Do Voters in Emerging Democracies Hold Candidates to a Higher Standard?

Transcription

1 Do Voters in Emerging Democracies Hold Candidates to a Higher Standard? Austin Hart Assistant Professor of Quantitative Methods School of International Service American University ahart@american.edu April 6, 2015 Abstract Are citizens in emerging democracies policy voters, performance voters, or merely rationalizers? Recent studies in established democracies show that voters hold officials accountable for their performance but rarely evaluate officials along policy lines. Instead, voters often do the opposite: adopting their preferred party s issue positions as their own (rationalization). Is the same true in emerging democracies where partisan identity is more fluid? I investigate this possibility by examining voters responses to three information upheavals in Mexico and Brazil. Based on analysis of panel surveys that span these sudden changes in the information environment, I find evidence of policy and performance voting. The surprising implication is that, at least in these cases, voters applied more rigorous constraints to public officials than their counterparts in established democracies. 1

2 After three failed bids for the Brazilian presidency, perennial Workers Party (PT) nominee Luiz Inácio Lula da Silva earned a decisive victory in Notably, Lula s victory came on the heels of a marked change in his policy positions. In a radical departure from the views espoused in prior campaigns, Lula redefined himself as a new breed of economic pragmatist. He cut his hair, sat down with business leaders, and, in a very public Letter to the Brazilian People, outlined his revised, even reversed, stances on issues like trade openness and debt repayment. But to what effect? Despite Lula s extraordinary efforts, there is reason to believe that his policy turn had little impact on his electoral success. Notably, recent studies in Western democracies show that voters hold governments accountable for their performance in office but not for their policy choices. Even when an issue suddenly becomes salient, individuals neither reward nor punish public officials for their stance on that issue (e.g. Lenz 2009). In fact, citizens regularly do the opposite, adopting the positions espoused by their preferred candidate or party (e.g. Gerber & Huber 2009; Lenz 2012). A gloomy jobs report might move vote choices but not public stances on social welfare or same-sex marriage. Instead of jumping ship in search of a more likeminded party, individuals rationalize their affinities by shifting their issue positions. The generalizability of these findings to new or emerging democracies like Brazil, however, remains unclear. On its face, the apparent prevalence of policy rationalization in Western electorates is discouraging. Citizens in emerging democracies often lack enduring political dispositions (e.g. McCann & Lawson 2003), and they cannot rely on strong party organizations to structure long-term political discourse. This makes the intake and sorting of new information about politics difficult, and it raises fears about manipulation from above. Absent stable attitudes, individuals may lack the capacity to sanction government performance much less constrain its policy choices (Almond & Verba 1962). Lula s surprising policy shift, therefore, may have inspired the kind of rationalization common in Western democracies: Brazilians who already supported Lula moderated their own economic positions while the vote tally remained unchanged. Although this points to the absence of policy voting and performance voting in emerging democracies, I offer a more optimistic view. Specifically, I argue that voters in these contexts engage in both behaviors precisely because they lack enduring political identities. Partisan attachments may structure attitudes and facilitate information processing, but they also play a key role in motivating rationalization. Without a strong identity to defend, voters are more likely, not less, to evaluate candidates on substantive grounds. Determining whether citizens in emerging democracies behave as policy voters, performance voters, or mere rationalizers is an important pursuit as each behavior paints a very different picture of democratic governance. Policy voters constrain and guide the political process. If individuals select candidates based on their policy preferences, and if officials are re-election minded, elections reflect the substantive will of the people. Performance voting offers a view of elections as referenda. While 2

