Aristotle believed that there were two kinds of science, one governing the natural world (physics), and the other the human world (politics). Until relatively recently, most practitioners of Aristotle’s brand of political science were theorists, and most were men. Over the past century, political science has been systematically formalized into a key discipline within the social sciences. Also in this time, women gained the right to formal education, with women now more likely than men to graduate college and increasingly more likely to attend graduate school.

Political science remains a majority-male discipline, but that might change soon. As illustrated below, women now represent about 25% of tenured political science professors, up from about 5% in 1980. And they represent 50% of all grad students, up from about 25%.

Shames1

Source: NSF

Yet a recent trend that worries me is the disproportionate focus on (and status of) highly complex statistical methodology, which is far more likely to be the province of men. For example, as illustrated below, the annual conference on political methodology (PolMeth) has consistently attracted far more men than women. Despite significant gains since the 1980s when there were virtually no women attending PolMeth, the proportion of female attendees at PolMeth has rarely exceeded 25% over the past twenty years. (The data used to create this graph was obtained from Kevin Quinn, the current president of the society for Political Methodology.)

Shames2

On a personal level, I am what might be called “methodologically omnivorous.” If you can show me that multiple types of data support your argument, I am far more likely to believe it. I use statistics regularly in my work. But even though quantitative data are important, they are inherently limited (as are all data, no matter how careful we are in collecting them). They are also often as easy to fudge as are the conclusions drawn from qualitative data (lying with statistics is an age-old art).

Quantitative data, as in large surveys or policy datasets, are extremely useful in answering certain types of questions, particularly about the generalizability of findings and big-picture or over-time changes. But interviews, ethnography, case studies, and other such methods often provide greater depth and particularity. They are wonderfully rich and complex in ways the quantitative data are often not. And according to a recent survey, these methods are more commonly used by women publishing in the field: Papers with only male authors used case studies (a qualitative method) 14.2% of the time, while papers with only female authors used case studies 23.5% of the time.

All methods have limitations, which should be discussed right up front by the scholars involved. My own preference would be to see both quantitative and qualitative approaches in any large project, so readers can be assured that the researchers involved attempted to reduce the biases that come from relying only on one method of analysis. Indeed, in the mid-1990s, some of the top scholars in political science (Gary King, Robert Keohane, and Sidney Verba) wrote in a widely-read book called Designing Social Inquiry that qualitative and quantitative data analysis should by all rights share a “unified logic of inference.” Both types of data, they explain, are excellent tools to help us as researchers do political science – that is, as they put it, both can “produce valid inferences about social and political life.”

Yet in the past few decades, quantitative data analysis seems to have gained disproportionate status and attention within the field. Students absorb this bias from professors, graduate students, and the texts they are assigned, as if by osmosis, and then tell me that their senior thesis must include statistics or it will not be “real” social science. They barely understand what “quantitative analysis” means, but they are determined to do it anyway. One student told me his chosen investigative method was regression analysis. His sample size? Five.

As a political science graduate student for the better part of a decade, as a job-seeker, and now as an assistant professor, I have felt the pressures to produce quantitative work. Yet my concern in this article is something more insidious: I fear that this cleavage in research methods is far from gender neutral.

The sexes are not equally distributed across subfields of political science. Women are more likely to study comparative government, and men continue to dominate the study of political theory, political economy, and security studies. While subfield gender breakdowns are hard to come by, publishing trends across the different journals gives us a sense of where women are concentrated. Women are more likely to publish in comparative politics while men are more likely to publish in other fields.

Furthermore, when I was a graduate student at Harvard, we did a systematic “climate” survey of all the students in my graduate department. We found that when male graduate students were admitted disproportionately in certain years (as in the several admissions committees that admitted 1 or 2 women compared with 7 or 8 men), the gender discrepancy was often explained in terms of statistical knowledge and experience. Men are more likely than women to study statistics, engineering and computer science in undergrad. If one is looking for applicants already versed in statistical methodology, one will likely get men more than women.

The unasked question, however, is why one would select graduate candidates based on previous statistical experience. The disproportionate status of quantitative methodology here has some probably-unintended, but no less real, gendered consequences. As our report on the survey pointed out to the graduate department, the imbalance among men and women, and the corresponding high status of statistical methods, had some negative consequences not just for women in the department (who often felt excluded and marginalized), but also for men who choose other, non-quant pathways of analysis.

Larry Summers’ opinions on women’s mathematical abilities notwithstanding, women are just as capable as men of engaging in quantitative methodologies if they so choose. Selecting students based on previous experience is a weak excuse; a program should select on potential and teach its students the skills they will need, not simply select those who already have them. We need to seriously examine the pervasive quantitative biases that prevent us from recognizing good data as good – even when non-quantitative – and bad data as bad, even when couched in fancy statistics. Doing so is not only an essential step toward gender equality in the discipline, it’s a step toward better science.

With special thanks to Tess Wise.


Further Reading:


About The Author

Avatar photo
Assistant Professor, Political Science, Rutgers University-Camden

Shauna Shames is an Assistant Professor in the Political Science Department of Rutgers-Camden. She received her Ph.D. from Harvard in May, 2014. Her primary area of academic interest is American political behavior, with a focus on race, gender, and politics. For her dissertation research, she analyzed data from an original survey and a set of in-depth interviews about potential candidates' expectations about politics, political campaigns, and serving in elective office. She has published articles, reports, and book chapters on women as candidates, black women in Congress, comparative child care policy, work/family conflict, abortion, and feminism.