Capturing Campaign Effects
The University of Michigan Press
Copyright © 2006University of Michigan
All right reserved. ISBN: 978-0-472-06921-7Contents
Acknowledgments............................................................................................................................ixThe Study of Political Campaigns Henry E. Brady, Richard Johnston, and John Sides.........................................................1I Voter Decision Making and Campaign EffectsThe Paradox of Minimal Effects Stephen Ansolabehere.......................................................................................29The Impact of Campaigns on Discrepancies, Errors, and Biases in Voting Behavior Patrick Fournier..........................................45Priming and Persuasion in Presidential Campaigns Larry M. Bartels.........................................................................78II Research Designs and Statistical Methods for Studying Campaign EffectsCampaigns as Experiments Stephen Ansolabehere.............................................................................................115Three Virtues of Panel Data for the Analysis of Campaign Effects Larry M. Bartels.........................................................134The Rolling Cross-Section and Causal Attribution Henry E. Brady and Richard Johnston......................................................164III Campaign Effects in Congressional and Senatorial Races: Information and IssuesMeasuring Campaign Spending Effects in U.S. House Elections Gary C. Jacobson 199Informational Rhythms of Incumbent-Dominated Congressional Elections Laurel Elms and Paul M. Sniderman....................................221Alternative Tests for the Effects of Campaigns and Candidates on Voting Behavior Benjamin Highton.........................................242IV The Rules of the Game and Election ResultsDo Polls Influence the Vote? Andr Blais, Elisabeth Gidengil, and Neil Nevitte............................................................263Strategic Learning in Campaigns with Proportional Representation: Evidence from New Zealand Richard Johnston and JackVowles.....................................................................................................................................280V The Role of the Mass MediaStudying Statewide Political Campaigns R. Michael Alvarez and Alexandra Shankster.........................................................307Gender, Media Coverage, and the Dynamics of Leader Evaluations: The Case of the 1993 Canadian Election ElisabethGidengil and Joanna Everitt................................................................................................................336Mass Media and Third-Party Insurgency Richard Jenkins.....................................................................................356Contributors...............................................................................................................................383Index......................................................................................................................................385
Chapter One
Voter Decision Making and Campaign Effects
The Paradox of Minimal Effects
Stephen Ansolabehere
Central to the study of campaigns and elections lies a paradox. Voters rely heavily on the information conveyed by campaigns in order to form judgments about who should govern. By most accounts, citizens in modern democracies know little about and feel distant from public affairs. Most people tune in during political campaigns, and what they see and hear can influence their opinions considerably, as shown by recent research on individual voting behavior (Ansolabehere and Iyengar 1995; Bartels 1993; Jacobson 1990; Johnston et al. 1992; Popkin 1991; Zaller 1992) and on aggregate voting patterns (Bartels 1987; Brady and Johnston 1987; Jacobson 1980, 1990). It should, thus, be hard to anticipate the outcomes of elections without knowing how the media cover the elections, what themes the candidates stress, which candidates advertise the most, and what is said during conventions and debates. Election outcomes are strongly predicted, however, by forces outside of campaign politics. Evidence from an extensive literature on economics and elections shows that the governing party's reelection depends overwhelmingly on whether the country is prosperous at home and at peace abroad. Campaigns seem to be inessential to understanding who wins and who loses.
The 1988 presidential election perhaps best exemplifies this puzzle. To many political scientists the 1988 presidential election represented an analytical triumph. George Bush won with nearly exactly the vote predicted by a simple model in which the vote is predicted by the percentage change in real personal disposable income. The lots of George Bush and Michael Dukakis, it seems, were drawn by the economic fates.
But Bush did not win with ease. Throughout the campaign the polls fluctuated dramatically, suggesting that large fractions of the electorate made up their minds in the months, even days, leading up to Election Day (Gelman and King 1993). Michael Dukakis even held a substantial lead in the polls late in the summer of 1988. Observers of the campaigns argue that Dukakis was brought down not by the sudden realization of many voters that the economy was doing well but by the themes and strategies of the competing campaigns. The Republicans' messages of crime, patriotism, and conservatism resonated well with the public, while the Democratic campaign lost its focus after the primaries. Bush used attack advertisements and the free media with cunning and success, while Dukakis refused to slug back (Germond and Witcover 1989; Jamieson 1992). Such factors are nowhere to be found in the macropolitical models. Even still, those models worked smashingly well, though in the end they looked a lot like voodoo political economics.
How is it that elections are highly predictable when voters rely heavily on what they see and hear in a particular campaign? The obvious, most common, and probably correct response is that the public discourse observed during campaigns confirms what most people privately believe about the state of the country and the political parties and candidates. Beginning with Paul Lazarsfeld's important studies of the elections of 1944 and 1948, social scientists have argued that campaigns reinforce the political orientations of voters and mobilize them to vote rather than convert large segments of the population to new ways of thinking or new patterns of behavior (Berelson, Lazarsfeld, and McPhee 1954; Key 1966; Campbell 1960).
