Behind the Jokowi s victory: did economic voting matter in the 2014 Indonesian presidential election?

Transcription

1 Asia-Pac J Reg Sci (2018) 2: ECONOMIC ANALYSIS OF LAW, POLITICS, AND REGIONS Behind the Jokowi s victory: did economic voting matter in the 2014 Indonesian presidential election? Nurdien Aji 1 Teguh Dartanto 2 Received: 8 February 2018 / Accepted: 30 April 2018 / Published online: 21 May 2018 The Japan Section of the Regional Science Association International 2018 Abstract The 2014 presidential election marked Indonesia s transition into a mature democratic Islamic country. Joko Widodo, known as Jokowi, was inaugurated as Indonesia s seventh president, defeated Prabowo Subianto by a margin of %, respectively. In the absence of an incumbent, voters evaluated both presidential candidates based on a mix of prospective and retrospective economic performance. This study merges data from the Village Potential Census (PODES) and the crowdsource data (Kawal Pemilu) from vote recap to investigate the existence of economic voting and the effect of socioeconomic conditions at the village level on voting behavior. Our study confirmed that economic access, conditions of infrastructure, and middle-class group played major roles in the Jokowi victory. The middle class tended to choose Jokowi, because his ideology best matched their own. Moreover, social cleavages related to religion and ethnicity are still dominant as villages with a Muslim majority tended to choose Prabowo that was associated with an Islamic representative. Unfortunately, media and black campaigns also significantly influenced voters decisions. This evidence provides valuable lessons learned for Indonesians in preparing for a better 2019 presidential election. Keywords Economic voting Presidential election Local economy Middle class Indonesia JEL Classification D72 H11 * Teguh Dartanto teguh.dartanto@ui.ac.id 1 Graduate Program in Economics, Faculty of Economics and Business, Universitas Indonesia, Ministry of Public Work and Housing, Depok, Indonesia 2 Department of Economics, Faculty of Economics and Business, Universitas Indonesia, Jl. Prof. Soemitro Djojohadikusumo, Campus UI Depok, Depok 16424, Indonesia

2 116 Asia-Pac J Reg Sci (2018) 2: Introduction Indonesia has been remarked as the world s largest Muslim-majority country and the world s third-largest democracy after India and the United States, because four general elections and three presidential elections after 1998 were held relatively fair and peaceful resulting in smooth power transitions and stability. Since the collapse of the Suharto government in 1998, Indonesia has undergone significant political change. The most visible move toward democracy in Indonesian politics was direct elections for both the executive and legislative branches. Direct presidential elections were held in 2004, 2009, and Susilo Bambang Yudhoyono (SBY) won the 2004 and 2009 elections, but Indonesia s Constitution prohibited him from running for a third term. Therefore, in the third direct presidential election held 9 July 2014, Joko Widodo defeated Prabowo Subianto and was sworn in as the seventh president of Indonesia 20 October In the 2014 presidential election, two coalitions formed after the results of the parliamentary election held 9 April 2014: the Great Indonesia Coalition (Koalisi Indonesia Hebat/KIH) led by the Indonesian Democratic Party of Struggle (Partai Demokrasi Indonesia Perjuangan/PDIP) with Joko Widodo and Jusuf Kalla as president and vice-president candidate, respectively (hereafter referred to as Jokowi) and the Red and White Coalition (Koalisi Merah Putih/KMP) led by the Greater Indonesia Movement Party (Partai Gerakan Indonesia Raya/Gerindra), which supported Prabowo Subianto and Hatta Rajasa for president and vice-president, respectively (hereafter referred to as Prabowo). Although the KMP received a larger legislative election vote (48.92% compared with 40.89%, see Table 1), Prabowo lost the presidential election to Jokowi %, respectively. The 2014 presidential election marked Indonesia s transition into a mature democracy, even though many indications to the contrary could be found. As is often the case with young democracies, various returns to authoritarian rhetoric and practices as well as a resurgence of populist rhetoric by some political stakeholders accompanied the 2014 presidential election (Mietzner 2014; Aspinall and Mietzner 2014, Lang 2015). However, within the context of rising populism and authoritarian rhetoric, one indicator of a matured democracy is citizens ability to think rationally in choosing who will be the best leader. Therefore, voter evaluation of the candidates or incumbents performance became the foundation for democracy s rational decision-making in choosing leaders (Orth 2001). The absence of the public s evaluation of political leaders economic performance means leaders with a poor performance may remain in power, and eventually, democracy will not generate benefits to society (Dassonneville and Lewis- Beck MS 2014). Nadeau et al. (2013) categorized the factors influencing the outcome of an election into three groups: economic, ideological, and social cleavages. Votes based on economic factors (economic voting) occurs when voters evaluate incumbents on economic performance, so the election is seen as a referendum on whether an incumbent s performance is sufficient to give them a chance to lead again or give a new candidate the chance to govern (Fiorina 1978). When there

