Myanmar's 2020 Election: Explaining the Strong Performance of the NLD and Some Ethnic Parties

Abstract Aung San Suu Kyi's National League for Democracy (NLD) won another landslide victory in the 2020 general election. Although there was widespread dissatisfaction with the government's poor management of the economy and ethnic conflicts, as well as with the pre-electoral coordination of ethnic parties in several states, opposition and ethnic-based parties had failed to gain more seats. Previous explanations had focused on the economy, electoral system bias, weak party institutionalization, and vote splitting among ethnic parties; however, they underestimated the significance of two contextual factors: military dominance of politics and ethnic conflict. This article argues that military dominance hindered normal political development in Myanmar. The anti-military sentiment favored the NLD, which made most Bamar voters disregard the party's poor economic performance. Despite the electoral system's bias, prolonged ethnic conflicts made ethnic parties that had fought for their community's causes more likely to maintain support. These arguments are verified by survey and electoral data sets. The military nullified the 2020 election claiming that electoral fraud was to blame, but the findings indicate that it was the political environment the military created that led to the victory of the NLD and some ethnic parties in the first place.

people were killed by the military in these demonstrations. Despite international sanctions and continuous protests, the power is firmly in the grip of the military, and the Myanmar people are again under military rule.
For the USDP and some ethnic parties, the NLD winning even more seats in the 2020 election was puzzling. The NLD's 2015 victory had been attributed to Aung San Suu Kyi's popularity and to the belief that she would bring much-needed economic and political changes to the country (Thawnghmung 2016). Ethnic parties, on the other hand, suffered from vote-splitting, weak party institutionalization, and disadvantage in the first-past-the-post electoral system (Aung 2018;Stokke 2019;Tan and Preece 2020). However, in the NLD government's first term, the country suffered from weak economic performance, stalled political reforms, and delayed peace talks (Callahan 2018). While former president Thein Sein was widely admired for his achievements during his reign (Zin 2016, 119), life under the NLD had not improved significantly, and some ethnic minority groups fled the country to avoid genocide, the Rohingyas being the most notable. Meanwhile, to overcome the electoral system's bias, several ethnic parties merged to avoid repeating the mistake of 2015 (Thant 2020). Because of these factors, the USDP believed the NLD would suffer at the ballot box (Soe 2020). The NLD swept most of the seats, however, making all the hopes and efforts futile. From the USDP's point of view, there was only one reasonable conclusion to explain the electoral outcome: It was rigged.
In this article, I argue that the election results should be analyzed from two broader perspectives: military dominance of politics and ethnic conflict. To begin with, NLD's continuous victories were not the result of electoral fraud as the military claimed, but should instead be understood in light of military dominance. After the promulgation of the 2008 constitution, the political environment had become more liberal, but the military still held significant power over the civilian government. This condition prolonged the political atmosphere commonly seen during the first post-authoritarian election when the pro-democracy and pro-authoritarian camps compete in a bipolar environment. The NLD benefitted from the anti-military sentiment, which made most Bamar voters overlook the party's mediocre performance on the economy. Furthermore, ethnic parties that had stood up for their community's interests had a greater chance of retaining stable support, despite electoral system bias. In those ethnic parties, the success was due to the social roots cultivated through struggles against the military. Therefore, it was the political climate the military and the USDP created that defeated them in the first place.
This article analyzes survey and electoral data and finds that Bamars' current and retrospective evaluations of the economy are not important in determining their vote choices between the NLD and USDP. All party identifiers are more likely to vote for the NLD than for the USDP. While party mergers tend to increase a party's seat share at the state level, parties with a rebel heritage (i.e. rebel groups-turned-parties) tend to gain more votes and have a higher chance of winning at the constituency level. Results suggest that the military and its affiliated parties would have a very difficult time defeating the NLD, since Bamar voters were less responsive to economic performance than voters are in a stable or consolidated multiparty democracy. On the other hand, some ethnic parties in Myanmar had managed to overcome the electoral system bias.
If not for the prolonged ethnic conflict in the country, rebel heritage would not be a source of social roots for these ethnic parties. The focus of the present article is to bring the Tatmadaw back to the center of Myanmar's electoral analysis, and to present empirical evidence of the ways in which such a political climate impacts voting behavior and party performance. Although not expected in the near future, if the military continues to try to gain legitimacy through free and fair elections, they should not underestimate the implications of military dominance and resistance to it.

