ETHNIC AND RELIGIOUS SENTIMENTS IN INDONESIAN POLITICS: EVIDENCE FROM THE 2017 JAKARTA GUBERNATORIAL ELECTION

Abstract Studies have documented how ethnic and religious sentiments shape the voting behavior of Indonesian Muslims. However, to date no studies have carefully measured the relative influence of these sentiments. I fill this gap in the literature by taking advantage of the candidacy of a Christian, ethnic Chinese candidate in the 2017 Jakarta gubernatorial election in Basuki Tjahaja Purnama (Ahok). Employing an original survey of Jakartan Muslims, I show through experimental and correlational analyses that Muslim voters are more opposed to Ahok than non-Muslim voters are and that this opposition is driven more by Ahok's ethnicity, as opposed to his religion. I also show that Muslim voters’ feelings toward ethnic Chinese shape their support for Ahok more than their feelings toward Christians. I discuss how these findings inform our understanding of the limits and extent of religious influence on Muslim voting behavior.

All these led studies to highlight Ahok's religion as the dominant point of contention in the election.
While convenient, this religion-focused explanation is limited in three ways. First, on a conceptual level, it fails to distinguish between voter religion and candidate religion. Second, it confounds political messaging and mobilization with voter attitudes. Third, it relies too heavily on voters' personal accounts of their voting rationales. The present study is an effort to improve these limitations and to test more carefully the extent to which Ahok's ethnicity and religion shaped voting behavior in the election.
Specifically, I analyzed an original survey of Jakartan Muslims that included both experimental and correlational modules. My analysis yields three main findings. First, Muslim respondents were less likely to support Ahok than non-Muslim ones. Thus, voter religion matters. Second, priming respondents with Ahok's ethnicity decreased their support for him. There was no change caused by the priming of Ahok's religion. In other words, there is only limited effect of candidate religion. Lastly, I find that respondents' feelings toward ethnic Chinese shaped their support for Ahok more than their feelings toward Christians. These findings, in turn, highlight the importance of differentiating the influence of voter religion and candidate religion and of more carefully examining the influence of different intersecting identities.

R E L I G I O U S A N D E T H N I C S E N T I M E N T S I N I N D O N E S I A N P O L I T I C S
Religious and ethnic sentiments have always played major roles in Indonesian politics. In terms of religion, ever since its founding, Indonesia has had to face a debate over whether or not it should enshrine sharia in its constitution (Elson 2009). Studies also have shown a growing influence of religious organizations. Conservative Islamic organizations have contributed to the increase of societal discrimination against religious minorities (Sumaktoyo 2020) and the conservative segment of Indonesian Muslim voters have become more influential as a result of better organizational capacity .
Among the many consequences of religious influence in politics, one that is of particular relevance for the present study is how it shapes voting behavior. There is ample evidence that religion shapes Indonesian Muslims' voting preferences. Surveys by the Indonesian Survey Institute (Lembaga Survei Indonesia 2018) find that in 2018, about 52 percent of Indonesians objected to having a non-Muslim governor, an increase of 12 points from 2016. Analyzing the 2014 presidential election results, Gueorguiev, Ostwald, and Schuler (2018) also find a strong religious influence in that conservative Muslim areas strongly favored Prabowo Subianto over Joko Widodo. Sumaktoyo, Ottati, and Untoro (2016) show that being religious improves voter support for a candidate and Pepinsky, Liddle, and Mujani (2012) find that an Islamic label affects voter support for a party under economic uncertainty. In a recent study, Fossati (2019) finds that religious ideology shapes Indonesian Muslims' voting and policy preferences.
Religion is far from the only factor that shapes Indonesian politics. Other studies examine another type of identity, namely ethnicity (Toha 2017;Aspinall 2011;Mavridis 2015). Of relevance to the present study are studies that examine anti-Chinese sentiments and their political consequences. Indonesia has a long history of anti-Chinese sentiment traceable to the Dutch colonial era (Purdey 2006). Under Dutch rule and with their encouragement, ethnic Chinese consolidated their influence in trade and commerce, separating themselves from the locals legally, economically and, to an extent, socially. This marked the beginning of deep-seated tensions between locals and ethnic Chinese.
Soeharto's 32 years of dictatorship exacerbated these tensions. He outlawed the celebration of Chinese culture and issued a policy requiring ethnic Chinese individuals to apply for a letter clarifying their citizenship. As an example of the absurdity of these policies, Indonesia's first Olympic gold medalists, Alan Budikusuma and Susi Susanti, who are of Chinese descent, were asked to prove their citizenship when registering their marriage.
One of the darkest episodes for Chinese Indonesians was the three days between the 13th and the 15th of May, 1998, shortly before Soeharto's resignation. Riots broke in the nation's capital, Jakarta, and mobs targeted Chinese Indonesians. At least 1,000 ethnic Chinese were murdered, 168 women of Chinese descent raped, and thousands of ethnic Chinese-owned businesses vandalized or looted (Wibowo 2001). Many business owners wrote "Belongs to natives" ("Milik pribumi") on their store fronts to avoid being looted. A fact-finding commission created by Soeharto's successor Habibie later issued a report which includes descriptions of possible military involvement. The report went unheeded by subsequent administrations, which took no effort to either find the perpetrators or rehabilitate the victims.
These anti-Chinese sentiments translate to the electoral arena. Except in the few regions where ethnic Chinese constitute a significant percentage of the population (e.g., Singkawang in Kalimantan, where 40 percent of the population is ethnic Chinese; Yew-Foong 2017), ethnic Chinese political candidates, if they even decided to run at the top of the ticket in the first place, face serious electoral challenges. Aspinall, Dettman, and Warburton (2011) document such an opposition against a non-Muslim, ethnic Chinese candidate running to be a mayor in North Sumatra. More generally, a nationwide survey finds that 64.4 percent of Indonesians were uncomfortable with an ethnic Chinese in a position of political leadership (Setijadi 2017a).
All these examples demonstrate the influence of religious and ethnic sentiments in shaping Indonesian politics. Unfortunately, to date no studies have analyzed the relative influence of the two sentiments simultaneously using a careful quantitative analysis. Such an analysis, however, promises insights into how identity politics plays out among Indonesian Muslims.
Because religious and ethnic sentiments are often strongly related (Chandra 2006), without a careful examination, what looks like religious influence in politics might actually be ethnic influence and vice versa. This is particularly true for Indonesian Muslims, who studies have shown have a history of tensions with Christians (Menchik 2016;Arifianto 2009) and ethnic Chinese (Purdey 2006;Azra 1994). The 2017 Jakarta gubernatorial election offered an opportunity to fill this gap in the literature. The presence of a Christian, ethnic Chinese candidate provided us with an opportunity to examine which of these two characteristics shape voter behavior more strongly.