3 governments face little constraint in the choices they make, they must produce results or face ouster. Lastly, in the worst-case scenario of a rationalizing public, democracy is a smokescreen. Elections are battles of image and branding, and power is wielded from the top down. Voters select a candidate and simply fall in line with their directives, offering little constraint. If emerging democrats are rationalizers, it fundamentally alters our understanding of, for example, the recent rise of the Left in Latin America. I evaluate these competing visions by examining voters responses to three information upheavals : Lula s economic policy transformation in Brazil s 2002 presidential election, the Mensalão vote-buying scandal that rocked Lula s PT prior to Brazil s 2006 election, and Felipe Calderón s sudden turn to economic issues (including a memorable wave of ideological attacks against his opponent, Andrés Manuel López Obrador) in Mexico s 2006 presidential election. I focus on these upheavals for three reasons. First, these sudden and public shifts commanded and focused voters attention on a specific set of issues. Second, all three occurred during a longitudinal study of public opinion, allowing me to isolate the direction of causality between issue opinions and vote choices. Finally, psychological attachments to political parties in Mexico and Brazil are weaker and less stable on average than in long-established democracies. The next section briefly outlines the policy voting, performance voting, and rationalization hypotheses, highlighting the difficulty of isolating these behaviors empirically. I then detail the three information upheavals under study and present the analyses. Contrary to findings in Western democracies, I find clear evidence of both performance voting and policy voting. I conclude by considering the surprising implication that voters in emerging democracies hold candidates to a higher standard than voters in established democracies. Competing Visions of the Western Voter After more than 60 years of research, profound disagreement about the underlying determinants of vote choice persists. 1 Despite the dissensus, there is wide-ranging empirical support for both the policy-voting hypothesis and the performance-voting hypothesis. Studies of proximity voting (e.g. Hinich & Enelow, 1984), for instance, find that voters select candidates whose policy views are closest to their own. 2 Evidence in support of the performance voting hypothesis (e.g. Fiorina 1981) is even more plentiful. Most notably, economic voting models identify an almost ubiquitous link between the 1 For a review of the literature on voting, see: Kuklinski and Peyton, Directional models (Rabinowitz & MacDonald 1989) reach the same conclusion in support of policy voting. Studies like these find that voters select candidates based on their positions on trade (e.g. Johnston et al. 1992; Scheve & Slaughter, 2001), foreign policy (e.g. Aldrich et al., 1989), social security (Johnston et al. 2004), and broader ideological leanings (e.g. Gelman & King 1993), to name only a few. Moreover, it appears that politicians feel the policy pressures from below. In Western democracies, shifts in the popular policy mood are reflected by shifts in public policy (e.g. Shapiro & Jacobs 1989; Erikson et al. 2001; Stevenson 2001). 3

4 state of the economy and the electoral success of incumbent-party candidates. 3 Thus, voters may not be choosing the candidate whose policy views are most proximate, but they are at least sanctioning or rewarding officials for their stewardship. Despite the wealth of evidence, there is a far less flattering interpretation of these findings. The connections between policy preferences or performance evaluations and vote choice may simply reflect the endogeneity of these opinions to vote choices. That citizens adopt the policy views of their party or preferred candidate is well known (e.g. Carsey & Layman 2006; Sears & Lau 1983). Voters may not, for instance, consult their opinion on school vouchers and then select a candidate. Instead, they select a candidate on trivialities like good looks (Todorov et al. 2005) and party affiliation (e.g. Campbell et al. 1960). If that candidate then speaks against school vouchers, the individual becomes less supportive of the policy. The same is true for performance evaluations, especially on the economy (e.g. Bartels 2002; Wlezien et al. 1997; Gerber & Huber 2009, 2010). How is it that, on such a fundamental question, empirical research points in opposite directions? Lenz (2009, 2012) shows that past studies cannot differentiate policy or performance voting from rationalization. The typical study compares the correlation between issue positions and vote choice before and after an issue is made salient. If the correlation strengthens, the researcher assumes that individuals are changing their vote choice to reflect their position on the salient issue (policy or performance voting). However, the evidence is also consistent with the opposite interpretation, that individuals changed their position on the salient issue to reflect their prior vote choice. In the confines of conventional study designs, these behaviors are indistinguishable: the researcher observes an increase over time in the correlation between votes and either policy positions or performance evaluations. Discerning whether issue opinions cause vote choices or the reverse is a matter of guesswork. Once the methodological problems are overcome (through the analysis of panel data), a clearer picture of voting behavior emerges. Both observational and experimental studies support the performance-voting hypothesis: voters do evaluate candidates on a wide range of issues, including the economy, education, and the environment (Bartels 2006; Hart & Middleton 2014). Yet, test after test reveals no evidence of policy voting (Lenz 2009). Instead, when policy issues become salient, voters regularly engage in rationalization (Lenz 2012). Rationalization without Borders? To what extent do these findings travel outside of long-established, Western democracies? Policy voting and performance voting are well documented in emerging 3 For a review of the economic voting literature, see: Lewis-Beck and Stegmaier From a comparative perspective, see: Lewis-Beck and Stegmaier Beyond performance voting on the economy, studies also show that voters also evaluate candidates for their integrity (e.g. Johnston et al. 2004) and handling, as they see it, of issues like national defense (Ladd 2007). 4

