THE ECONOMIC MOOD BEFORE THE ELECTION

A standard complement of evaluative items taps the time dimension (retrospective v. prospective) and the target dimension (sociotropic v. egotropic) of the economic vote. (See the recent review by Lewis-Beck and Stegmaier Reference Lewis-Beck, Stegmaier, Dalton and Klingemann2007). Four such items appear in Table 1. The retrospective sociotropic item, the most widely used, asks about whether the national economy has gotten better or worse over the past year. By this measure, the economic picture was “darker than dark” before the 2008 contest (column 1, Table 1). Fully 94% of respondents concluded economic performance over the last year was worse, and within that group over twice as many said “much worse.” (That is, those declaring “worse” = 31%, those declaring “much worse” = 63%). Clearly, the dire objective economic conditions registered a strong impression on these voters.

Table 1.

Economic Perceptions in the 2008 Presidential Election

Socio-retro: “Would you say that over the past year the nation's economy has gotten much better, gotten better, stayed about the same, gotten worse, or gotten much worse?”

Ego-retro: “Would you say that over the past year your own personal finances have gotten much better, gotten better, stayed about the same, gotten worse, or gotten much worse?”

Socio-prospect: “What about the next 12 months? Do you expect the economy as a whole, to get much better, get somewhat better, stay the same, get somewhat worse, get much worse?”

Ego-prospect: “And in the next 12 months? Do you expect your own personal finances to get much better, get somewhat better, stay the same, get somewhat worse, get much worse?”

First two and last two categories are recoded. Don't know answers and refusals are excluded. Source: Cooperative Campaign Analysis Project (October wave).

These leading perceptual economic measures, for example, the retrospective sociotropic item, routinely treat the economy as a valence issue. There is assumed consensus among the electorate that a healthy economy is good. However, the economy may be viewed as a position issue, especially when the item deals with economic policy. (On the electoral difference between valence and position issues, see the classic statement by Stokes [Reference Stokes1963].) There is dissensus among the electorate over the particular policies that will achieve a healthy economy. With respect to these position issues, we would expect economic voting to be more “policy-oriented” than “incumbency-oriented” (See the useful distinction on these two types in Kiewiet [Reference Kiewiet1983, chap. 2].) In particular, an elector may vote with party X regardless of whether it is incumbent, if that party is closer to his or her position on an economic policy. The classic example is a voter who sticks with the Democrats through thick and thin, because that party is closer to his or her preference on unemployment policy.

We explored a number of such position issues, finding two of particular interest: macroeconomic policy priorities and fairness in tax policy. Consider the former, with respondents asked to rank in terms of national importance the following economic conditions: unemployment, growth, trade, and inflation. As can be seen, unemployment, by far, is the top priority (Panel A, Table 2). Of course, not everyone agrees. This is to be expected, since it is a position, rather than a valence, issue. To illustrate, an unemployed line worker might rank unemployment as the top priority, while a plant manager with much seniority might rank inflation as the top priority. At bottom here are issues of economic distribution.

Table 2.

Voters' Position Issues in the 2008 Presidential Election: Macroeconomic Policy Priorities and Fairness in Tax Policy

Panel A. Percentages of who have ranked unemployment, inflation, growth, or trade as first priority. Question: “Drag your choices onto the number boxes on the right, to rank each of these economic conditions in terms of their national importance.”

Panel B. Distribution of answers to the following question: “Suppose a rich person has one dollar, and a poor person has one dollar. How much tax should each pay on that dollar? Rich person pays 60 cents, poor person pays 10 cents, rich person pays 50 cents, poor person pays 20 cents, rich person pays 40 cents, poor person pays 30 cents, rich person pays 30 cents, poor person pays 30 cents.”

Source: see Table 1.

The second position issue is very clearly distributional, for it deals with tax equity. The respondents were told to imagine that a rich person and a poor person each have a dollar. They are then asked to consider the tax rate applied to each, from highly unequal (60 cents to 10 cents) to equal (30 cents taken from each). We observe a wide difference of opinion here about what is “fair” (Panel B, Table 2). Surprisingly, perhaps, less than half thought that the rich person should pay more out of his or her dollar than the poor person. The notion of the “flat tax” appears to have gained much ground in America. Despite the fact that the income tax was implemented as national policy in the United States by the 1913 passage of the Sixteenth Amendment to the Constitution, its progressive form commands a minority view in contemporary America.

ECONOMIC ISSUES AND THE VOTE: A FIRST LOOK

How do these economic valence and position issues relate to the vote? The expectation, from an extensive literature, is that the national economy as valence would be rather strongly related to the vote choice (Nadeau and Lewis-Beck Reference Nadeau and Lewis-Beck2001). This expectation receives support in the cross-tabulation of retrospective sociotropic assessments with candidate preference (Panel A, Table 3). Those who perceived the national economy as “much worse” were much more likely to vote Obama, when compared to those who simply perceived it as “worse.” The traditional relationship continues to hold, then, with worsening economic perceptions associated with a heightening of the anti-incumbent vote (r = .44).

