Retrospective and Prospective Voting Research Paper

Research carried out in the ensuing decades demonstrated the unrealistic nature of this ideal. Politicians blur, equivocate, and traffic in vague generalities. When they do proffer comprehensible proposals, the centrifugal pull of the median voter induces them to offer much the same thing (Black 1958). Voters, in turn, typically lack the information, stable preferences, and cognitive sophistication that ‘issue voting,’ i.e., choosing on the basis of competing policy proposals, would require (Converse 1964).

While many observers fretted that these findings bode ill for the practice of democracy, others reasoned that it was this particular model of electoral competition, and not the voters and politicians, that was deficient. They argued that, rather than consider the merits of nebulous, hypothetical policy alternatives, voters decide to either retain or reject incumbent politicians on the basis of the objective conditions that obtained during their tenure in office (Downs 1957, Key 1966, Fiorina 1981). This theory of retrospective voting has three major features. First, voters need to acquire little information over and above that which they assimilate in their daily lives. Second, it is the results of policies that voters evaluate, not the policies themselves. Third, elections are primarily referenda on the performance of incumbent officeholders, and so provide little guidance as to the future course of public policy. As Key (1966) put it, ‘(the electorate) judges retrospectively; it commands prospectively only insofar as it expresses either approval or disapproval of that which happened before.’

There is any number of indicators of well-being to which voters might respond—the crime rate, environmental quality, combat deaths, etc. Research on retrospective voting, however, has focused primarily upon economic conditions. Virtually all of it can be placed in one of two categories: Time-series analyses of aggregate economic and voting data, and analyses of responses to opinion surveys, in either cross-sectional or panel form.

1. Time-Series Analyses

In one of the most influential studies in the history of political science, Kramer (1971) demonstrated that the better the recent performance of the American economy, the larger the share of votes won by congressional candidates of the incumbent president’s party. Subsequent time-series studies of aggregate economic and voting data found that the electoral fortunes of the president himself depend importantly upon economic fluctuations (Fair 1978), as do the fates of candidates of his party running in senatorial and gubernatorial elections.

Most subsequent research followed Kramer’s lead in several important ways. Incumbency is defined by the party of the president, not by whether those running for lesser office are themselves officeholders. The analyses assume voters have short time horizons—support for the incumbent party is a function of economic performance in the year or so running up to the election. Finally, the economic variable they key upon is usually per capita income change, the variable Kramer found to have the largest impact upon election outcomes.

These studies have produced overwhelming evidence as to the existence of economic retrospective voting. But it is not a uniform, ubiquitous phenomenon. In the USA, favorable economic conditions clearly benefit the incumbent president, and in on-year elections, congressional candidates of the president’s party also benefit . The economy has a smaller impact upon midterm elections, however, as many voters appear to be seeking to counterbalance the ideological orientation of the incumbent president (Alesina et al. 1993, Erikson 1990). Economic retrospective voting is also stronger and more consistent in some countries than in others (Paldam 1991). According to Powell and Whitten (1993), the strength of economic retrospective voting is a function of ‘clarity of responsibility’ in the institutions of government. Where it is lacking, voters have difficulty in deciding whom to reward and whom to punish, thus undermining their ability to vote on the basis of economic performance. Clarity of responsibility is obscured by multiple parties in the governing coalition, minority governments, weakly cohesive parties, opposition party control of a policy-making institution (either a separately elected presidency or house of the legislature), and opposition party chairs of parliamentary committees. These features arise most frequently in proportional representation regimes.

Despite the vast amount of research that has been done on aggregate voting and economic data, there are weaknesses in this approach that make the findings less robust, less definitive, and less cumulative than one would suppose. First, although the economic measures used in these studies are perceived to be quite precise, the more historical data—particularly for years prior to World War II—are not. Kiewiet and Udell (1998) report that the electoral impact of unemployment more than doubles when a more accurate, revised series is used in place of the official Bureau of Labor Statistics series.