T H E J A K A R T A E L E C T I O N I N A N U T S H E L L
The 2017 Jakarta gubernatorial election was the third for the city, after elections in 2007 and 2012. The first round of the election saw three pairs of candidates competing for the office: Agus Yudhoyono-Sylviana Murni, Basuki Tjahaja Purnama (Ahok)-Djarot Saiful Hidayat, and Anies Baswedan-Sandiaga Uno. Despite being the incumbent governor, Ahok was not elected to the position. He had been the vice governor for Joko Widodo when they ran together in the 2012 Jakarta election, and he replaced Widodo as governor when the latter won the presidency in 2014. This means that the 2017 election was the first time Jakartans had a minority candidate at the top of the ticket.
While other localities adopt a plurality system, Jakarta uses a majority system due to its status as capital. If no candidate wins the majority of votes, a runoff will pit the top two candidates against one another. The first round of the election was held on February 14, 2017. Table 1 presents the candidates and their party coalitions in this first round. We see in the table that Ahok's vote share was lower than what we would have expected had the  voters voted along party lines. A runoff was held on April 12, 2017, pitting Ahok against Baswedan. Of the parties that supported Yudhoyono in the first round, the National Awakening Party and the United Development Party lent their support to Ahok, the National Mandate Party supported Baswedan, and the Democrat Party chose to be neutral. Baswedan ended up winning the runoff with 57.96 percent of the votes, well above the vote share of his coalition (26.8 percent). Ahok, therefore, lost the election despite the fact that the two new members of his coalition (the National Awakening Party and the United Development Party) are Islamic parties and that at least 65 percent of the voters were actually satisfied with his work as governor (e.g., own survey; SMRC 2016).
What explained Jakartan Muslims' voting behavior in the election? Why did Ahok lose despite his strong approval rating? Most accounts of the election focus on the mobilization of religious sentiment against Ahok (Arifianto 2017;Aspinall 2017;Chaplin 2017;Fealy 2016;Lindsey 2016;Mietzner, Muhtadi, and Halida 2018), suggesting the importance of religion and religious sentiment. Three pieces of evidence support this account of the election.
The first relates to the fact that Ahok is the only non-Muslim in the ballot. This was bound to ignite controversies. Conservative Muslim groups held demonstrations to oppose Ahok's candidacy on this ground. The issue of non-Muslim political leaders is also a controversial topic in Indonesia, where the majority of Indonesian Muslims are opposed to having a non-Muslim leader (Lembaga Survei Indonesia 2018).
The second piece of evidence comes from polls that show how religious similarity was a major voting consideration in the election. A survey by the Indonesian Survey Circle (Lingkaran Survei Indonesia) finds that 71 percent of respondents deemed it important to choose a governor who shares their faith (Andreas 2017). A survey by Saiful Mujani Research and Consulting (2016) also finds that almost 12 percent of Baswedan would-be voters said they supported him out of religious similarity.
The third piece of evidence is related to the blasphemy charge against Ahok. In October 2016, Buni Yani, a former journalist, uploaded to his Facebook account a video of Ahok giving a speech in which he referred to the Qur'an's Surah Al-Maidah Verse 51. The verse was relevant to the election as some interpret it as justifying opposition to a non-Muslim leader. Yani captioned the video with a misleading transcription that made it appear as though Ahok was saying that the voters are "lied to by Surah Al-Maidah Verse 51," while he was actually saying "lied to by the use of Surah Al-Maidah Verse 51." The incorrect transcription implied that Ahok was suggesting that the Qur'an lies, whereas the actual statement indicates that Ahok was suggesting that some people used the Surah to lie for political gains. Yani's video led to a blasphemy charge against Ahok and created an opportunity for conservative Muslims to field a more organized opposition against him. Two demonstrations demanding Ahok's imprisonment were held on November 4 and December 2, 2016, each attended by over 300,000 people. These demonstrations were unprecedented in the post-Soeharto era, leading observers to argue that moderate groups are losing their influence (Arifianto 2017). More relevant to the election, however, is that these demonstrations changed the face of the anti-Ahok sentiments, from what had been mostly political statements and survey numbers to an overtly religious movement.
On May 9, 2017, three weeks after his electoral defeat, Ahok was sentenced to two years in prison for the blasphemy charge. Interestingly, the court also later found Buni Yani guilty of hate speech through his distribution of the misleading transcription and sentenced him to one and a half years in prison.
The preceding review illustrates how religious sentiments colored the election. I argue, however, that it is premature to highlight Ahok's religion as the dominant factor that shaped Jakartan Muslims' voting behavior unless we explicitly compare it to other antecedents of voting behavior. This is particularly true considering that the pieces of evidence that highlight the primacy of religion, and the studies that rely on them, are limited by conceptual and measurement challenges.
The first limitation of the studies that emphasize the importance of religious sentiments in the election is their lack of distinction regarding whose religion matters, the voter's or the candidate's? Voter religion matters to the extent that we observe different voting patterns between Muslims and non-Muslims. Candidate religion matters to the extent that we observe voter opposition to Ahok being driven largely by his Christian identity.
Existing studies have conflated these concepts by assuming that conservative Muslim voters' opposition to Ahok (i.e. that voter religion matters) also means that it was Ahok's religion who brought about the resistance in the first place (that candidate religion matters). Even in studies that discuss Ahok's ethnicity as an important factor that affected voter support for him (e.g., Aspinall 2017; Fealy 2016; Setijadi 2017b), Ahok's religion is still regarded as the more dominant factor. Aspinall (2017), for example, draws a parallel between the Jakarta election and an election in Medan that also fielded an ethnic and religious minority candidate, arguing how religion seemed to trump ethnicity in both cases.
Evidence for the importance of voter religion (i.e. evidence that Muslim voters are less supportive of Ahok than non-Muslim ones), however, does not automatically mean that candidate religion matters. The lower support for Ahok among Muslim voters may be driven by various factors. It may be driven by religious sentiment or the fact that Ahok is Christian and a Christian should not lead a Muslim-majority city. But it also may be driven by anti-Chinese resentment or the fact that Ahok is of Chinese descent.
The possibility of anti-Chinese resentment driving the anti-Ahok opposition is plausible considering the deep-seated tensions between the Muslim majority and the ethnic Chinese (Azra 1994). Purdey (2006, 6) notes the founding of Sarekat Islam in 1911 as a significant episode in the history of tensions between the indigenous Muslims and the ethnic Chinese. The founding of ICMI or the Association of Indonesian Muslim Intellectuals is another example of indigenous Muslims' organizing against what they perceive as ethnic Chinese's disproportionate influence in Indonesia's social and economic spheres (Sidel 2006;Hefner 1993). That we cannot conclude that Ahok's religion matters just because voter religion matters, in turn, means that a systematic inquiry that seriously considers both Ahok's religion and ethnicity is needed.
The conceptual challenge is compounded by measurement challenges. These challenges affect both studies that take more qualitative approaches and studies that rely on public opinion surveys, albeit in different ways.
The measurement challenges in qualitative studies are evident in approaches that confound political mobilization and voter attitudes. Religious mobilization and rhetoric indeed colored the election. But mobilization and rhetoric tell us more about the strategy that elites use than what factors drive voters' behavior. One may not examine the former (i.e. highlighting the religious demonstrations) and use it as evidence for the latter (arguing that the demonstrations evidence Ahok's religion as the factor that drove Muslim voters' opposition of him).
To illustrate this point we can look at studies on Brexit and the election of Donald Trump. Pro-Brexit politicians often argue for Brexit using economic reasons, such as that Brexit would give the UK more independence to manage its economy. Recent studies, however, have showcased the importance of ethnic or religious sentiments in driving pro-Brexit votes (Iakhnis et al. 2018;Burnett 2017). Similarly, although Trump's America First tapped into economic grievances of the American working class, studies show how anti-immigrant and anti-Muslim sentiments drove support for Trump, often more strongly than economic grievances (Green and McElwee 2019; Lajevardi and Abrajano 2019). What these examples suggest is that we may not use politician rhetoric to make inferences about voter behavior. The two may or may not overlap, which necessitates an explicit study of the voters themselves.
The measurement challenges are also relevant to studies that rely on public opinion surveys. There were two strategies through which surveys gauged the prevalence of religious sentiments in the election. The first strategy used open-ended questions. The surveys asked respondents who had already set their minds on a candidate to mention their primary reasons for supporting the candidate. Religion was consistently one of the top reasons, especially among supporters of Anies Baswedan (SMRC 2016; Warburton and Gammon 2017). Another survey agency, Charta Politika (2017a;2017b), asked respondents who said they would not vote for Ahok to provide a reason for their opposition. The issue of religion was again one of the top reasons offered by the respondents. The second strategy was more straightforward, asking respondents directly whether religious similarity was an important consideration for them. A survey by the Indonesian Survey Circle found that 71.4 percent of respondents believed that it was important to vote for a leader with similar religion (Andreas 2017).
These surveys provided insights into the election and, at least on the face of it, they all highlight the importance of Ahok's religion. However, such a direct measurement Ethnic and Religious Sentiments in Indonesian Politics 147 use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jea.2020.35 approach can be problematic for two reasons. First, the approach demands too much from the respondents. In the case of the approach that asked respondents to name a reason why they supported a candidate, for example, the respondents had to sort through all possible reasons and choose one that was most important for them. Given the nature of humans as cognitive misers who rely on heuristics (Fiske and Taylor 1991;Tversky and Kahneman 1973) and the fact that voters rarely think about politics as deeply as social scientists like to imagine they do (Lupia 2016), the respondents' top-of-the-mind answers were likely shaped more by cognitive biases than by an elaborate thinking process.
If cognitive biases mattered, then why did religion consistently show up as one of the most mentioned considerations? This is because of the religious messaging in the election. Theories of survey responses highlight the importance of elite rhetoric and salient events (Zaller 1992). When respondents were asked why they supported or opposed a candidate, rather than going through all the considerations in mind and carefully weighting their importance, the respondents simply mentioned the most salient one. What justification was most salient, in turn, was shaped by political rhetoric and mobilization at that time, which predominantly targeted Ahok's religion.
The second reason why such a direct approach can be problematic relates to social desirability. Human beings are motivated to portray themselves in a positive light and avoid negative impressions. Given the high level of religiosity and the growing embrace of public piety in Indonesia (Pepinsky, Liddle, and Mujani 2018;Hasan 2009), it is possible that the same social desirability in the direction of providing a religious response affected which voting rationale was offered by the survey respondents. This is particularly true considering that religious rhetoric employed by the elites likely projected a norm that a good Muslim should support a Muslim candidate, motivating respondents to provide a socially desirable religious answer.