Myanmar elections explained
Following years of military rule, Myanmar held elections again in 2010, and observers and scholars had closely examined the election results. NLD and some ethnic parties, which participated in the nullified 1990 election, refused to accept the 2008 constitutional framework and boycotted the election. As a result, the USDP won overwhelmingly, holding 388 of the 498 seats in parliament. Other newly established parties also gained seats. Among them were the National Democratic Force (NDF), the Shan Nationalities Democratic Party (SNDP), and the Rakhine Nationalities Development Party (RNDP), all splinter groups from the 1990 old movement parties (Stokke Forthcoming,10). The electoral outcome indicates that the splinter group of the NLD (i.e. the NDF) was unable to lead the pro-democracy forces to defeat the USDP. Meanwhile, a significant number of seats were won by ethnic parties in parliament, and the results gained some attention. Ethnicity as a psychological ideology should influence political behavior, and the geographic concentration of ethnic groups will make it easier for minority ethnic parties to enter parliament even when using the first-past-the-post (FPTP) electoral system (Brown 2003, 7-8;Selway 2015;Moser 2001, 4). In theory, FPTP with single-member districts tends to create a two-party system when voters are incentivized to defect to large parties to avoid wasting their votes on small ones (known as strategic voting) (Duverger 1954). In Myanmar, ethnic parties are likely to suffer when their supporters side with a large national party in order to defeat the other least liked one. However, in the absence of the NLD, ethnic voters seemed to vote sincerely in 2010, and ethnic parties won because the ethnic groups they represented made up the majority of voters in single-member districts.
With the 2010 election results serving as an encouragement, ethnic parties and elites were enticed to enter electoral competitions. In the 2015 election, ethnic parties proliferated. A total of 91 parties registered to compete, of which 59 were ethnic parties (Transnational Institute 2015, 2). They included old movement parties that competed alongside their splinter groups. For ethnic parties, the results were disappointing. Not only did the NLD win a landslide victory among Bamar voters, it also won great support from ethnic minorities. Except for the Arakan National Party, all other ethnic parties performed poorly. Those results are attributed to three factors: the charisma of Aung San Suu Kyi, the organization of the NLD, and the fragmentation of small ethnic parties (Thawnghmung 2016). As a symbol of democracy's struggle against military rule, Aung San Suu Kyi was seen by many as the candidate who would bring about much-needed change to the country, including a better economy (Welsh, Huang, and Chu 2016). A good number of supporters mobilized to the NLD in response to her popularity, and the party provided financial and organizational support to local campaigns. Small ethnic parties became victims of strategic voting when ethnic voters supported the NLD which also nominated ethnic candidates in state constituencies. Additionally, these local ethnic parties lacked distinct programs to attract voters; party institutionalization was weak among most of them (Stokke 2019). The presence of so many ethnic parties also resulted in a split in the votes of those inclined to cast ballots based on ethnicity (Tan and Preece 2020).
Following the failure of the 2015 election, many ethnic parties rethought their strategy and party mergers became the most popular choice. There had been a merger of parties in Chin, Kachin, Kayah, Kayin (Karen), and Mon States to prevent vote splitting (Knirsch and Heugas 2020, 6). Meanwhile, peace negotiations with various ethnic armed groups had reached a standstill and the NLD government had performed poorly, which was likely to lead to ethnic voters not supporting the NLD government as they did in 2015 (Callahan 2018;Transnational Institute 2020, 9). With limited information of the 2020 election, it appears that the merger strategy worked for ethnic parties in Mon and Kayah States, but failed to change the results in Chin, Kachin, and Kayin States (The Irrawaddy 2020; Zin 2020). In fact, the 2015 electoral data indicates that in many constituencies, even with these ethnic parties combined, they could not defeat the NLD (Aung 2018). Nevertheless, the NLD secured 399 seats in parliament, surpassing its previous landslide victory (390 seats) (Zin 2020). As a result, USDP officials accused the election of fraud, and General Min Aung Hlaing joined the chorus of warnings about the legitimacy of the vote (Thura 2020). The Tatmadaw finally intervened in February 2021 to invalidate the election, and the Myanmar people reacted with protests and general confrontations with the military.
