One-Night Stands and Long-Term Commitments. Vote Intention and Presidential. Approval for Sebastián Piñera in Chile,

One-Night Stands and Long-Term Commitments. Vote Intention and Presidential Approval for Sebastián Piñera in Chile, 2009-2014 Patricio Navia patricio.navia@nyu.edu New York University/Universidad Diego Portales Lucas Perelló perel531@newschool.edu The New School/Universidad Diego Portales Paper prepared for the 2016 ECPR General Conference, Prague, September 7-10, 2016 September 1, 2016 (8922 words) Shifts from vote intention to presidential approval under favorable economic conditions cannot be solely explained by economic vote variables. Distinguishing between core constituencies and swing supporters helps explain fluctuations in presidential approval. In 2010, Sebastián Piñera (2010-2014) became the first right wing President of Chile since democracy was restored in 1990, ending 20 years of consecutive center-left Concertación governments. With logit models and predicted probabilities, based on electoral and presidential approval polls, we test two competing hypotheses. First, that Piñera won as a result of voters punishing the Concertación. Second, that his government s policies alienated the swing supporters that helped him win. We find evidence for both. Piñera s approval among younger Chileans, the upper and middle class and the nonidentified in the political scale swing supporters declined immediately after he took office before the 2011 student protests. Support for Piñera fell across the board after the student protests, except among conservatives his core constituency and centrists. Neither favorable economic conditions nor government policies explain the rapid decline in Piñera s approval after he took 1

office. Voters who wanted to punish the Concertación and voted for Piñera abandoned him as soon as he became president. Key words: presidential approval, vote intention, economic vote, retrospective vote, prospective vote, ideological orientation, Chile 2

One-Night Stands and Long-Term Commitments. Vote Intention and Presidential Approval for Sebastián Piñera in Chile, 2009-2014 In January of 2010, Sebastián Piñera (2010-2014) became the first right wing President of Chile since the end of military rule in 1990 and after two decades of center-left Concertación governments. Two years later, Piñera reached unprecedented low approval ratings. In 2011, the unequal access and high cost of higher education mobilized thousands of students against his government. The spontaneous barricades, tear gas and widespread support for the student demands brought back memories of the protests held in 1980s against the dictatorship. After a protest on August 4 th, Piñera allegedly confided to his advisors: How did we let this get out of control? (La Tercera, March 8, 2014). After peaking at 63 percent in October 2010, with the successful rescue of 33-trapped miners, Piñera s approval fell to 23 percent in December of 2011. Piñera s low approval has been associated both to the student protests and to his alleged accidental victory caused by a punishment vote against the Concertación. With logit models and predicted probabilities, based on pre-electoral and presidential approval polls, we test those two hypotheses. We find evidence for both. Piñera s approval among younger Chileans, the upper and middle class and the non-identified in the political scale swing supporters declined immediately after he took office before the 2011 student protests. Support for Piñera fell across the board after the student protests, except among conservatives his core constituency and centrists. Neither favorable economic conditions nor government policies explain the rapid decline in Piñera s approval after he took office. Voters who wanted to punish the Concertación and voted for Piñera abandoned him as soon as he became president. 3

Presidents who enjoy high approval in their honeymoons and lose support as their terms progress are a common occurrence in democracies. In countries with low levels of ideological identification, weak party systems, or in nations that experience drastic shifts in economic conditions, presidential approval volatility should not be surprising. However, in countries with institutionalized party systems and economic stability, presidential approval volatility warrants an explanation. In studying countries with stable party systems that undergo favorable economic conditions, we can identify the determinants of rapidly fluctuating presidential approval. By identifying swing supporters who threw their support behind a candidate to punish the previous government from voters who abandoned the president as a result of the policies the government implemented. Presidential approval has varied in Chile since 1990, but no prior president had such a huge fluctuation in approval, under favorable economic conditions, as President Piñera. Given Chile s stable multiparty system, with a left and a rightwing coalitions, changes in presidential approval under favorable conditions cannot be explained by ideology or economic vote variables alone. Shifts among swing supporters should explain presidential approval volatility. In what follows, we discuss the determinants of voting preferences and presidential approval, advancing an explanation of presidential approval volatility. We then examine the context of the 2009-10 presidential election and summarize the main events of Piñera s administration. With electoral and presidential approval polls, we present logit models and predicted probabilities to show continuity and change in his electoral and presidential approval support base. Determinants of voting preferences and presidential approval 4

Vote intention and presidential approval can be explained by short, medium or long-term variables. Three major models are associated to each set of variables (Bartels 2011). The Columbia Model, inspired on sociological explanations based on Lipset and Rokkan (1967), explains behavior by long-term variables cleavages such as race, social class, religion and other permanent traits (Lazarsfeld, Berelson, and Gaudet 1944, Lau and Redlawsk 2006). Under the Columbia Model, voting patterns are stable and predictable. The Michigan, or socialization, model uses a socialpsychological approach based on medium-term preferences (Campbell et al. 1960, Converse 1964, Jennings and Niemi 1968, Ventura 2001, Richardson 1991). Voters are politically socialized by parties or societal organizations and have stable, but not immutable, preferences. New socialization processes can alter electoral preferences (Bartels 2011). The Columbia and Michigan models predict stable electorates. Variations in electoral preferences result from changes in the composition of society or in socialization processes. These explanations are less appropriate to account for short-term fluctuations in electoral behavior. However, in a new democracy that undergoes its first alternation in power such as Chile in 2010 changes in electoral preferences may be observed when voters experience firsthand the effects of having the former opposition in power. A third model, Rochester, adopts an economic vote or rational choice approach based on shortterm preferences (Downs 1957, Duch and Stevenson 2008, Lewis-Beck and Stegmaier 2007, Erikson, MacKuen, and Stimpson 2002). Economic performance explains electoral behavior. When there is economic growth, low inflation and low unemployment, the incumbent candidate/party has more probabilities of winning (Duch and Stevenson 2008, Erikson, MacKuen, 5

