Effect of gender role threat on vote preference

This is a really important finding, and indicates something that we will very much need to find a way to fight, should Clinton be the Democratic candidate.


 Volumes of research in sociology have shown how men respond to perceived threats to their masculinity: in the face of personal or societal threats to their masculine identity, some men become more likely to endorse anti-gay stances, pro-gun policies, or anti-abortion views…

In the study, a randomized experiment was embedded in an otherwise normal political survey of likely voters in New Jersey. Half of the respondents were asked about the distribution of income in their own households – whether they or their spouse earned more money – before being asked about their preference in the Presidential general election. The other half were only asked about the distribution of income in their household at the end of the survey. This question was designed to remind people of disruption to traditional gender roles, without explicitly mentioning Clinton or a female president, and simulate the sorts of subtle gender-based attacks that can be expected when Clinton is a general election candidate.

The effects of the gender role threat question are enormous. As Figure 1 shows, men who weren’t asked about spousal income until after being asked about the Presidential election preferred Clinton over Trump, 49 to 33. However, those who were reminded about the threat to gender roles embodied by Clinton preferred Trump over Clinton, 50 to 42. Concerns about gender role threat shifted men from preferring Clinton by 16 to preferring Trump by 8, a 24 point shift…

The case that this is really about Clinton’s gender, rather than her party is made clearer by the fact that the same experiment has almost no effect on support for Sanders in the match-up with Trump.

This seems pretty compelling, and very worrying.


8 thoughts on “Effect of gender role threat on vote preference

  1. The linked piece doesn’t give any information about who the author is or any details of the poll (other than the fact that it was done in New Jersey). So I did some digging. Here’s the author: https://twitter.com/DanCassino

    And the study methods and numerical results: http://view2.fdu.edu/publicmind/2016/160323/

    The total survey had about 700 respondents, for a margin of error of less than 4 percentage point. The gender x gender priming groups each had between 166 and 193 respondents, for a margin of error of 7-8 percentage points. In the Clinton vs. Trump table (that’s the first one), for Clinton the difference between the priming and no-priming groups are within the margin of error for women, men, and both combined. For Trump, for men, the difference is about a percentage point larger than two margins of error, meaning it’s just barely statistically significant; the differences for women and for women and men combined aren’t significant.

    So I don’t think there’s very strong evidence of a gender priming effect here.

    Also, even if this study found compelling evidence of a gender priming effect, it’s not clear what impact this effect might have on the election. The effect would only be a difference between men who were very recently exposed to a gender-salient stimulus and men who were not so recently exposed. A short-term experiment doesn’t give us information about a long-term cumulative effect. And — unless GOP campaigners are shouting at voters about intrahousehold income distribution just outside the no-campaigning line — it’s the long-term cumulative effects we should be concerned about.

    I agree that thinking strategically about gender will be very important if Clinton is the nominee, in the same way that race was strategically important in 2008 and 2012. I’m not so sure that gender priming specifically is something to worry about.

  2. This is terrifying; thanks for sharing.

    At the end of the linked LSE blogpost, it says: “Sanders’ advantage over Clinton in the general election polls doesn’t necessarily have anything to do with his appeal to young people, or his policies: his advantage can be explained entirely by the fact that he’s a man.”

    I think that might be an over-reach: what the study showed was that IF men were exposed to the “gender prime” before they were asked whether they preferred Clinton or Trump, then those men were much more likely to say they supported Trump. But the men who were NOT exposed to the gender prime favored Clinton by 16 points, only 5 points less than women (in the non-gender-primed condition) favored Clinton.

    So the key thing is: are voters being gender-primed right before they actually cast their ballots? If not, then these results don’t terrify me quite so much. (Nobody asked me about me and my spouse’s incomes the last time I stepped inside a polling booth.) Are there things that can be done to de-emphasize gender, and make it less salient, at the moment voters make political decisions?

    (Also, it is perhaps worth noting that the gender prime *helped* Hillary among women — but not as much as it hurt her among men. So gender priming is a double-edged sword… it’s just that one side is sharper than the other, for Clinton.)

  3. Thanks, Dan, for checking up on that. That’s very reassuring!

    You also raise an interesting point about priming and long-term effects. Political campaigns tend to repeat effective messages over and over, so this makes me wonder about the cumulative effect of repeated priming over a long period. Do you know of anything on this?

  4. I don’t know the priming literature very well at all (basically just that there are significant problems replicating priming effects — it’s one of the standard examples discussed in the replication crisis literature in psychology). Cumulative priming studies would be relatively slow and expensive, so relatively uncommon. Quick searches of Google Scholar and Scopus are turning up mostly papers on language priming with small samples. Maybe there are more valuable studies in the literature, but I don’t know enough to find them quickly.

  5. I think there are political-science reasons to be a little sceptical:

    1) most people – including most likely voters – don’t follow politics until quite close to a general election. So polls this far out are fairly us informative.

    2) conversely, by the time the election comes around, the great majority of people who vote do so on very predictably partisan lines, fairly independently of nominee details. The campaigns are very effective at convincing each party’s voters of the merits of their candidate and the demerits of the other side’s. (Look at how widely popular Romney was with the GOP base during the campaign, as against before or after.)

    So – I think we have good reason to (a) expect declared voting preferences this far out to be sensitive to all manner of effects; (b) doubt that this will actually be the case in November.

    I concede that Trump is anomalous enough to make any argument based on precedent and past evidence shakier than in “normal” cycles.

  6. […] Recent research indicates that if Hillary Clinton is the democratic candidate, sexist bias against her could cost her a significant number of votes.  This should not come as a surprise after the months and months of vilification she has received; contrary to what is said, she is not the most the immoral, dangerous person to run for presidency.  What can we do?   Well, since we can’t roll back the times, perhaps there is little we can do now. […]

Comments are closed.