3 Asia-Pac J Reg Sci (2018) 2: Table 1 The 2014 parliament election results and coalitions. Source: Election Commission (Komisi Pemilihan Umum/KPU) Party Vote Percentage (%) Prabowo coalition (Koalisi Merah Putih/the Red and White coalition) Partai Gerakan Indonesia Raya/Great Indonesia Movement Party (Gerindra) 14,760, Partai Golongan Karya/Party of the Functional Groups (Golkar) 18,432, Partai Amanat Nasional/National Mandate Party (PAN) 9,481, Partai Keadilan Sosial/Prosperous Justice Party (PKS) 8,480, Partai Persatuan Pembangunan/United Development Party (PPP) 8,157, Partai Bulan Bintang/Crescent Star Party (PBB) 1,825, Jokowi coalition (Koalisi Indonesia Hebat/the Great Indonesian coalition) Partai Demokrasi Indonesia Perjuangan/Indonesian Democratic Party of Struggle (PDIP) 23,681, Partai Nasional Demokrat/National Democratic Party (Nasdem) 8,402, Partai Kebangkitan Bangsa/National Awakening Party (PKB) 11,298, Partai Hati Nurani Rakyat/People s Conscience Party (Hanura) 6,579, Partai Keadilan dan Persatuan Indonesia/Indonesian Justice and Unity Party (PKPI) 1,143, Neutral Partai Demokrat/Democratic Party (Demokrat) 12,728, PBB and PKPI are non-parliamentary party due to unable to meet the parliamentary threshold of 3.5%

4 118 Asia-Pac J Reg Sci (2018) 2: is no incumbent, as in Indonesia s 2014 presidential election, voters evaluated the new candidates performance on their previous duties or similarity of candidates proposed programs during the campaign. Voters may also compare their socioeconomic conditions to the candidates programs, and then evaluate which programs are more rational or executable. Jokowi s core program can be simplified into improved government services for the population, while Prabowo s core program was strongly nationalist in its platform of a Great Indonesia by promising to protect the nation s resources and wealth and restore pride in the country (Aspinall and Mietzner 2014; Mietzner 2014). Nadeau et al. (2013) s second factor is ideology. Voters tend to choose the candidate who best matches their own (Downs 1957). In the 2014 presidential election, there was little difference between Jokowi s and Prabowo s ideologies. Both presented themselves as populists but in different styles. Jokowi s populism was pragmatic, moderate, and inclusive, while Prabowo s populism was similar to that of Chavez and Thaksin that proposed more on economic nationalism (Aspinall and Mietzner 2014; Mietzner 2014). In terms of an Islamic supported party, Jokowi was only supported by the National Awakening Party (PKB), which is a traditionalist Muslim party. Prabowo was endorsed by four conservative Islamic parties: Prosperous Justice Party (PKS), United Development Party (PPP), National Mandate Party (PAN), and Crescent Star Party (PBB). The third factor is social cleavages, which is the tendency to choose political candidates or party based on a similar demographic group, such as religion, race, ethnicity, geographic location, education level, income level, or social class. This tendency of vote is the result of a long-term process (Anderson 2000). Aspinall and Mietzner (2014) and Mietzner (2014) examined the contrasting public personas between Jokowi and Prabowo. Jokowi looked and sounded like an average lowermiddle-class Indonesian, in contrast to the typical Indonesian elite politicians. He wore cheap clothes and spoke in a casual and unrefined manner with a Javanese accent, ate at roadside food stalls, and looked villagey (ndeso). Prabowo, on the other hand, personified a highly educated person and a Javanese aristocrat, replicating Sukarno s campaign style with uniforms, marching, and a highly theatrical character. Nonetheless, in a diverse society like Indonesia, voters perceived the differences between Jokowi s and Prabowo s style and campaign programs from many points of view. For example, rural poor voters might favor Jokowi due to his earthy personal style, while urban voters might favor Prabowo s self-confidence, experience, and sophisticated program. While Indonesia will have the fourth direct presidential election in 2019, lessons learned from 2014 will be precious resources for those who want to run for president as well as for the society in preparing for a more mature presidential election. This study, therefore, aims at evaluating the crucial factors affecting the Jokowi s victory in the 2014 presidential election. Did economic voting exist in the 2014 presidential election? Were socioeconomic conditions at the village level factors in Jokowi s victory? Did ideology and social cleavage strongly affect voting behavior? There are many types of research on economic voting (Lewis-Beck and Stegmaier 2013), but most research uses economic conditions at the country level as the determinant variable. Research on the effect of economic conditions at the local

5 Asia-Pac J Reg Sci (2018) 2: level is quite limited, however. Recent studies include Higashikata and Kawamura (2015), Healy and Lenz (2017), Johnston et al. (2000), Books and Prysby (1999), and Burbank (1997) who all analyzed the economic conditions at the city, region, or state levels. This study will combine the 2011 Village Potential Census (PODES) data and the crowdsource data of vote recap at the polling station ( pemil u.org) to analyze whether economic voting and socioeconomic conditions at the village level affected the Jokowi s victory. 1 The main contribution of this study is combining the PODES data and the crowdsource data, which allows us to analyze the relationship between economic voting, social cleavage, and socioeconomic factors at the village level and the election result. In a diverse socioeconomic country like Indonesia, however, an election analysis at the district or provincial level might result in misleading conclusions. The rest of the paper is organized as follows. Section 2 presents the conceptual framework on factors influencing the outcome of an election that can be grouped as follows: economic, ideological, and social cleavages. Section 3 elaborates the literature review and descriptive analysis of the 2014 presidential election in Indonesia. Section 4 explains empirical strategy including econometric modeling and merging data of Podes and crowdsource data. The discussion and analysis of econometric estimations are shown in Sect. 5; then, Sect. 6 summarizes conclusions and policy implications. 2 Conceptual framework of voting behavior This section will discuss a conceptual framework on an economic choice and economic voting in the election. Downs (1957) described the relationship of voters to candidates as principal and agent, where voters choose the candidate who can provide maximum utility to them and their candidates, with the sole purpose of being in charge (new candidate) or maximizing their period in charge (for an incumbent). Thus, the voting is also a form of consumption. Unlike the consumption of goods, which is in the form of continuous things, political consumption is a discrete-choice election: consumers can only consume one choice from all available options. Our approach is taken from McFadden s (1973) qualitative discrete choice. Each voter has the expected utility in selecting a candidate; for example, U j is the expected utility to choose Jokowi and U p is the expected utility in choosing Prabowo; therefore U j = Z j + ε j = X j β + ε j U p = Z p + ε p = X p β + ε p (1) (2) 1 This is possible because the Election Commission (Komisi Pemilihan Umum/KPU) uploaded the C1 document (a document of vote recap at the polling station level) and then Kawal Pemilu (the Election Monitoring Group) conducted a crowdsource to digitize and recapitalize the C1 document from polling station to national level, created alternative election result data beside an official data from KPU. The result from the Kawal Pemilu is only slightly different from KPU (0.14%).

