Bringing the President Back In: The Collapse of Lehman Brothers and the Evolution of Retrospective Voting in the 2008 Presidential Election

Article excerpt


Despite prevailing negative conditions, initial analyses of the 2008 presidential election, including this one, find significant but not particularly strong economic voting effects during the fall campaign. In this article, the authors pay special attention to how the economic information context changed during the campaign and how those changes affected the evolution of retrospective voting. The findings show that there were two distinct phases of the fall campaign, that retrospective voting was nonexistent prior to the collapse of Lehman Brothers but was strong following the collapse. In effect, the collapse of Lehman Brothers turned the election into a referendum election.


economic voting, 2008 presidential election, retrospective voting, economic evaluations, presidential approval

Since the publication of Fiorina's (1981) seminal work, studies of retrospective voting, especially economic retrospective voting, have assumed a prominent role in electoral studies in the United States and around the globe. The central tenet of retrospective voting is that voters make simple assessments about current conditions, broadly defined, and then support or oppose the incumbent administration based on those assessments. While voters may judge the incumbent administration on any number of criteria- foreign policy, the economy, corruption, and so on-much of the focus in studies of retrospective voting has been on the impact of the economic evaluations (Lewis-Beck and Stegmaier 2000, 2007). For instance, election-forecasting models are almost always retrospective in nature and almost always include measures of economic performance as a key retrospective indicator, often in conjunction with a measure of presidential approval (Holbrook 2010).1

At first blush, the 2008 presidential election would seem to offer the perfect context for strong retrospective voting: a declining economy and an unpopular president should both have sent strong signals to the electorate. In short, the economic and political context in 2008 was so negative, in comparison to other years, that forming and using retrospective evaluations should have been a relatively easy task, one that would be expected to result in an Obama victory. However, as we discuss below, there are good reasons to have had more tempered expectations for retrospective voting, and economic voting in particular. More important, however, is that the campaign did not unfold amid a single contextual backdrop; instead, the context of the election changed dramatically with the focus on economic turmoil that began in earnest when Lehman Brothers Inc., a major financial services firm, collapsed on September 15, 2008.2 We take advantage of this important shiftto test ideas about the impact of context on retrospective voting in the 2008 presidential election. Specifically, we examine the evolution of retrospective voting during the fall campaign, paying special attention to the muchdiscussed collapse of the Lehman Brothers financial services corporation as a potential catalyst for retrospective voting. Specifically, we examine whether the collapse of Lehman Brothers served a priming function, leading voters to rely more heavily on economic perceptions following the collapse. We also consider whether the collapse may have helped turn voters' attention to national conditions more generally and led them to assign more weight not just to economic evaluations but also to presidential performance evaluations.

Studies of the 2008 Election

Interesting and important scholarship on the 2008 election is now beginning to emerge, including work on race (Hutchings 2009; Pasek et al. 2009; Philpot, Shaw, and McGowen 2009), campaign effects (Masket 2009; Osborn, McClurg, and Knoll 2010), and the economy (Erikson 2009; Holbrook 2009; Lewis-Beck and Nadeau 2009; Linn, Moody, and Asper 2009). There are a number of findings from these and other studies that provide some insight into the role of the economy. For instance, forecasting models had a relatively accurate year,3 coming within 2.9 percentage points, on average, of the actual vote (Holbrook 2010).4 In addition, a number of other studies focused on economic and other sources of change in candidate support during the campaign (Erikson 2009; Linn, Moody, and Asper 2009) as well as on how the economy affected individual votes (Holbrook 2009; Lewis-Beck and Nadeau 2009). Erikson (2009) explored a number of different aspects of the economic underpinnings of the election and found that daily aggregate readings of economic attitudes tracked very well with changes in candidate support in the post-Lehman (September 15) period of the campaign but that changes in the performance of the stock market during the same period had a much more modest effect on vote share. Linn, Moody, and Asper (2009) also explored candidate support in trialheat polls but took a longer view, using data as far back as early January 2008. Their results are intriguing and provide a mixed picture of the role of the economy. On one hand, they find that a cumulative measure of economic crisis events is significantly related to candidate support, with Obama gaining votes as the events unfolded. On the other hand, they find that candidate support is unresponsive to the performance of the stock market as well as to the tone of news coverage about the economy. In addition, the authors found that another important retrospective element, presidential approval, was significantly related to candidate support, with Obama gaining votes as Bush's approval rating slipped.

Evidence of economic voting has also been found at the individual level. Lewis-Beck and Nadeau (2009) utilize data from the Cooperative Campaign Analysis Project to isolate the role of economic attitudes, paying special attention to how attributing responsibility of the economy can condition the effect of these attitudes. Their analysis, based on interviews conducted during October of the election year, finds clear evidence of classic sociotropic economic voting, with especially pronounced effects from those who attributed responsibility to the president. Most similar to our study is Holbrook's (2009) analysis, which focused on putting the 2008 election in context by comparing the level of economic voting in 2008 to that found in earlier elections. Using data from the American National Election Study (ANES) 2008-9 panel study, Holbrook found a significant, but perhaps weaker than expected, relationship between economic attitudes and vote choice in 2008. Holbrook argued that even with this modest level of economic voting at the individual level, the aggregate impact was quite pronounced because of the decidedly negative distribution of economic perceptions.

