More Misinformed than Myopic: Economic Retrospections and the Voter’s Time Horizon

This includes Fiorina’s (1981) seminal work on retrospective voting in American elections. As the above quote implies, he theorizes that voters incorporate information on a set of performance criteria which are evaluated over the incumbent’s entire term. Fiorina’s empirical tests, however, pair electoral data with economic indicators one year previous and thus assume myopic voters.

Duch and Stevenson (2008, p. 46) indicate that some of their surveys they use in their book may have posited a retrospective evaluation of two years rather than one year. However, they give no indication of the prevalence or implications of this two-year item in their database of surveys.

Using survey data from Poland, Owen and Tucker (2010) compare the effects of a conventional 12-month item to an item evaluating economic performance since the beginning of the transition from communism. Mondak and Gearing (1998) also tie retrospections to time since transition in their study of presidential approval in Romania, though they do not compare effects across time horizons. More generally, we are aware of only one other national election study series—Denmark’s—in which the sociotropic economic voting question asks respondents to recall economic conditions for something other than 12 months previous (e.g., Stubager et al. 2013).

TAPS is an online survey of a national probability sample of 2,000 American adults. The sample is recruited using a random selection of residential addresses, stratified on age and ethnic group. Respondents who do not have access to a computer or online services were provided access by TAPS. The survey is conducted for the Weidenbaum Center at Washington University by Knowledge Networks. The cooperation rate was 78 %.

As described below, respondents are asked to evaluate economic performance across four indicators: growth, unemployment, inflation, and the size of the budget deficit. Question wording is provided in the Online Appendix.

The choice of “since December 2008” as long-term referent was intentional. Framing retrospections by using a more salient referent, such as “the height of the Great Recession,” would have conjured up strong images but would have also likely biased responses in a positive direction. And while we wished to ask respondents to think about the economy’s trajectory since the beginning of the Obama Administration, we shied away from referring to the precise historical event (e.g., “since the 2012 presidential election” or “since Obama took office”) because doing so would have confounded ‘pure’ economic evaluations with political attitudes unrelated to performance outcomes (e.g. partisan rationalizations, ideology, race, leader effects, and so on) (Palmer and Duch 2001).

Informed by Table 1, correct assessments of economic trends are as follows: Growth ST frame increased (a lot or a little); Growth LT increased (a lot or a little); Unemployment ST decreased (a lot or a little); Unemployment LT increased (a lot or a little); Consumer prices ST stayed the same or increased a little; Consumer prices LT increased (a lot or a little); Deficit ST stayed the same or decreased; Deficit LT increased (a lot or a little).

We also examined whether perceptual accuracy varies across individual attributes in terms of political sophistication and partisanship. These analyses did not yield any consistent trends across indicators (See Tables A2-A5 in the Online Appendix).

Overreporting in studies of voter turnout is a case in point. Keeter et al. (2002, p. 2) point out that “research that both validates voting claims and asks the same question at different points in time has found over reporting increases the farther away the question is asked from the target election.”.

We also examined the effect of respondents’ objective political and economic knowledge levels on the probability they accurately estimate the change in economic conditions. Not surprisingly, those with higher levels of knowledge are more likely to make the correct assessments of change in economic conditions. However, there is no systematic conditioning effect of temporal frame on the relationship between objective knowledge and accuracy of economic assessments.

The economic perceptions items are coded such that “increased a lot” = 2, “increased a little” = 1, “stayed the same” = 0, “decreased a little” = −1, and “decreased a lot” = −2.

According to an April 2012 CBS News/New York Times National Poll, 54 % of Americans viewed the economy, jobs, or the deficit as being the most important problem facing the country. No respondents listed inflation or prices as the most important problem.

This conclusion parts ways with Healy and Lenz (2014). The difference may be attributable to research design. Healy and Lenz supply voters with hypothetical annual figures of growth; thus their experimental setup takes for granted that voters dispose of such economic information when evaluating performance. Our design makes no such assumption but rather compares voter knowledge of macroeconomic performance across the short and long spans.

Recall that our survey was in the field in March 2012, less than five months before the November 6 vote. Thus, our dependent variable is vote intention. Duch and Stevenson (2008) show that reliance on vote intention or vote choice yields similar survey responses.

In our sample, 53.8 percent intended to vote for Obama and 43.4 for Romney. We omit from analysis the 2.8 % who did not signal their vote intentions.

To check the robustness of the interaction effect, we estimated its coefficient and standard error with the Inteff package in Stata (Ai and Norton 2003). The results are consistent with those reported here.

We assign values that best approximate the direction and magnitude of macroeconomic conditions: Growth SR +1, Growth LR +2, Unemployment SR +1, Unemployment LR −1, Consumer Prices SR +1, Consumer Prices LR +2, Deficit SR −1, Deficit LR +2.

We are mainly interested in the magnitude rather than the direction (+/−) of this effect.

Exceptions are found with respect to sophistication and growth: For high sophisticates 12-month growth evaluations had a stronger effect on their vote intent, a conditional result displayed in Fig. A1 in the Online Appendix. We also find a myopic effect on vote intent for evaluations of unemployment, in this case shaped by partisanship. Those voters who identify with the Democratic Party gave greater weight to 12-month assessments of unemployment (appendix Fig. A2). Despite this partial evidence for voter myopia, we observed mixed evidence in the case of the budget deficit. In evaluating the rate at which the national budget deficit had changed, Democrats did not consistently weigh their assessments in the expected direction; instead they gave differing weights to each reference frame as a function of the direction of change of their evaluations of the deficit (Appendix Fig. A3). All told, no consistent pattern is revealed with respect to the conditioning factors of sophistication or partisanship.