T H E P R E S E N T S T U D Y A N D I T S B R O A D E R I M P L I C A T I O N S
The above review explains why the conventional wisdom concerning the Jakarta election emphasizes the importance of religious sentiment and how conceptual and measurement challenges warrant a deeper examination of this conventional wisdom. The primary goal of the present study, therefore, is to address these limitations in the existing studies of the Jakarta election and contribute to the broader debate over how religious and ethnic sentiments shape Indonesian politics. A secondary, but not less important, goal is to contribute to and connect two currently detached lines of research-one on Muslim voting behavior and one on minority politics.
Studies on Muslim voting behavior largely focus on the question of whether or not Muslims unconditionally support an Islamic candidate or party. There is little evidence for such unconditional support (Kurzman and Naqvi 2010;Mujani and Liddle 2009).
To the extent that such support exists, it is evident only when voters are uncertain about the candidate's or the party's performance (Pepinsky, Liddle, and Mujani 2012). Studies on minority politics, on the other hand, examine how and why majority voters are less supportive of minority candidates. In the comparative context, studies have documented how minority candidates generally underperform in British elections (Fisher et al. 2015;Thrasher et al. 2017) or how, in India, non-Muslims are less likely to vote for a Muslim candidate than Muslims are (Heath, Verniers, and Kumar 2015).
An opportunity exists to connect these two research lines. Political behavioral studies that examine how Muslims in a Muslim society respond to a minority candidate are rare. On the other hand, such an examination is normatively important. If we are to improve the quality of democracy in Muslim countries, then how the Muslim majority respond to a minority candidate will be an aspect of political representation that warrants more study. A healthy democracy requires not only free elections, but also an adequate representation of its minority groups.
Given these goals, the present study is both an effort to understand ethnic and religious influence in Indonesian politics more specifically and an exploration of the limits and extent of religious influence in Muslims voters' evaluation of a minority candidate more broadly. Specifically, using an experimental approach that circumvents the aforementioned conceptual and measurement challenges, it tests four predictions related to how Ahok's ethnic and religious identities influenced Muslim voters' support for him.
The first three predictions are derived from the above review concerning Indonesian Muslims' tension with non-Muslims and Chinese Indonesians. First, we should expect voter religion to matter-Muslim voters should be less supportive of Ahok than non-Muslim ones. For the second and third predictions, we expect Ahok's religion and ethnicity to negatively affect his electoral prospect among Muslims voters. If Ahok's religion influenced the voters' support for him, then it would evidence that candidate religion matters-an assertion that existing studies have made, but actually conflate with voter religion. The fourth prediction is related to the relative importance of Ahok's ethnicity and religion. Would Muslim voters' support for him be more negatively affected by his ethnicity or religion? As such a comparison has never been explored before, I do not have an a priori prediction for this question and instead expect one of three scenarios to happen.
The first scenario is that the two would matter equally. The second scenario is that religion would matter more than ethnicity. As we concern ourselves with Muslim voters, then religion is the one divide that separates Ahok and the voters. The third scenario is that ethnicity would matter more than religion. This could happen given the deepseated resentment against ethnic Chinese and given the fact that this resentment was institutionalized by the state under Soeharto.  The curious reader might wonder why H2, H3, and H4 focus on Muslim voters. The reasons are both theoretical and pragmatic. First, Muslim voters, as the majority religious group, are the most important voting bloc. They are also the focus of the existing studies of the election. Focusing on Muslim voters, therefore, puts the present study in direct dialogue with these studies. Second, in the context of examining resistance against a minority candidate, ethnic and religious sentiments arguably shape the majority group's attitudes more because the minority groups likely would overwhelmingly support the candidate out of a feeling of minority solidarity (Glasford and Calcagno 2012). Lastly, there are very few non-Muslim respondents in the sample, limiting the types of analysis that I can carry out with these respondents.