6 120 Asia-Pac J Reg Sci (2018) 2: Economic Voting National Economy Local Economy JOKOWI Voters (Village) Utility Maximization Uj : Up VOTE Political Ideology Middle Class Long-Term Factors Social Cleavage Religion, Ethnicity Location, Education PRABOWO Fig. 1 Conceptual framework of voting behavior. Source: Authors where Z j (Z p ) is a representation of utility functions that affect voter preferences, U j (U p ) is not the same as Z j (Z p ), and Z j (Z p ) is linear with X j β s (X p β s) parameter. When a voter chose Jokowi, they believed that the utility obtained from their choice was greater than if they chose Prabowo, and then, U j > U p. Thus, the probability is Pr(U j > U p )=Pr ( (Z j + ε j ) > (Z p + ε p ) ) = Pr ( (ε p ε j ) > (X j β X p β) ) (3) Pr(U j > U p )=F(ε(Xβ)) where ε = ε p ε j and Xβ = X j β X p β. F is a joint cumulative distribution function. This equation could be written in the form of a latent utility model index as Y = U j U p where Xβ is the factor that influences consumer choice in choosing a presidential candidate at the presidential election (for the detail explanation, see McFadden 1973). Nadeau et al. (2013) refined their original categorization of factors influencing the vote: economic voting, where the voting is based on voters evaluation of incumbent s economic performance; ideology, where candidate selection is based on ideology; and, social cleavages, in which the choice is based on a long-term relationship Y = X j β X p β + ε j ε p Y = Xβ + ε (4)

7 Asia-Pac J Reg Sci (2018) 2: between communities and the party/candidate. Figure 1 illustrates the conceptual framework showing how all factors influenced voter choice in the election. Moreover, all factors (economic vote, political ideology, and social cleavage) will affect voter utility. Voters will evaluate the candidates for the one that could maximize their utility, and then, they will choose either Jokowi or Prabowo. 3 Descriptive analysis of the 2014 presidential election 3.1 Economic voting Voters, as the principal, will choose a candidate who provides maximum utility, while a candidate as agent aims to officiate or maximize a length of service (Downs, 1957). Because a performance can be evaluated based on the incumbent s performance, an election is a referenda mechanism to decide whether the incumbent s performance was strong and deserving of another chance to serve or a new candidate should be given the opportunity (Fiorina 1978). The incumbent s economic performance is also low-cost information for voters to predict the economic conditions under which the incumbent will be in charge in their second term, so the retrospective view is another form of prospective (future orientation) voting (Downs 1957; Fiorina 1978; Kiewiet and Rivers 1984). Lewis-Beck and Stegmaier (2013) reviewed economic voting articles and found that retrospective (focus on the past performance as the material for evaluation) and sociotropic (macroeconomic conditions) play an important role in rational voting. In the Indonesian context in the 2014 presidential election, there was no incumbent candidate, because SBY could not run for the third period and no incumbent party because SBY s Democrat Party decided to stay neutral. Campbell et al. (2010) found that economic voting can still occur even without an incumbent. The closeness between the incumbent and a competing candidate can offer voters the chance to see the candidate closest to the incumbent as the incumbent s successor. However, the level of economic voting depends on the closeness between the incumbent and the candidate. During the 2014 presidential election, most members from the SBY coalition party (Golkar, PKS, and PPP) joined the KMP to support Prabowo. Meanwhile, Hatta Rajasa who was the coordinating minister in the SBY era becomes Prabowo s vice presidential candidate. This circumstance caused an image of closeness between Prabowo and the incumbent SBY. However, Gerindra actively opposed the government in the SBY administration s second term. The Prabowo s campaign offered, however, a radical change in policy. Prabowo campaigned against corrupt political elite selling the nation to foreign countries (Aspinall and Mietzner 2014). This campaign position can also be seen as a contradiction that complicated Prabowo s relationship with the SBY. As for the sociotropic analysis, existing research is usually conducted at the national level (Lewis-Beck and Stegmaier 2013). Many sociotropic researchers analyzed the impact of the local economic condition, such as Burbank (1997) who found that observing the local economy is a form of low-cost search information for

