[A response by the author of "Latino Voting in the 2008 Election," Prof. James G. Gimpel]

A few critical comments on the recent CIS Backgrounder, "Latino Voting in the 2008 Election: Part of a Broader Electoral Movement," (see here, for instance) have surfaced on various blogs in the last few days. This is a good thing, as it draws attention to the fact that the stories that are being told about the GOP losses in the 2006 and 2008 elections need to be challenged. Elections are complicated, and always subject to multiple interpretations as to why the results came down the way they did. The best way to gauge which interpretation is correct is to examine evidence for alternative explanations, not shout louder than the other person.

As an important sidebar to this discussion, I don't think anyone is arguing against the importance of the Republicans continuing to broaden their base. No one that I know of believes that one party should be all-European American (since many Latinos consider themselves white, it doesn't make sense to say "all-white" since that would not exclude Latinos). There are a few on the fringe who may believe in an all-European American party, but they have neither a following nor an audience.

That makes the real argument about the means for broadening the GOP's base. I don't believe that survey or other data have made a very convincing case that this should be accomplished by adopting libertarian positions on immigration policy, lenient views of border enforcement, or by adopting an amnesty for illegal immigrants. The data simply do not show what many observers want them to show.

Critical to the accumulation of knowledge on any subject is the importance of comparison. What we often lack in news and interest group use of survey data are the comparisons of other groups that would help adjudicate claims such as: Latinos voted Democratic because of Republican position-taking on immigration. How can we possibly know this is true if we do not look at other electoral groups to see how they voted?

A group might look unique in its response to a particular campaign when there are no comparisons. When the relevant comparisons are considered, this uniqueness fades. This is what the CIS Backgrounder tried to provide; a comparison of Latinos with other groups in the electorate. These comparisons show that Latinos were not at all alone in moving to the Democratic side in 2008 relative to 2004. Every group considered in the paper moved in that very direction. This makes it much less likely that the election was about immigration, since we know that many of these groups don't care much about immigration policy. Considered not just in isolation, but alongside the movement of so many other demographic groups, it appears that the election was about the economy and fears of joblessness and foreclosure, not about immigration.

The other comparisons we need are comparisons across issue salience. Certainly a survey tabulation may indicate that immigration is "an important" issue for Latinos, but there are probably many issues that are ranked as "important": education, health care, jobs, crime, perhaps even gun rights, national security, and gay marriage. But that doesn't mean that any of these issues influenced the voting decision or steered a voter toward one party or the other. Voters will regularly express a strong preference for Coke over Pepsi, or strawberry over vanilla. But that does not mean that these entrenched views have direct relevance for the vote decision.

Comparisons across elections, such as the House and Senate elections of 2008, are important too. One might carelessly to point to some Republican candidates who lost running on reform platforms as a repudiation of a particular policy view. But this ignores not only the Republican candidates who won with the same record, but the large number of Republicans who lost adhering to more liberal views on the subject, or whose views were unarticulated. Considered selectively and in isolation, observers will make a false inference from incomplete data. All of the contested outcomes need to be considered, those races where immigration was mentioned and those where it was ignored.

We do know from decades of research that party identification is relevant for vote choice. It is the dominant consideration by virtually every account of political behavior. At the very least, then, when considering the impact of an issue position on the vote, pundits and commentators should control for the influence of party identification as a significant rival explanation. Most of the Latinos who consider immigration a highly salient issue are committed Democrats, plain and simple. They would not have voted for the McCain/Palin ticket under any circumstances. There are always a few exceptions, but the data suggest that they are a small number. Other control variables should include socioeconomic status, and possibly religion. Concerns about the economy should be included as a rival explanation for the vote.

Voting decisions are often overly intellectualized by pundits, interest groups, and yes, even academics like myself. We probably want to believe that voters are more attentive to what's happening here in Washington than they really are. But the evidence for an elaborate issue-oriented worldview among anyone but the most elite voters is quite weak. Numerous investigators have found that respondents provide "off-the-top-of-the-head" answers to survey questions that do not have a deep anchorage in a lucid framework, but reflect temporary impressions, usually structured by a pollster's multiple-choice response options, that may change with the next survey. The resulting response instability doesn't mean we should abandon survey research, but we should be well aware of what simple survey tabulations can tell us.

Does this understanding of the survey response mean that voters, Latino or otherwise, are stupid or politically illiterate? Not at all, but they do live busy lives, concerned primarily about family, work, and school. The amount of attention that they devote to political affairs is finite. They frequently use information shortcuts, called heuristics, to make judgments, and party identification is one of the most convenient of those. Voters also consult others in their environment, which is the reason I have highlighted the role of location of residence as an important influence on political learning (see here, for instance). They have a sense of whether the economy is doing poorly or doing well. We may all wish that voters had the time to pay attention to politics as closely as an interest group advocate in Washington does, but if they did, we wouldn't need these interest group advocates trying to tell them what they should think.

In conclusion, we have every reason to doubt that immigration policy influenced the 2008 general election outcome. More generally, alternative explanations for election results go untested by the Washington advocacy and punditry corps. Comparisons with other groups, and comparisons with other issues are essential for evaluating specific claims. People express strong opinions about many issues, but this does not mean that their opinion is a determinant of their vote choice. Finally, interpretations of survey results should reckon with the knowledge we have accumulated about survey responses. Voters in open-ended discussion and deliberation on an issue will often shade and nuance responses that look clear-cut in a typical survey. This does not prove that voters are ignorant, but quite the opposite: they are preoccupied with more pressing things, and are often thoughtful.