![Sample our Politics & International Relations journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5947/TOC-Subject-Sample-2021-Politics-International-Relations.jpg"})
Abstract
The 2008 American presidential contest occurred amidst economic conditions unlike any seen in decades. Media assessments have often attributed Barack Obama's victory to the faltering economy, particularly the financial crisis that erupted just seven weeks before election day. In this article we assess the role of the economy and other political issues on vote choice, and find that the impact of the economic crisis is more nuanced than is often assumed. We find that while the economy did matter for the general election, so too did social issues. More interestingly, the collapse itself seemed to have only a minor impact because so many people had already made up their minds before the collapse. Finally, we show that while Obama benefited from the economy in the general election, it may have actually worked against him in the primary phase of the contest.
Acknowledgments
We would like to thank Joshua Clinton, Christian Grose, Dick Niemi, Harold Stanley, Lynn Vavreck, Herb Weisberg, Simon Jackman, and Richard Johnston for helpful feedback. We would also like to acknowledge Trevor Tompson, Knowledge Networks, the Associated Press, and Yahoo News! for the data used in our analysis.
Notes
1. Public statement, 19 September 2008. Full transcript available at ⟨http://georgewbush-whitehouse.archives.gov/news/releases/2008/09/20080919-2.html⟩.
3. See, for instance, CNN analysis at ⟨http://www.cnn.com/2008/POLITICS/11/05/mccain.anatomy.loss/index.html⟩. A closer look at the polling numbers suggests that McCain's numbers were in decline even before the Lehman Brothers' collapse; the negative media coverage of Sarah Palin also seems to have contributed to his downturn in aggregate support (Johnston & Thorson, Citation2009).
4. The graph reports only the “the economy in general” category. Measured in this way, the economy does not overtake Iraq in importance until February 2008 (see ⟨http://www.gallup.com/poll/104464/Economy-Surpasses-Iraq-Most-Important-Problem.aspx⟩ for details). If we combine that category with “fuel/oil prices” and “unemployment”, etc., the “economy” overtakes Iraq three months earlier in November 2007, before any caucuses or primary elections have been held, but still well after the invisible primary is underway (see ⟨http://www.gallup.com/poll/103699/Americans-Economic-Issues-Countrys-Top-Problem-Today.aspx⟩ for details). Although it is difficult to identify the exact point at which public attention shifted, it is clear that the economy became increasingly important over the course of the nomination stage of the election. A Google Insights word search similarly finds a sharp increase in the incidence of “economy” over the course of February 2008.
5. Survey by Gallup Organization, 6–9 December 2007. Retrieved 16 October 2009 from the iPOLL Databank, The Roper Center for Public Opinion Research, University of Connecticut. ⟨http://www.ropercenter.uconn.edu/ipoll.html⟩.
6. The study was a collaboration between The Associated Press and Yahoo Inc., with support from Knowledge Networks and collaboration with faculty from Harvard and Stanford Universities. The KnowledgePanel® panel members are chosen via a probability‐based sampling method and using known published sampling frames that cover 99% of the US population. Sampled non‐internet households are provided a laptop computer or MSN TV unit and free internet service. The wave‐one survey (baseline) was fielded on 2 November 2007 to a sample of 3,548 panel members age 18 years or older who represented a general population sample. The total number of completed interviews at the baseline was 2,714, and 76.5% of those who were fielded the survey completed the baseline interview. This represents a cumulative response rate (CUMRR1) of 11.2%, using the formula specified in Callegaro and DiSogra (2009). This rate is a multiplicative combination of the panel recruitment response rate (AAPOR3), the household profile rate and the survey completion rate, but excludes the household retention rate. The study attempted to re‐interview each of the baseline cases for a total of 11 waves. 1,068 respondents completed all 11 waves of the survey.
7. Beyond lack of substantial variation in 2008, some scholars have argued that these items could be endogenous to vote preference, with supporters of the incumbent claiming the economy is in great shape and supporters of the challenging party claiming it is in the doldrums (Bartels, Citation2002).
8. Includes only respondents who voted for Obama or McCain. Figure and all other descriptive statistics are computed using post‐stratification weights provided by Knowledge Networks. The post‐stratification variables include age, race, gender, Hispanic ethnicity and education using distributions from the most recent data from the Current Population Survey (CPS).
9. The question was asked in the seventh wave of the survey which went into the field on 6 September, two days after the close of the Republican National Convention. Although the survey remained in the field until October, 89.9% of its respondents completed the survey before 15 September when Lehman Brothers declared bankruptcy, Bank of America acquired Merrill Lynch, and the federal government took on AIG. The results in Tables and as well as the model in Table are based on the sample of respondents completing wave seven prior to this date. Our substantive conclusions and model results do not change if we restrict our analysis to those respondents who completed the wave‐seven survey before the takeover of Fannie Mae and Freddie Mac.
