” permitting eligibility for the Minority Student Recruitment Program (APSA 2018b). However, it is unclear how underrepresentation is determined, given that Mealy (2018, 2) reported that, in US-based APSA data, “East Asian or Asian American membership is slightly higher in percentage than the US Asian American population.”

Political science can be improved if its members and leaders adopt practices that help avoid negative consequences of political asymmetry in its membership and leadership. It is hoped that this includes the practices listed here and other practices that can be agreed upon by political scientists of good faith.