and Stimpson 2002, Duggan and Fey 2005). However, macroeconomic indicators do not always coincide with people s opinions (Morgan 2003). Public opinion polls allow us to assess people s economic perceptions. Citizens may vote in function of socio-tropic or ego-tropic perceptions and using prospective or retrospective reasoning. Economic vote approaches assume that voters attribute to incumbents the responsibility for managing the economy, judge economic conditions and vote accordingly (Lewis-Beck and Stegmaier 2007). But rationality is not exclusively reduced to economic concerns. When a country undergoes economic stress under a working class party government, a working class-minded voter is being rational when supporting the party she was born or socialized into (Knutsen 2007). Yet, most rational choice approaches tend to use economic variables to frame voters utility functions precisely because, even if at the margin, different elections produce different results. Thus, some pivotal voters are influenced by short-term variables. Early economic vote studies used inflation, unemployment, and economic growth to suggest that voters respond to fluctuations in the economy (Kramer 1971). Later, Fiorina (1981) and Lewis- Beck (1988) combined economic indicators with institutional variables to account for vote intention. Systematic cross-national analysis of individual-level survey data that addresses how economic voting varies across political contexts has reported differences in the way voters condition their vote choice on the economy in different political and economic contexts (Duch and Stevenson 2008, 8). 6

When voters are unsure about which parties are responsible for economic policy making, they are less able to punish or reward politicians (Powell and Whitten 1993). When voters cannot cast an economic vote, as they do not know whom to hold accountable for the economy, the relationship between economic performance and electoral outcomes is more difficult to prove (Duch and Stevenson 2005, 2). Tucker et al. (2006) evaluate the effect of the economy on vote intention by supposing that voters have no information about how parties are likely to handle the economy. Their Referendum Model entails that people concerned with the state of the economy are more likely to punish the government than people who think the economy is doing well, and that the worse economic conditions are in a particular area, the more likely that more people will be dissatisfied with the state of the economy. The economic vote is also applied to studies of presidential approval. Presidential approval polls are used as a proxy for electoral support (Erikson, MacKuen, and Stimpson 2000, Erikson, MacKuen, and Stimpson 2002, Canes-Wrone and de Marchi 2002, Nadeau et al. 1999, Edwards 1997). When the economy is doing well, presidential approval is likely to be higher and the incumbent party is more likely to be re-elected. Under certain institutional party systems, presidential approval can also be used as an independent variable to predict vote intention for the incumbent party. Nannestad and Paldam (1994) identified the widely popular Vote-Popularity (VP) function. When the party system is more fluid, the VP-function is less effective to predict vote intention (Nannestad and Paldam 1997). Though economic vote models allow for fluctuations in presidential approval, they should respond to changes in economic conditions or in perceptions about the economy. When economic 7

conditions and perceptions are stable, there should be no significant fluctuation in presidential approval. In discussing presidential approval volatility, Kriner and Schwartz (2009) argue that developments that conflict with a group s partisan predispositions increase the volatility of approval (2009: 609). Those developments affect groups differently. Here, we contend that hardliners are more immune to those shocks while swing supporters will show higher presidential approval volatility under those circumstances. In Latin America, the presence of economic voting has been well documented. Economic crises undermine support for incumbents and induce electoral volatility (Remmer 1991). Also, the greater the fragmentation of the party system, the greater the effect of economic voting on incumbent candidate/party vote (Remmer 1991). Roberts and Wibbels (1999) found pro-incumbent vote shifts. Stokes and her collaborators analyzed public support for economic reforms in the 1990s, pointing to economic performance as independent variables that explain presidential approval (Stokes 2001). Studying countries with low party system institutionalization, Gélineau (2007) reports that institutional characteristics constrain the ability of voters to hold the incumbent party responsible for economic performance. Lewis-Beck and Ratto, with polls from 12 countries, showed that governments are rewarded or punished according to the economic performance they command (Lewis-Beck and Ratto 2013, 489). Kitschelt et al. (2010), condition vote intention to the types of linkages that parties develop with their electoral bases. Carlin and Singer (2015) confirm the usefulness of the economic vote but qualify it by particular traits and conditions in different Latin American countries. 8

At the country level, Remmer and Gélineau (2003) and Cataife (2011) showed the importance of the economic vote in Argentina. Presidential approval for the outgoing president in Mexico helps explain vote intention for the incumbent party candidate (Romero 2009). The effect of economic performance on presidential approval has also been reported in Venezuela (Weyland 1998) and Peru (Morgan 2003, Arce 2003, Arce and Carrión 2010, Weyland 2000). The availability of polling data has allowed associating economic perception to democratic consolidation, political culture, democratic legitimacy, trust in institutions and other indicators of democratic quality (Seligson and Smith 2010, Córdova and Seligson 2010, Booth and Seligson 2009, Hawkins et al. 2008, Segovia et al. 2008, Morales 2008, 2009, Luna 2008). In Chile, economic variables have also been used to explain presidential approval (Cabezas and Navia 2010, Morales 2008, Perelló 2015) and during the transition to democracy (Panzer and Paredes 1991). Yet, until 2010, Chile had only experienced center-left Concertación governments. Piñera was the first right-of-center democratically elected president since the democratic transition. His election called into question the stability of electoral preferences that favored the Concertación. Since support for Piñera fell after he took office, when economic conditions were positive 5.3 percent GDP growth, 3.8 percent inflation and 6.9 percent unemployment between 2010 and 2014 the economic vote alone is insufficient to explain fluctuations in presidential approval. Piñera s approval declined from 45 percent in mid-2010 to a low of 23 percent in 2011. Presidential approval recovered at the end of his term, when it reached 32 percent. Though he was not the first Chilean president to see his approval decline, he was the first to experience such a decline under favorable economic conditions thus ruling out the possibility of an economic vote explanation. 9