Table 3.

Economic Perceptions, Macroeconomic Priorities, Fairness in Tax Policy, Attribution of Responsibility, and the Obama Vote

Entries represent the percentages supporting Obama. For the definition of the variables, see Tables 1 and 2.

The presence of a valence link to the vote comes as no surprise. Any suspense rests with the relatively under-investigated positional links. With respect to the unemployment issue, it must first be asked if this is simply partisan projection. That is, voters who favor the Democratic candidate may assert that unemployment has highest priority because they know that is a top Democratic issue. Thus, any correlation between priority for unemployment and vote for Obama would essentially be spurious. Put another way, there would be an endogeneity problem. If such spuriousness exists, then we would expect little objective link between the real unemployment rate and the unemployment priority (since the latter is shaped by personal partisanship rather than acquaintance with the economic facts). For example, respondents would tend to be ignorant of the actual unemployment rate—but they are not.

When asked to estimate the current unemployment rate (of 6%), 74% got it right. In sum, these respondents were very aware of this objective macroeconomic problem, and gave it their highest policy priority. It is reasonable to ask, then, how this real and widespread concern about unemployment relates to vote intention. The answer appears in Panel B, Table 3. Those who saw unemployment as a first priority were overwhelmingly for Obama. Candidate Obama maintained his majority even among those who gave unemployment second priority. Nevertheless, among those who gave it low priority (third or fourth), McCain was the big winner. Obviously, this positional issue, which can be cast as a left-versus-right preference on unemployment policy, sharply sorts voters into two camps (r = .45).

The second position issue, on tax equity, again arrays respondent preferences on a continuum, from strong support to no support for income redistribution. Clearly, this invokes the hypothesis of “policy-oriented,” over “incumbency-oriented” economic voting. During the campaign, Obama was sometimes charged with being “too liberal” regarding tax policies, seeking to take from the wealthy and give to the working class. This criticism, or interpretation, seemed to resonate with 2008 voters (Panel C, Table 3). Among respondents who heavily favored redistribution (shares of 60/10, or 50/20, from rich to poor), over three-quarters were Obama supporters. Even with those who favored only mild redistribution (shares of 40/30), Obama found a majority. In vivid contrast, among those who favored no redistribution at all (equal shares of 30/30), McCain won in a landslide. The political division between the two groups—pro-income-redistribution v. anti-income-redistribution—is stark. This question—what is a “fair” tax burden?—stands as a defining positional issue distinguishing the economic views of the Obama and McCain camps. Of all the economic issues under study here, this one yielded the highest correlation with vote, r = .50.

AN ECONOMIC MODEL OF THE OBAMA VOTE

These strong relationships between different economic issues and 2008 vote intention tantalize. Nevertheless, they could appear much more modest after the imposition of statistical controls, in a properly specified multivariate model. The theoretical structure of our model comes from the “funnel of causality” postulated in The American Voter (Campbell et al. Reference Campbell, Converse, Miller and Stokes1960, chap. 2; see also Lewis-Beck et al. Reference Lewis-Beck, Jacoby, Norpoth and Weisberg2008, chap. 2). The funnel of causality traces the temporal flow of four sets of variables influencing voting behavior, two long term (socio-demographics and party identification), and two short term (issues and candidates). From this framework, we draw the following baseline conceptual specification: Vote = f (Socio-Demographics, Party Identification, Economic Issues). The logistic regression results appear in Table 4 (at the bottom of which the measures are fully detailed).

Table 4.

Multivariate Logistic Regression Models of the Obama Vote

* ρ ≤ .05, ** ρ ≤ .01, one-tailed tests.

Dependent variable = 1 if voting intention for Obama, 0 otherwise. Party Id = 0 for Republicans, .5 for Independents, 1 for Democrats. Economy = 0 for much better, better and same, .5 for worse and 1 for much worse. Control variables: Age (in years), Gender (Female), Education (seven categories), Income (14 categories), Race (Black). Ideology = Scale ranging from 0 (very conservative) to 1 (very liberal; DKs = .5). Equity = 0 for 30/30 share, .5 for 40/30 and DKs, and 1 for 50/20 and 60/10. Unemployment = 0 for fourth priority, .33 for third, .67 for second and 1 for first. Responsibility = 0 for Congress and business leaders, .5 for Federal Reserve Board and DKs, and 1 for president. The pseudo-R² is McFadden.

