While the United States is commemorating the 94th Anniversary of women gaining the right to vote, we still have questions about whether women are adequately represented in Congress. We know that the United States elects fewer women to higher office than virtually all Western democracies, but does this affect policy? According to our newest study, men and women represent their constituents very differently.

Most of the research on women’s representation looks at the types of policies Congress passes, and whether women vote for more “pro-women” policies than their male counterparts. Academic research on this topic is mixed, with some scholars finding overall differences between women and men in Congress, others finding differences on specific issues, and others finding no difference at all between the two groups. While this has been a hot topic in political science for several decades, we still don’t have a definitive answer for whether men and women represent their constituents differently.

This research typically relies on roll call votes – that is to say whether women are more likely to vote “yes” on legislation that men vote “no” on. Roll call data have a number of limitations, and just because they are the easiest data to access doesn’t mean they are the right data. To answer this unresolved question, we looked in a different place: speeches.

Speeches, specifically Senate floor speeches, give us better leverage on the question for  three reasons.

  1. They can be on any topic the Senator chooses
  2. They can be measured on virtually infinite dimensions rather than a forced “yes/no” dichotomy
  3. Politicians speak far more than they vote, giving us bigger data

By looking at all Senate floor speeches in the 112th Congress (2011-2012), and comparing men and women in a high-dimensional model of word co-occurrence called HiDEx, we were able to compute a “semantic differential” score for a number of different words representing women’s issues. We found a few significant differences between men and women.


We present our findings in the figure, and the full working paper is available on Academia.edu. We show that women were more likely to describe other women and healthcare positively than men were, while they were more likely to describe guns negatively. They were also more likely to describe abortion as being about “choice” than “life.” But most strikingly, women linked other more closely to “work” rather than “home.” This shows a fundamental difference in worldviews between men and women, and a fundamental difference in how the different genders view the role of women in the modern world. These findings hold even when controlling for partisanship.

By applying a novel analysis to an old question we were not only able to help bring new evidence to a debate in the academic literature, but we were able to learn more about the differences between men and women than we would have if we only looked where everyone else was looking. Finding a new dataset and applying a more sophisticated, high-dimensional analysis gives us the advantage and shows differences that weren’t otherwise apparent.