To explain Piñera s approval, we must first turn to the different reasons advanced to explain his victory. Declining turnout, with slightly over half eligible voters casting ballots and with even lower turnout rates among the youth, weakened the effect of the authoritarian-democratic cleavage that characterized elections in the 1990s (Contreras and Navia 2013, Luna and Toro Maureira 2010). As a result, Piñera had an opportunity to build a stronger support base for rightwing parties attracting moderate voters. In this reading, Piñera s victory might have been partially caused by the Concertación, but depending on the policies implemented by his government, Piñera could induce a new alignment in the Chilean electorate. Pointing to self-inflicted wounds by the Concertación, Navia and Morales (2010) imply that, despite his clever campaign strategy and efforts to reinvent the right, Piñera would not have won had the Concertación avoided mistakes in the candidate nomination process. In discussing the prospects for the 2013 election, Elacqua and Aninat (2013) argued that pragmatic middle class voters abandoned the Concertación in 2009 because Frei put too much emphasis on state solutions: Frei focused on the conflicting interests of the state and the private sector ignoring the fact that Chile's new middle class was by then deeply invested in the market and received little help from the state (Elacqua and Aninat 2013). We suggest that the high presidential approval volatility in Chile under Piñera is explained by the presence of swing voters who abandoned him after they helped elect him. Hence, models that predict stable electorates or those that associate shifts in approval to economic conditions fail to explain presidential approval volatility in Chile. Since elections were consistently favorable to the Concertación between 1989 and 2005, we test if Piñera s victory was simply the result of an antiincumbent punishment vote. However, since the 2011 student protests were also a defining 10

moment for Piñera s presidency, we also test if that was the time when swing-voters who were instrumental for Piñera s election deserted him as President. We formally propose two hypothesis. Hypothesis 1: Chileans voted for Piñera in 2009 because they wanted to punish the Concertación. As soon as Piñera became President, Chilean voters abandoned him. Hypothesis 2: Regardless of why they voted for Piñera, Chilean voters abandoned him in the wake of the student protests of 2011. Presidential elections and presidential popularity in Chile In 1990, the party system combined elements of the pre-1973 party system and the particularities of Chile s transition to democracy. Before 1973, parties were aligned on a left/center/right continuum. Under military rule, a democratic/authoritarian divide defined the transition to democracy, particularly in the 1988 plebiscite. Pro-democracy parties formed the Concertación, a coalition comprised of centrist and leftist parties. In the right, the Alianza, with Renovación Nacional (RN) and Unión Demócrata Independiente (UDI) grouped those who supported Pinochet. The 1988 plebiscite cleared the way for a democratic election in 1989 that resulted in the victory of the Concertación presidential candidate. After badly losing in 1989 and 1993, the Alianza forced a runoff in 1999, when the country was undergoing an economic recession. Yet, the Concertación prevailed for the third consecutive time. In 2005, the combined vote for the two Alianza presidential candidates, Sebastián Piñera (RN) and Joaquín Lavín (UDI), was higher than the Concertación vote for the first time since 1989. In the runoff, Concertación s Michelle Bachelet 11

defeated Piñera, though the latter became the frontrunner for the 2009 election. Thus, between 1989 and 2009, although the rightwing candidates became increasingly more competitive, the Concertación always ended up winning the presidential election. In 2009, there were 4 presidential candidates. Piñera was the sole rightwing candidate. Former president Eduardo Frei (1994-2000), a moderate Christian Democrat (PDC) was the Concertación candidate. The Communist Party, and its Juntos Podemos Más coalition, recruited Jorge Arrate, a former minister and prominent member of the Socialist Party (PS), a leading Concertación party, as its candidate. Another former PS leader, outgoing member of the Chamber of Deputies, Marco Enríquez-Ominami (ME-O) also ran as an independent. The presence of two former Concertación leaders as alternative presidential candidates reflected the weakening of the center-left coalition. Yet, because the authoritarian-democratic divide was still relevant (former dictator Pinochet died only in late 2006), many expected all leftwing forces to coalesce around a single candidate presumably Frei in the runoff, just as it had happened in 1999 and 2005. Precisely because the perception that the authoritarian democratic divide was so strong, Piñera ran a campaign highlighting his moderate positions and his opposition to Pinochet in the 1988 plebiscite. If the Concertación had historically won because the median voter had opposed Pinochet in 1988, Piñera appealed to the median voter that had grown tired of 20 years of consecutive Concertación governments. The world economic crisis of 2008 sent Chile into a recession for the first time in 10 years. Even though outgoing President Bachelet remained highly popular, support for the Concertación 12

declined to an all-time low in polls. In the 2008 municipal election, the Alianza obtained its first electoral victory in such contests since they were first held in 1992. In 2009, the Concertación further alienated many supporters when it rejected open primaries to select its presidential candidate (that decision led Enríquez-Ominami to resign from the PS and run as an independent). Thus, there are several reasons that can explain the Concertación s electoral vulnerability the weaning after 20 years in power, unfavorable economic conditions, a divisive presidential nomination process and a unified opposition. As Table 1 shows, Piñera received a plurality in the first round vote and won the runoff by a slim margin. Table 1. Presidential election results, 2009 Candidate First round vote N First round vote % Runoff vote N Runoff vote % S. Piñera 3.074.174 44.1 3.367.790 51.6 E. Frei 2.065.061 29.6 3.591.182 48.4 M. Enríquez-Ominami 1.405.124 20.1.. J. Arrate 433.195 6.2.. Valid Votes Total 6.977.544 100 6.958.972 100 Null and Blank votes 286.592. 244.399. Turnout total 7.264.136. 7.203.371. Voting age population 12.277.915. 12.277.915. (VAP) Valid votes as % of VAP 56.8 56.7 Source: authors with data from Servel Piñera s narrow victory reflected his ability to attract support from sectors traditionally associated with the Concertación. As Table 2 shows, using the highly cited Centro de Estudios Públicos (CEP) poll of October 2009, Piñera had strong support from rightwing and upper class voters, two strong support groups for the Alianza. However, Piñera also did well among the youth. In the runoff, Frei did better among older Chileans those most likely to be influenced by the authoritarian- 13

democratic cleavage. Piñera won in every other age group, but his support was strongest among those aged 34 and younger. Piñera also gained support among the middle class another traditional Concertación voting bloc and the well-to-do. Frei only won among the poor. Since poverty had declined under Concertación governments and the middle class expanded, Piñera s ability to attract support from middle class voters gave the right a great opportunity to forge a new cleavage to replace the authoritarian/democratic divide. Piñera also benefited from the Concertación s infighting. One of every three voters that supported Enríquez-Ominami in the first round voted for Piñera in the runoff. Though Arrate s voters were more loyal to the Concertación, Piñera also attracted some support from that group. 14