5 democracies (e.g. Baker 2009; Weyland 1998; Remmer 1991). Yet Lenz s (2012) claim of observational equivalence is equally applicable. Moreover, there is reason to believe that rationalization should be more likely in emerging democracies, where political attitudes are often unstable. If citizens lack enduring dispositions, the foundations for policy and performance voting are shaky at best. Moreover, because attitudes are not entrenched, voters may be more likely to follow the lead of a charismatic candidate, backfilling issue positions and performance evaluations as they go. Of special concern is the weakness of political parties in emerging democracies. Longstanding party systems enable voters to think critically about politics and to process p0litical news effectively. Parties play a critical role in structuring political discourse. They shape the boundaries of public debates and define alternative viewpoints. When voters are exposed to new information about politics, then, they have a simple means of organizing it and connecting it with existing knowledge. For instance, simply learning a candidate s party affiliation clues us in to a wide range of policy positions she likely holds. Even the mention of a simple buzzword might allow a voter to lump a policy in with a larger bundle of political ideas. However, making sense of new information becomes more difficult where parties are unstable and where political discourse is not well defined. This suggests, then, that the task of constraining policymakers is even more difficult in emerging democracies than in their more established counterparts. Neither policy voting nor performance voting seem likely. Against this expectation, I argue that both policy and performance voting are more likely to occur in new or emerging democracies. While they lack the established party systems that help structure information flows, voters in new democracies also lack the psychological motivation to rationalize support for a candidate. Partisan attachment, so deeply embedded in established democracies, colors how voters see the political world, influences how they collect information, and especially what they do with that information (e.g. Lodge & Taber 2013). Partisans want to defend their team. So when new information uncovers an inconsistency between the party line and individual beliefs, voters are motivated to change their positions to get back in line (e.g. Kunda 1990). Without widespread partisanship, however, voters in emerging democracies are more open to persuasion (Greene 2011) but not out-and-out manipulation (McCann & Lawson 2003). As parties rise, crumble, and reorganize from election to election, voters are left to shop for a candidate that reflects their views. Moreover, because they do not identify strongly with a particular party, there is little reason to maintain party loyalty in light of new policy stances or poor performance. They can simply find a new candidate. I expect, therefore, that policy voting and performance voting are more prevalent in emerging democracies than in established democracies. Study Design and Data To evaluate the extent to which voters in emerging democracies evaluate candidates based on policy positions and performance in office, I evaluate priming and persuasion 5

6 effects in response to three information upheavals. Information upheavals occur when political campaigns or the media suddenly raise the salience of an issue. I focus on these dramatic events because voters in the real world are notoriously inattentive to politics (e.g. Zaller 1992), especially in emerging democracies. Yet upheavals capture the public eye, even if only in the short term, and thrust new political information onto voters. We expect opinions to change, even dramatically, in these contexts. Therefore, because information upheavals engender strong responses from voters, they present an ideal opportunity to differentiate performance and policy voting from rationalization. I focus specifically on voters responses to upheavals in the 2002 and 2006 Brazilian presidential elections and the 2006 Mexican presidential election. 4 In 2006, Mexican presidential candidate Felipe Calderón from the incumbent National Action Party (PAN) made a drastic change in his message. He fired his chief advisors, and, with the election less than four months away, left behind the identity he built in the PAN primaries as the conservative candidate with traditional values. He rebranded himself a competent economic leader who could continue Mexico s recent economic success. He also unleashed a memorable and aggressive attack against the leftist challenger, Andrés Manuel López Obrador (colloquially called AMLO). Calderón labeled AMLO a radical whose new economic model would plunge Mexico into a cycle of debt and decline. As Hart (2013) and Singer (2009) detail, the election, which had seen little economic rhetoric prior to Calderón s shift, suddenly became a battle over the state of Mexico s economy and whether an ideological leftist could be a competent economic steward. Consequently, Calderón overcame a double-digit deficit to win the election. Lula s policy reformation in Brazil s 2002 election was equally dramatic. At the behest of his media strategist Duda Mendonça, the thrice-defeated PT nominee and committed socialist broke away from his radical past. He reinvented both his image and his policy profile. He trimmed his scruffy beard and traded his tee-shirts for business suits. And, in an open and well-publicized Letter to the Brazilian People, Lula outlined the new economic policy positions that would define his campaign. Rather than criticize the developed world for taking advantage of Brazil through trade, he criticized it for not trading enough with Brazil. He backed away from his prior blanket-critiques of privatization and softened his stance on land reform. Most notably, he stopped advocating for default on Brazil s international debt. Instead he insisted that Brazil honor its financial obligations and repay its creditors in full. Despite suffering two of the most lopsided defeats in Brazil s history in the prior elections, Lula the moderate earned a first and decisive victory for the PT. 4 All three events have been covered extensively elsewhere. Here I provide a brief overview of the upheaval of interest, leaving behind the context in which it occurred and the events that may have precipitated it. For detailed information on the 2006 Mexican presidential election, see especially: Shirk (2009), Langston (2009), and Bruhn (2009). For information about Calderón s sudden shift in particular, see: Hart (2013). On the 2002 and 2006 Brazilian elections, including the Mensalão scandal, see, for instance: Baker (2009), Hunter (2010), Samuels (2004), Power and Taylor (2011), and Rennó (2011). 6