Plausible Expectations for Economic Voting in 2008

There are plausible arguments, grounded in research from studies of information processing, for expecting strong economic voting in 2008 as well as for more tempered expectations. The crux of the strong economic voting argument is based on research showing that while people pay relatively little attention to policy making and have relatively little information about the issues of the day, there is a strong tendency to rely on relatively accessible information shortcuts for decision making (Delli Carpini and Keeter 1996; Popkin 1994; Zaller 1992; Zaller and Feldman 1992). Generally, the issues, events, and personalities that receive the most media coverage are the most accessible, or are perceived as more important, and can form the basis of candidate evaluations in presidential elections (Iyengar and Kinder 1987; Valentino, Hutchings, and White 2002; Miller and Krosnick 2000; but see Malhotra and Krosnick 2007). Studies of economic voting have provided evidence for just this sort of media effect. While some studies have focused on the tone of economic news-positive or negative-and have found that good news is good for incumbents (Holbrook 2002; Nadeau et al. 1999; Shah et al. 1999), the more directly relevant studies are those that have shown that more news about the economy increases the weight people assign to economic considerations when they are either evaluating presidential job performance (Edwards, Mitchell, and Welch 1995) or casting votes in presidential elections (Hetherington 1996; Vavreck 2009). Given an estimate that the amount of news coverage about economic issues in 2008 was ten times greater than in 2004 and more than four times greater than in 2000 (Holbrook 2009), findings from this line of research would seem to augur for strong economic voting effects in 2008.

While all of this suggests heightened levels of economic voting, there were reasons to have had tempered expectations for the level of retrospective voting in 2008. Chief among them was the open nature of the race. The absence of an incumbent president running for reelection should have made it more difficult to frame the election as a referendum, and retrospective voting should have been weaker than if President Bush had been on the ticket (Campbell 2008; Holbrook 2008; Nadeau and Lewis-Beck 2001). Moreover, given that 2008 was the first election since 1952 in which the incumbent party was not represented on the ballot by a sitting president or vice president, it certainly seems plausible that Barack Obama would have had more difficulty making the anti-Bush case stick. In addition to the open nature of the contest, the skewed nature of economic evaluations may be responsible for weaker-than-expected economic voting (Holbrook 2009). In effect, with virtually everyone (90 percent) agreeing that the economy was "somewhat" or "much" worse than in the previous year,5 voters on both sides of the partisan divide generally agreed on the miserable state of the economy. With this type of restricted variance on a key independent variable, the individual-level relationship was not as strong as might have been expected in a year when the economic vote was primed so heavily by media coverage.

Taking Another Look at 2008

In this article, we take a closer look at the role of retrospective considerations in shaping the 2008 presidential vote. As mentioned earlier, our primary departure from the existing literature is that we examine the evolution of retrospective voting during the fall campaign. While the existing research on the 2008 election focuses on the potential for the economic collapse or for general economic conditions to have stimulated economic voting, none of the research examines this phenomenon the way we do in this article, that is, with an explicit focus on the Lehman Brothers collapse as a conditioning influence on retrospective voting. Linn, Moody, and Asper (2009), for instance, examine the additive influence of economic crisis events on changes in vote share, and Erikson (2009) examines a similar influence of aggregate economic attitudes in the aftermath of the Lehman collapse. However, neither study examined how economic events changed the weight assigned to economic attitudes. Lewis-Beck and Nadeau (2009) were limited in what they could say on this aspect of economic voting because their analysis was based on interviews conducted in October, after many of the key events had already occurred. Similarly, as a consequence of using the ANES 2008-9 panel study (which did not ask the economic assessment question in September), Holbrook (2009) was able to examine the role of economic attitudes only in October, thus precluding an analysis of how the collapse of Lehman Brothers might have primed economic voting. We provide just such an account in this article.

The Potential Impact of the Collapse of Lehman Brothers

To be sure, the collapse of Lehman Brothers was not the only economic event or issue in 2008. In the year preceding the collapse, there were multiple other indicators of impending economic doom, all of which might have primed the economic vote. And in the weeks before Lehman's collapse, there were other important warning signs, such as the rescues of Fannie Mae and Freddy Mac. But the big event, the triggering event, seems to have been the announcement that Lehman Brothers would declare bankruptcy. This sentiment is reflected in a New York Times article published the day after the election, chronicling how the 2008 presidential campaign unfolded:

It was the third week of September, and Senator John McCain was speaking to a nearly empty convention center in Jacksonville, Fla. Lehman Brothers had collapsed that day, a harrowing indicator of the coming financial crisis and a reminder that the presidential campaign was turning into a referendum on which candidate could best address the nation's economic challenges. (Nagourney, Rutenverg, and Zeleny 2008)

The potential impact of the Lehman Brothers collapse is perhaps best understood by examining media coverage of economic events in the fall of 2008. Toward that end, we present data in Figure 1 that account for the number of stories about the economy published in the front section of the New York Times or broadcast on the NBC Nightly News. These data make clear that coverage of the economy before the Lehman collapse was considerably different from coverage in the post-Lehman period. To be sure, there was an uptick in stories in the week or so before the collapse, reflecting events discussed above, but the real spike was immediately after Lehman Brothers declared bankruptcy, and the level of coverage remained generally very high throughout the rest of the preelection period.