M U S L I M V O T E R S A N D S U P P O R T F O R A H O K
To test these hypotheses, I fielded an original survey in Jakarta in November and early December of 2016. 1 A stratified random sampling was used to obtain a sample of Jakartans who were older than seventeen years old or had been married (i.e. eligible voters). Details of the sampling are available in the Online Appendix along with characteristics of the sample. A total of 1,195 respondents were interviewed face-to-face, of which 1,047 were Muslim.
When asked about vote choice, 14.44 percent of the Muslim respondents said they would vote for Agus Yudhoyono, 14.72 percent said they supported Ahok, and 14.15 percent said they supported Anies Baswedan. About 43 percent said they had not decided and about 14 percent refused to answer the question. Among non-Muslim respondents, 8.84 percent intended to vote for Agus Yudhoyono, 59.18 percent for Ahok, and 3.40 percent for Anies Baswedan. About 21 percent said they had not decided and 7.48 percent refused to answer the question. I restricted the analysis to those who indicated a vote choice (Yudhoyono, Ahok, or Baswedan) and dichotomized the responses, assigning score of 1 if the respondent indicated that they would vote for Ahok and 0 otherwise. A simple cross-tabulation indicates that Muslims were less likely to vote for Ahok than non-Muslims (χ 2 (1) ¼ 82:94, p < :01). A regression model also arrives at the same conclusion and is available in the Online Appendix. These results, in turn, provide support for the voter religion hypothesis.

