Articles and Analysis


The Gender Gap Vanishing Act

Topics: 2008 , Barack Obama , Hillary Clinton , Pew Research Center , The 2008 Race

The recent Pew poll of Democrats and Democratic-leaning registered voters nationwide finally provides us with some crosstabs of the Democratic presidential primary of gender by other demographic variables. We can now observe what I hypothesized last week from exit polls-the gender gap in the Clinton vs. Obama race can be substantially explained by other demographic variables. Age, socioeconomic status, and ideology may be driving the Democratic primary more than gender.

Overall, the Pew study shows that among Democrats, Clinton has a larger lead over Obama with women (+21) than she does with men (+5). The gender gap, defined as the gender difference in support for the winning candidate, is 8 points. The demographic breakouts are in the table below. (Pew surveyed 1515 adults, including 621 Democrats. We have no additional subgroup size information, nor have we performed significance testing.)


The gender gap almost disappears with older voters. Clinton has nearly the same large lead with older women (+26) as with older men (+21). However, among younger voters, she leads with women (+17) but trails with men (-9).

Socioeconomic status is a larger cleavage. The race is exactly the same with women in households earning over $50,000 as with their male cohorts (41% Clinton, 36% Obama). Clinton trails with college educated voters, regardless of gender (women: -3; men: -11).

But the most dramatic gender difference is along ideological fault lines. Obama leads with liberal women (+5 Obama), but trails with liberal men (+15 Clinton). Clinton has a strong lead with non-liberal women (+37 Clinton), but ties with non-liberal men.

These data suggest a few possibilities. First, ideology, socioeconomic status, and age are likely all more important drivers of the vote than gender. Second, there appears to be an interaction with ideology and gender. Of all the demographic groups of women examined, Obama does best with liberal women, and worst with non-liberal women.

This is not to say that the issue of Clinton's gender isn't important. But its importance is external; voters say others may respond differently to Clinton's gender. The same Pew study also shows more Democrats say Clinton's gender will hurt her if she is the nominee (34%) than say Obama's race will hurt him (29%). Similarly, more Democrats say Obama's race has not been a factor so far (57%) than say the same about Clinton's gender (38%).

I have been arguing this point here for a while. While the press continues to assume voters view Clinton through the lens of their own gender, perhaps it is the press's own lens that is clouded. At a minimum, we should recognize that gender patterns in the Democratic primary are more nuanced than the simplistic "Women Support Hillary" frame.


Daniel T:

The problem I have with your analysis is that it can easily be twisted the other way. Lower income women are more likely to identify with Clinton (and support her) based upon their gender than women who have had a college education. Lower income women are more likely to have expereinced discrimination based upon their gender (think 9 to 5) than college educated women (who are more likely also to have higher incomes). In short, more educated and higher earning income women are more likely to support Obama because they identify with something other than race/sex while lower educated, lowever earning women are more likely to support Hillary based upon her gender because that is all they can identify with.

I am not arguing that the above is true. But it is as equally a plausible response to the data you provide as your own arguement is.


Daniel T:

I might also suggest that difference between liberal men and women is that liberal men might feel compelled to support a woman because, in their minds, that is an important part of the liberal plank while liberal women don't feel they same (Hillary is just another woman to them).

In any event, it is possible to argue that socioeconomic/ideology factors are in pretextual for actual gender bias. Your data doesn't really help tease that out.



Agree with Daniel T. here. This is the second time this analyst has bent over backwards to avoid recognizing what the data she is looking at suggest is possible: that lower-income, less-educated (lower SES) white women are supporting Clinton while their higher SES counterparts are supporting Obama.

Now I'm starting to be interested not just in learning what's driving those differences in candidate support among women, but in why analysts might be avoiding discussing the differences at all.

What is so scary about exploring the way gender, race and class are driving voting behavior in the Democratic race for president this year? Those factors have always been considered significant influences in voting behavior.

Is it off limits to look at whether lower SES Democratic white women are influenced by race and/or gender more than other voters?



Sometimes there is just noise - there is no signal. Perhaps we are trying too hard to find a pattern.

I think that what we have here are cross-currents. For example, take a progressive 60 year old professional woman. She would be more prone to agree with Obama, but could intellectually identify with (and vicariously share a victory by) Hillary. Nobody can predict how individuals will break when the candidates differ on race, gender, age, political orientation, etc.


Margie Omero:

I think there are a few things to consider:

First, yes, there could be "noise" that is not particularly meaningful. But since these results confirmed much of what we found from the NH exit polls, that explanation seems less likely to be true.

Second, most public polling addresses the "who" & "how" of voting, rather than the "why." I'm sure the candidates' internal polling could tell us quite a deal more. This means we can develop lots of theories as to why people are voting the way they are, without, unfortunately, a very good way to test those theories. Others' theories are certainly as plausible as mine.

Lastly, there's certainly nothing scary about talking about gender &/or race having an impact on the race. (Indeed, I talk about it a lot here!) But the gender discussion about HRC in the national press is quite overwhelming. At least it is to me. And the story seems to stay the same regardless of what the data show. If the data suggest an alternative--that voter gender is not the most important demographic--it is important to remind others to keep an open mind.

Given the complexity of this presidential race, i think it's crucial that press, bloggers, & DC talkers use caution when making assumptions about gender and race. It's easy to write a "men are turned off by HRC" story or "women are conflicted about voting for the first woman" story, based on a few choice quotes. It's harder to examine the public polling and wonder if there is anything else at work. Unfortunately, we just don't know the actual answer just yet.



I agree with Margie that a major problem with the analysis is a lack of "why". Daniel T can speculate that older, liberal men making >$50,000 (that would be me)see voting for HRC as part of a liberal plank. On the other hand, we may just think that Hillary is the better choice for president (that would also be me). Were the two choices Clinton and Richardson, I would favor Richardson.



I have a problem with the lack of any additional variables -- race, sexuality, etc. "Liberal men" can include a substantial number of gay men, who are probably disproportionately a part of the Dem base. Some surveys have showed that HRC enjoys wide support among glbt voters. At the end of the day, I think it is abundantly clear that multiple factors contribute to a candidates support. To erase gender from the equation is just plain folly.


Post a comment

Please be patient while your comment posts - sometimes it takes a minute or two. To check your comment, please wait 60 seconds and click your browser's refresh button. Note that comments with three or more hyperlinks will be held for approval.