8 122 Asia-Pac J Reg Sci (2018) 2: the voters to assess the incumbent s performance. Healy and Lenz (2017) argued that voters often over-estimated the authority of the central government. Some responsibilities under the local government s authority can be mistakenly credited to the central government s authority. Therefore, voters may attribute positive regional economic conditions to the president s performance and re-elect the incumbent to maintain good economic conditions. In the 2014 presidential election, with little proximity between the two candidates and the incumbent, village economic conditions may not have much effect on the results. 3.2 The role of media in voting behavior Voter assessment of economic conditions is not only conducted by observing their nearest environment, but also through the media (Vigna and Kaplan 2007). In the context of the 2014 presidential election, the media s role was undoubtedly important, because most media owners, especially the nation-wide broadcast networks, were also politicians involved in the election. TVOne and ANTV were owned by Aburizal Bakrie, the leader of Golkar party, and MNCTV, RCTI, and Global TV were owned by Hary Tanoesudibyo, both of whom were Prabowo supporters. Jokowi supporters included Surya Paloh, the founder of Nasdem party and owner of MetroTV. The TV stations candidate endorsements, including black campaigns, were abundant, with the result that Indonesia Broadcast Commission (Komisi Penyiaran Indonesia/KPI) and the Election Supervisory Board (Badan Pengawas Pemilu/ Bawaslu) issued a warning to TV stations against the broadcasting of biased information. However, biased information eventually affected voters who watched television. Since more television stations supported Prabowo than endorsed Jokowi, television access likely influenced voters to choose Prabowo. In addition to television, the Internet, especially social media, was also a significant phenomenon in the 2014 presidential election. Kushin and Yamamoto (2010) found that Internet access is positively and significantly related to political engagement. The power of the Internet can be seen in Facebook s prediction analysis of the election based on mentions, where the resulting turnout was close to the actual election results. 2 Approximately 83 million Indonesians are Internet users, highlighting the important of the Internet influencing voters decision. Both Jokowi and Prabowo use the Internet as part of their media campaigns, but Prabowo had a more organized Internet campaign team than Jokowi s which tended to be sporadic. 3 Prabowo s campaign team was both well-funded and well-organized, while Jokowi s campaign team was dominated by volunteers, and had to use crowdsource fundraising to fund his campaign. The campaign machine s effectiveness was shown in Prabowo s approval rating which increased dramatically from 20 to 45% within weeks (Aspinall and Mietzner 2014). Therefore, media exposure from both television and the Internet 2 Indonesia Election Tracker, Suara Indonesia ( ook.com/faceb ookin dones ia). 3 Serial Tweets 2014.

9 Asia-Pac J Reg Sci (2018) 2: Table 2 Analysis of social media on smear campaign and negative campaign. Source: indicator Indonesia and Politikawave Jokowi Prabowo Indicator Indonesia Politikawave Indicator Indonesia Politikawave Bus procurement Figurehead candidate 1998 abduction Psychopat Traffic congestion Case suspension First lady Violence Image oriented Teacher certificate Citizenship Citizenship Megawati s puppet Communist a Stock transaction Foreigner minion Christian a Call down Esemka Car Supported by Jewish a Christian a Could not ablution a Shiite a Could not pray a Chinesse a Could not cite Koran a a Religion- and ethnicity-related issues likely influenced voters to choose Prabowo on the strength of Prabowo s media control. Hence, the village with media exposure is likely to choose Prabowo. During the 2014 Indonesia presidential election, media also became a tool for negative and smear campaigns. A smear campaign slanders the competitors with inaccurate data. 4 In the Indonesian context, the smear campaigns often used religious and ethnic sentiments, two social factors that have historically caused considerable conflict in Indonesia. Indicator Indonesia 5, which analyzed social media during the campaign, noted that both presidential candidates were subjected to negative and smear campaigns (Table 2). Table 2 shows that smear campaign with religion and ethnicity-related issues (Christian, Communist, Shiite, Chinese could not pray/ablution/could not recite the Koran and supported by Jews) was extended only toward the Jokowi. Chinese and Jewish issues are about ethnicity, but since most Indonesian Chinese were not Muslim, it would be a religious issue. The target of the religious smear campaign was designed to change the voter choice of the Muslim community. For this reason, Muslim communities with access to media, which was largely biased in favor of Prabowo, were expected to be influenced by this smear campaign and to choose Prabowo. Hence, the Muslim community not exposed to information may not be much affected by the smear campaign, making it more likely to choose Jokowi. 4 A smear campaign is an effort to damage someone s reputation. A black campaign or propaganda is an effort to damage someone s reputation by spreading false or incorrect information, while a negative campaign is the process of spreading negative information about someone to worsen the public image. 5 esiai ndica tor.org/.

10 124 Asia-Pac J Reg Sci (2018) 2: a x b p q r s t Left ½x ½x Right Fig. 2 Ideology spatial. Source: re-drawn from Downs (1957) 3.3 Ideology Downs (1957) adopted Hotelling s (1929) explanation of the spatial effect of ideology on the vote. Ideology is described as a one-dimensional line, where the ends of each are the left, which describes socialism, where the state plays a significant role in the economy, and the end of the right, which is capitalism, where the state plays a minor role in the economy. If there are two candidates (a and b), where a is on the left and b on the right, then voters on the left side of a will vote a (p and q), and voters on the right side of b will choose b (t). While voters who are in between a and b will choose who has the closest distance (Fig. 2). Thus, the candidates are trying to get to the midpoint to obtain the most possible votes. The voter group in the middle (median voter) is described as the middle-class based on income level (Ananta et al. 2004, 1978). Some studies analyze the role of the middle class in the political process, among of them are Inman (1978) and Manza et al. (1995). The middle class is defined as an individual with income level from of US$2 to US$20 per capita per day; in 2013, Indonesia s middle class reached 56.7% of the population (Kriswasana S 2014). The middle class is often referred to as the main driver of the economy because of high productivity and financial capabilities that allow for high consumption. They play a major role in the economy with a decisive role in a country s process of democratization (Barro 1999). In the 2014 presidential election, the background of the candidates main party supporter tended to target a lower class of the community. Thus, the Gerindra campaign tried to associate with farmers and fishermen, since Gerindra s leader Prabowo Subianto was the HKTI (Himpunan Kerukunan Tani Indonesia/Organization of Farmer of Indonesia) leader, and PDIP s slogan was the Partai Wong Cilik (Grassroots Party). Likewise, both candidates offered economic platforms with some populist economic programs, such as agrarian reform, as well as self-sufficiency on energy and food. However, the two candidates show different approaches to realizing their visions. For example, in the case of food self-sufficiency, Jokowi s approach emphasized supply side, which focused on increasing agricultural production to meet local demand. Prabowo, however, gave greater weight to state intervention on prices and distribution. Moreover, in the energy sector, to reduce fuel subsidies, Jokowi used a price approach by raising subsidized fuel prices (premium), while Prabowo