A H O K T H E C H I N E S E , A H O K T H E C H R I S T I A N , A N D V O T I N G L I K E L I H O O D
The next three hypotheses concern Ahok's own religion and ethnicity. Do these characteristics affect Muslim voters' support for him? If so, which one is more consequential? I embedded in the survey a module that employed an experimental design. Respondents were randomly assigned to one of the five groups: a control group, an ethnicity group, a religion group, an ethnicity-religion group, and an ulema group. Table 2 presents the treatment wording along with the original and the effective size of each group.
A group's original size indicates how many Muslim respondents were assigned to that group, whereas a group's effective size indicates how many respondents in that group were analyzed. The attrition rates are not significantly different across the groups (see Online Appendix). This experimental approach circumvents the conceptual and measurement challenges suffered by the existing studies in two ways. First, it addresses the conceptual problem by explicitly focusing on Ahok's ethnicity and religion, holding voter religion constant. Second, it circumvents the methodological problems by not relying on respondents' own reports of how much Ahok's ethnicity or religion affected their vote preference. Rather, the approach provided information about Ahok's ethnicity (or religion) to a group of respondents and then compares the group's level of support for Ahok to a control group that did not receive this information. With the randomization rendering the two groups probabilistically balanced (see Online Appendix), any difference in the levels of support could then be attributed to the treatment. Figure 1 presents the percentages of respondents who said they would vote for Ahok in each group along with their 95 percent confidence intervals. There is no statistically significant difference in the percentage of respondents who said they would vote for Ahok between the religion and control groups (d = .06, s.e = .06, n.s., two − tailed t − test). On the other hand, there is a statistically significant difference between the ethnicity and control groups. Respondents who were reminded that Ahok is of Chinese descent were 10 points less likely to vote for him than respondents in the control group (d = −.10, s.e. = .05, p < .10, two − tailed t − test).
The effect of Ahok's ethnicity can also be estimated by comparing the religion group to the ethnicity-religion group. Since both groups are identical, except in the latter's priming of Ahok's ethnicity, such a comparison would give us an estimate of the effect of priming Ahok's ethnicity, holding Ahok's religious identity salient. I find that respondents in the ethnicity-religion group were 15 points less likely to vote for Ahok than respondents in the religion group. (d = −.15, s.e. = .06, p < .05, two − tailed t − test).
A generalized linear test further suggests that the estimate of the effect of Ahok's ethnicity obtained from comparing the ethnicity and the control groups and the estimate obtained from comparing the religion and the ethnicity-religion groups are not statistically different (F (1,679) = .41, n.s.). Lastly, the minimal effect of Ahok's religion is also  (Ahok) is one of the candidates running in the Jakarta gubernatorial election in February. As we know, Ahok is of Chinese descent. Will you vote for Ahok as governor of Jakarta? Religion Group (Original: 216, Effective: 141) Basuki Tjahaja Purnama (Ahok) is one of the candidates running in the Jakarta gubernatorial election in February. As we know, Ahok is a Christian. Will you vote for Ahok as governor of Jakarta? Ethnicity-Religion Group (Original: 196, Effective: 121) Basuki Tjahaja Purnama (Ahok) is one of the candidates running in the Jakarta gubernatorial election in February. As we know, Ahok is of Chinese descent and Christian. Will you vote for Ahok as governor of Jakarta? Ulema Group (Original: 217, Effective: 155) Basuki Tjahaja Purnama (Ahok) is one of the candidates running in the Jakarta gubernatorial election in February. As we know, Ahok is Christian and several ulema prohibit voting for non-Muslim leaders. Will you vote for Ahok as governor of Jakarta?

Ethnic and Religious Sentiments in Indonesian Politics 151
use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jea.2020.35 evident when we compare the ulema group to the control group. About the same percentage of respondents in the ulema group indicated they would vote for Ahok as those in the control group (d = −.01, s.e. = .05, n.s., two − tailed t − test).
These results support the Candidate Ethnicity Hypothesis, but not the Candidate Religion Hypothesis. In addition, the results also support the Primacy of Ethnicity Hypothesis. Making Ahok's ethnicity salient decreased Muslim voters' support for him. Priming voters with Ahok's religion, on the other hand, had no effect on his vote share.
That Ahok's being a non-Muslim did not seem to matter for the respondents, who were all Muslim, is perhaps counterintuitive. To further test the robustness of the findings, I examine two additional hypotheses. The first hypothesis relates to feelings toward ethnic Chinese and Christians and how they affect support for Ahok. If Ahok's ethnicity indeed mattered more, then we should expect how the respondents felt toward ethnic Chinese to affect their support for Ahok more than their feelings toward Christians. The second hypothesis relates to heterogeneous treatment effects. If ethnicity trumped religion in the election, then we should see the effect of the ethnicity priming to be invariant across voter characteristics.
H5 (Ethnic Feeling): How respondents felt toward ethnic Chinese affected their support for Ahok more than how they felt toward Christians.

H6 (Voter Characteristics):
The negative effect of the ethnicity priming should hold across voter characteristics.

