« Dems' Chances in '06 Senate Races (Part II) or Pundits vs. 'Six-Year Itch' | Main | Zogby Poll: Americans Want Action in Darfur »

Demographic Groups and Turnout in the 2004 Election

There was a general rise in turnout in the 2004 election–around 6 percentage points (the exact figure varies depending on which denominator you use for eligible voters–see the Committee for the Study of the American Electorate and Michael McDonald’s United States Elections Project for two different estimates). But which demographic groups’ turnout went up the most and which went up the least in the election?

That interesting question is typically addressed right after the election by using data from the national exit polls. Of course, since the exit polls, by definition, include only voters, there is no direct measurement of any group’s turnout in the exit poll. But it is possible to take a given group’s representation in the exit poll sample and combine that with Census information about that group’s representation among all eligible voters plus the overall number of votes cast and derive an indirect turnout estimate. When you see turnout estimates for various demographic groups shortly after the election, that’s where they come from.

But there are problems with this procedure and it is certainly not one for which the exit polls were designed. These polls are designed to give information about the political preferences and attitudes of different demographic groups–information which is available nowhere else--not to generate reliable information about the demographic composition and turnout rates of the voting pool and how they have been changing over time.

For this reason, I always like to take a careful look at the Census Voter Supplement, administered as part of the November Current Population Survey (CPS) in every election year (Presidential and off-year). The Voter Supplement collects basic information about whether respondents voted, whether they were registered and a very small number of other items (for example, what time of day did the respondent vote?). No information is collected about who the respondent voted for and what the respondents’ political attitudes and partisan preferences are.

The lack of political information means the Voter Supplement is useless for examining what any given election is about. But, its huge size—90,000 to 100,000 respondents 18 and over, combined with the rich demographic information always collected by the CPS makes it a superb source for analyzing how the demographics of the voting pool have changed over time.

The Census Bureau has now released the data from the 2004 Voter Supplement. Here are some findings of interest from these data:

1. Young voters had, by far, the strongest turnout surge of any age group. As shown in this nice chart provided by the Center for Information and Research on Civic Learning and Engagement (CIRCLE), turnout among 18-24 year-old citizens surged 11 points, compared to only 5 points among 25-34 year olds, 4 points among 35-44 year-olds and even less among older age groups. Note that the overall turnout increase in the Voter Supplement data is just a little over 4 points, so the measured 11 point increase among 18-24 year-olds is even more impressive.

2. By race, turnout rose the most among nonhispanic white citizens (5.3 points), compared to 3.1 points among blacks, 2.1 points among Hispanics and 1.2 points among asians. These data also provide a higher estimate of the nonhispanic white proportion of voters in 2004 (79 percent vs. 77 percent in the NEP exit poll) and a lower estimate of the minority proportion of voters (21 percent vs. 23 percent).

3. The exit polls claimed that 26.1 percent of voters had a high school degree or less, 31.7 percent had some college or an associate degree and 42.1 percent had a four year college degree or more. The Voter Supplement, on the other hand, pegs the education distribution of voters at 36.6 percent with high school or less, 31 percent with some college and 32.4 percent with a four year degree or more. These are big differences.

Whose estimate is more likely to be right? In this case, it’s a pretty easy call to make. If we accepted the exit poll estimate that 42.1 percent of voters had a four year degree or more, that would imply that the turnout rate of college-educated citizens in 2004 was 101 percent.

That seems, well, just a wee bit on the high side. I suggest we all go with the Voter Supplement estimates on this one.