The initial cut, with its leaner specification, offers no surprises (see column 1 of Table 4). First, the model fit, as judged by the pseudo R-squared, compares well with models estimated on other presidential election surveys (Nadeau and Lewis-Beck Reference Nadeau and Lewis-Beck2001). Further, socio-demographics work as expected, with younger, lower income, and black voters significantly more likely to favor Obama. Also, party identification also fulfills its usual role as the voter's decisive psychological anchor. Lastly, retrospective sociotropic evaluations register a statistically and substantively significant impact on vote choice. For example, when a voter's economic evaluation deteriorates from “better or ok” to “much worse,” the probability of an Obama vote jumps by 43%.

As impressive as the simple valence force of the economy appears to be, it is only the beginning of the story. How does this valence issue actually travel to the ballot box? Recall that economic voting theory rests on the Responsibility Hypothesis: electors hold the incumbent accountable for the economy (Lewis-Beck and Paldam Reference Lewis-Beck and Paldam2000). Thus, if the economy falters, the voters in 2008 would be expected to blame the incumbent Republicans and vote against McCain. Given this logic, it becomes useful to demonstrate empirically to whom the respondents attributed economic responsibility. (There is, perhaps surprisingly, relatively little direct evidence on this question. See the pioneering study of Peffley [Reference Peffley1984]; and more recently Rudolph and Grant [Reference Rudolph and Grant2002].)

In the fall of 2008, the task of economic attribution was not especially easy. President Bush, after a largely hands-off approach to the market, suddenly declared it imperative that Congress enact an emergency bailout. The Federal Reserve Board, under the appointment power of the president, had acted unsuccessfully to halt the economic slide through lowering interest rates. Congress itself, although under shaky Democratic control since 2006, passed the bailout bill under a bipartisan label. Finally, there stood the Wall Street financiers and corporate executives, who were pilloried from all sides. In this confusing time, asked who was in charge of the economy, some (13%) responded they did not know. The rest scattered their answers thusly: president (11%); Federal Reserve Board (23%); Congress (33%); business leaders (20%).

Given the tremendous state of flux economic actors and their policies found themselves in, were voters able to turn attributions for responsibility into an economic vote, as theory would dictate? Yes. In the fourth panel (Panel D) of Table 3, one sees economic responsibility related to vote. As economic blame falls more clearly on the Republican incumbent, vote intentions toward Obama increase, r = .36. Indeed, among those who blamed the president himself, 84% are declared for Obama, followed by 63% who blamed the Federal Reserve. However, if instead of blaming Bush directly or indirectly, they blamed Congress or business, they were overwhelmingly for McCain. Thus, economic attribution, targeted specifically at the president, reinforces the general strength of the retrospective sociotropic vote. A voter who perceives economic trouble, and lays that unambiguously at the president's feet, is much more likely to engage in anti-incumbent voting. This interaction effect needs modeling, and we do so. As expected, it is significant, appropriately modifying the main economic effect reported (see column 2, Table 4).

What of the impact of the positional economic issues—unemployment policy and tax equity? Both manage highly significant coefficients, especially the latter. (See column 3, Table 4.) Further, substantive effects appear strong. For one example, a respondent who gives lowering unemployment a No. 1 priority, in contrast to a No. 4 priority, has a 31 percentage point greater probability of voting Obama. For another example, a respondent who favors serious income redistribution (i.e., a 60/10 tax share), in contrast to no income redistribution (i.e., a 30/30 share), has a 38 percentage point greater probability of voting Obama. Observe that these are effects comparable to a major shift in sociotropic judgment (i.e., from “ok” to “much worse”), which varies between 34 to 45 percentage points depending on respondents' views about the responsibility of the president concerning the prevailing economic situation.

These results are theoretically quite important. They suggest an important dimension of economic voting—from positional issues—has been neglected. Further, they even imply that, at least in this contest, positional economic issues were jointly more important than valence economic issues. Moreover, economic issues taken together, valence and positional, clearly add greatly to the explanation of the vote. Finally, note, in particular, the large pseudo R-squared increase (plus seven points) when the “new” economic variables (i.e., positional and the responsibility interaction) are added to the model. (Compare the equations in columns 1 and 3, Table 4.)

With respect to the economic positional issues, a concern has to be put to rest. Possibly, these issues register significance simply because they are tapping into ideological thinking along a liberal-conservative continuum. After all, lower unemployment and greater income redistribution are classic, if not central, “liberal” positions. It may be, then, that these two positional issues are merely proxies for an omitted ideology variable. To test this possibility, we add a standard liberal-conservative self-placement measure to the equation (column 4, Table 4). One observes, as expected, that this ideology variable is significant, with those more liberal favoring Obama. However, one continues to observe that the positional economic issues still exhibit strong independent effects. Finally, we note that ideology does operate as a surrogate for additional issue stances, for example, on welfare, lifestyles, foreign policy. (On this point, see Lewis-Beck et al. Reference Lewis-Beck, Jacoby, Norpoth and Weisberg2008, 223–29.) Ideology, in this sense, acts as a “super-issue,” controlling on a range of issues. Overall, its presence in the model strengthens our confidence in the robustness of the economic effects.