Table 2. Vote intention for Frei and Piñera among different groups, 2009 Variables Frei Piñera First round Runoff First round Runoff Gender Men 23.9 35.4 39.3 46 Women 23.9 37.6 32.1 41.8 Age 18-24 16.7 37.2 35.2 46.8 25-34 16.5 30.3 41.1 51.6 35-54 26.8 38.3 34.9 42.5 55+ 30.5 38.3 32.5 37.4 Region Santiago 18.1 31.8 36.8 47.3 Others 28 39.8 34.7 41.4 Socio economic status (SES) High 12 21.2 59.7 62.9 Middle 17.7 31.9 37.6 46.9 Low 31.7 43 30.8 38.5 Ideology Right 9 15 76.4 79.4 Center 33.4 45.9 30.2 43 Left 38.9 64.3 10.9 16.8 None 19.8 29 31.6 42 ME-O voters (18.7%)*. 40. 32.6 Arrate voters (4.5%)*. 59. 17.4 *Vote intention for ME-O and Arrate. Source: Author s elaboration of data based on CEP Survey # 61 (www.cepchile.cl) Upon taking office, Piñera s approval remained in the mid-40s. In his second year, a wave of student protests weakened the government. Piñera s approval fell to 26 percent in mid-2011 and 23 percent in December 2012. Though the protests eventually lost strength, Piñera s approval never fully recovered. By late 2013, at the end of his administration, his approval was at 34 percent. 15

Figure 1. Presidential approval and vote intention for ruling coalition candidate, 1989-2013 100 75 50 25 0 Jun-90 Dec-90 Jun-91 Dec-91 Jun-92 Dec-92 Jun-93 Dec-93 Jun-94 Dec-94 Jun-95 Dec-95 Jun-96 Dec-96 Jun-97 Dec-97 Jun-98 Dec-98 Jun-99 Dec-99 Jun-00 Dec-00 Jun-01 Dec-01 Jun-02 Dec-02 Jun-03 Dec-03 Jun-04 Dec-04 Jun-05 Dec-05 Jun-06 Dec-06 Jun-07 Dec-07 Jun-08 Dec-08 Jun-09 Dec-09 Jun-10 Dec-10 Jun-11 Dec-11 Jun-12 Dec-12 Jun-13 1st Round 2nd Round Approves Disapproves Neither Source: Author s elaboration of data based on www.cepchile.cl Since 1990, presidential approval has not been closely associated with election results. As Figure 1 shows, when Eduardo Frei won in 1993, outgoing Concertación president Patricio Aylwin had slightly over 50 percent approval. In 1999, Ricardo Lagos won despite Frei s 29 percent approval. In 2005, Bachelet won as approval for Lagos bordered 60 percent. In 2009, despite Bachelet s 80 percent approval, the Concertación candidate lost. In 2013, when outgoing President Piñera s approval was growing, the Alianza presidential candidate received less than 30 percent in the first round vote. Thus, in only two of the five presidential elections, the result was consistent with presidential approval levels. 16

Figure 2. Presidential approval and economic outlook, 1990-2013 100 90 80 70 60 50 40 30 20 10 0 Jun-90 Dec-90 Jun-91 Dec-91 Jun-92 Dec-92 Jun-93 Dec-93 Jun-94 Dec-94 Jun-95 Dec-95 Jun-96 Dec-96 Jun-97 Dec-97 Jun-98 Dec-98 Jun-99 Dec-99 Jun-00 Dec-00 Jun-01 Dec-01 Jun-02 Dec-02 Jun-03 Dec-03 Jun-04 Dec-04 Jun-05 Dec-05 Jun-06 Dec-06 Jun-07 Dec-07 Jun-08 Dec-08 Jun-09 Dec-09 Jun-10 Dec-10 Jun-11 Dec-11 Jun-12 Dec-12 Jun-13 President: Approves Outlook: Good Outlook: Neutral Outlook: Bad Source: Author s elaboration of data based on www.cepchile.cl Presidential approval has evolved more closely with the economic outlook. CEP polls ask about the economic situation on a 5-point scale, from very bad to very good. We group good and very good as positive, bad and very bad as negative and consider stagnant as neutral (though the wording stagnant rather than the same for the neutral value might have a negative connotation). Figure 2 shows the evolution of presidential approval and positive economic outlook between 1990 and 2013. Positive outlook has the lowest variance, but moves in the same direction as presidential approval. Under Piñera, positive outlook was higher than in previous administrations, but presidential approval was lower. This is inconsistent with economic vote predictions and different from what was observed in Chile before Piñera came into office. 17

In the first decade of democratic rule, presidential approval moved closely with neutral economic outlook, but under Piñera, it was lower. Since 1990, negative outlook and presidential approval have been inversely related. The negative outlook has not always been correlated with actual economic conditions. In 2009, under an economic recession, negative outlook was on the decline. When Piñera first came into office, and despite the expansion of the economy, negative outlook was increasing. Under Piñera, negative outlook was inversely correlated with presidential approval. Chileans have grown more pessimistic and have approved of their president less in ways that are not always consistent with economic performance. Data and Methodology We use the widely available and highly respected Centro de Estudios Públicos (CEP) polls. We analyze both the longitudinal evolution of vote intention in a pre-electoral poll in 2009 and of presidential approval in 9 polls conducted under President Piñera s term. Though we have 10 surveys, the pre-electoral poll included questions on vote intention for the first-round vote and for the likely runoff between Piñera and Frei. We use logistic regression models and predicted probabilities to identify continuity and change in the determinants of vote intention and presidential approval. The dependent variables are VOTING and APPROVING which consist of vote intention and presidential approval, taking the value of 1 if the respondent votes for Piñera or approves of the president. The variable SEX is 1 for women. AGE cohorts have been divided into 4 groups: 18-24, 25-34, 35-54 and 55+ (the reference category). REGION takes the value of 1 for those who lived in Santiago and 0 for those who reside elsewhere. SOCIOECONOMIC STATUS distributes 18