7 Lula s PT had long been known as the party above the corruption that characterized other parties and, notably, the government of President Fernando Henrique Cardoso. Yet the party with the clean hands was rocked in 2005, just a year before Lula s reelection bid, with the revelation of a massive vote-buying scheme. The Mensalão scandal erupted on June 6, 2005 when Deputy Roberto Jefferson revealed in a newspaper interview that the PT was paying large, monthly sums to buy support for legislative votes. The scandal quickly broadened, engulfing much of the senior PT leadership. Though President Lula faced no direct accusations, sustained and farreaching corruption had emerged on his watch. If voters select candidates based on policy positions or performance in office, these information upheavals should have two potential effects: priming and persuasion. Priming refers to a change in the criteria voters use to evaluate candidates, and it is driven by issue salience (e.g. Iyengar & Kinder 1987; Iyengar et al. 1984). If a voter is exposed to a wave of adds about the economy, for instance, as in Mexico 2006, she comes to evaluate the incumbent or incumbent-party candidate based on his handling, as she sees it, of the economy. If she believes the economy was doing well, she responds to the new information by adopting a more positive view of the president. Of course, new information might also persuade voters, for instance, that the ruling party is corrupt. If voters constrain governments performance and policy choices, this change in opinion should then drive change in candidate evaluations. Information upheavals, therefore, produce a two-step, persuasion effect: changing opinions on the salient issue and then changing votes. These effects should be pronounced (unless voters are rationalizers) and targeted to the small set of issues thrust into the limelight. Note that priming and persuasion effects are quite similar in that both are responses to a spike in the salience of a particular issue. Voters receive political stimuli and react in a way that causes them to update their evaluations of political leaders. However, they differ in an important way. Priming occurs when the stimulus raises the salience of an issue attitude but where the attitude itself remains stable. The voter reevaluates the candidate because she is reminded that the economy is doing poorly, not because the stimulus convinces her of that. Persuasion occurs when the stimulus raises the salience of an issue attitude and provides an impetus for a change in that attitude. The voter reevaluates the candidate because the stimulus convinces her that the party is corrupt. Whether an upheaval induces priming effects, persuasion effects, or both, depends on the upheaval itself. Where the upheaval is revelatory, I expect persuasion. Where the upheaval simply changes the terms of the debate, I expect priming. Of course, upheavals are multifaceted, and they may contain both. However, I do not expect to observe both priming and persuasion effects on the same issue at the same time. Table 1 presents an overview of each case, the timing of the upheaval, the specific issues made salient in the upheaval, and the data used evaluate the extent to which voters were either primed or persuaded as a result of the upheaval. 7

8 Table 1. Summary of the Cases, Data, and Target Issues Cases: 3 Information Upheavals Timing Dataset Panel Waves Target Issues Brazil 2002: Presidential hopeful Lula makes a pragmatic shift in economic policy. He outlines his surprising moderate swing in a Letter to the Brazilian people. June 23, 2002 Two-City, Six- Wave Panel Survey, Brazil 1: Apr : Aug : Oct (p) - Privatization - Land reform - Free trade - Political ideology Brazil 2006: The 2005 Mensalão vote-buying scandal breaks in advance of the 2006 election, revealing that the ruling PT, built on an image of honesty, is corrupt. June 26, 2005 Two-City, Six- Wave Panel Survey, Brazil 4: May : Jul : Oct (p) - Corruption Mexico 2006: Presidential hopeful Felipe Calderón rebrands himself as a competent economic leader and attacks AMLO as a radical leftist who endangers Mexico s growth. March, 2006 Mexico 2006 Panel Study 1: Oct : Apr : Jul (p) - Political ideology - Econ. evaluation - Econ. competence Note: (p) signifies a post-electoral wave of interviews. Analysis and Results This section proceeds in two parts. The first evaluates possible persuasion effects, and the second evaluates priming in the three information upheavals. For each, I describe the empirical test and then present the estimates. Care to reconsider? Persuasion To test for persuasion, I following the procedure outlined by Lenz (2012) and evaluate the effect of opinion change on issue x during the upheaval on future change in vote choice, y. 5 This necessitates three waves of panel data, one pre-upheaval interview and two post-upheaval interviews. I model this prior persuasion effect as: y! = π! x! x! + b! x! + π! (y! y! ) + ρ! y! + u!, where y! is vote choice at wave 3 and π! is the effect of opinion change on issue x from wave 1 to wave 2 controlling for prior issue opinion, x!, the change in vote choice between waves 1 to 2, (y! y! ), and wave 1 vote choice, y!. If the estimate of π! is statistically significant, there is evidence of either policy voting or performance voting, depending on the issue. This is a conservative test of persuasion. Yet I choose it because it ensures that the change in issue opinion preceded a change in vote choice (ruling out rationalization as an alternative explanation). Given the conservative nature of the test, significant results provide strong support for the hypothesis that voters in emerging democracies are both policy voters and performance voters. 5 Finkel (1993) and Greene (2011) use a similar test to evaluate campaign persuasion. 8