E T H N O -R E L I G I O U S F E E L I N G S A N D V O T I N G F O R A H O K
I test the Ethnic Feeling Hypothesis by regressing the binary vote choice variable on feelings toward Christians and ethnic Chinese and a set of control variables. I measured feelings toward both groups by using the standard feeling thermometer questions, asking respondents whether they disliked or liked Christians and individuals of Chinese descent, respectively.
Responses were placed on a five-point scale ranging from strongly dislike to strongly like. The two measures are strongly correlated (r = .74, p < .01). Those who are positive toward Christians are also positive toward ethnic Chinese. This strong correlation highlights the importance of carefully teasing out the effects of Ahok's ethnicity and religion. Muslim voters' antipathy toward Ahok may be driven more by his ethnicity than by his religion. But this ethnic voting could be mistaken as religious voting if the researcher looks only at the religious mobilization by Islamic organizations.
I employed logistic regression models to estimate the relationships between ethnic and religious feelings, on the one hand, and support for Ahok, on the other. I included in the models a set of control variables: gender, age, education, income, ethnicity, knowledge about the election, and level of satisfaction with Ahok as the incumbent governor. I measured knowledge about the election by asking respondents when the first round of the election would be held. I gave respondents score of 1 if they provided the correct answer and 0 otherwise. Satisfaction with the incumbent captures one of the most important determinants of vote preference (Kramer 1971). I asked respondents whether they were satisfied with Ahok's performance as governor and placed the responses on a four-point scale ranging from very dissatisfied to very satisfied. The measure's correlations with feelings toward Christians and toward ethnic Chinese are of equal magnitudes (r = .23, p < .01 and r = .24, p < .01, respectively), suggesting that including the variable is unlikely to differentially bias the estimates of both feeling thermometers.
I ran two basic models and a full model (see Online Appendix). The first basic model includes feelings toward Christians whereas the second basic model includes feelings toward ethnic Chinese as the main predictor. The full model includes both feeling thermometers. All the models include the aforementioned covariates. I plot the predicted probability from each model as Figure 2.
The basic models suggest that moving from the most negative feeling toward Christians to the most positive one is related to a 45-point increase in the probability of voting for Ahok and moving from the most negative to the most positive feeling toward ethnic Chinese is related to a 57-point increase in the probability. The 57point increase across the range of feelings toward ethnic Chinese is also evident in the full model where feelings toward Christians is controlled. To the contrary, the predicted probabilities across the values of feelings toward Christians in the full model are flat, which highlights how the explanatory power of feelings toward Christians disappeared once we control for feelings toward ethnic Chinese. This result is consistent with the experimental finding and affirms our Ethnic Feeling Hypothesis. In addition, the limited effect of the feeling toward Christians further supports the Primacy of Ethnicity Hypothesis.

D O T H E R E S U L T S V A R Y B Y V O T E R C H A R A C T E R I S T I C S ?
The preceding two analyses suggest it was Ahok's ethnicity, as opposed to his religion, that decreased voter support for him and that the respondents' feelings toward ethnic Chinese mattered more than their feelings toward Christians in shaping support for Ahok. A natural question that follows these findings is whether these results vary by voter characteristics. Specifically, would we see the effects of Ahok's religion and feelings toward Christians to be stronger among more religious voters?
To answer this question, I examine how levels of satisfaction with Ahok, personal religiosity, religious tolerance, and support for sharia on behalf of the voters affect their support for Ahok. I measured personal religiosity by asking respondents how often they practiced shalat on a five-point scale ranging from never to regularly. I measured religious tolerance by asking respondents if they agreed or disagreed with the statement "Christians have the right to be elected as regent, mayor, or governor, even in areas that are Muslim majority" on a four-point scale ranging from strongly disagree to strongly agree. Lastly, I measured support for sharia by asking respondents if they agreed with the statement "Islamic laws must be thoroughly applied in Indonesia" on a four-point scale ranging from strongly disagree to strongly agree. Figure 3 presents interaction results from the experimental study. There are three main findings from this exercise, all consistent with the patterns in Figure 1. First, in none of the models does the religious priming significantly decrease respondents' support for Ahok. If anything, the religious priming actually increased support for Ahok among those opposed to Islamic law (d = .17, s.e. = .09, p < .10). Second, the ethnicity FIGURE 2 Support for Ahok based on Religious and Ethnic Feelings priming significantly decreased support for Ahok among those with low satisfaction (d = −.09, s.e. = .04, p < .05), low religiosity (d = −.14, s.e. = .09, p < .10), low tolerance (d = −.08, s.e. = .05, p < .10), and high tolerance (d = −.21, s.e. = .10, p < .05). Lastly, the negative effect of the ethnicity priming is also evident when we compare the religious priming to the ethnicity-religion priming. Comparing these two groups, I find that support for Ahok decreased among respondents high in religiosity (d = −.15, s.e. = .06, p < .05), low in tolerance (d = −.13, s.e. = .05, p < .05), and low in support for sharia (d = −.24, s.e. = .09, p < .01). Figure 4 presents the marginal effects of feelings toward ethnic Chinese and toward Christians derived from interaction models in which I interacted both feeling thermometers with each potential moderator. Our interest is to examine whether in any of these models feelings toward Christian takes primacy over feelings toward ethnic Chinese.
The figure suggests two findings. First, among respondents who are predisposed to political Islam (low tolerance and high support for sharia), neither feeling thermometer matters in shaping support for Ahok. Second, the only instance when religious sentiment is influential while ethnic sentiment is not is among respondents with low satisfaction toward Ahok. Among these respondents, feelings toward Christians is positively related to support for Ahok while feelings toward ethnic Chinese is not. In the case of Jakarta, however, these voters are in the minority, which might explain why we see only minimal evidence of religious voting.

A N O T E O N P O T E N T I A L C O N C E R N S
Before discussing how these findings inform our understanding of Indonesian Muslims' voting behavior, I address three remaining concerns about the limited effect of Ahok's religion.