Previous studies argued that Myanmar's elections were mainly influenced by three factors. Aung San Suu Kyi and the electoral strategy of the NLD were the first. Despite her fall from grace as a democratic icon in the international community, Daw Suu remained popular with the people of Myanmar (Liu and Kyaw 2020). The NLD presented itself as multi-ethnic and nominated ethnic candidates in constituencies in which ethnic parties also participated. FPTP also gave the party an advantage through strategic voting when ethnic voters abandoned their co-ethnic parties. There were many new parties formed for electoral purposes and had very weak institutionalization, making them hard to compete with the NLD and the resource-rich USDP. Several parties representing the same ethnic group also contributed to vote-splitting. Although these explanations are still capable of explaining the 2020 election results, they are not sufficient to explain why NLD support strengthened despite poor performance. According to survey data, 20 percent of respondents rated the current economy as bad or very bad in 2015 under the Thein Sein regime, but that number increased to 27 percent in 2020 when the NLD was in power. 1 Meanwhile, 18.5 percent of respondents thought the economic condition had deteriorated in 2015, but 24.8 percent believed it worsened in 2020. Considering that most Myanmar people had high expectations for the NLD, the numbers do indicate that there was disappointment among ordinary citizens. Scholars, NGO activists, and even an NLD member of parliament interviewed by the author in summer 2018 in Yangon expected the NLD to receive fewer seats in the 2020 election, largely due to the poor economic conditions. Some observers also predicted that ethnic voters would lower their support for the NLD due to party mergers and dissatisfaction with how the government dealt with ethnic conflicts (Callahan 2018;Knirsch and Heugas 2020). Additionally, not all ethnic parties were affected by electoral system bias; some had reliable and stable support, indicating they had achieved party institutionalization, but no explanation is offered as to how such institutionalization occurs. In this article, I argue that these elections should be analyzed from the perspective of two contextual factors specific to Myanmar: the military's dominance over politics and prolonged ethnic conflicts. Both conditions upheld the NLD's electoral strength as well as strong local ethnic parties.

Missing links: Military dominance and ethnic conflict
Despite the military using electoral fraud as the justification for the coup, international observers generally characterized the 2020 elections as free and fair (Liu 2020). However, it is puzzling how the NLD managed to win even more seats in 2020 despite the government's failure to deliver its promises. One possible explanation is that a majority of Myanmar people have no other alternative but to vote for the NLD to avoid being ruled by the military junta. This fear is common among voters in countries in transition, especially when it comes to the first post-authoritarian election. The first election (known as the transitional election) is usually marked by a contest between those opposed to and in favor of the authoritarian regime (Pettai 2012, 14-15). Once such conflict is settled in the transitional election, the subsequent elections will gradually bring the electoral process back to a reliable and generalizable state (Bogdanor 1990, 288). Evidence suggests that many parties participate in founding elections (i.e. the first few elections after authoritarian rule), reflecting newly earned political liberalism, but then disappear over time (Turner 1993). It is also likely that the largest and second-largest parties will lose votes in the following elections (Reich 2004). Democratization often gives people the false belief that democracy can cure all ills, but in fact, countries in transition have to overcome many authoritarian legacies, making effective governance and economic improvement challenging (Schmitter and Karl 1991). As a result, the pro-democracy parties tend to win a majority of seats in the transitional election, thanks to a bipolar environment that creates a centripetal force as the pro-democracy and pro-authoritarian camps join forces to defend their causes, resulting in the vote concentrating on the leading parties of the conflicting camps (Przeworski 1991, 66-67). However, after the initial excitement of democratization wears off, these parties may lose their support in subsequent elections. Moreover, the bipolar pattern will disintegrate as ideological and tactical differences erode both camps' unity when the transition process grows deeper (Klingemann 2019, 505;Bielasiak 1997).
Myanmar's 2020 election may well fit into this scenario, as voters gradually lose trust in pro-democracy parties. The USDP also expected that voters would recognize the achievements of the Thein Sein government's political and economic reforms and punish the incumbent to assert their accountability. However, the military's dominance over key aspects of politics, as well as its reserve seats in parliament, stymied the process of transition. In consequence, the regular democratic pattern of electoral accountability had been disrupted, since voters were locked into the political climate of the transitional election. NLD, a leading anti-military party, gained from this situation as voters placed defeating the military above the performance of the NLD government. Therefore, the NLD's increased number of seats in 2020 is not due to its campaign strategy nor to FPTP protection, but to the political climate the military created to maintain its power.
Prolonged ethnic conflict, on the other hand, has made many ethnic parties strive for protecting cultural and political rights. However, only parties that have demonstrated that they have fought for such rights will gain support from ethnic minorities. These groups are more "institutionalized" in the sense that they provide values and identity to connect with their supporters. Party institutionalization refers to "the process by which the party becomes established in terms of integrated patterns of behavior and of attitudes, or culture" (Randall and Svåsand 2002, 12). Generally, weak party institutionalization is thought of as lacking a value system or ideology that guides policy formulation (Stokke 2019, 327;Moser 1999), but value infusion is more than establishing policies; it fosters the spirit of loyalty to the survival of an organization (Levitsky 1998, 79). In developing countries, political parties are usually created from above around a charismatic leader, with little resources and full of factionalism (Randall and Svåsand 2002, 18-19). Under certain conditions, however, new parties can establish a distinctive culture that fosters cohesion. Parties derived from rebel groups meet this requirement. Rebel groups that achieve their goals are more likely to turn to politics to gain credibility and more likely to win elections due to the established organizational structures (Acosta 2014;Allison 2006). This explains why some ethnic parties could resist the NLD's appeal to their co-ethnic voters and maintained stable support, since co-ethnic identity is not enough to secure majority support, credibility matters as well. Credibility is especially high for rebel groups-turnedparties which have proven themselves committed to the cause.