respondents according to high, middle and low groups (reference category). POLITICAL IDEOLOGY divides respondents into those who self-identify with the right, center and left, and those who do not identify on the ideological scale (our reference category). Predictors SOCIOTROPIC and SOCIOTROPIC PROSPECTIVE identify economic outlook. Respondents are divided into positive, neutral and negative (reference category) assessments. The same applies for the variables EGOTROPIC and EGOTROPIC PROSPECTIVE. The model for vote/approval is: VOTE/APPROVE = B 0 + B 1 SEX + B 2 AGE + B 3 REGION + B 4 SOCIOECONOMIC STATUS + B 5 POLITICAL IDEOLOGY + B 6 SOCIOTROPIC + B 7 SOCIOTROPIC PROSPECTIVE + B 8 EGOTROPIC + B 9 EGOTROPIC PROSPECTIVE (1) Voting for Candidate Piñera and Approving of President Piñera As Figure 3 shows, respondents in the high socio-economic strata (SES) were twice as likely to vote for Piñera as those in the lowest group for the first round vote. Support for Piñera among the middle class would grow faster than among the poor in case of a runoff. Ideology was a strong predictor of vote intention, as those in the right strongly supported Piñera. Among those not identified ideologically, Piñera had more support than the other candidates. Piñera was also more popular among men and younger voters. His vote intention in the runoff would increase more markedly among younger voters. Piñera s lowest support, in the first run and runoff, was among 19

oldest voters. Piñera was also marginally more popular in Santiago, and even more so in case of a runoff. Once Piñera took office, presidential approval declined, but not uniformly across groups. Piñera always had higher approval among those identified with the right. When comparing vote intention with presidential approval, support for Piñera was more stable among conservative and upper class Chileans, whereas among younger voters and Santiago residents, support for candidate Piñera did not translate into high presidential approval. At the same time, support for candidate Piñera was strongest among young men between 25 and 34 years of age who lived in Santiago, belonged to higher socioeconomic groups and identified with the right. However, President Piñera struggled to maintain support among younger voters and Santiago residents. 20

Table 3. Vote intention and presidential approval for Sebastián Piñera by selected indicators, 2009-2013 (%). Month Gender Age Region Men Women 18-24 25-34 35-54 55+ Santiago Others % Oct-09 1R 39.9 32.9 43.7 43.3 36.1 32.1 38.3 34.9 36.3 Oct-09 2R 44.2 41.0 47.8 50.4 43.3 37.2 46.6 39.8 42.5 Jun-Jul 10 48.8 48.0 49.7 47.3 47.5 49.8 44.3 51.1 48.8 Nov-Dec 10 49.5 44.1 38.4 52.5 43.8 52.3 42.5 49.9 47.1 Jun-Jul 11 33.0 23.1 29.3 28.3 28.4 26.8 28.0 28.2 28.6 Nov-Dec 11 24.2 22.6 20.5 25.0 22.3 25.6 21.4 24.9 23.3 Apr 12 25.5 23.5 24.6 20.9 22.9 29.7 23.4 25.3 24.4 Jul-Aug 12 32.1 24.4 27.5 27.1 25.6 33.0 25.8 29.5 28.1 Nov-Dec 12 34.7 30.5 31.0 36.4 29.0 35.8 31.7 33.2 32.6 Jul-Aug 13 31.8 32.5 30.8 31.5 29.4 37.2 31.3 32.7 32.1 Sep-Oct 13 35.1 34.6 29.0 39.2 38.4 30.0 33.9 35.5 34.4 Month Socioeconomic Status (SES) Political Ideology Low Middle High Others Right Center Left % Oct-09 1R 28.9 39.8 58.6 32.5 77.5 32.3 8.8 36.3 Oct-09 2R 34.5 47.0 62.1 40.1 80.8 43.2 12.6 42.5 Jun-Jul 10 44.6 5.0 72.4 42.1 81.3 49.9 21.3 48.8 Nov-Dec 10 43.3 48.6 60.6 43.7 80.2 57.0 17.5 47.1 Jun-Jul 11 20.7 27.8 71.6 23.8 64.1 27.9 4.1 28.6 Nov-Dec 11 16.7 26.1 48.6 18.7 64.1 23.0 7.0 23.3 Apr 12 20.5 24.9 54.0 16.6 62.8 30.4 8.3 24.4 Jul-Aug 12 26.1 28.4 41.6 24.8 73.0 30.7 6.1 28.1 Nov-Dec 12 26.8 34.4 50.7 28.3 76.3 35.4 14.0 32.6 Jul-Aug 13 26.5 34.7 52.2 23.7 82.1 41.8 15.7 32.1 Sep-Oct 13 29.5 36.8 54.5 29.3 84.0 39.8 15.7 34.4 Source: Author s elaboration of data based on CEP Surveys # 61-70 (www.cepchile.cl) 21