There are a number of ways in which this change in media attention could have affected the election. One sensible expectation in the face of such dramatically negative news about the economy would be an equally dramatic shiftin economic evaluations. If such a negative shiftdid occur, then the advantage presumably would have gone to the Obama campaign. Thus, the collapse of Lehman Brothers could be credited with helping Obama win, via its impact on economic perceptions. This is the anticipated additive effect. However, 2008 was anything but a normal year. As discussed earlier, attitudes toward the economy were already at historically negative levels. As a result, something like a ceiling effect might have prevented those attitudes from growing dramatically more pessimistic in the post-Lehman period. These limits might also have blunted the additive impact of the collapse on candidate support, as people had already developed negative impressions of the economy. The data summarized in Figure 2 illustrate how much things changed in the face of the collapse of Lehman Brothers. Here we use responses to the standard retrospective economic evaluation question from the preelection wave of the 2008 ANES to compare the distribution of evaluations in the thirteen interview days up to and including the announcement of the intent of Lehman Brothers to declare bankruptcy (September 15) to the distribution in the remaining fifty preelection interview dates after the collapse.6 Two important patterns emerge from this figure. First, the distribution of economic attitudes did grow more negative in the post-Lehman period. In fact, the difference in the mean evaluation is statistically significant (t = 6.59). However, the substantive magnitude of the shiftis not dramatic, as a change of -0.24 points on a 5-point scale does not represent a sea change of the sort that might be expected in the face of such a dramatic series of events. In fact, the correlation between economic attitudes and a dummy variable separating the two periods is just -.14, suggesting a weak relationship between the Lehman collapse and economic evaluations. Of course, the primary explanation for this is that the pre-Lehman distribution was already extremely negative and could not grow very much more negative. In both cases, there is limited variation in responses, with the share that saw the economy as somewhat worse or much worse growing from 86 percent in the pre-Lehman period to 92 percent in the post-Lehman period. Perhaps, given these limits, the change that did occur is more impressive than it appears.

We test for an additive effect on vote from the Lehman collapse, but we are much more interested in the impact of that interruption on the relationship between economic attitudes and vote choice. Drawing on the literature described above, we anticipate that the collapse of Lehman Brothers and the increased shiftin media attention following it had a priming effect on economic voting. Specifically, the expectation is that respondents assigned greater weight to economic attitudes in the post-Lehman period than in the pre-Lehman period. This is very much in keeping with the logic underlying Holbrook's (2009) comparison of economic voting in 2008 to earlier years when the economy received much more limited exposure in the media. However, our approach is to consider the impact of a single interruption-the collapse of Lehman Brothers-on the weight assigned to economic attitudes over the course of a single fall campaign. In other words, rather than comparing economic voting in 2008 to that in other years, we examine changes in economic voting during the 2008 campaign.

Lehman Brothers and the President

We add an important wrinkle, however, something that is not quite an alternative hypothesis but more a complementary one. Recall that one of the reasons offered for potentially weaker retrospective voting is that the incumbency cue is weaker in the absence of an incumbent president on the ticket (Campbell 2008; Holbrook 2008, 2009; Nadeau and Lewis-Beck 2001). This idea was originally applied to economic voting (Nadeau and Lewis-Beck 2001) but has also been applied to the broader retrospective indicator, presidential approval (Holbrook 2008). We argue here that the triggering event (Lehman Brothers) does not just make economic attitudes more accessible but turns attention to national conditions, more generally, and makes the performance of the incumbent administration more central to the campaign. In effect, we are saying the collapse of Lehman Brothers served to bring the president back in to the campaign, that the broader retrospective argument was easier to make and more likely to resonate in the post- Lehman period than before the collapse. Certainly, all presidential elections are national elections; but key events of the magnitude of the collapse of Lehman, which have explicitly national consequences and become a focus of national and international media, must serve to get the voting public to focus even more intently on national conditions. So in addition to testing for the priming effect on economic evaluations, we are also interested in a more direct test of activating presidential performance, or "bringing the president back in." Toward that end, we also analyze the impact of the Lehman collapse on the relationship between presidential approval and vote choice. Consistent with the logic described above, we expect that the collapse of Lehman made presidential approval a more relevant consideration for voters, that is, that voters had an easier time making a connection between their assessments of President Bush's job performance and their vote following September 15, the date of Lehman's collapse, than before, and this should be reflected by a stronger relationship between approval and vote intention during the later period than in the earlier period.