Second, the data series are very short, and many studies have examined even smaller subsets of the available observations (e.g., postwar mid-term elections). Researchers also omit ‘unusual’ election years, and make idiosyncratic choices as to the beginning and ending years of their series. Even small changes in the data series can affect results dramatically. For these same reasons, aggregate-level economic retrospective voting models, though touted as forecasting tools, are not. In such models, error bars are usually around plus or minus 5 percent, and due to sample selection effects this is surely an under-estimate. While an improvement over guesses based upon historical averages, forecasts containing that much error are neither impressive nor useful.

A reasonable response to this problem is to analyze pooled, cross-section time series of disaggregated data. Few studies have adopted this approach, but those that have underscore the value of more data even when observations are not independent. Rosenstone’s (1982) analysis of state-level data finds the electoral effects of income change to be close to the average of the effects reported in various studies using aggregate time series data, but his estimates are far more precise. Peltzman’s (1990) also analyzes state-level data, but his findings are at odds with the bulk of the aggregate-level findings. Specifically, he concludes that voters base their decisions on what transpires economically during the entirety of the incumbent president’s entire administration, and not just the last year or so.

Peltzman’s study points to another potential deficiency in the time-series literature. Most studies posit simple retrospective voting rules on the part of the electorate, the null hypothesis being that recent trends in economic conditions have no impact upon voting decisions. A few studies have found support for alternative hypotheses that imply greater ‘rationality,’ either in terms of expectations formation, as in the Peltzman study, or that voters punish incumbents for deviating from a prescriptive fiscal and monetary policies (Chappell and Keech 1985). However, in an explicit comparison of the ‘naıve’ retrospective model vs. a ‘rational’ model positing that voters differentiate between systematic growth and exogenous income shocks, Alesina et al. (1993) conclude that it is the naıve model that best comports with the data.

2. Opinion Survey Analyses

Survey researchers have sought to account for the time-series evidence of economic retrospective voting by identifying the precise ways in which individual voters’ economic evaluations influence their voting decisions. The hypothesis receiving the most initial attention was that voters ‘voted their pocketbooks,’ reacting primarily to trends in their own personal economic circumstances. This seemed to be a natural extension of the simple retrospective voting hypothesis that has dominated the aggregate-level time series analyses, as it posits (a) low information requirements—voters base their decisions to support or oppose incumbents on the basis of information acquired in the course of their daily lives, (b) what matters to voters are objective conditions, not campaign promises, and (c) voters are most reactive to what they have experienced recently.

What researchers found, however, was that whether or not voters were personally doing well or poorly economically had a surprisingly weak impact upon their reported voting decisions. The choice to retain or to reject the present incumbents was instead conditioned much more strongly upon their assessments of recent trends in the nation’s economy (Kinder and Kiewiet 1979). Although many scholars understood these findings as showing that voters vote according to what they perceive to be the interests of the country as a whole rather than on the basis of narrow self-interest, this inference is unfounded. Voters may be motivated purely by concern for their own individual welfare, but realize that the performance of the economy overall is a superior indicator of the efficacy of the incumbent’s policies. Peltzman (1990) puts it this way:

To a crude approximation, the president’s policy affects mainly aggregate income rather than its personal distribution. If voters understood this, they would respond mainly to aggregate income surprises and ignore personal deviations from the aggregate. Voters care about these personal deviations, but they are mainly irrelevant in evaluating the policy of their agent.

Many studies of survey data surmised that the relationships they observed between economic concerns and voting decisions were too weak to account for the aggregate-level, time-series findings. Seldom, however, did they make the calculations necessary to determine if this were actually the case. The few studies that do detect no gross inconsistencies. Indeed, after correcting for measurement error in the survey data, Markus (1988) identifies individual-level effects that would produce interelection vote swings even larger than those observed by Kramer (1971) and Fair (1978).