Table 3 (continued). Vote intention and presidential approval for Sebastián Piñera by selected indicators, 2009-2013 (%). Month Sociotropic Sociotropic Prospective Positive Neutral Negative Positive Neutral Negative % Oct-09 1R 27.1 40.2 38.4 41.3 32.8 27.8 36.3 Oct-09 2R 31.9 45.1 48.8 47.4 39.6 31.4 42.5 Jun-Jul 10 62.3 50.9 31.7 68.8 36.2 25.1 48.8 Nov-Dec 10 74.6 44.3 25.2 73.3 31.4 12.8 47.1 Jun-Jul 11 49.0 33.4 8.6 56.6 20.0 10.4 28.6 Nov-Dec 11 45.0 24.7 9.9 52.5 17.6 9.9 23.3 Apr 12 46.7 25.3 9.6 43.2 20.5 9.0 24.4 Jul-Aug 12 45.8 27.4 16.8 47.4 23.6 13.6 28.1 Nov-Dec 12 61.0 25.8 13.7 50.5 26.1 21.0 32.6 Jul-Aug 13 65.4 26.6 17.0 38.9 27.8 45.2 32.1 Sep-Oct 13 62.6 30.7 16.0 39.7 31.1 45.2 34.4 Month Egotropic Egotropic Prospective Positive Neutral Negative Positive Neutral Negative % Oct-09 1R 37.1 35.4 38.5 40.6 35.1 23.0 36.3 Oct-09 2R 43.8 41.3 45.3 46.7 41.6 32.0 42.5 Jun-Jul 10 62.0 48.4 36.4 56.9 41.8 32.8 48.8 Nov-Dec 10 60.7 45.8 32.5 62.6 38.1 10.8 47.1 Jun-Jul 11 44.9 24.0 12.6 39.8 22.6 8.7 28.6 Nov-Dec 11 33.7 24.7 10.7 33.4 20.4 9.7 23.3 Apr 12 38.6 24.1 12.4 29.6 23.8 9.7 24.4 Jul-Aug 12 39.0 26.4 20.3 36.4 25.0 8.8 28.1 Nov-Dec 12 50.5 26.8 22.3 44.2 25.4 19.4 32.6 Jul-Aug 13 50.2 27.9 22.6 35.7 29.0 34.9 32.1 Sep-Oct 13 49.1 31.2 28.6 41.6 29.5 35.3 34.4 Source: Author s elaboration of data based on CEP Surveys # 61-70 (www.cepchile.cl) 22

Inferential Analysis We use logit regressions to determine support for Piñera as a candidate and as a president. Table 3 and Figure 4 show those results. The first four columns present vote intention for Piñera in the first round and runoff respectively as the dependent variables. The following eight columns have presidential approval for each year of Piñera s presidency as the dependent variable. Vote Intention Results for vote intention for candidate Piñera in the first round vote and runoff are statistically significant for sex, socioeconomic status, ideology, and to a lesser extent for age and economic outlook. In the first round scenario, women were less likely to vote for Piñera than men (β=.43, p 0.5), but not for the runoff election. Age did not have a significant effect. As table 3 shows, there are no statistically significant differences when contrasting the preferences of respondents aged 55 or older. There is a similar distribution of results for the runoff election. The only significant difference is that candidate Piñera gained more support from respondents in the 25-34 age group (β=.62, p 0.5). There is marked variance in socioeconomic terms. Vote intention for Piñera grew with income. In the first round, both middle (β=.48, p 0.5) and high SES (β=.73, p 0.1) respondents displayed greater support for him in comparison to the low-income group. A similar outcome also takes place for the middle SES group in the runoff election (β=.58, p 0.01). Political ideology is also a robust predictor. Not surprisingly, respondents that identified with the right were more likely to support Piñera than those who did not identify with a political ideology, in both hypothetical scenarios, the first round (β= 1.18, p 0.01) and runoff (β= 1.51, p 0.01). The opposite is observed for 23

respondents that self-identified with the left, which displayed the least support for Piñera (β= 2.12 and β= 1.96, p 0.01). Interestingly, respondents that identified with the center were not statistically different from the non-ideological group in the first round vote, despite showing a slightly negative vote intention for Piñera in the runoff (β=.41, p 0.1). Economic outlook had a limited effect on Piñera s vote intention. Respondents with a positive sociotropic view were less likely to support Piñera in the runoff (β= 1.07, p 0.01), while those with a positive sociotropic prospective were more inclined to do so (β=.58, p 0.1). Egotropic outlook is a weak predictor: only respondents with a positive assessment of their personal situation were significantly less inclined to support Piñera than those with a negative outlook (β=.55, p 0.5) in the first round vote. Presidential Approval Table 3 also presents a series of models on presidential approval in the 9 CEP polls in the period. Contrary to Piñera s vote intention, sex fails to be a significant predictor across all polls in explaining presidential approval. Age varies as a predictor of approval. Though there are no marked differences between age cohorts in the first two years of Piñera (with the exception of respondents aged between 25 and 34 in November-December 2010), there is an age gap in the next two years. In April 2012, respondents aged 25-34 began displaying growing disapproval in comparison to the reference group (β=.78, p 0.01). The July-August 2012 survey shows similar results (β=.55, p 0.1). Yet, starting on November-December 2012, the gap is extended to respondents aged 18-24 (β=.68, p 0.5) and 35-54 (β=.68, p 0.01), as well as those aged 25-34 (β=.52, p 0.5). The survey of July-August 2013 continues to show how Piñera was disapproved by respondents aged 25-34 (β=.6, p 0.5) and 35-54 (β=.82, p 0.01). Interestingly, there are no 24

statistically significant differences across age cohorts in the last presidential survey under Piñera, in 2013. Socioeconomic status also displays coefficients that show a change in Piñera s original bases of support. Overall, SES is a weak predictor of presidential approval an outcome that contradicts the strong differences observed for Piñera s vote intention. There are no statistically significant differences between middle and high SES respondents when contrasted to the lowest SES group. The only exception is in June-July 2011, when high SES respondents displayed greater approval (β= 1.27, p 0.01). Hence, the high and middle SES segments that voted for Piñera almost immediately stopped approving of him as president, as can be seen starting in June-July 2010. Political ideology is the most consistent predictor of presidential approval. Right-leaning respondents supported President Piñera the most, while respondents that self-identified with the left supported him the least. Centrists only sporadically displayed statistical significance. Starting in June-July 2010, support for Piñera was greater among right-leaning (β= 1.59, p 0.01) and centrist respondents (β=.41, p 0.1) and lower among those self-identified with the left (β=.79, p 0.01), all in comparison to respondents that did not identify on the left-right scale. Centrist respondents displayed greater support for candidate Piñera, but almost immediately stopped approving of President Piñera evidencing a rapid change in Piñera s electoral base. The remaining polls show similar results. Support for Piñera was greater among rightists and lower among leftists. This trend is stable throughout his presidency. However, there are fluctuations among centrists in April 2012 (β=.78, p 0.01) and July-August 2013 (β=.71, p 0.05), both polls 25