Data and Analysis

The data used in this article come from the 2008 ANES. Because we are interested in the effects of economic evaluations and presidential approval on vote choice during the course of the 2008 presidential campaign, we rely solely on the preelection interviews, which were conducted from September 2 through November 3, 2008.7 In the first set of models, we examine the additive effects of the Lehman collapse and retrospective evaluations (economy and presidential approval) on vote intention. In a second set of models, we examine whether the Lehman collapse led voters to assign more weight to economic evaluations and/or presidential approval than they did before the collapse occurred. To test whether the fall of Lehman affected the relationship between economic evaluations and vote, we interact economic evaluations with a dummy variable measuring whether a respondent was interviewed before or after the Lehman collapse.8 Similarly, we interact approval and the same Lehman dummy variable to capture the impact of the collapse on evaluations of presidential performance more broadly. In addition, all models include a full complement of relevant control variables, including party, ideology, attitudes toward the Iraq war and gay marriage, religious attendance, union memberships, race, sex, and a dichotomous variable identifying residents of the South. By integrating the date of the interview and its associated interactions (along with a battery of control variables), we are able to gauge how retrospective voting changed in the wake of the collapse of Lehman Brothers.

Additive Models

The additive models are presented in Table 1, where the first column includes economic evaluations and the second column replaces those evaluations with presidential approval. At this point, we should acknowledge that approval and economic evaluations are of course related to each other, though the relationship is not as strong as one might expect (r = .28). In this part of the analysis, we treat the two in separate models, but we examine their joint impact later.

Focusing specifically on the key variables of interest, the results for the additive model are something of a mixed bag. First, the Lehman interruption is not significantly different from zero, indicating that in the presence of multiple controls, respondents interviewed after the collapse of Lehman Brothers were no more or less likely to vote for McCain than those interviewed prior to the collapse. This does not square with the popular notion that the flood of economic information following the collapse translated into a boon for Obama. At the same time, both retrospective measures-economic evaluations and presidential approval-are significantly related to vote choice in the anticipated direction. Those effects, however, are less than overwhelming. For example, the overall shiftin the probability of casting a McCain vote changes from .14 for those who viewed the economy as "much worse" to .40 for those who saw it as "much better," representing a change in probability of just .26.9 This effect is similar to that found in Holbrook's (2009) analysis of the ANES panel data. At the same time, those who disapproved strongly of the job President Bush was doing had a .11 probability of supporting McCain, while those who approved strongly had a probability of .47, a shiftin probability of .36. These effects are potentially very important, but perhaps not as strong as expected given the economic context.10

Of course, as suggested earlier, we think the full story of retrospective voting in 2008 requires examining effects conditioned by the collapse of Lehman Brothers. If the shiftin attention to economic affairs following the collapse primed retrospective voting, then there should have been a relatively weak connection between retrospections and vote intention prior to September 15 and a stronger connection after that point. By not accounting for this difference in Table 1, we are in effect "averaging" the weak and strong relationships and ending up with significant but not overwhelming relationships.

We turn to the conditional analysis in Table 2, where economic evaluations and presidential approval are first evaluated separately and then together in a single model. The findings from the separate models look just like we expected them to look. For both economic attitudes and presidential approval, there is no relationship between retrospections and vote intention in the period prior to the collapse but a relatively strong, positive relationship in the period following the collapse. The nonsignificant slopes for the "economy" and "approval" and the significant and positive slopes for the interaction terms document these effects. But these terms do not tell the whole story, as the full effect of the retrospective evaluations, along with the variance and significance of those effects, is given by the combination of the additive and conditional terms and their underlying error structure (Friedrich 1982).

Estimates of the effects of retrospective evaluations pre- and post-Lehman are summarized in Figure 3. The patterns that emerge from these two graphs are virtually identical. In the top panel, we plot the relationship between economic evaluations and vote intention in both the preand post-Lehman periods. The difference between the pre- and post-Lehman slopes could hardly be starker. In the early period, the slope was virtually flat (b = 0.005, t = 0.03), whereas the relationship in the latter period is quite strong (b = 0.552) and statistically significant (t = 3.52). Contrary to the change in probability over the range of economic attitudes from the additive model (.26), the change in the probability of casting a McCain vote increases by .44 as you move from the most negative to the most positive economic attitudes. Similar results are obtained with presidential approval. Once again, there is no relationship between approval and vote in the pre-Lehman period (b = 0.105, t = 0.490) but a strong positive relationship in the post-Lehman period (b = 0.868, t = 6.29). In this case, the probability of casting a McCain vote increases by .48, appreciably higher than the change found in the additive model (.36), as approval changes from the most negative to the most positive values. Thus, in both cases we have strong support for an activation effect from the collapse of Lehman Brothers.11

Is September 15 the Right (Best) Day?