11 Asia-Pac J Reg Sci (2018) 2: Prabowo Jokowi Left Right Fig. 3 Ideology position of Prabowo and Jokowi. Source: authors advanced the idea of imposing a tariff on the rich (apparently rich people meant car owners). While Prabowo planned to nationalize foreign mine companies, Jokowi held that Indonesia should respect government contracts with companies and proposed renegotiating contracts with a better deal for Indonesia when current contracts ended (Aspinall and Mietzner 2014). Based on these examples and other economic programs proposed by the two candidates, when Jokowi and Prabowo are compared, Prabowo aspired to more state intervention into the economy, thus Prabowo s ideological position sits more to the left than Jokowi s (Fig. 3). Therefore, it is estimated that the middle class will tend to choose Jokowi, while the lower income group will choose Prabowo. 3.4 Long term factors Beyond the economic factors, which are short-term, another factor with a long-term effect, and embedded in the community for a long time, is social cleavage (Anderson 2000; Nadeau, et al. 2013). This factor is the result of a long history in relations between the community group and a political party or other political affiliation. Social cleavage includes religion, race/ethnicity, education level, and area characteristics (i.e., urban/rural area). According to Baswedan (2004), the role of religion in Indonesian politics can be illustrated as a spectrum of political parties, which varies from secular parties at one end (PDIP), followed by Golkar, PAN, PKB, PKS, and PPP, to the most religious parties (in the Indonesian context: most Islamist) at the other end, the PBB. In the 2014 presidential election, new parties threw their hat into the political ring, namely: Democrats, Gerindra, Nasdem, Hanura, and PKPI. These parties can be placed on Baswedan s spectrum and coincide with Golkar, given that the parties ideologies are similar and their founders are either former Golkar members or former army generals from the new order era. In the coalition process that could only produce two alliances of candidate supporters, the Jokowi coalition tends toward the secular, because only one party in the Jokowi coalition can be grouped as a religious party (PKB), while the Prabowo coalition is more likely to include religion (Islamic) because of its many religious parties (PKS, PPP, PAN, and PBB). The Prabowo coalition was supported

12 126 Asia-Pac J Reg Sci (2018) 2: Jokowi (PDIP, Nasdem, PKB, Hanura, PKPI) Prabowo (Gerindra, PAN, Golkar, PKS, PPP, PBB) Secular Islamist Fig. 4 Religious spectrum of the 2014 presidential election candidates. Source: modified from Baswedan (2004) by a well-proven Islamist political network owned by PKS, but Prabowo was also endorsed by an Islamist militant group, the Front Pembela Islam (Islamic Defenders Front) (Mietzner 2014). Based on this analysis, the two coalitions can be plotted into Baswedan s religious and political spectrum in Fig. 4. Thus, the Muslim-majority villages hypothetically tend to choose Prabowo, because the KMP coalition character is more Islamic than the Jokowi s coalition. By following Liddle and Mujani (2007), in addition to religious variables, this study also includes the variable ethnicity, with level of education and urban as control variables. In the case of parliamentary election in Indonesia, religion as well as ethnicity significantly influenced the election outcome (Ananta et al. 2004; Higashikata and Kawamura 2015). 4 Empirical strategy 4.1 Econometric model This study seeks to capture whether there is a close relationship between economic voting as well as socioeconomic factors at the village level and a village voter s behavior in choosing a candidate in the 2014 presidential election. This study transformed the conceptual framework of Fig. 1 into an empirical econometric model. The model adopts Nadeau et al. s (2013), framework where the vote is determined by economic voting, information, ideology, and social cleavage. The econometric model is as follows: PJ i = α + J K β j EconVote ji + γ k Info ki + j=1 k=1 L M γ l Ideo log y li + δ m Social mi + ε i l=1 m=1 where i is a village, i = 1, 2, 3,, 51,503; PJ is a dummy variable, where 1 is the vote for Jokowi at the village-i is more than 50% (Jokowi won), while 0 is the vote for Prabowo at the village-i is more than 50% (Prabowo won). Based on the nature of the dependent variable, this study applies both logistic and probit estimation. Complementing logistic and probit estimation, we perform ordinary least square (OLS)