displaying greater support among centrists when compared to the non-ideological. Yet, the general trend is that centrists did not show significant differences with the reference group. Economic outlook was an important predictor of presidential approval, unlike vote intention. Generally speaking, sociotropic views outweigh egotropic ones in their explanatory power, while positive views are more important than neutral or pessimistic assessments. The trend shows how Piñera s approval was higher among respondents that judged the sociotropic situation as being the same or better (there is an exception in July-August 2012). The sociotropic prospective variable also has explanatory power for presidential approval, though to a lesser extent than sociotropic views. Results show how respondents with a positive sociotropic prospective tended to approve more of the President. Presidential approval lost significance among those with a neutral assessment (in comparison to the reference group: those with a negative view). Interestingly, results display a change towards the end of Piñera s presidency (July-August 2013 and September- October 2013). As table 3 shows, both neutral and positive prospective assessments negatively explain presidential approval. This might have to do with the expectations that the leftwing Concertación (then called Nueva Mayoría) would return to power. Egotropic views are more limited in explaining presidential approval, though when significant, they tend to present similar results as the sociotropic outlook. Egotropic assessments fail to display strong explanatory power. Exceptions can be seen in June-July 2011 and April 2012, when respondents with neutral or positive outlooks had a statistically and positive effect on presidential approval. Meanwhile, egotropic prospective views are more significant in explaining 26

Piñera s support: approval was greater among respondents with a positive or neutral perspective of their personal economic situation. 27

Table 3. Logit regression model. Vote Intention Presidential Approval Predictors Oct 09 1R Oct 09 2R Jun-Jul 10 Nov-Dec 10 Jun-Jul 11 Nov-Dec 11 Apr 12 Jul-Aug 12 Nov-Dec 12 Jul-Aug 13 Sep-Oct 13 Sex -0.430** -0.247 0.214 0.273-0.273 0.0572 0.154-0.283 0.0536 0.179 0.304 Women (Men) (0.176) (0.177) (0.165) (0.217) (0.217) (0.203) (0.181) (0.199) (0.189) (0.205) (0.199) -0.265-0.177-0.0876-0.659-0.488-0.444-0.373-0.332-0.681** -0.159-0.152 18-24 (0.372) (0.371) (0.239) (0.425) (0.431) (0.304) (0.281) (0.333) (0.291) (0.263) (0.299) Age 0.210 0.626** -0.210 0.0939-0.188-0.270-0.783*** -0.555* -0.523** -0.603** 0.133 25-34 (55+) (0.286) (0.296) (0.249) (0.303) (0.296) (0.359) (0.273) (0.287) (0.264) (0.261) (0.310) -0.0470 0.0618 0.0391-0.569** -0.287-0.333-0.317-0.311-0.685*** -0.820*** 0.243 35-54 (0.205) (0.203) (0.196) (0.236) (0.236) (0.238) (0.220) (0.247) (0.212) (0.242) (0.224) Region -0.0767 0.105-0.364** -0.201-0.200-0.261-0.115-0.212-0.345* 0.100-0.237 Santiago (Others) (0.180) (0.185) (0.167) (0.218) (0.224) (0.232) (0.187) (0.208) (0.189) (0.206) (0.194) 0.483** 0.581*** -0.00773 0.216 0.237 0.366-0.109-0.305 0.0245 0.0789 0.216 Middle SES (0.189) (0.193) (0.166) (0.236) (0.222) (0.229) (0.188) (0.199) (0.184) (0.206) (0.218) (Low) 0.734* 0.680 0.794 0.424 1.272*** 0.458-0.294 0.501 0.00951 0.781 0.223 High (0.396) (0.415) (0.549) (0.519) (0.481) (0.482) (0.424) (0.409) (0.592) (0.505) (0.484) 1.811*** 1.519*** 1.559*** 1.204*** 1.386*** 1.718*** 2.146*** 1.995*** 1.795*** 2.492*** 2.456*** Right (0.237) (0.261) (0.219) (0.386) (0.243) (0.249) (0.244) (0.259) (0.244) (0.288) (0.369) Ideology -0.327-0.415* 0.415* 0.204 0.298 0.0291 0.784*** 0.330 0.340 0.713** 0.445 Center (None) (0.243) (0.242) (0.222) (0.301) (0.348) (0.315) (0.242) (0.266) (0.276) (0.277) (0.273) -2.124*** -1.971*** -0.791*** -1.373*** -1.873*** -1.072** -0.701*** -1.601*** -1.081*** -0.586** -0.870*** Left (0.268) (0.257) (0.228) (0.299) (0.390) (0.449) (0.271) (0.357) (0.251) (0.271) (0.313) 0.154-0.408* 0.714*** 0.439* 1.057*** 0.939*** 0.728*** 0.377 0.773*** 0.615** 0.890*** Neutral Sociotropic (0.235) (0.243) (0.199) (0.226) (0.256) (0.333) (0.215) (0.254) (0.245) (0.292) (0.254) (Negative) -0.459-1.072*** 1.145*** 1.260*** 0.997*** 1.430*** 1.549*** 0.670** 2.153*** 2.138*** 2.169*** Positive (0.288) (0.289) (0.272) (0.305) (0.346) (0.411) (0.271) (0.302) (0.276) (0.331) (0.292) -0.119 0.300 0.253 0.495 0.285 0.274 0.537 0.371 0.163-0.597* -0.709** Neutral Sociotropic pro (0.368) (0.368) (0.275) (0.363) (0.389) (0.277) (0.346) (0.324) (0.366) (0.333) (0.337) (Negative) 0.359 0.606 1.123*** 1.758*** 1.232*** 1.478*** 1.156*** 1.048*** 0.564-0.0857-0.802** Positive (0.377) (0.382) (0.278) (0.397) (0.431) (0.379) (0.369) (0.377) (0.394) (0.417) (0.357) -0.300-0.175 0.261 0.0566 0.195 0.386 0.472** 0.00506-0.370 0.184-0.188 Neutral Egotropic (0.233) (0.239) (0.228) (0.226) (0.293) (0.267) (0.227) (0.252) (0.263) (0.261) (0.253) (Negative) -0.384-0.0323 0.398 0.483 0.616* 0.316 0.795*** 0.353-0.00257 0.209-0.300 Positive (0.288) (0.291) (0.289) (0.356) (0.321) (0.293) (0.278) (0.339) (0.296) (0.306) (0.324) 0.368 0.0317 0.116 1.312** 0.536 0.569* 0.888** 0.676* 0.198-0.0247-0.257 Neutral Egotropic pro (0.483) (0.532) (0.298) (0.524) (0.421) (0.326) (0.406) (0.366) (0.335) (0.520) (0.344) (Negative) 0.351 0.0956 0.325 1.604*** 0.842* 0.798** 1.002** 1.002*** 0.892** -0.0420 0.257 Positive (0.493) (0.549) (0.308) (0.554) (0.437) (0.359) (0.416) (0.372) (0.352) (0.526) (0.361) -0.447 0.0813-2.029*** -3.094*** -3.287*** -3.569*** -4.063*** -2.266*** -1.943*** -1.501*** -1.347*** Constant (0.561) (0.589) (0.418) (0.640) (0.538) (0.431) (0.560) (0.471) (0.432) (0.550) (0.450) Observations 844 790 1.366 1.293 1.395 1.433 1.419 1.375 1.364 1.353 1.328 Source: Author s elaboration of data based on CEP Surveys # 61-70 (www.cepchile.cl) 28