9 For clarity, Table 2 (above) presents OLS estimates of the prior persuasion effects in the wake of Lula s radical moderation in Again, the coefficient estimates for the change scores represent the effect of opinion change during the upheaval on future change in vote choice. In contrast to studies in established democracies, I find clear evidence of policy voting. Voters who changed their opinion about the issues in the wake of an upheaval go on to reevaluate the candidate. In this case, Lula s pragmatic shift in policy and the surrounding debate drove voters to revise their own opinions on the issues and then reevaluate Lula as a candidate between August and October. Specifically, I find that shifts in ideological self-placement and support for the privatization of services and industry translated into new opinions of Lula (π!"#$ = and π!"#$ = 0.062). It is unlikely that I observe these relationships by chance alone (p < 0.05, two-tailed). By contrast, I find no evidence of persuasion on the issue of land reform. Whether or not the upheaval primed these opinions instead is taken up in later analysis. Table 2. Estimated Prior Persuasion Effects, Brazil 2002 DV: Vote for Lula (Oct 2002: Post Elec) Variable Coef. SE Δ Lula vote intention, Aug Apr * (0.028) Lula vote intention, Apr * (0.034) Δ Ideology, Aug Apr * (0.035) Ideology, Apr * (0.038) Δ Privatization, Aug Apr * (0.032) Privatization, Apr * (0.033) Δ Land reform, Aug Apr (0.031) Land reform, Apr * (0.034) Δ Lula: intelligent, Aug Apr (0.055) Lula: intelligent, Apr (0.059) Δ Lula: leader, Aug Apr * (0.050) Lula: leader, Apr * (0.058) Δ Lula: honest, Aug Apr (0.057) Lula: honest, Apr (0.060) Constant * (0.052) N = * p < Ordinary least squares estimates. All variables are scaled to range from 0 to 1 and recoded so that coefficient estimates should be positive. 6 For the sake of simplicity, I estimate all of the models using OLS, even when the dependent variable is binary. Logit estimates and, in the Mexican case, multinomial logit estimates yield identical substantive conclusions. 9

10 Figure 1 presents these results with estimates of the prior persuasion effects for the other cases under study (full model estimates for the other cases are given in the appendix). The black circles represent the point estimates for the target issues and the bars surrounding them are 95% confidence intervals. In addition to the evidence of policy voting in the 2002 Brazilian election, I find strong support for the performancevoting hypothesis in the other cases. While voters in Mexico were not persuaded to adopt new ideological stances, Felipe Calderón s economic turn did move voters to reassess the relative economic competence of the candidates. Consequently, they used the new information to adjust their overall preference for the leading candidates. Those who were convinced that Calderón was the more competent economic steward came to rate him more favorably in contrast with AMLO. Figure 1. Prior Persuasion Effects by Issue Type BR 02: Ideology MX 06: Ideology BR 02: Priva3za3on BR 02: Land Reform BR 06: Corrup3on MX 06: Handle Econ Performance Policy PosiGons Effect of prior opinion change on change in vote /candidate evaluagon Similarly, the shocking revelation of widespread corruption in the Brazilian PT led to dramatic changes in opinions about how well President Lula was handling corruption. The hit to his reputation as the leader of the party with the clean hands also translated into a hit to his vote tally. While this was unfortunate for Lula, it is an encouraging sign for voters in emerging democracies. Struck by unfolding political events, they took in new information, shifted their opinions accordingly, and then rewarded or sanctioned elected officials. This is a far cry from the consistent evidence of rationalization observed in longstanding Western democracies. 10

11 Let s talk about something else. Priming I test for priming effects in these information upheavals by comparing the effect of preupheaval issue opinion on pre-upheaval candidate evaluations against the effect of preupheaval issue opinion on post-upheaval candidate evaluations. 7 If the weight voters attach to pre-approval opinion increases in response to the upheaval, it signifies a priming effect: voters shifted their candidate evaluation to match their prior opinion on the issue. For this two-wave test, I specify the wave 1 salience weight for issue x, (b!,! ) as: y! = b!,! x! + u!. Similarly, I specify the wave 2 salience weight for the wave 1 measure of x, (b!,! ) as: y! = b!,! x! + u! The magnitude of the priming effect for issue x, θ!, is the difference between the salience weights: θ! = b!,! b!,!. If voters are engaging in policy voting or performance voting after an upheaval makes an issue salient, this difference should be positive and significantly different from zero. Table 3 presents the full estimates of the salience weights for each study wave as well as the estimated priming effects in the 2006 Mexican presidential election. As with the persuasion tests, the results provide clear support for the policy voting and performance voting hypotheses. Calderón s sudden turn to the economy primed evaluations of Mexico s past economic performance (θ!"#$ = = 0.066). The barrage of attacks he levied against López Obrador s new economic model, also primed political ideology (θ!"#$ = = 0.070). Both effects are statistically significant and demonstrate that, in the wake of the information upheaval, individuals changed their comparative assessment of the candidates to reflect their ideological leanings and their retrospective economic evaluations. Figure 2 presents complete results of the priming tests. As with the prior persuasion tests, the evidence supports both the policy voting and performance voting hypothesis. In fact, I find significant priming effects for 4 of the 5 policy issues tested. Only ideological leanings in Brazil s 2002 election failed the test. It is important to note that this does not imply that ideology was unrelated to vote choice. In fact, the estimate for ideology was statistically significant in both the wave 1 and wave 2 models. Rather, this is evidence that the issue weight did not increase over time. As discussed above, there is also evidence of economic priming in Mexico s 2006 election. One important piece to note is that there is evidence of either persuasion or priming for each of the issues targeted by the upheaval. However, only for one issue 7 This builds on Lenz s (2012) two-wave test for priming effect. 11