13 Asia-Pac J Reg Sci (2018) 2: estimation for consistency and robustness check. The OLS estimation uses Jokowi s vote share (percentage vote for Jokowi in village-i) as dependent variable. OLS estimation uses robust standard error estimations to deal with heterocedasticity issues. Following the conceptual framework in Fig. 1 that the individual vote is a discrete choice either 1 (choose Jokowi) or 0 (choose Prabowo), then this study will analyze the result of logistic model instead of OLS result. The estimation results from both logistic model and OLS model should theoretically be similar in the magnitude. Another reason to use a logistic regression result as a basis of analysis is a practical reason, since a candidate and campaign team concerns how to win the election instead of how many votes that can be collected at the village level. There is a possibility losing some information when we converted continuous data (vote share) into a discrete data. However, as our concern is in elaborating the marginal effect (probability) of some variables on the Jokowi s victory, then the best method is the logistic model. EconVote represents a vector of economic voting indicators. Many economic voting studies use income level, unemployment, and inflation as indicators. Due to data limitation at the village level, this study uses socioeconomic conditions at the village communities: (1) the proportion of the poor; (2) availability of economic access indicated by the availability of banks at the village; and (3) availability of infrastructure indicated by paved roads and markets at the village. Based on economic voting theory, villages with good economic conditions (a small proportion of poor people, good economic access, and sufficient infrastructure) are expected to choose Prabowo as incumbent (in this case as a candidate Prabowo is closer to the incumbent). Due to a complicated relationship between Prabowo and SBY as well as unclear association between KMP and incumbent, however, voters in village with good infrastructure and economic conditions may also choose Jokowi. Info represents access to information. To determine the impact of information on the election results, with a partisan television channel, national television broadcast access is used as an information variable. However, because the government-owned station television TVRI tends to be neutral, we excluded TVRI from the variable. Hence, the variable is: whether the village gained access to private owned television with nation-wide broadcast (dummy). To measure the role of the Internet as part of media, with 60% of Indonesians accessing the Internet through a mobile device 6, the existence of a strong cellular signal (so they can access the Internet) in the village serves as a proxy to access information via the Internet. With Prabowo s mastery over the media, it is expected that the village with access to private television broadcasting and a strong cellular signal will tend to choose Prabowo. Ideology represents the ideological position among the villagers. However, it is quite difficult to assess the villagers ideology. Therefore, this study uses middleclass characteristics as a proxy for ideological position (see Appendix 1 for the detailed explanation). Characteristics of the middle class in Indonesia are landowner farmers, entrepreneurs, and white-collar workers, while lower class characteristics are agricultural laborers and industrial workers. For the empirical evidence, 6 Ministry of Communication and Information.

14 128 Asia-Pac J Reg Sci (2018) 2: a middle-class proxy is the percentage of land-owner farmers and the number of SMEs (Small Medium Enterprises which represent entrepreneurs) in the village. The proxies for the lower class are a proportion of farm workers and the dummy whether the majority of the population work as industrial labor. With Jokowi s ideological position more akin to that of the middle class, we hypothesized that the village with a high number of SMEs and a high proportion of entrepreneur farmers tend to choose Jokowi, while villages with a high proportion of farm laborers and the majority of the population are working, as industrial labors tend to choose Prabowo. Social represents social cleavage. Social cleavage includes Muslim-majority population village (dummy). Because the Prabowo coalition is more Islamic than Jokowi, it is assumed that the Muslim-majority village would choose Prabowo. Other variables are ethnicity (whether most of the population is Javanese), urban or rural (dummy), and the level of education shown by the ratio of total population compared with some elementary schools in the village. To determine the influence of the religious smear campaign-related issues to the Muslim community vote, a village with a Muslim majority, but without access to private television broadcasting and cellular signals (dummy), is used as a dependent variable. Because Jokowi is a candidate who becomes a target for religion-related smear campaign issues, it is assumed that the black campaign did not reach the Muslim-majority village without media access. Thus, the villagers will show some different behaviors within a Muslim-majority village with media access with greater exposure of religion-related smear campaign. 4.2 Village potential census and crowdsource election data This study estimates the econometric model using a cross-sectional data of merging PODES and the crowdsource data. The dependent variable of vote data uses the 2014 presidential election results from Kawal Pemilu (Election Monitoring Group), while explanatory variables use the 2011 Village Census (PODES) published by the Central Statistics Agency (BPS). BPS conducts a PODES census every 3 years, which most likely captures infrastructure and limited socioeconomic conditions at the village level. Yet, Kawal Pemilu s crowdsource data are a vote recap at the polling station level, consisting of 478,828 polling stations data across Indonesia. The polling station-level data were then aggregated at the village level, and up to 81,138 of the vote, data were obtained at the village level in 497 district/cities in Indonesia. Soon after the presidential election was over, pollsters released their quick count prediction results. Most predicted that Jokowi won the election, albeit with a small gap. In this critical time, after election day until the KPU release of the official results, four pollsters (which owned or were funded by Prabowo s allies) released quick count result showing that Prabowo had won the election. In that uncertain time, immediately after the polls closed, the Election Commission s policy to uploaded election results from every polling station made the 2014 presidential election the most transparent election in Indonesian history (Mietzner 2014). Kawal Pemilu, which is a small number of founders and

15 Asia-Pac J Reg Sci (2018) 2: Table 3 Gap vote between data before and after combination. Source: authors calculation No. Province Before combination (Kawal Pemilu) After combination (Kawal Pemilu & PODES) Gap before and after (%) Number of vote Jokowi vote (%) Number of vote Jokowi vote (%) 1 Aceh 1,860, ,277, Bali 2,147, ,656, Banten 5,576, ,577, Bengkulu 956, , Yogyakarta 2,208, ,312, DKI Jakarta 5,377, ,425, Gorontalo 599, , Jambi 1,764, ,136, West Java 23,545, ,947, Central Java 19,389, ,088, East Java 21,501, ,529, West Kalimantan 2,598, ,044, South Kalimantan 1,860, ,476, Central Kalimantan 1,161, , East Kalimantan 1,850, ,473, Bangka Belitung 612, , Riau Islands 819, , Lampung 4,317, ,706, Maluku 850, , North Maluku 559, , West Nusa 2,523, ,057, Tenggara 22 East Nusa 2,129, ,584, Tenggara 23 Papua 1,747, ,089, West Papua 502, , Riau 2,674, ,846, West Sulawesi 619, , South Sulawesi 4,242, ,646, Central 1,398, , Sulawesi 29 South East 1,124, , Sulawesi 30 North Sulawesi 1,249, , West Sumatera 2,327, , South Sumatera 4,144, ,971,

