One of the stories attracting attention this week is a Gallup finding that a significant percentage of Democrats would vote for McCain if their candidate was not the nominee. I've blogged a couple of times already this week (here and here) in an attempt to put these findings into perspective. One question I have not yet been able to address, however, is the extent to which voters actually do defect from their party's nominee after a contested election.
Fortunately, the National Annenberg Election Study (NAES) can provide some insight into this question. In 2000, the NAES conducted interviews throughout the entire year focusing largely on the presidential election. Fortunately, even in their Fall interviews, the NAES continued to ask respondents who they had voted for in the presidential primaries. This isn't necessarily an ideal measure; after all, the Gallup poll is asking respondents who they support, not who they voted for. In 2000, the nomination campaigns were over by early March, meaning that many never had a chance to vote in the primaries when they were still meaningful (59% of NAES respondents reported that they did not vote in the presidential primaries). Nevertheless, the results are interesting, and worth taking a look at. The figure below presents the intended presidential vote for four different groups: those who voted for Bush in the primaries (15% of respondents), those who voted for McCain (5%), those who voted for Bill Bradley (4%), and those who voted for Gore (16%). This is based on over 10,000 interviews that the NAES conducted between September 4th and October 2nd.
How common were defections? Fairly common, actually. Even in September, only 49% of former McCain voters intended to vote for Bush and 29% were planning on casting their ballot for Gore (in March of 2000, a Pew Survey reported that 51% of McCain supporters planned to vote for Gore). McCain supporters were also far more likely to be undecided late in the race as 11% of this group reported that they still did not know who they intended to vote for.
Former Bradley supporters were also divided. While 52% of this group planned on voting for Gore, another 28% intended to vote for Bush.
Perhaps unsurprisingly, both former Bradley and former McCain supporters appeared more likely to prefer 3rd party candidate Ralph Nader. About 10% of Bradley supporters and 7% of McCain supporters expressed their intent to vote for Nader.
Overall, this suggests that it is not necessarily uncommon for voters to vote for the other party's nominee when their favored candidate has lost the nomination. In 2000, about three out of every ten McCain or Bradley supporters did so. Of course, the relative size of the former McCain/Bradley voters was not very big. Thus, ultimately, the defectors did not make up a large share of the electorate. The figure below presents the makeup of those who reported voting for Bush and Gore in the general election.
The largest part of Bush and Gore's coalitions were citizens who had not voted in the primaries. This is not surprising since turnout in primaries is usually far less than it is in general elections and in 2000 the nomination races were over after less than half the states had voted. What is really important here is how big a share of Bush and Gore's support the McCain and Bradley defectors accounted for. Based on this survey, former McCain voters accounted for 4% of those who intended to vote for Gore while former Bradley supporters accounted for 2.6% of those who voted for Bush. (Interestingly, citizens who voted for Gore in the primaries accounted for 1.8% of those who intended to vote for Bush and Bush primary supporters were 1% of those who intended to vote for Gore).
Ultimately, the relatively small size of the former McCain and Bradley supporters meant that the defectors did not account for a sizable share of either Bush or Gore's general election constituency. In 2008, two things will be different. On one hand, the size of the group of citizens who voted for the candidate who loses the Democratic nomination will be much larger than it was in 2000. This will make the potential group of defectors far larger and more significant. On the other hand, the 2000 election, while close, was not terribly polarizing. Many voters did not seem to think that it would make a major difference who won the presidential election. In fact, according to the National Election Study, 24% said they didn't care who won the election in 2000, compared with just 15% who said the same thing in 2004. If 2008 continues to be as polarizing an election as 2004, it seems less likely that a large share of Democrats or Republicans would defect since a polarizing environment generally means that partisanship will trump other concerns. Nevertheless, this analysis does suggest that defections do happen and there may be some cause for the eventual Democratic nominee to be concerned.
UPDATE: I should note that I also looked at the distributions for interviews NAES conducted after the general election. In these interviews, respondents told interviewers how they voted in the general election, not how they intended to vote. Essentially, I found very similar percentages of McCain and Bradley voters did, in fact, defect. This is not surprising since most voters have made up their minds by October. I used the October interviews in this post because there were vastly more interviews conducted in that month (providing a large sample size to work with).
UPDATE 2: Upon checking the data from my post, I discovered that I had been reporting that the data above came from October interviews when, in fact, it was from September interviews. I have now changed the information in the post to match the data that I was analyzing. I should note that there was little change in the October interviews. For example, McCain voters were only slightly more likely to say that they were going to vote for Bush (53% rather than 49% in September). My apologies for the initial mistake.