Former secretary of state Hillary Clinton meets voters during a Nov. 4 campaign rally in Detroit, during her failed presidential bid. (Melina Mara/The Washington Post)

On May 2, during the Women for Women International’s annual luncheon, CNN’s Christiane Amanpour asked Hillary Clinton whether misogyny contributed to her defeat in the 2016 presidential election. The Democratic nominee replied: “Yes. I do think that it played a role.” On May 31, Clinton reiterated the claim in a widely publicized interview at Recode’s Code Conference 2017 event.

In making this claim, Clinton asserted what many political commentators, and no doubt millions of Americans also believe: Negative attitudes toward women affected voters in 2016, and the impact of these attitudes influenced the outcome of the election.

We bring fresh data, and a surprising finding, to this topic.

In the fall of 2016, we asked six questions about the role and status of women on a national survey called the Cooperative Congressional Election Study (CCES). Respondents could agree or disagree with these six statements:

  • “There is a lot of equality between men and women in the United States”
  • “The proper role for women is to be good wives and good mothers.”
  • “New parents — both mothers and fathers — should get paid leave from their jobs to take care of their newborns.”
  • “Women often earn lower salaries than men for doing the same jobs.”
  • “Most people generally feel threatened by women who are strong.”
  • “Qualified women can get promoted to higher jobs as easily as qualified men.”

These questions are intended to capture whether survey respondents have “progressive” or “traditionalist” attitudes toward women’s roles and statuses, without any reference to Clinton, Donald Trump, political parties or the election.

As the graph below shows, for the most part, progressive attitudes are more prevalent than traditionalist ones, but sizable minorities of those answering the survey expressed traditionalist attitudes, especially among men.

Moreover, as the next graph illustrates, people with more progressive attitudes were more likely to vote for Clinton than were those with more traditionalist attitudes. On average, 65 percent of those with relatively progressive attitudes said that they voted for Clinton, compared to 24 percent of those with relatively traditionalist views.

Of course, many other factors also affect voting behavior. Via a statistical model, we accounted for not only attitudes toward women’s roles — combining the six questions into a single index — but also judgments about President Barack Obama’s performance on issues voters identify as especially important, evaluations of the economy, attitudes toward the Affordable Care Act, attitudes toward immigration, sense of racial resentment, gender and demographics (age, education, income, race and ethnicity).

But even after accounting for other factors, attitudes toward women’s roles were still correlated with how people voted. For example, if we imagine that the index runs from zero (most progressive) to 100 (most traditional) the average voter scores roughly a 40. Holding other factors equal, a shift from a relatively progressive position (20) to a relatively traditional position (60) would reduce the chance of voting for Clinton from 57 percent to 17 percent. The finding is robust — the impact of attitudes toward women’s roles was consistent in statistical models with many different combinations of factors that might influence how people voted.

One interesting question is whether attitudes about women’s roles were more strongly related to the votes of men or women. We didn’t find evidence of any difference. These attitudes mattered similarly for both men and women.

Another important question is whether attitudes about women’s roles mattered more in 2016 than in 2012? If so, this suggests that there really was something distinctive about 2016, when a female candidate ran against a male candidate who had made many crude comments about women.

The 2016 CCES asked respondents whether they supported Obama or Romney in 2012. If we apply the same statistical model to people’s 2012 vote choice, we find that attitudes toward women did not have a meaningful association with whether people supported Obama and Romney, despite the Obama campaign’s attacks on Romney and Republicans for waging a “war on women.” Attitudes toward women’s roles and statuses did not have the same traction in 2012 that they did in 2016.

In short, our analysis suggests that Hillary Clinton is correct: Attitudes toward women’s roles and statuses influenced presidential voting in 2016. If fewer voters had held traditionalist attitudes toward women’s roles and statuses, Clinton’s national popular vote total (already a plurality) would have increased. Even small shifts in these attitudes could have affected the outcomes in states like Michigan, Pennsylvania and Ohio, where Clinton lost by an average of only 0.57 percent.

That said, there is another important implication of our findings — one more surprising and actually more favorable to the Clinton campaign. Our survey clearly shows that attitudes toward women’s roles and statuses were tilted in a progressive direction, so the salience of women’s roles in voter decision-making likely helped Clinton more than it hurt her. She had more votes to “gain” from people with progressive attitudes than she had votes to “lose” from those with traditionalist views.

Thus, playing what some observers might call the “woman card” may have been good politics for Clinton in 2016 — even if it was not enough to bring her to the White House.

Harold Clarke and Marianne Stewart are professors in the school of economic, political and policy sciences at the University of Texas at Dallas.