Recent social science research has further refined our understanding of reinforcement. The stylized facts that emerge from this research are that campaign communications are reinforcing for different groups within the electorate, depending on their partisan orientation and prior levels of information and interest (Zaller 1992). Partisans, especially low information partisans, seem to be particularly susceptible to messages from their own party's candidates. Partisans seem resistant to messages that are sponsored by the opposing party or that favor the opposing party. Nonpartisans seem to behave somewhat differently. Low information nonpartisans seem to tune out political messages of all sort, but the high information nonpartisans seem to be particularly influenced by political communications.
But there is a fundamental problem with reinforcement arguments. People seem to throw away the public information that campaigns produce, or they use it inefficiently or not at all. Why would people use information this way? Do people just fool themselves?
Two sorts of explanations of the reinforcing effects of campaigns are commonly expressed. First, political psychologists argue that these effects arise because people use cognitive shortcuts, such as priming, cognitive dissonance, and acceptability biases. As a result, individuals believe information that agrees with their beliefs and ignore information dissonant with their beliefs (Iyengar and Kinder 1987; Popkin 1991; Zaller 1992). By this account, elections are predictable because individuals hear what they want to hear, which is consistent with their private assessments of the economy. Second, political economists and many political scientists argue that people use their private information and discount publicly provided information, such as news and ads, because it is not as credible or as reliable as each individual's personal and private experiences (Downs 1957; Key 1966; Fiorina 1981). By this account elections are predictable because aggregate indicators of national well-being are the sum of many individual experiences, and the political campaigns matter little to those judgments.
Each explanation has complementary strengths and weaknesses. Psychological accounts emphasize the importance of public information, such as ads and news, that all people see. These arguments arise out of attempts to understand the ups and downs that occur during elections, as many of the essays in this volume attest. Psychological explanations, though, have difficulty explaining the behavior of key groups within the electorate. In the advertising case, it is hard to think of a priming or framing story that explains the behavior of both partisans and independents and in particular why high information independents appear to move in the direction of the sponsor of the ads. Finally, these stories imply an eventual convergence of beliefs that is rarely observed. If all information were truly public and common, then over time we would observe complete consensus in public opinion on topics where all people agree about the objectives, such as greater economic growth or presidential job approval.
Economic explanations emphasize private information and provide a coherent way to understand why people have different beliefs and preferences. Individuals' beliefs and preferences are rooted in their disparate experiences in the economy and with the government, parties, and specific politicians. However, such explanations discount the publicly conveyed information. This hardly seems consistent with the basic conjectures of rational choice, namely, that individuals make decisions in the most efficient (lowest cost) way or with empirical reality (Erikson et al. 2003). Economic man seems consistent with the observed predictability of elections but not with the variation that occurs during campaigns.
This essay offers a simple theory of campaign information and learning that is entirely consistent with the basic observations made in empirical research. I build on the insights of political psychologists that campaigns fundamentally involve public information and of political economists that individuals' beliefs and preferences derive considerably from their private experiences and information. I assume that people use all information and that they value it equally. It is the mix of the public information from campaigns and the private information from personal experiences that creates the dynamics social scientists have observed. Private information generates heterogeneity in people's beliefs; public information creates coordinated shifts in those beliefs, which look identical to the patterns observed in survey data and experiments.
The account offered here raises what I consider to be a very important set of methodological concerns in the study of campaigns. What is an adequate model of voter reasoning and behavior? The intellectual move made here is to introduce public as well as private information into a simple model of economic voting and to show that many of the basic individual-level behavioral changes can be understood in these terms. The challenge for political economists, then, is to develop models that incorporate more subtle notions of information. Several empirical studies have done so and have found that both the information in the mass media and private information matter (Erikson et al. 1993). Little attention has been given to establishing the differential effects that these factors have on the total vote.
The challenge for political psychologists is much greater-to show that specific psychological phenomena actually influence electoral behavior. The important possibility exists that much of what cognitive psychologists have uncovered simply averages out in the process of counting votes. In short, many of the subtle processes of cognition and learning noted by survey and experimental researchers may be inessential to understanding election outcomes. To judge the importance of such factors requires that political psychologists have a clear standard or null model against which to judge the cumulative effects of psychological factors. The political economic model offered here provides a baseline against which to measure the importance of such effects.
In the first section of this essay, I develop an economic model of election. In the second section I introduce campaign information as public information and consider the possible effects that it may have. The concluding section offers some thoughts on how this model might be used to integrate questions of strategic voting, issue voting, and other topics touched on in other essays in this volume.
A Model of Economic Voting
Models of economic voting (e.g., Downs 1957; Kramer 1973; Fiorina 1981; Alesina, Londregan, and Rosenthal 1995) make four assumptions about individuals: (1) individuals pursue their self-interest, (2) they rely on private information, (3) they make judgments about the comparative competence of the parties or candidates rather than about policies, and (4) they use rules to update information.