12 Table 3. Estimated Issue Weights and Priming Effects, Mexico 2006 Variable (October Measures) DV: Calderón Thermometer Adv. Priming Effect (θ) Oct Apr b (x,apr) b (x,oct) A Ideology * * * (0.019) (0.023) (0.030) Economy (past year) * * (0.019) (0.023) (0.030) Calderón Econ. Competence * * (0.024) (0.030) (0.039) PAN identification * * (0.014) (0.017) (0.022) PRI identification (0.014) (0.017) (0.022) PRD identification * * (0.013) (0.017) (0.021) Constant (0.027) (0.033) N = 977. Ordinary least squares estimates with standard errors in parentheses. * p < 0.05 A: The priming effects are estimated directly by stacking the October and April data and interacting the independent variables with an indicator for the post-upheaval, April wave. The coefficient estimates on the interaction terms equal the difference in issue weights from October 2005 to April (privatization in Brazil 2002) is there evidence of both effects. This is consistent with my earlier expectation. Where the upheaval caused strong shifts in opinions, it is hard to imagine how the initial opinion could become increasingly predictive of later candidate evaluations. Similarly, where the information upheaval raised the salience of an issue without moving individuals issue positions, there is little reason to expect persuasion. One challenge for future research, then, is to identify these differences ex ante. Conclusion This study evaluated the extent to which voters in emerging democracies evaluate candidates based on their policy choices and performance in office. While recent studies in long-established democracies conclude that voters do hold governments responsible for their performance but not their policy choices, I argued that we should expect to observe both policy voting and performance voting in emerging democracies. Although scholars have raised concerns in the past that the instability of political attitudes in these countries would hinder voters capacity to constrain governments, I contend that the absence of strong partisan identities (which motivate rationalizing) allows voters to interpret new information critically and sanction or reward governments accordingly. 12

13 Figure 2. Priming Effects by Issue Type BR02: Ideology BR02: Priva3za3on BR02: Land Reform BR02: Free Trade MX06: Ideology MX06: Econ Eval MX06: Handle Econ BR06: Corrup3on Perfromance Policy PosiGons Difference in Issue Weights Change in the effect of prior opinions on vote choice/candidate evalua3on Based on the analysis of priming and projection effects in Mexico and Brazil, I find clear evidence in support of both the policy voting and performance voting hypotheses. In three cases of information upheaval, in which the media or political candidates suddenly raised the salience of an issue, voters changed their candidate preferences to reflect their opinion on the issue. This was evident in tests of priming and persuasion effects. Unlike past studies, which could not rule out rationalization as an alternative explanation for their findings, the tests presented here confirm that the causal arrow runs from issue positions and performance evaluations to vote choice, not the reverse. These findings lead to a surprising conclusion: voters in emerging democracies provide a greater level of constraint on policymakers than do voters in long-established democracies. By no means does this imply that citizens in emerging democracies approximate Jefferson s idealized yeoman farmers. Nor does it imply that emerging democracies enjoy a higher overall quality of governance. However, it does demonstrate that elections in emerging democracies are about more than the economy, stupid, and about more than popularity and persona. Candidates are elected, at least in part, for the substantive positions they take. 13

