The election has long passed, but debate over what happened continues in full force. In the data-driven and social science corner, it will be fueled by large-scale surveys of Americans conducted after the election asking the public about their voting behavior and a myriad other things. Chief among these sources are the Congressional Cooperative Election Study and the American National Election Study. The latter, the CCES, was released last Friday, and can shed important new light on how people decided to cast their vote, which can be broken down by a host of different demographic and political variables.
Before taking a look at some interesting findings, there’s an important caveat to this to bear in mind: surveys such as these are plagued by respondents’ propensity to over report turnout. That holds for ones from the Census Bureau and the ANES, as well for the CCES; the 2016 CCES survey indicates a 93.2 percent self-reported turnout rate, which cannot be right given actual vote data pointing to that number being closer to 60 percent.
It means that at this point, there’s no way around including some people who said they vote but didn’t actually do so in the analysis. Does this introduce bias? When non-voters—who trend more Democratic and liberal—report their preferences as voters, this tilts things more blue. While the CCES applies weights to make its overall vote closer to the 48-46 percent Clinton margin, it might not solve this problem at subgroup levels. This bias will be corrected when vote validation data is released this summer. Thus, at this point, the results here might be more Democratic, but the bias is not too large.
With that key caveat, I wanted to check some voting dynamics within different racial groups that the CCES can speak to. While race is the strongest demographic predictor of vote choice, other variables, such as age and gender, add some nuance to voting behavior breakdowns. That’s what I’ll explore here, getting at the following question: when broken up into different racial/ethnic groups, how different do Americans vote along gender and age lines?
I’ll try to answer that with three graphs below for three different racial groups: whites, blacks, and Hispanics/Latinos. In each of them, I’ll show reported vote choice—for Hillary Clinton, Donald Trump, or for a third party under “Other”—by whether someone is female or male (going across each row) and by whether their age is between 18 to 29, 30 to 44, 45 to 54, 55 to 64, or 65+ (going down each column).
Let’s start with white Americans.
As you can see in the above graph, there’s a relationship between both vote and age and vote and gender among whites. Going from left to right on each row, Clinton vote generally decreases (while Trump’s increases) as you progress into older age groups. Splitting that pattern by gender reveals an interesting dynamic however: the age gap between the youngest and oldest groups is larger among men than it is among women. Among women, 65+ year olds voted Trump by 18.7 points more than 18-29 year olds did. But among men, that same gap in Trump vote share is 29.1 points.
In some part, that’s due to young white men being much more likely vote for a third-party candidate (such as Gary Johnson). At a 16.1 percent vote for a non-major party candidate, 18-29 year male whites voted third party much more than any other subgroup.
Going down each column in this graph confirms a well-known dynamic: even among whites, women are more inclined to vote for Democrats (in this case Clinton blue) than their male counterparts. Interestingly, because of white male youth’s propensity for third-party voting in this election, white females voted Trump at about the same rates as white males did (i.e. their difference in their rates is not statistically significant). It’s only once you move to older ages that a consistent eight to nine point gap emerges where men vote Trump much more than women do.
Let’s now move to African Americans.
It’s not surprising that among all subgroups here, African Americans vote Democrat at overwhelmingly high rates. That’s something that this election cycle has certainly not overturned (I’d recommend this great book for understanding why this persists). However, there are some really interesting dynamics that develop here. Most notably, black male youth voted Clinton significantly less than every female age group did, and significantly less than the two oldest male age groups. While the oldest age group among females only voted Clinton by 3.4 points than did the youngest female age bracket did, the difference among black men was very different: 65+ year olds voted Clinton by 14.8 points more than 18-29 year olds did. Indeed, 18 to 29 year old blacks—and those between 18 and 54 more broadly—voted for Trump much more than many of these other black subgroups. Thus, as you move from left to right in this graph, the female-male gap in Clinton vote shrinks as you get to older age groups; among 65+ blacks, their propensity for voting Clinton is indistinguishable.
One more thing is worth taking note of here in order to place it in a cross-race context: while Clinton vote decreases with older age among whites, it increases a bit more—as seen here—among blacks.
Finally, let’s take a look at these same dynamics but among Hispanics/Latinos.
The same breakdown for Hispanics here follows the same pattern by gender (women vote Clinton more than men do) as for whites and blacks, but matches more closely to the age relationship among whites: the older you get, the less Democrat the vote becomes among Hispanics. The drop-off in Clinton vote from the youngest to the oldest age group appears a little larger among females (15.9 points) than among males (11.7 points), but not by much. A more considerable jump in Trump vote occurs among women Hispanics going from under 55 years old to over. Meanwhile, it’s a more gradual upward trend in Trump vote with each older age group when looking at men. The gender gap in Trump vote among Hispanics is not that different across most age groups (i.e. not passing the test for a statistically significant difference). In this sense, the gender gap is not as large among Hispanics compared to whites and blacks after “controlling” for age.
A survey like the CCES with 64,600 respondents makes this type of analysis and findings possible. Drilling down to gender by age dynamics within the three largest racial/ethnic groups in the United States, it’s clear there is a gender and age relationship with vote choice even after controlling for race. Yet, that plays out a bit differently across races, making for gender and age gaps in some cases and not in others, all of which is always crucial to acknowledge as we continue to make sense of what happened in the 2016 election.