Figure 3. Coefficient plot Source: Author s elaboration of data based on CEP Surveys # 61-70 (www.cepchile.cl) 29

Predicted Probabilities We computed predicted probabilities to illustrate continuity and change in Piñera s support. As Figure 4 shows, support for Piñera falls almost across all groups after he was elected. Yet, there are differences in the extent and timeframe of the decline. Hence, some groups stopped backing Piñera almost immediately after he was sworn into office, while others did so when the student mobilizations erupted providing evidence in support of both the electoral punishment and student mobilization hypotheses. Piñera s support among men fell in the first post-electoral survey (June-July 2010), while remaining relatively stable among women. Then, starting in 2011, Piñera s approval drastically fell among men and women, evidencing an effect that coincided with the rise of the student mobilizations. Thus, the protest vote was stronger among men while the effect of the student movement was equally damaging for Piñera among men and women. Age wise, results show how Piñera s strongest electoral support group, respondents aged 25-34, quickly abandoned him in the first presidential approval poll. All other age groups (18-24, 35-54 and 55 and more) show a similar probability of supporting Piñera in the pre-electoral and first postelectoral surveys. Though respondents aged 25-34 years old, alongside the oldest cohort, had the highest presidential approval in November-December 2010, approval fell across all groups in June- July 2011 pointing to the effect of student mobilizations. From then onwards, Piñera s support was strong among the oldest respondents and weak among the younger cohorts. Only at the end of his presidency (July-August and September-October 2013) did the composition of approval change, with those aged 35-54 displaying higher approval. Thus, after leading voting for Piñera in 30

the first round and runoff, the 25-34 age group quickly abandoned Piñera when he became president. That age group can be characterized as a protest vote against the Concertación. In the wake of the student protests, support for Piñera fell among all age groups, falling more drastically among the two youngest cohorts (which is consistent with the fact that protests were led by students). The region of residence shows continuity in pre and post-electoral support for Piñera among respondents that resided in the rest of the country and change for those in Santiago. Those residing in the capital immediately abandoned Piñera and can be characterized as a protest vote against the Concertación. Yet, starting in 2011, approval for Piñera declined drastically among those residing in Santiago and elsewhere. Socioeconomic status provides further evidence of continuity and change in support for Piñera. Figure 4 shows that Piñera s electoral support was greater among middle and high SES respondents and significantly lower for low-ses respondents. Yet, in the first presidential approval poll, Piñera s support fell among middle and high SES respondents. Both groups abandoned Piñera soon after his term began. Approval for Piñera in early 2010 was the same in all SES groups. The remaining eight presidential approval polls show similar levels of support for Piñera across all SES groups. Thus, high and middle income Chileans were more likely than low income Chileans to be among those who voted for Piñera as a protest vote against the Concertación. Ideology also shows variance in when different groups abandoned Piñera. Though conservatives were always more likely than the rest to support Piñera, support for the first rightwing president 31

declined markedly among conservatives in the first presidential poll. Conversely, the probability that a self-identified leftist approved of President Piñera was higher than the probability that a leftist had voted for Piñera. For the rest of the term, Piñera s approval among conservatives remained stable, even in the wake of student protests in 2011. Thus, there is some evidence that some rightwing voters who supported Piñera abandoned him once Piñera came into office presumably because he disappointed them when he adopted moderate policies as a President especially on human rights and other issues associated with the legacy of the Pinochet dictatorship. Ideology also shows a distinct pattern among the non-identified. After strongly supporting Piñera lower only to right wingers the non-identified quickly abandoned Piñera as a President. The non-ideologically identified went from being the second most important group among Piñera s electoral base to becoming one of the least supportive groups, higher only than those in the left. Centrists displayed similar pre and post-electoral levels of support. There is strong evidence that points to the non-ideological Chileans as being consistent with a protest vote against the Concertación. As soon as Piñera became President, this group abandoned him and stayed away until the end of Piñera s presidency. The student protests of 2011 seemed to have had little impact on the probability of approving of Piñera across different ideological groups. Economic outlook variables provide further insights into fluctuations in Piñera s support base. They show differences in the way respondents view their personal and societal situations given the expectations about who will win the next election. Starting with sociotropic outlook, though a negative assessment increased the probability of voting for Piñera, it decreased the probability of approving of his presidency. The opposite is true for respondents with a positive sociotropic 32