14 Bibliography Aldrich, John H., John L. Sullivan, and Eugene Borgida Foreign Affairs and Issue Voting: Do Presidential Candidates Waltz before a Blind Audience? American Political Science Review 83(1): Almond, Duane F., and Sidney Verba The Civic Culture. Boston: Little, Brown. Baker, Andy The Market and the Masses in Latin America. New York: Cambridge University Press. Bartels, Larry M Beyond the Running Tally: Partisan Bias in Political Perceptions. Political Behavior 24(2): Priming and Persuasion in Presidential Campaigns. In Capturing Campaign Effects, ed. Henry E. Brady and Richard Johnston. Ann Arbor, MI: University of Michigan Press. Bruhn, Kathleen López Obrador, Calderón, and the 2006 Presidential Campaign. In Consolidating Mexico s Democracy, ed. Jorge I. Domínguez, Chappell Lawson, and Alejandro Moreno. Baltimore, MD: The Johns Hopkins University Press. Campbell, Angus, Phillip E. Converse, Warren E. Miller, and Donald E. Stokes The American Voter. New York: John Wiley & Sons, Inc. Carsey, Thomas M., and Geoffrey C. Layman Changing Sides or Changing Minds? Party Identification and Policy Preferences in the American Electorate. American Journal of Political Science 50(2): Erikson, Robert, James Stimson, and Michael MacKuen The Macro Polity. Cambridge: Cambridge University Press. Finkel, Steven E Reexamining the Minimal Effects Model in Recent Presidential Campaigns. Journal of Politics 55(1): Fiorina, Morris Retrospective Voting in American National Elections. New Haven: Yale University Press. Fiorina, Morris Retrospective Voting in American National Elections. New Haven: Yale University Press. Gelman, Andrew and Gary King Why are American presidential election campaign polls so variable when votes are so predictable? British Journal of Political Science 23(4): Gerber, Alan S. and Gregory A. Huber Partisanship and Economic Behavior: Do Partisan Differences in Economic Forecasts Predict Real Economic Behavior? American Political Science Review 103(3): Partisanship, Political Control, and Economic Assessments. American Journal of Political Science 54(1): Greene, Kenneth Campaign Persuasion and Nascent Partisanship in Mexico s New Democracy. American Journal of Political Science 55(2): Hart, Austin Can Candidates Activate or Deactivate the Economic Vote? Evidence from Two Mexican Elections. Journal of Politics 75(4):

15 Hart, Austin and Joel A. Middleton Priming under Fire: Reverse Causality and the Classic Media Priming Hypothesis. Journal of Politics 76(2): Hinich, Melvin J., and James M. Enelow The Spatial Theory of Voting: An Introduction. Cambridge: Cambridge Univerisity Press. Hunter, Wendy The Transformation of the Workers Party in Brazil, New York: Cambridge University Press. Iyengar, Shanto & Donald R. Kinder News That Matters. Chicago: The University of Chicago Press. Iyengar, Shanto, Donald R. Kinder, Mark D. Peters and Jon A. Krosnick The Evening News and Presidential Evaluations. Journal of Personality and Social Psychology 44(4): Johnston, Richard, Andre Blais, Henry E. Brady, and Jean Crête Letting the People Decide: Dynamics of a Canadian Election. Stanford, CA: Stanford University Press. Johnston, Richard, Michael G. Hagen, Kathleen Hall Jamieson The 2000 Presidential Election and the Foundations of Party Politics. New York: Cambridge University Press. Kuklinski, James H. and Buddy Peyton Belief Systems and Political Decision Making. In The Oxford Handbook of Political Behavior, ed. Russell J. Dalton and Hans-Dieter Klingemann New York: Oxford University Press. Kunda, Ziva The Case for Motivated Reasoning. Psychological Bulletin 108(3): Ladd, Jonathan Predispositions and Public Support for the President during the War on Terrorism. Public Opinion Quarterly 71(4): Langston, Joy The PRI s 2006 Presidential Campaign. In Consolidating Mexico s Democracy, ed. Jorge I. Domínguez, Chappell Lawson, and Alejandro Moreno. Baltimore, MD: The Johns Hopkins University Press. Lenz, Gabriel S Learning and Opinion Change, Not Priming: Reconsidering the Priming Hypothesis. American Journal of Political Science 53(4): Follow the Leader? How Voters Respond to Politicians Policy and Performance. Chicago, IL: University of Chicago Press. Lewis-Beck, Michael S. and Mary Stegmaier Economic Models of Voting. In The Oxford Handbook of Political Behavior, R.E. Goodin (ed.). Oxford: Oxford University Press The Economic Vote in Transitional Democracies. Journal of Elections, Public Opinion, and Parties 18(August): Lodge, Milton and Charles S. Taber The Rationalizing Voter. New York: Cambridge University Press. McCann, James A and Chappell Lawson An Electorate Adrift? Public Opinion and the Quality of Democracy in Mexico. Latin American Research Review 38(3): Power, Timothy J., and Matthew M. Taylor Corruption and Democracy in Brazil: The Struggle for Accountability. South Bend: University of Notre Dame Press. 15