The arguments and evidence presented above make a strong case for September 15, the date that Lehman Brothers announced their intent to file for bankruptcy, as an important demarcation point in the fall campaign: the economy took center stage in the media and retrospective voting was much stronger after the announcement than before. But we have to consider the possibility that this date is not magical, that if we had chosen any other date, we would have found a significant interaction effect, simply because retrospective voting was stronger at the end of the campaign than at the beginning. Certainly, because of some sort of lagged effect, we might find that a dummy variable taking a value of 1 a few days after the collapse would provide a slightly better explanation, but that is not what concerns us here, as we would view such evidence as largely consistent with our argument. Instead, we are concerned with the possibility that the September 15 demarcation (or a few days later) is not terribly special and that it is outperformed by alternative demarcations before or significantly after. In other words, we want to be confident that the events surrounding the collapse of Lehman Brothers are the right signal and that we are not just picking up the effects of other events that occurred later in the campaign.

To get at this issue, we created dummy variables for each of the sixty-three preelection interview dates and interacted them separately with both economic evaluations and presidential approval. Following this, we tested the impact of those interactions in sixty-three models that also included all of the control variables used in the original model. We then compared the daily model performance to get a sense of which daily interactions provided the best fit. To make these comparisons, we took our cue from Raftery (1995), who suggests using the Bayesian information criterion (BIC) as a way to compare model fit. There are several accepted methods for estimating BIC, but they all tend to produce the similar results (Long and Freese 2006, 112). The estimation we used is illustrated in Equation 1.

... (1)

If BIC k - BIC m > 0, then model m is preferred to model k , but if BIC k - BIC m < 0, then model k is preferred to model m . In short, a model with a lower BIC is preferred to one with a larger BIC.

When comparing models, Raftery (1995) suggests that a BIC difference of 0 to 2 is "weak" evidence that one model fit is better than another (139). A difference of 2 to 6 implies "positive" evidence, a difference of 6 to 10 is "strong" evidence, and a difference greater than 10 implies "very strong" evidence. Figure 4 shows the BICs for the economic evaluation models and the presidential approval models. The horizontal lines divide the models according to BIC differences-based on the model with the lowest BIC. Points that fall beneath the lowest horizontal line (2) represent models that have a "weak" difference from the model with the best fit. In other words, they are not much different from the best fit. Those above the bottom line but below the middle line (6) have a "positive" difference with the best-fitting model. In other words, they do not perform quite as well as the best-fitting model, but they are not a lot weaker. Models above the middle line but below the top line (10) have a "strong" difference. And those above the top line have a "very strong" difference. These last two categories represent models that are appreciably weaker than the best-fitting model.

As can be seen in the BIC graphs, the evidence for a uniquely important Lehman Brothers effect is much stronger for the approval model than for the economic evaluations model. If the collapse of Lehman Brothers distinctively primed economic attitudes, it is not evident from the economic evaluation models. In fact, according to Figure 4, there is little evidence to suggest that the models fit best at any particular time, though the models with the worst fit were those with interruptions before the Lehman collapse. Moreover, there does not appear to be any abrupt change on or around the time that Lehman Brothers collapsed. Instead, it looks like there is a slight tendency for interactions to become more important as time goes on, and we seem to be picking up part of this with the Lehman interaction. However, plots of the BICs from the presidential approval models illustrate a decisive change around the time of the Lehman collapse: the fit is much worse for early dates, the best-fitting (lowest-BIC) daily interaction models all fall between one and five days after the collapse, and the fit deteriorates steadily from that point on. We view this as strong evidence that the collapse was an important triggering event that served to "bring the president back in" to voters' minds, something that, given the distribution of presidential approval, benefitted the Obama campaign.12

Thus far, we have examined economic evaluations and presidential approval in separate models. Although it is important to examine the effects of these variables in separate models, it is also important to consider the joint impact of approval and economic evaluations, especially if we are interested in a more complete picture of retrospective voting. The third column of Table 2 presents the results of a model that includes both economic evaluations and presidential approval along with the interactions of each of these variables with the Lehman dummy. Contrary to the separate economic evaluations model, the full model in Table 2 shows that the interaction between economic evaluations and the Lehman dummy variable is not significant when the interaction between presidential approval and the Lehman dummy is also included.13 However, the interaction between approval and the Lehman dummy remains statistically significant in the full interaction model. Thus, when we take into account both economic evaluations and presidential approval, the data make a stronger case that it was presidential approval that was activated by the Lehman collapse. This supports the pattern presented above (Figure 4), which showed that the approval models fit best around the time of the Lehman collapse while the economy models did not, as well as the more general theory that the Lehman collapse helped bring the president into the campaign and into the voters' minds.14

Shall we conclude from this that economic evaluations had no influence on vote intention, even in the aftermath of Lehman Brothers? We think not. One important consideration is that presidential approval became more responsive to variation in economic evaluations after the Lehman collapse, in effect that the robust effect from presidential approval may be channeling the seemingly moribund effect from economic evaluations. We test this idea in Table 3, which presents presidential approval as a function of economic evaluations, interacted with the Lehman interruption. We also include the core variables party identification, political ideology, and opinions on the Iraq War. The significant interaction between economic evaluations and the Lehman collapse in Table 3 indeed shows that economic evaluations were more strongly tied to presidential approval after the Lehman collapse than they were before the collapse. Prior to the collapse, the slope for the e ffect of economic evaluations on approval was .32, but it was twice that size (.64) in the period following the collapse. We take this as strong evidence that although economic evaluations had no independent direct effect in the full vote model in Table 2, they did have an important indirect effect, through presidential approval, and that effect grew stronger following the collapse of Lehman Brothers.