I N T E R S E C T I O N A L I T Y O F R E L I G I O N A N D E T H N I C I T Y
The first concern relates to the intersectionality of religion and ethnicity and has an empirical and a theoretical aspect. The empirical aspect of the concern is based on the fact that 70 percent of Chinese Indonesians are Christian (Brazier 2006) and less than 5 percent are Muslim (Hew 2018). This means priming Ahok's being Chinese might result in the respondents' realizing that Ahok is a Christian and vice versa. I certainly cannot rule out the possibility of such a contagion effect where two connected schemas are made salient by the activation of one of them. However, this concern cannot explain the patterns in Figure 1.
To use the concern to account for the findings that Ahok's ethnicity mattered but his religion did not means that one has to be willing to accept that the religion priming actually primed ethnicity, but not religion, and that the ethnicity priming primed religion, but not ethnicity. This is an implausible scenario. Studies that examine a contagion effect (e.g., Nickerson 2008) find that an effect resulting from a treatment contagion is likely weaker than the one resulting from the actual treatment. In other words, the main effect itself has to be significant first for the contagion effect to be significant.
The theoretical aspect of the concern is related to Ahok's ethnicity and religion in the eyes of the Muslim voters. How come we find only little effect of Ahok's religion among voters defined by their religion? This is again related to the subtle but important distinction between voter religion and candidate religion. Existing studies have conflated these two concepts, incorrectly assuming that because Muslim and non-Muslim voters react to Ahok differently, it must mean that it was Ahok's religion that drove this divide-in effect subjugating the effect of Ahok's ethnicity to that of his religion. This would be true if Ahok had not had an ethnic minority background. But Ahok's Chinese background complicates the story.
That there is a history of tensions between ethnic Chinese and the Muslim population (manifested, among other ways, in the founding of Sarekat Islam and ICMI ;Hefner 1993;Purdey 2006) means that Ahok's ethnicity might be as powerful as (or even more powerful than) his religion in shaping the Muslim voters' opposition toward him. Such an influence of Ahok's ethnicity should not be surprising considering the stickiness of stereotypes against ethnic Chinese. Even Muslim ethnic Chinese sometimes face discrimination and stereotypes due to their ethnicity (Hew 2018;Muzakki 2010). This distinction between voter religion and candidate religion, in turn, enables the present study to estimate the effect of Ahok's ethnicity independently of his religion, hence explicitly acknowledging that Ahok's ethnicity has important consequences of its own.

A H O K ' S R E L I G I O N C A N N O T B E M A D E M O R E S A L I E N T T H A N I T A L R E A D Y I S
The second concern relates to the survey timing., which was between two major anti-Ahok demonstrations. It is perhaps the case that the protests already made religion very salient that the religion priming provided the respondents with no new information because they already knew that Ahok is Christian. This is a possible explanation and future studies of similar topics should take this possibility into account in their designs. However, there are three reasons why this explanation does not jeopardize the present study's conclusion that Ahok's ethnicity mattered more than his religion.
First, the experimental design used "Ahok," which is a Chinese name. In addition, the respondents likely knew that Ahok is of Chinese descent from his physical appearance. As such, it was as likely that the respondents knew that Ahok is of Chinese descent as they knew that Ahok is Christian. Yet, despite the possibility that both Ahok's ethnicity and religion might not be new information for the respondents, we still see a significant effect of Ahok's ethnicity, but not his religion. This suggests that Ahok's ethnicity mattered more than his religion.
Second, while this account may be able to explain results from the experimental design, it cannot explain why the effect of feelings toward Christians disappears once feelings toward ethnic Chinese is accounted for in the correlational analysis. This suggests that the explanatory power of Ahok's religion (proxied by feelings toward Christians in general) is to a large extent shaped by Muslim voters' feelings toward Ahok's ethnicity (proxied by feelings toward ethnic Chinese).
Third, this explanation implies that political knowledge matters. Less knowledgeable respondents should be less likely to know that Ahok is a Christian. Consequently, if political knowledge mattered, then we should expect significant interaction effects between knowledge about the election and the treatment assignment. I find no such interactions (see Online Appendix) and the patterns in Figure 1 hold up regardless of respondents' levels of political knowledge.

C L A S S , N O T E T H N I C , R E S E N T M E N T
The third concern relates to the possibility that it is class, not religion or ethnicity, that shaped Muslim voting behavior in the election. One of the drivers of resentment against Chinese Indonesians is a perception that they are more economically successful than the average Indonesian (Setijadi 2017a). That the present study finds a negative effect of the ethnicity priming, therefore, may be driven by a (mis)conception about ethnic Chinese being the economic elite. This concern, however, does not contradict the present study's findings as it is more about why ethnicity matters (a question about mechanism) than about whether ethnicity matters (a question about effect).
On the other hand, it is less likely that class per se, independent of ethnicity, fully explains the results presented. Several studies have highlighted the limited role of social class in shaping Muslim political behavior in Indonesia (Liddle and Mujani 2007). In addition, I also find no interaction effects between respondent income and treatment assignment. The patterns presented in Figure 1 are found across respondents regardless of their income levels.