10. We lose some respondents for this analysis who had only responded to either wave seven or wave nine since the cross‐tab necessarily includes only those answering both.
11. We collapse into a single category any respondents who say they trust neither candidate or both candidates equally. While there may be some substantive differences between these two groups, we are interested in relative trust (i.e. trusting one candidate more than the other) so the responses are equivalent – they trust neither candidate more than the other.
12. Like the trust in handling the economy item, the trust in handling terrorism item is included in the model as a series of two binary indicators for trusting Obama and trusting neither candidate more than the other. The excluded category is trusting McCain.
13. As with the regression model presented in Table , Table contains differences in predicted probabilities. The full sets of coefficient estimates for all models are contained in the Appendix.
14. The multivariate models are estimated without weighting but with controls for all variables used in calculating the weights (except race because of perfect prediction in some cases) (Gelman & Carlin, Citation2001). Indeed, some of our post‐estimated simulations preclude the use of weights. Nonetheless, when estimating just the weighted model, we find nearly identical coefficients and any changes in standard errors do not change our key conclusions.
15. The probabilities presented in the following two paragraphs (as well as in Table ) are simulated from the model results with covariates set at particular values. The covariates of interest are candidate support in early September, trust in handling the economy in early September, and change in trust in handling the economy. All other covariates are set to mean or modal values with trust in handling terrorism set to trusting neither candidate more than the other. To summarize sets of cases (e.g. Obama supporters in early September regardless of whom they trusted to handle the economy), probabilities are averages of the predicted probabilities within the set weighted by the proportion of cases in the set that fall into each category. It is worth noting that there is not data support for 4/27 combinations and there are a very small number of cases in many of the others
16. In fact, there are no cases that supported Obama in September, trusted him on the economy in September, and came to trust McCain on the economy in October, so this probability refers to the 10 cases who supported Obama in September, trusted him on the economy in September, and came to trust neither candidate more than the other in October.
17. The one exception is the very small group of voters who supported McCain in September but came to trust Obama to do a better job on the economy after the collapse – their probability of remaining loyal to McCain is 0.23 (although the confidence interval overlaps 0.50 – the natural cutoff for predicting direction of the vote).
18. Of course, this exercise does not give us a true causal effect since the counterfactual scenario in which no change in trust occurs is not synonymous with a comparison between what actually happened and what might have happened if there had been no economic crisis at all.
19. Iowa entrance polls found that 35% of respondents said the economy was the most important issue and 35% said Iraq was the most important issue: ⟨http://www.cnn.com/ELECTION/2008/primaries/results/epolls/#IADEM⟩.
20. We use two questions with identical wording except the name of the relevant issue. The exact wording was “How important is each of the following issues to you personally … [the economy/the situation in Iraq]?” Respondents had five response options ranging from “Extremely important” to “Not at all important”. Since very few respondents place themselves in the two categories indicating least importance, these two categories are collapsed with the middle category to reduce skew and generate a three‐point scale of concern.
21. Although endogeneity is a concern with most observational studies, it is theoretically less clear why those who favored Clinton would justify this support by saying the economy was extremely important while those who favored Obama would make no such justification of their choice. Even if voters engage in post‐hoc rationalization of their preference by believing their preferred candidate is favored on the issues that are put before them and then deeming those issues important, it remains theoretically unclear why this sort of justification would show up among one candidate's supporters but not the other's. In any event, our aim is simply to identify if there is a relationship, whatever the direction of the causal arrows.
22. A large majority of these non‐Democrats are those respondents who identified with neither major party; only 14.5% of them are Republican identifiers (constituting just 5.7% of the sample in this analysis).
23. The 4.6% of leaners (based on follow‐up question) were also included in the analysis. It is important to note that the candidate preference question used for defining this is distinct from the question that asked respondents' actual primary vote. We conduct a similar analysis using reported primary vote as the outcome, restricting the sample to only those primary voters who cast ballots for either Obama or Clinton. The results are nearly identical.
24. Predicting counterfactual election outcomes for nomination contests is exceptionally tricky for at least two reasons, one having to do with our analysis sample and the other with the nature of presidential primary contests. First, because our analysis sample is non‐black likely Democratic primary voters only, Clinton has higher levels of support among this sample than she would have among all likely Democratic primary voters. Black respondents choosing between Clinton and Obama in the April survey broke for the latter 82% to 18%. In the sample used in this analysis Clinton has a “vote share” of 62%. Second, these predicted outcomes cannot actually refer to vote shares at all. Analyses of candidate preferences from a nationally representative sample in April are difficult to map back on to possible behavior in primary elections because the separate contests were held in different places at different times. The matter is further complicated by the lack of certainty of what the actual vote totals received by each candidate were (see Balz & Johnson, Citation2009: 273).