The 2008 election provided a fascinating context in which to examine retrospective voting. Despite conditions (clear negative economic cues and an unpopular president) that might have been expected to lead to strong retrospective voting, early reports (Holbrook 2009) suggested that the individual-level relationship between economic evaluations and vote choice, while significant, was not as impressive as might be expected. The absence of an incumbent candidate and, ironically enough, conditions that were so negative that there was very little variation in economic perceptions were thought to have led to the weaker than expected relationship. While this made a great deal of sense to us, we reject the assumption that there was a single economic (or for that matter incumbency) context during the fall campaign. Instead, we argue that there were two distinct phases of the fall campaigns: those days prior to the collapse of Lehman Brothers, on September 15, as the first phase, and the days following the collapse as the second phase.

Our argument is that the collapse of Lehman Brothers activated economic retrospective voting and also served to bring the president back in to the campaign. Voters are influenced by the accessibility and perceived importance of information, and we think that the collapse of Lehman Brothers made economic considerations more accessible and also increased the perceived importance of that information. At the same time, we argued that the collapse made it easier for the Obama campaign to frame the election as a referendum on the Bush administration, even though neither President Bush nor Vice President Cheney was on the ticket. To be sure, the Obama campaign made a concerted effort in this direction before the collapse, but we argue that the renewed focus on economic issues following the collapse provided a backdrop against which it was easier to make anti-Bush sentiment more relevant to vote choice. This sentiment is reflected in postelection comments from David Axelrod, Obama's campaign manager, who, commenting on the importance of the Lehman collapse and related events of the following weeks, said, "It fed our narrative that he [McCain], like the President, was completely out of touch with the reality of the economy. We drove it hard in our ads" (Jamieson 2009, 75).

To be clear, our argument is not that the Lehman collapse produced important direct additive effects but that it generated important conditional effects-that the relationships between retrospective economic and presidential evaluations were weaker pre-Lehman than post-Lehman. Of course, given the negative skews of both economic and presidential evaluations, this heightened relevance would be expected to result in important gains for the Obama campaign. These are all fine ideas, but they remain just that-ideas-without support from the data. On this front, we found mostly supportive evidence. First, we used simple media coverage data to illustrate a clear and theoretically compatible shiftin the information environment at almost the precise date of the Lehman collapse. Second, the analysis of the preelection wave of the 2008 of ANES shows that both economic evaluations and presidential approval had no significant relationship to vote choice before the collapse but were both strong predictors of the vote after the collapse. However, a closer look at model fit statistics shows that the Lehman interruption is a more uniquely important date for presidential approval than for economic evaluations. In addition, when entered in the same model, it is clear that the important conditional effect from the collapse is on presidential approval, with economic evaluations losing their significance. Finally, we also demonstrate that part of the impact of presidential approval post-Lehman comes from the increased connection between economic evaluations and approval in the post-Lehman period. This finding indicates a very important, albeit indirect, way in which economic perceptions, activated by the collapse of Lehman Brothers, influenced voting behavior in the 2008 election.

In the end, we think our findings have a broader relevance than just the 2008 election. Instead, we think the research presented here points to the importance of context in models of retrospective voting. The simple task of retrospective voting is easier and, we think, more likely in some contexts than in others. While some attention has been paid to this idea in studies of U.S. (Nadeau and Lewis-Beck (2001).) and other (Powell and Whitten 1993; Stevenson and Vavreck 2000) elections, we encourage continued efforts in this direction. Perhaps most important, however, is the recognition that, in some cases, the campaign is not accurately described as having a single context. Instead, events that occur during the campaign period can have important contextual effects on voting behavior.


We would like to thank Kathy Dolan, Dave Armstrong, and the anonymous reviews for their thoughtful comments on this article. Any errors are ours.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interests with respect to the authorship and/or publication of this article.

Financial Disclosure/Funding

The authors received no financial support for the research and/ or authorship of this article.


1. In fact, in a group of nine preelection forecasts published prior to the 2008 presidential election in the disciplinary journal PS: Political Science & Politics, eight included at least one measure of economic performance, and five of those eight also included a measure of presidential approval (Campbell 2008)

2. At the time the company filed for bankruptcy, Lehman Brothers had $691 billion in assets, making it the largest bankruptcy in U.S. history.

3. This includes the twelve forecasts published in the October 2008 issue of PS: Political Science & Politics as well as Ray Fair's model ( vot1008.htm).