D I S C U S S I O N
The present study contributes to the discussion on ethnic and religious influence in Indonesian politics by examining the 2017 Jakarta election. Analyzing a survey experiment that circumvents the limitations of other studies of the election, I find that voter religion matters in that Muslim voters are less likely to support Ahok compared to non-Muslims. However, it was Ahok's ethnicity, as opposed to his religion, that actually drove the opposition among Muslim voters.
These findings carry methodological and theoretical implications. On the methodological level, the present study highlights the importance of employing more indirect approaches to study voter behavior. The use of public opinion surveys to understand voter behavior in Indonesia is an important first step. However, this step needs to be followed by survey designs that take into account research on how voters may not be aware of their own attitudes and simply report what is socially desirable or salient in society (Zaller 1992). List experiments, vignette experiments, and conjoint experiments are some tools that Indonesian survey researchers can employ to understand voter behavior better without asking voters the difficult question of why they do what they do.
On a theoretical level, the present study highlights the need to distinguish carefully the two different levels on which religion matters: the voter level and the candidate level. The importance of religion on the voter level corresponds to its effects in creating different voting blocs between Muslim voters who are more opposed to Ahok and non-Muslim voters who are very supportive of him. This is the clearest evidence that religion plays an important role in Indonesian politics and this is also the focus of many studies of the election and Indonesian politics in general Mujani, Liddle, and Ambardi 2018).
That voter religion matters is also likely to be the reason why religion remains a popular political mobilization tool and why, in the case of the Jakarta election, we observed religiously motivated anti-Ahok mobilization. Politicizing religion enables politicians to forge "a maximal coalition" as the Islamic identity serves as the lowest common denominator among Muslim voters who come from different backgrounds. Politicizing religion also offers plausible deniability in a religious country like Indonesia. If the religious mobilization backfired and the politician was accused of being intolerant, for example, they could say that all they did was follow religious teaching.
Just because voters are divided by their religions and politicians opt to engage in religious mobilization, however, it does not mean that the Muslim voters are swayed more by the candidate's religion. This is particularly true if the candidate also has another characteristic that is viewed negatively by these same voters. In the case of the Jakarta election, there are two reasons that can explain why Ahok's ethnicity mattered more than his religion.
The first reason is psychological and concerns expectations. As a result of Soeharto's anti-Chinese policies, public officials of Chinese descent are uncommon. This means that Indonesians are less exposed to ethnic Chinese public servants, let alone popular and successful ones. Less exposure, in turn, means less information and more opportunities for the negative stereotypes about ethnic Chinese to shape what voters expect from an ethnic Chinese official (Setijadi 2017a).
These expectations, however, were largely violated by Ahok's success in office and high approval rate. The expectancy violation theory suggests that disconfirming information is more influential because it "is more salient, more memorable, more attention grabbing, and instigates more processing time" (Johnston, Lavine, and Woodson 2015, 476). What this means is that voters would regard Ahok's accomplishment in office as even more impressive exactly because it disconfirms their expectations and negative stereotypes about ethnic Chinese. Ahok's ethnicity serves as a unique piece of information that voters can use in their decision making. The same mechanism did not apply to Ahok's religion because Christian officials with a good reputation are a common occurrence. Given good performance, Ahok's status as a member of a religious minority was no longer a powerful piece of information, which might explain why we find a significant effect of Ahok's ethnicity, but not of his religion.
The second reason why Ahok's ethnicity mattered more than his religion is related to counter-mobilization from moderate Muslim groups. The blasphemy charge against Ahok led to demonstrations demanding his imprisonment. These demonstrations provided space for conservative clerics to advise Muslims not to elect a non-Muslim leader. Moderate Muslim clerics responded to this development by arguing for the opposite: That Ahok did not blaspheme Islam and that Muslims may vote for a non-Muslim. For example, Ahmad Ishomuddin, one of the leaders of Nahdlatul Ulema, a moderate Muslim organization, even served as an expert witness for Ahok in his blasphemy trial. These narratives and counter-narratives might cancel each other out, resulting in a minimal effect of religion.
On the other hand, virtually no such systematic counter-mobilization happened when it came to addressing attacks against Ahok's ethnicity. The ethnic sentiment was largely swept under the rug. Everyone knew it was there, but only few talked about and countered it. This might explain why Ahok's ethnicity mattered more in the election: It mattered precisely because the relative absence of debates on ethnicity left voters to resort to their own ethnic stereotypes.
C O N C L U S I O N Confirming findings of previous studies, I find that Muslim voters are less supportive of Ahok than non-Muslim voters. Diverging from these studies, however, I show that it is necessary to differentiate between voter religion and candidate religion. Analyzing a survey experiment that circumvents conceptual and methodological limitations of existing studies, I find that Ahok's ethnicity, but not his religion, negatively shaped Muslim voters' support for him. This, in turn, highlights the importance of intersecting identities. The present study shows that when two salient identities are so intertwined (in this case Ahok's religion and ethnicity), scholars need to be much more careful in highlighting one of the identities as the identity that is more dominant in shaping voter preference. Conservative Muslims' opposition to Ahok gave the impression that this was mainly because of his religion. But the complex intersectionality of Ahok's ethnicity and religion means that we must not take this impression for granted.
One potential research avenue on this topic, and one that promises even deeper insights into ethnic and religious sentiments in Indonesian politics, is to build a database of elections and their candidates, along with information about the candidates' characteristics. From such a database, we can engage in inquiries about the importance of religion across the elections (by comparing two candidates from the same ethnic background, but one of which is a religious minority) or the importance of ethnicity (by comparing two Muslim candidates, one of which an ethnic majority and the other an ethnic minority).
One may hope that in the future there will be more political candidates with minority backgrounds running for office in Indonesia. These candidates likely would face strong opposition from the dominant group. The empirical approach outlined here provides a template for scholars interested in gaining a better understanding of how different identities such as ethnic and religious identities on behalf of the candidate overlap and interact in shaping voter behavior. A better understanding of how ethnicity and religion shape Indonesian politics, in turn, would enable us to respond to identity-based mobilization more effectively and improve the quality of Indonesian democracy by making it more representative of minorities. Nathanael Sumaktoyo (nathanael.sumaktoyo@fulbrightmail.org) is a postdoctoral fellow with the Politics of Inequality Project at the University of Konstanz, Germany and an incoming Assistant Professor of Political Science at the National University of Singapore (starting August 2021). His research interests are on the topics of political behavior, minority politics, intergroup relations, and causal inference. He maintains a personal website www.nathanael.id.

A C K N O W L E D G M E N T
The author thanks the editor, the two anonymous reviewers, participants of the Indonesia Study Group workshop at the Australian National University, participants of the 2019 SEAREG Conference at the University of British Columbia, and participants of research seminars hosted by the Centers for Southeast Asian Studies at Northern Illinois University and at the University of Wisconsin-Madison for valuable feedback. All errors remain the author's.

C O N F L I C T S O F I N T E R E S T
The author declares no conflicts of interest.
N O T E 1. The survey did not ask about the blasphemy accusation to avoid raising a controversial subject matter that might jeopardize other studies in the survey. Furthermore, there is also a reason to suspect that the blasphemy charge's electoral impact would be minimal. Surveys prior to the blasphemy case already showed that Ahok's vote share was around 45 percent, which was within margin of error from his eventual vote share of 42 percent (SMRC 2016;Charta Politika 2016).