16 130 Asia-Pac J Reg Sci (2018) 2: Table 3 (continued) No. Province Before combination (Kawal Pemilu) After combination (Kawal Pemilu & PODES) Gap before and after (%) Number of vote Jokowi vote (%) Number of vote Jokowi vote (%) 33 North Sumatera 6,292, ,148, Total 130,533, ,135, ORIGINAL DATA (million) COMBINED DATA (million) ACEH BALI BANTEN BENGKULU YOGYAKARTA DKI JAKARTA GORONTALO JAMBI WEST JAVA CENTRAL JAVA EAST JAVA WEST KALIMANTAN SOUTH KALIMANTAN CENTRAL KALIMANTAN EAST KALIMANTAN BANGKA BELITUNG RIAU ISLANDS LAMPUNG MALUKU NORTH MALUKU WEST NUSA TENGGARA EAST NUSA TENGGARA PAPUA WEST PAPUA RIAU WEST SULAWESI SOUTH SULAWESI CENRAL SULAWSI SOUTH EAST SULAWESI NORTH SULAWESI WEST SUMATERA SOUTH SUMATERA NORTH SUMATERA Fig. 5 Comparison of vote data before and after combined. Source: authors comparison hundreds of independent volunteers, digitized the uploaded documents and created alternative data that could be monitored online. These data became an important reference between two different quick count predictions before the official results were released by KPU. It turns out that Kawal Pemilu s count was close to the Election Commission s result (53.01% Jokowi vote versus 53.15% Jokowi vote). There was a gap between the original and combined data, because some village names are not the same in the Kawal Pemilu data and the PODES data. However, missing data are relatively evenly distributed throughout the provinces in Indonesia. Table 3 and Fig. 5 describe the comparison between the original Kawal Pemilu data and the merged data of PODES and Kawal Pemilu in each province in Indonesia. When comparing the percentage of votes for each candidate, the differences before and after the merging in each province are relatively small, whereas the most significant difference is in the province of North Maluku (3.89%). On average, however, the difference is equal to 0.41% (Table 3). The merged data are not that much different from the original data, meaning that the attrition data are randomly distributed.

17 Asia-Pac J Reg Sci (2018) 2: Table 4 Estimation results. Source: authors estimation Dependent variable Dummy Jokowi won (1 = vote > 50%, 0 = other) Vote for Jokowi (%) Independent variables Logit SE Probit SE OLS* Robust SE Economic voting variables Poverty (%) (0.002) (0.001) (0.013) Economic access (1 = availability of bank; 0 = other) *** (0.018) *** (0.011) *** (0.131) Infrastructure (1 = availability of paved roads and market; 0 = other) *** (0.018) *** (0.011) * (0.133) Access to information Cellular signal (1 = good signal; 0 = other) *** (0.032) *** (0.019) *** (0.241) Television broadcast access (1 = availability of private national television broadcast; 0 = other) *** (0.042) *** (0.015) *** (0.288) Ideology (middle class) Agriculture entrepreneur (%) *** (0.000) *** (0.000) *** (0.003) SMEs (number) *** (0.000) *** (0.000) *** (0.001) Agriculture laborer (%) (0.001) (0.000) *** (0.003) Industry laborer (1 = majority of the population work as industrial labor; 0 = other) *** (0.056) *** (0.035) *** (0.355) Social cleavages Moslem majority (1 = Moslem majority; 0 = other) *** (0.042) *** (0.022) *** (0.244) Java majority (1 = Javanese majority; 0 = other) *** (0.025) *** (0.015) *** (0.169) Moslem majority without media (1 = Moslem majority but lack access of media; 0 = other) *** (0.046) *** (0.027) *** (0.312) Urban (1 = urban area; 0 = other) *** (0.036) *** (0.022) (0.264) Education (%) (0.000) (0.000) (0.000) Constanta *** (0.061) *** (0.034) *** (0.399) Pseudo R 2 (R 2 for OLS) Wald χ 2 (F for OLS) 8, , , 515 Number of observations 51,503 51,503 51,503 *** Significant in 1%; ** significant in 5%; * significant in 10%. Figures in parenthesis are standard error (*robust standard error in OLS)