4. However, this average encompasses a wide range in forecasts, ranging from Lockerbie's prediction of McCain losing with just 41.8 percent to Campbell's prediction of a McCain victory with 52.7 percent of the two-party vote (Campbell 2008; Lockerbie 2008). Still, as a group, this collection of largely retrospective studies predicted an outcome that was quite close to what actually emerged.

5. Taken from the 2008 American National Election Study Time Series Study.

6. "How about the economy in the country as a whole? Would you say that over the past year the nation's economy has gotten better, stayed about the same or gotten worse?" This question was then branched to five response categories: much worse, somewhat worse, same, somewhat better, much better.

7. Strictly speaking, since we are looking at preelection vote intention rather than postelection vote report, we are not looking at actual voting behavior. Nevertheless, since the correlation between preelection vote intention and postelection reported vote is .91, with only 4 percent of all respondents changing their vote, we will continue to use terms such as economic voting and retrospective voting when describing our results.

8. The Lehman variable is a dichotomous variable based on interview date: 0 = interviewed through Sept. 14th, 1 = interviewed on Sept. 15th or after. The number of people interviewed before the collapse is 643. The number of people interviewed after is 1,679.

9. All other variables are set to their mean values.

10. It is worth reminding the reader that these findings should not be read as saying there were weak aggregate effects. Even with modest individual-level effects, the aggregate impact can be quite pronounced if the distributions of the independent variables are skewed dramatically, which was the case for both economic evaluations and approval in 2008.

11. Following the lead of Malhotra and Margalit (2010), we also tested for differences in how partisan groups reacted to the economic crisis. Although we found some partisan differences, they were not consistent, nor always theoretically expected. For instance, the interaction between economic attitudes and the Lehman dummy variable was most important for independents (including leaners), while the approval interaction was most important for independents and Republicans.

12. We also explored whether the Lehman interruption was unique to presidential approval and economic attitudes. After all, it is possible that any number of variables, such as party identification, ideology, demographic characteristics, or other issues, might take on greater importance as the campaign progresses, and a dummy variable such as the one we use here would pick up those effects. Generally speaking, this is not the case. However, the one variable that did behave in a similar fashion was evaluations of whether the Iraq war was "worth the cost." This variable fits squarely into the retrospective framework, so we see this finding as supportive of our thesis about bringing the president back into the election.

13. When testing for the joint significance of economic evaluations and the interaction term, the pair of variables does not add significantly to the model (χ2 = 2.30, p = .317).

14. We plotted Bayesian information criterion (BIC) values from the full model. The plot of BIC statistics from the full model (approval and economic evaluations) is strikingly similar to that for approval in Figure 4, once again verifying the importance of the Lehman interruption, relative to other days.



Campbell, James E. 2008. "The Trial-Heat Forecast of the 2008 Presidential Vote: Performance and Value Considerations in an Open-Seat Election." PS: Political Science & Politics 41:697-70.

Delli Carpini, Michael, and Scott Keeter. 1996. What Americans Know about Politics and Why It Matters. New Haven, CT: Yale University Press.

Edwards, George C., William Mitchell, and Reed Welch. 1995. "Explaining Presidential Approval: The Significance of Issue Salience." American Journal of Political Science 39:108-34.

Erikson, Robert S. 2009. "The American Voter and the Economy, 2008." PS: Political Science & Politics 42:467-71.

Fiorina, Morris. 1981. Retrospective Voting in American National Elections. New Haven, CT: Yale University Press.

Friedrich, Robert. 1982. "In Defense of Multiplicative Terms in Multiple Regression Equations." American Journal of Political Science 26:797-833.

Hetherington, Marc J. 1996. "The Media's Role in Forming Voters' National Economic Evaluations in 1992." American Journal of Political Science 40:372-95.

Holbrook, Thomas M. 2002. "Forecasting with Mixed Economic Signals: A Cautionary Tale." PS: Political Science & Politics 34:39-44.

Holbrook, Thomas M. 2008. "Incumbency, National Conditions, and the 2008 Presidential Election." PS: Political Science & Politics 41:709-12.

Holbrook, Thomas M. 2009. "Economic Considerations and the 2008 Presidential Election." PS: Political Science & Politics 42:473-78.

Holbrook, Thomas M. 2010. "Forecasting U.S. Presidential Elections." In The Oxford Handbook of American Elections and Political Behavior, edited by Jan E. Leighley,. Oxford, UK: Oxford University Press.

Hutchings, Vincent L. 2009. "Change or More of the Same? Evaluating Racial Attitudes in the Obama Era." Public Opinion Quarterly 73:917-42.

Iyengar, Shanto, and Donald R. Kinder. 1987. News That Matters. Chicago: University of Chicago Press.

Jamieson, Kathleen Hall. 2009. Electing the President, 2008: The Insiders' View. Philadelphia: University of Pennsylvania Press.

Lewis-Beck, Michael S., and Richard Nadeau. 2009. "Obama and the Economy in 2008." PS: Political Science & Politics 42:479-83.