There are other important ways in which survey analyses of economic retrospective voting have paralleled the aggregate-level time series literature. In a study of five major Western European democracies, Lewis-Beck (1988) finds that the strength of economic retrospective voting was inversely related to the number of parties in the governing coalition. Like Powell and Whitten (1993), he reasoned that multiparty regimes obscure responsibility, making it difficult for voters to determine to whom to attribute credit or blame. Second, other survey researchers have looked beyond simple retrospective voting for evidence of other ways in which economic concerns might affect voting decisions. The most prominent of these alternatives is the prospective voting hypothesis, i.e., that voters base their decision upon how they expect the economy to perform in the near future (MacKuen et al. (1992)). But here, too, direct comparisons tend to favor the retrospective hypothesis. Specifically, Anderson and Wlezian (1997) report that much of the correlation between economic expectations and vote choice arises from respondents’ attitudes concerning the incumbent president coloring their expectations about the future. Retrospective evaluations, in contrast, appear to be more resistant to cognitive consistency pressures. Many supporters of Jimmy Carter may have been confident that the economy would turn around in 1981, but few could convince themselves that 1980 had been a good year.

Finally, survey research on economic retrospective voting, as in the case of time-series analyses, has been hampered by deficiencies in the data. Here, however, the problem is the small number of survey questions available for analysis rather than a small number of observations. The only individual-level data concerning economic evaluations and vote choice that have been gathered on an ongoing, systematic basis or those contained in the National Election Studies, which began in the early 1950s. The questions they ask are reasonably good, as survey questions go, and we are indeed extremely fortunate to have them. Nonetheless, we are left with the nagging doubt that differently posed questions might reveal more accurately the extent of economic retrospective voting.

Bibliography:

1. Alesina A, Londregan J, Rosenthal H 1993 A model of the political economy of the United States. American Political Science Review 87: 12–33
2. Black D 1958 The Theory of Committee and Elections. Cambridge University Press, Cambridge, UK
3. Chappell H W, Keech W R 1985 A new view of accountability for economic performance. American Political Science Review 79: 10–27
4. Converse P 1964 The nature of mass belief systems in mass publics. In: Apter D E (ed.) Ideology and Discontent. The Free Press, London
5. Downs A 1957 An Economic Theory of Democracy. Harper & Row, New York
6. Erickson R S 1990 Economic conditions and the congressional vote: A review of the macrolevel evidence. American Journal of Political Science 34: 373–99
7. Fair R 1978 The effect of economic events on votes for president. Review of Economics and Statistics 60: 159–73
8. Fiorina M P 1981 Retrospective Voting in American National Elections. Yale University Press, New Haven, CT
9. Key V O 1966 The Responsible Electorate. Belknap Press of Harvard University Press, Cambridge, MA
10. Kiewiet D R, Udell M 1998 Twenty-five years after Kramer: An assessment of economic retrospective voting based upon improved estimates of income and unemployment. Economics and Politics 10: 219–48
11. Kinder D, Kiewiet D R 1979 Economic discontent and political behavior: The role of personal grievances and collective economic judgments in congressional voting. American Journal of Political Science 23: 495–527
12. Kramer G B 1971 Short-term fluctuations in U.S. voting behavior, 1896–1964. American Political Science Review 65: 131–43
13. Lewis-Beck M S 1988 Economics and Elections: The Major Western Democracies. University of Michigan Press, Ann Arbor, MI
14. MacKuen M, Erikson R, Stimson J 1992 Peasants or bankers: The American electorate and the US economy. American Political Science Review 86: 597–611
15. Markus G B 1988 The impact of personal and national economic conditions on the presidential vote: A pooled cross-sectional analysis. American Journal of Political Science 32: 137–54
16. Paldam M 1991 How robust is the vote function? A study of seventeen nations over four decades. In: Norpoth H, LewisBeck M S, Lafay J-D (eds.) Economics and Politics: The Calculus of Support. University of Michigan Press, Ann Arbor, MI
17. Peltzman S 1990 How efficient is the voting market? Journal of Law and Economics 33: 27–64
18. Powell G B, Whitten G D 1993 A cross-national analysis of economic voting taking account of the political context. American Journal of Political Science 37: 391–414
19. Rosenstone S J 1982 Forecasting Presidential Elections. Yale University Press, New Haven, CT
20. Wlezien C, Franklin M, Twiggs D 1997 Economic perceptions and vote choice: Disentangling the endogeneity. Political Behavior 19: 7–17