The first assumption holds that individuals pursue their self-interest. They act in ways that will make them better off economically. A single indicator best captures the economic voter's interest: real personal disposable income. Kramer (1973) and subsequent researchers have shown that changes in real personal disposable income predict election outcomes best. Folded into this measure, of course, are inflation (real), taxes (disposable), wages, growth and employment (income), and other factors that the government may influence. Whatever the factors involved, all people are assumed to want and to benefit from the same thing: income growth.
How citizens pursue their interests in the electoral arena is more difficult to formulate. Economic voting usually implies that the electorate assesses the abilities or competence of the competing parties or politicians more than the policies they promise. People do not control the policy levers directly, nor do they appear to know much about the policies pursued by the government. Even if the candidates and parties could clearly promise specific policies, it is not evident that policies that make sense today, such as minimum wages or low interest rates, would be appropriate in two or four years. As a result, it is often supposed that economic voters really want the most able governance of the economy (DiIullio and Stokes 1993; Alesina, Londregan, and Rosenthal 1995).
The second assumption holds that people rely on private information, their own experiences in the economy and with the government, to form electoral judgments. Forecasting models simplify this further: people simply vote their pocketbooks, their personal finances (Kramer 1973; Tufte 1975). Pocketbook voting is surely a strong version of this assumption. More generally, private information consists of information that each individual observes independent of what others observe. Political economists commonly model private information as a signal, the realization of a random variable, that an individual receives (Alesina, Londregan, and Rosenthal 1995). Different people see different signals, different features of the economy. Some people are highly attentive to wages, others to Buctuations in the stock market. Importantly, unlike information that all people see, such as the unemployment rate, private information leads people to hold different beliefs about which party or politician governs best.
This assumption is subtly at odds with notions of information found in much political psychology and survey research. To most political scientists, an individual is informed if he or she knows a lot of facts about the candidates involved in a race, follows current affairs closely, or has weighed the issues carefully. Measures of knowledge include the number of correct identifications of prominent politicians, newspaper readership, and the number of answers to survey questions. Such measures capture the heterogeneity in levels of factual information within the electorate. By this metric, the public does not stack up well against the classical image of a democratic citizenry.
Private information, as it is formulated in economic models of voting, implies that each person receives the same amount of information. All persons live under the government for the same amount of time and, thus, draw inferences about politics and public affairs on the same amount of personal experiences. Am I better off now than I was four years ago? Or last year? The sorts of experiences people have is a different matter.
The third assumption defines the nature of the signal or private information that people receive. Private information may indicate that (1) the incumbent candidate or party governs well, (2) the incumbent candidate or party governs badly, or (3) public affairs are irrelevant to the person's life or there is insufficient information to make a judgment. In Kramer's original model of economic voting, an individual judges the incumbent party favorably if the individual's income increases substantially; he or she judges the incumbent party unfavorably if personal income decreases substantially. Modest changes in income give ambiguous information about the government's performance.
The fourth assumption holds that people update their beliefs over time. Following Fiorina (1981), I assume that individuals keep a running tally of their information. People express support for a party or politician if the sum of their past experiences with that party or politician is favorable. More sophisticated models suppose that voters revise their beliefs using Bayes's rule (Calvert 1985; Bartels 1993) or some other more complicated learning model. Such models lead to more subtle predictions than those offered here. The assumption of a simple moving average of information has the strength of simplicity. People try to reduce the effort they put into making decisions, and more sophisticated learning strategies require more information and effort.
These four assumptions describe voters as narrowly focused on their own interests and their own experiences. These are not the citizens of classical democratic theory. But the beauty of democracy is that it adds up these disparate experiences pretty effectively. Each individual doesn't have to know what others think and feel or how the entire macroeconomy works in order to reach a decision and voice his or her own preference.
The final piece to economic models of voting shows that public decisions indeed aggregate individuals' private information and judgments. In the terms developed in social choice theory, theories of economic voting describe a situation in which people have the same preferences (economic growth) but disparate beliefs. Condorcet's 1787 jury theorem addresses precisely this problem: majority rule will accurately measure the private information of jurors the larger the size of the jury. Contemporary political scientists have developed a rich literature relating to this theorem, applying it not only to juries but also to elections (Austen-Smith and Banks 1996; Fedderson and Pessendorfer 1997). Under the assumptions discussed here, the vote of the majority will accurately reflect the extent to which the population has prospered, as claimed by forecasting models.
(Continues...)
Excerpted from Capturing Campaign Effects Copyright © 2006 by University of Michigan . Excerpted by permission.
All rights reserved. No part of this excerpt may be reproduced or reprinted without permission in writing from the publisher.
Excerpts are provided by Dial-A-Book Inc. solely for the personal use of visitors to this web site.
