Partisan Defection and Change in the 2008 US Presidential Election

Abstract

Party identification remained an important determinant of vote choice in the 2008 election. Indeed, the extent to which people voted according to their partisanship remained as exceptionally high as it had been in the 2004 election. The Democrats led in partisanship, with a greater lead than in 2004. The ANES four‐wave panel survey shows that some change occurred in the Democratic direction during 2008. The Democrats gained among most population groups, with the exception of older citizens. Obama's victory margin was due to his carrying pure independents and the growth in strong Democrats as opposed to strong Republicans. Both candidates lost the votes of some partisans who disagreed with them ideologically. The rate of defection among major‐party identifiers to the other major party hit post‐1950 lows in 2004 and 2008, reflecting increased polarization in the electorate. The partisanship shifts of young people and Hispanics could portend realignment, although that depends on their satisfaction with the Obama administration.

Acknowledgments

We thank Lynn Vavreck, Richard Niemi, Harold Stanley, and other participants for comments on an earlier version of this paper at the Geurin‐Pettus conference on the 2008 election at SMU‐in‐Taos, and Marc Hetherington for his comments at the 2009 American Political Science Association meetings.

Notes

1. As usual, this report owes considerable gratitude to the American National Election Studies. Their pre‐ and post‐election surveys have been weighted by the post‐election weight (v080102) to adjust for the intentional oversample of African Americans and Latinos in the survey design.

2. Figure 1 shows that the trends are quite similar regardless of whether or not Independents who “lean” towards one of the party are included as partisans.

3. It would seem simple to look at panel attrition in the ANES panel survey to estimate the extent to which partisans demobilized in 2008 out of dissatisfaction with the major party nominees. Attempts to do so, however, are confounded with the usual lower interest in politics and turnout of people with less education, lower income, etc., which typically translates into lower turnout for Democrats than Republicans and greater panel attrition for Democrats. Unfortunately there is no obvious way to remove this confounding effect.

4. There are many plausible explanations for the shift toward Democratic Party identification in 2008. A novel one, suggested by an anonymous reviewer, posits that party primary/caucus registration rules might have “forced” changes in party identification. Specifically, people who had not been registered previously as Democrats might have changed their party identification in order to participate in a closed, or semi‐closed, Democratic Party primary/caucus. Presumably, such changes in party identification would have been more common in the Democratic direction, because the 2008 Democratic Party presidential nomination process continued to be competitive long after John McCain effectively had secured the Republican Party nomination. Thus, some voters might have viewed voting in a Democratic Party primary/caucus as the only way to make their votes count. Since the Democratic Party in 25 states allowed only registered Democrats to vote in their 2008 primaries/caucuses, and 5 states allowed only registered Democrats and independent/unaffiliated voters to do so, it is possible that many voters became registered Democrats or Independents strictly in order to vote in a competitive primary or caucus.

Interestingly, further examination of the ANES panel data indicates the opposite relationship between Democratic Party identification and the type of Democratic Party primary or caucus (closed, semi‐closed, or open) in which respondents voted; the correlation between increasing Democratic Party identification across successive panel waves and the extent to which Democratic Party primary/caucuses were closed to voters not registered as Democrats, was statistically significant in the negative direction. In other words, movement toward Democratic identification was much less common among respondents voting in closed and semi‐closed Democratic primaries/caucuses than those voting in open primaries/caucuses.

As best we can interpret this information, the act of voting in any Democratic Party primary/caucus made many individuals think of themselves more as Democrats than otherwise would have been the case. Since, by definition, it is easier for non‐party members – individuals not previously registered as Democrats – to participate in open primaries/caucuses, these effects should have been, as in fact they were, more common in states where the Democratic Party held open versus closed or semi‐closed primaries/caucuses.

5. The top frame uses variable wgtc09 as the weight variable, the middle frame uses wgtL10, and the bottom frame uses wgtL11, while wgtc11 was used for Table 3. This allows the September–October and the October–November comparisons to benefit from the extra respondents added to the panel study in the autumn to make up for panel attrition, while the January–September comparison in Table 4 and the January–November comparison in Table 3 are necessarily restricted to the original respondents who stayed in the panel.

6. The January wave was conducted before it became clear who the major candidates would be, but it included some hypothetical candidate pairings. While it did not happen to include an Obama‐McCain pairing, it did ask an approval voting question for a large number of possible candidates. We consider people who indicated they would vote for Obama but not McCain to be Obama supporters, and people who said they would vote for McCain but not Obama to be McCain supporters, with everyone else coded as undecided between the two candidates.

7. As Hillygus and Henderson (2010) show, there was very little change in candidate support between September and the election, with only one‐sixth of their Associated Press/Yahoo News panel respondents showing any change.

8. As Johnston et al. (2010) points out, the economic evaluations of Democrats and Independents were already quite negative before September, so the people whose evaluations turned most negative when the meltdown hit were Republicans, but they were least susceptible to Obama's appeal.

9. Additionally we cannot tell whether the difference between the first two waves and between the autumn waves is due to the greater time interval between the January and September waves, but we doubt that is the critical factor.

10. Liberals were less likely to vote than conservatives, but that reflects the usual tendency of pro‐Democratic social groups, like the high school educated, to vote at lower rates than the rest of the population.

11. See the discussion of the Democratic primaries in Jackman and Vavreck (2010) and Grose et al. (2010) for analysis of the Clinton–Obama race and its implications for the general election.

12. Leaners are combined with partisans in this comparison.