16 Rabinowitz, George, and Stuart Elaine Macdonald A Directional Theory of Voting. American Political Science Review 83(1): Remmer, Karen L The Political Impact of Economic Crisis in Latin American in the 1980s. American Political Science Review 85(3): Rennó, Lucio R. Corruption and Voting. In Corruption and Democracy in Brazil: The Struggle for Accountability ed. Timothy J. Power and Matthew M. Taylor. South Bend: University of Notre Dame Press. Samuels, David From Socialism to Social Democracy: Party Organization and the Transformation of the Workers Party in Brazil. Comparative Political Studies 37(9): Scheve, Kenneth F., and Matthew J. Slaughter What Determines Individual Trade-Policy Preferences? Journal of International Economies 54(2): Sears, David O., and Richard R. Lau Inducing Apparently Self-Interested Political Preferences. American Journal of Political Science 27(2): Shapiro, Robert, and Lawrence Jacobs "The Relationship Between Public Opinion and Public Policy." In Political Behavior Annual, ed. Samuel Long. Boulder: Westview. Shirk, David Choosing Mexico s 2006 Presidential Candidates. In Consolidating Mexico s Democracy, ed. Jorge I. Domínguez, Chappell Lawson, and Alejandro Moreno. Baltimore, MD: The Johns Hopkins University Press. Singer, Matthew Defendamos lo que hemos logrado. Política y Gobierno 15(Special Issue on 2006 Mexican Election, Fall): Stevenson, Randolph T The Economy and Policy Mood: A Fundamental Dynamic of Democratic Politics? American Journal of Political Science 45(3): Todorov, Alexander, Anesu N. Mandisodza, Amir Gorn, and Crystal C. Hall Inferences of Competence from Faces Predict Election Outcomes. Science 308(5728): Weyland, Kurt Peasants or Bankers in Venezuela? Presidential Popularity and Economic Reform Approval, Political Research Quarterly 51(2): Wlezien, Christopher, Mark Franklin and Daniel Twiggs Economic Perceptions and Vote Choice: Disentangling the Endogeneity. Political Behavior 19(1): Zaller, John The Nature and Origins of Mass Opinion. Cambridge: Cambridge University Press. 16

17 Table A1. Estimates of Prior Persuasion Effects, Mexico 2006 DV: Calderón Therm. Adv. (July 2006: Post-Election) Variable Coef. SE Δ Calderón Therm. Adv., Apr Oct * (0.041) Calderón Therm. Adv., Oct * (0.052) Δ Ideology, Apr Oct (0.026) Ideology, Oct (0.028) Δ Calderón Econ. Competence Adv., Apr Oct * (0.036) Calderón Econ. Competence Adv., Oct * (0.045) Δ PRI Identification Apr Oct (0.021) PRI Identification, Oct (0.020) Δ PRD Identification, Apr Oct (0.021) PRD Identification, Oct * (0.021) Δ PAN Identification, Apr Oct * (0.020) PAN Identification, Oct * (0.020) Constant (0.041) N = 707. * p < Ordinary least squares estimates. Estimates for the change scores are estimates of the prior-persuasion effect. 17

18 Table A2. Estimates of Prior Persuasion Effects, Brazil 2006 DV: Vote for Lula (Oct. 2006: Post Elec) Variable Coef. SE Δ Lula Vote Intention, Jul 06 May * (0.036) Lula Vote Intention, May * (0.043) Policy Positions Δ Ideology, Jul 06 May (0.045) Ideology, May * (0.045) Performance/Character Evaluations Δ Lula: Handling Corruption, Jul 06 May * (0.048) Lula: Handling Corruption, May * (0.078) Δ Lula: Intelligence, Jul 06 May (0.069) Lula: Intelligence, May (0.085) Δ Lula: Leader, Jul 06 May (0.071) Lula: Leader, May (0.091) Δ Lula: Honest, Jul 06 May (0.077) Lula: Honest, May (0.100) Constant * (0.066) N = 583. * p < Ordinary least squares estimates. Estimates for the change scores are estimates of the prior-persuasion effect. 18

19 Table A3. Estimates of Priming Effects, Brazil 2002 DV: Lula Vote Intention (1) Apr 2002 (2) Aug Priming Effect (θ) b (2) b (1) A Apr Aug Ideology (Apr) * * (0.024) (0.027) (0.036) Privatization (Apr) * * * (0.022) (0.025) (0.034) Free Trade (Apr) * * (0.026) (0.029) (0.039) Land Reform (Apr) * * * (0.023) (0.025) (0.034) Lula: Intelligent (Apr) * * (0.036) (0.040) (0.054) Lula: Leader (Apr) * * (0.035) (0.039) (0.052) Lula: Honest (Apr) * * (0.037) (0.041) (0.056) Constant (0.035) (0.039) N = Ordinary least squares estimates with standard errors in parentheses. * p < 0.05 A: Due to rounding, differences may not reflect the exact difference between the estimates displayed in columns 1 and 2. 19

20 Table A4. Estimates of Priming Effects, Brazil 2006 DV: Vote for Lula (1) May 2004 (2) July 2006 Priming Effect (θ) b (2) b (1) A Jul May Lula: Handling Corruption (May) * * (0.068) (0.071) (0.099) Lula: Intelligence (May) * * (0.070) (0.072) (0.100) Lula: Leadership (May) * * (0.070) (0.072) (0.101) Lula: Honesty (May) * * (0.082) (0.085) (0.118) Ideology (May) * (0.038) (0.040) (0.055) Constant * (0.065) (0.068) N = 960. Ordinary least squares estimates with standard errors in parentheses. * p < A: Due to rounding, differences may not reflect the exact difference between the estimates displayed in columns 1 and 2. 20