18 132 Asia-Pac J Reg Sci (2018) 2: Discussion and analysis Table 4 shows results of three methods using logit, probit, and OLS estimations. The OLS estimation result is similar to the logit and probit estimations. The only difference sign was in the poverty variable, but it can be ignored, because the poverty variable is insignificant. These three estimations confirm the robustness of estimations. For the deeper analysis, we typically rely on the logit result, because this method underlies the theoretical micro approach that the dependent variable is a discrete choice whether Jokowi won or not. Table 5 shows the marginal effect of logit estimation. This result showed that the villages with positive economic conditions, such as the availability of good infrastructure (paved roads) and economic access (availability of banks), tended to choose Jokowi. A strong tendency that villages with better economic conditions prefer Jokowi raises the question of whether economic voting took place. PDIP, as the main Jokowi supporter party, was consistently acting as the opposition during SBY s two presidential periods making it difficult to see the closeness of SBY with Jokowi. The explanation may be found in Fiorina s (1978) retrospective prospective concept. Fiorina (1978) explained that retrospective analysis (economic conditions of the past) is a way for voters to predict what will happen in the future (prospective). It is also in line with Miller and Wattenberg (1985) who argued that in an election without an incumbent, voters evaluate economic performance in the form of mix between prospective and retrospective. The analysis of economic programs offered by the two candidates showed Jokowi more accommodating toward the market and a preference for a smaller portion of government intervention in the economy than Prabowo. Jokowi also was more pragmatic, offering bureaucracy and system improvements to create better policy and services from government. Yet, Prabowo proposed a radical change to the system that would change many aspects of government and Indonesians daily life (Aspinall and Mietzner 2014). SBY s economic program also tended to be pro-market with less state intervention into the economy (by Indonesian standards). Thus, the public might not have noticed drastic changes if the next president was Jokowi, while the public expected a significant change in economy if Prabowo became president. Therefore, the village with positive economic conditions is likely trying to maintain the level of utility obtained during the SBY period and to ensure little change in the next administration s economic policies. Thus, villagers will prefer Jokowi to Prabowo. While there was some disappointments over SBY s policies during his second term, the Indonesian economy was quite strong in Thus, voters did not expect radical change (Aspinall and Mietzner 2014). The economic voting behavior in Indonesia has not much changed since Higashikata and Kawamura (2015) using the 2004, 2009, and 2014 election data indicated the existence of retrospective economic voting in the parliamentary elections, in which voters assess government performance based on the economic conditions that voters have perceived. Using district & city Gross Domestic

19 Asia-Pac J Reg Sci (2018) 2: Table 5 Marginal effect of logit model. Source: authors estimation Variables independents dy/dx Std. error Economic voting variables Poverty (%) Economic access (1 = availability of bank; 0 = other) *** Infrastructure (1 = availability of paved roads and market; 0 = other) *** Access to information Cellular signal (1 = good signal; 0 = other) *** Television broadcast access (1 = availability of private national television broadcast; 0 = other) *** Ideology (middle class) Agriculture entrepreneur (%) *** SMEs (number) *** Agriculture laborer (%) Industry laborer (1 = majority of the population work as industrial labor; 0 = other) *** Social cleavages Moslem majority (1 = Moslem majority; 0 = other) *** Java majority (1 = Javanese majority; 0 = other) *** Moslem majority without media (1 = Moslem majority but lack access of media; 0 = other) *** Urban (1 = urban area; 0 = other) *** Education (%) *** Significant in 1%; ** significant in 5%; * significant in 10%

20 134 Asia-Pac J Reg Sci (2018) 2: Regional Product (DRDP) per capita, they found that districts experiencing a higher economic growth gave more support to the ruling parties. Therefore, those who enjoyed a good economic performance of SBY era tends to choose Jokowi due to non-radical change. This is also consistent with the results of Mujani et al. (2018) that use micro-models on four elections since They found that voters tend to favor incumbent if economic conditions are good enough, while voters who consider poor economic conditions tend to vote opposition. These studies are in line with our results, indicating that regions (villages) with better economic conditions tend to prefer incumbent (associated with Jokowi), and vice versa. If our explanation is correct, interestingly, the variables, such as the proportion of poor people, the level of access to the village economy, and infrastructure conditions, explained economic voting, even though the variables are not commonly employed to describe economic voting; usually, the variables used include unemployment rate, inflation rate, growth rate (Lewis-Beck and Stegmaier 2013) and debt default conditions, level of employment, and wage levels (Healy and Lenz 2017). In contrast, the variables used in this study are tangible and more easily seen directly by the villagers, such as banks, markets, and the condition of roads. This study shows that village facilities like banks, markets, and infrastructure conditions in the village play important role in the voter s evaluation at the time of choosing a president. The probability of Jokowi getting more than 50% votes increases by almost 2% in the village with good infrastructure. While local road construction and economic facilities like the market belong to the district/city s authority, they still play a role in the presidential election. These findings confirm Healy and Lenz s (2017) finding that the public often over-estimates the central government s authority. The voter considers the good performance of district/city governments in developing economic and infrastructure facilities in the villages to be a central government s (i.e., president s) achievement. The ideology represented by middle-class variables, such as agricultural entrepreneurs (landowners, farmers) and number of SMEs, shows a stable positive result. The villages with a sizeable middle-class population tended to choose Jokowi. The village dominated by middle class tended to vote Jokowi, while lower class society (village dominated by agricultural laborers and industrial workers) tended to prefer Prabowo. Aspinall and Mietzner (2014) maintained that Prabowo s proposal would be interesting to poor voters with economic disadvantages and less access to government institutions. Inman (1978) confirmed that the median population determines the victory in voting. Candidates ideology in the form of economic programs became an essential element in voters assessment of the candidates. The 2014 presidential election also showed that media control is a noteworthy factor. Villages with television broadcasts or good signals showed a significant and robust tendency to choose Prabowo. The probability of Jokowi won in the village with a good signal of private national television broadcast and cellular signal decreases 11 and 5.4%, respectively. This finding is similar to Vigna and Kaplan s (2007) study, which found that biased media can influence political decisions. Kushin and Yamamoto (2010) also confirmed that the Internet, especially social media, has a significant role in shaping a political choice. With 82% of Indonesia s