Lewis-Beck, Michael S., and Mary Stegmaier. 2000. "Economic Determinants of Election Outcomes." In Annual Review of Political Science, edited by Nelson Polsby. Palo Alto, CA: Annual Reviews.

Lewis-Beck, Michael S., and Mary Stegmaier. 2007. "Economic Models of Voting." In The Oxford Handbook of Political Behavior, edited by Russell J. Dalton and Hans-Dieter Klingemann. Oxford, UK: Oxford University Press.

Linn, Suzanna, Jonathan Moody, and Stephanie Asper. 2009. "Explaining the Horse Race of 2008." PS: Political Science & Politics 42:459-65.

Lockerbie, Brad. 2008. "Election Forecasting: The Future of the Presidency and the House." PS: Political Science & Politics 41:713-16.

Long, J. Scott, and Jeremy Freese. 2006. Regression Models for Categorical Dependent Variables Using Stata. 2nd ed. College Station, TX: Stata Press.

Malhotra, Neil, and Jon A. Krosnick. 2007. "Retrospective and Prospective Performance Assessments during the 2004 Election Campaign: Tests of Mediation and News Media Priming." Political Behavior 29:249-78.

Malhotra, Neil, and Yotam Margalit. 2010. "Short-Term Communication Effects or Longstanding Dispositions? The Public's Response to the Financial Crisis of 2008." Journal of Politics 72:852-67.

Masket, Seth E. 2009. "Did Obama's Ground Game Matter? The Influence of Local Field Offices during the 2008 Presidential Election." Public Opinion Quarterly 73:1023-39.

Miller, Joanne M., and Jon A. Krosnick. 2000. "News Media Impact on the Ingredients of Presidential Evaluations: Politically Knowledgeable Citizens Are Guided by a Trusted Source." American Journal of Political Science 44:295-309.

Nadeau, Richard, and Michael S. Lewis-Beck. 2001. "National Economic Voting in US Presidential Elections." Journal of Politics 63:159-81.

Nadeau, Richard, Richard G. Niemi, David P. Fan, and Timothy Amato. 1999. "Elite Economic Forecasts, Economic News, Mass Economic Judgments, and Presidential Approval." Journal of Politics 61:109-35.

Nagourney, Adam, Jim Rutenverg, and JeffZeleny. 2008. "Near- Flawless Run Is Credited in Victory." New York Times, November 5. 05recon.html?fta=y.

Osborn, Tracy, Scott D. McClurg, and Benjamin Knoll. 2010. "Voter Mobilization and the Obama Victory." American Politics Research 38:211-23.

Pasek, Josh, Alexander Tahk, Yphtach Lelkes, Jon A. Krosnick, B. Keith Payne, Omair Akhtar, and Trevor Tompson. 2009. "Determinants of Turnout and Candidate Choice in the 2008 U.S. Presidential Election: Illuminating the Impact of Racial Prejudice and Other Considerations." Public Opinion Quarterly 73:943-94.

Philpot, Tasha S., Daron R. Shaw, and Ernest B. McGowen. 2009. "Winning the Race: Black Voter Turnout in the 2008 Presidential Election." Public Opinion Quarterly 73:995-1022.

Popkin, Samuel. 1994. The Reasoning Voter. Chicago: University of Chicago Press.

Powell, G. Bingham, Jr., and Guy D. Whitten. 1993. "A Cross- National Analysis of Economic Voting: Taking Account of the Political Context." American Journal of Political Science 37:391-414.

Raftery, Adrian S. 1995. "Bayesian Model Selection in Social Research." Sociological Methodology 25:111-63.

Shah, Dhavan V., Mark D. Watts, David Domke, David P. Fan, and Michael Fibison. 1999. "News Coverage, Economic Cues, and the Public's Presidential Preferences, 1984-1996." Journal of Politics 61:914-43.

Stevenson, Randolph T., and Lynn Vavreck. 2000. "Does Campaign Length Matter? Testing for Cross-National Effects." British Journal of Political Science 30:217-35.

Valentino, Nicholas A., Vincent L. Hutchings, and Ismail K. White. 2002. "Cues That Matter: How Political Ads Prime Racial Attitudes during Campaigns" American Political Science Review 96:75-90.

Vavreck, Lynn. 2009. The Message Matters: The Economy and Presidential Campaigns. Princeton, NJ: Princeton University Press.

Zaller, John. 1992. The Nature and Origins of Mass Public Opinion. Cambridge: Cambridge University Press.

Zaller, John, and Stanley Feldman. 1992. "A Simple Theory of the Survey Response: Answering Questions versus Revealing Preferences." American Journal of Political Science 36:579-616.

[Author Affiliation]

Thomas M. Holbrook1, Clayton Clouse1, and Aaron C. Weinschenk1

1University of Wisconsin, Milwaukee, Milwaukee, WI, USA

Corresponding Author:

Thomas M. Holbrook, University of Wisconsin, Milwaukee,

Department of Political Science, P.O. Box 413, Milwaukee, WI 53201



An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.