On Tuesday, the Pew Hispanic Center released a report suggesting that Hispanic turnout will likely drop significantly in the upcoming midterm elections. The report has generated a fair amount of news coverage, but the coverage has also been criticized for failing to provide sufficient context for interpreting the poll results. In this post, I hope to provide some of that context.
First some background. The Pew Center report finds that only 51% of registered (the reason for the italics will become clear below) Hispanics say they are "absolutely certain" they will vote in this year's midterm elections, compared to 70% of all registered voters who report the same. It is somewhat informative to learn that registered Hispanics appear to be less likely to vote in this contest than non-Hispanics, but that would not be particularly surprising given that Hispanics generally turn out at lower rates than non-Hispanics. The important contextual information that is missing is how these vote intentions compare to previous midterm elections. Or, as Joshua Tucker writes, "I don't know if 51% of Hispanics planning on voting in this election is better or worse for the Democrats than in previous elections." This is the key piece of information necessary to really judge what the Pew report is telling us.
To put the 51% figure into the proper context, I downloaded data from the Pew Center's early October turnout report from 2006. A word of caution about using this 2006 survey as a baseline for comparison: while the question wording and timing of the 2006 survey is basically the same as for the 2010 poll, the sampling approach in the 2010 survey is somewhat different. Both surveys used a RDD approach, but the 2010 sample is stratified to ensure an over-representation of Hispanic respondents. (More on the 2010 approach in the appendix to this report.)
Even with that in mind, the differences are actually pretty significant. In the 2006 study, 68% of registered Hispanics said that they were "absolutely certain" they would vote (compared to 71% among non-Hispanics). If we assume these figures can be compared, the 2010 survey would be reporting almost an 18 percentage point drop in the share of registered Hispanics saying that they were certain they would vote.
An 18 percentage point decline in the percentage of registered Hispanics intending to vote strikes me as quite large. To confirm that this drop was real, I looked for another poll that I could use to confirm that vote intentions were lower among registered Hispanics this year. I was able to access a June 2010 Gallup poll through the Roper Center Archives. While this poll was conducted a few months ago, it also asked respondents whether they were "absolutely certain" they would vote in the midterm elections. My analysis of the data from this survey reveals that 61% of registered Hispanics said that they were certain they would vote, compared to 75% of registered non-Hispanics. So the values are a bit higher, but the gap is still there. Recall that the 2006 Pew survey indicated no such gap between registered Hispanics and registered non-Hispanics.
Based on this bit of polling triangulation, it appears as though registered Hispanics are, in fact, less likely to vote in 2010 than they were in 2006. The question is why this might be the case. Most explanations for this drop are likely to focus on the numerator (the number of Hispanics intending to vote), with the assumption that something is making Hispanics who participated in 2006 less likely to participate in 2008. But it is also worth considering how much of this may actually be driven by a change in the denominator (the number of Hispanics who are registered to vote). After all, we know that the 2008 campaign did a lot to register citizens to vote, and many of these new registrants were Hispanics. Furthermore, most of the new registrants are also likely to be only occasional voters who would generally skip midterm elections. Thus, it is possible that the same number of registered Hispanics intend to vote in 2010 as did in 2006, but there are simply more registered Hispanics now than there were four years ago (or more of the Hispanics who now end up in surveys are registered to vote). In other words, the percentage who intend to participate might have gone down at least partially because the numerator grew, not because the denominator shrunk.
We can gain some preliminary insight into this hypothesis by returning to the 2006 Pew survey and the 2010 Gallup poll to see what percentage of Hispanics in those surveys reported being registered to vote. In the 2006 Pew poll, only 66% of Hispanic respondents reported that they were registered to vote; in the 2010 Gallup poll the figure was 80%. While I hesitate to go too far with these figures, they do suggest the possibility that the number of Hispanics voting in 2010 may not actually drop; instead, it may simply be the case that more non-voting (or infrequently voting) Hispanics are now registered to vote. Thus, as the political world continues to mull over the reasons for the expected drop in Hispanic participation this year, we should make sure that this decline is real, and not simply a function of an increase in Hispanics who are registered to vote.