The above discussion explains why the NLD could enjoy continuous support and why ethnic parties had varying levels of electoral strength. In the context of Myanmar, military dominance created a bipolar environment that drove people to pour their support behind the NLD to change the status quo. Therefore, improving the economic situation is not the most important factor in determining votes. Voters from ethnic minorities should be more concerned about their cultural and political rights. However, only parties that have demonstrated that they will protect such rights will gain support from ethnic minorities. Two hypotheses result from this discussion.
Hypothesis 1: Myanmar voters' vote choice is not determined by economic performance.
Hypothesis 2: Parties with a rebel heritage tend to have more electoral success than ethnic parties without the heritage.

Data and method
This article uses two sources of data to test the hypotheses. The first is the survey data conducted by the Asian Barometer Survey (ABS) in Myanmar in two waves, one in 2015 and the other in 2020. 2 These surveys ask two questions that are the dependent variables of this article: vote choice in the last general election and vote intention for the next. Since both surveys were conducted before the general election, we can compare whether support for a particular party increased between the last and upcoming election. Figure 1 presents the distribution of vote choice and vote intention in survey data. The majority of respondents to the surveys did not reveal their vote choice or vote intention. It appears that Myanmar voters are still reluctant to reveal their political orientations, likely due to the long-term political restrictions under the military rule. Our hypothesis testing is, therefore, challenged by the high percentage of invalid answers. As a mitigation of the data shortfall, we classify and analyze the invalid answers using multinomial logistic regression to determine the respondents' socioeconomic backgrounds and, subsequently, their political orientations. Because the NLD and several other parties boycotted the 2010 election, about one-fourth of respondents voted for ethnic parties and splinter groups of the 1990 movement parties. When asked about the vote intention for the 2015 election, the NLD was the clear winner. Ethnic parties still secured a large number of votes among minority voters, however ( Figure 1, upper right panel). The percentage of respondents who actually voted for the NLD in the 2015 election (self-reported in the 2020 survey) is significantly higher than their voting intentions in the 2015 survey (22.5 percent versus 48 percent). Although voters might still vote for the NLD, they were reluctant to reveal their vote choice ahead of the election. Thus, the actual votes for the NLD in the 2020 election would be much higher than what was revealed in the survey for vote intention. Election results certainly confirm this. Overall, as compared with voting intentions for the 2015 election and the 2020 election, it is clear that NLD support increased while support for the USDP and ethnic parties decreased significantly.
The survey data are used to test Hypothesis 1. Current economic evaluation, retrospective economic evaluation, and preference for federalism are the explanatory variables. Myanmar people have reason to vote for parties offering economic improvement, since their living standards are low under the long-running military rule. Myanmar's 2015 election was said to have been influenced by the economic environment at the time, and the people overwhelmingly chose economic development over democracy (Welsh, Huang, and Chu 2016, 138). As a result of poor economic performance, the NLD was expected to win fewer seats before the 2020 election, which indicates that vote choice is influenced by perception of economic management. When voting, ethnic voters should support federalism in order to protect their cultural and political rights. As standard controls, partisanship, satisfaction with the government, and socioeconomic indicators are also included. Since vote choice and vote intention are categorical variables, we use multinomial logistic regression to analyze the data (more details in the appendix).
Electoral data collected from websites is the second source of data. Each election has different information available. Votes received by each party at the regional/ state level are included in the 2010 general election results. As constituency-level information is available for 2015, the data are the most complete. Because the military nullified the 2020 election, reports on election day are the only source of information available. We know only how many seats each party won in each region or state in this election. Figure 2 shows the seat share of ethnic parties at the state level across three elections. In Chin, Shan, and Rakhine states, seat shares of ethnic parties appear to be relatively stable. Kachin, Kayah, Kayin, and Mon states then show varying degrees of change. There are two main sources of support for ethnic parties. First are parties with rebel backgrounds and we refer to literature, websites, and other sources to identify them. Despite the fact that many ethnic parties may have connections with rebel armed groups, we use strict criteria to identify parties with rebel heritage. This refers to whether the party has/had a military wing or the founding leaders had connections with armed groups. They include the Pa-O National Organization The second source of ethnic party support are parties competing in the 1990 election. In re-democratized elections, parties that participated in elections prior to authoritarian rule are more likely to win elections (Manning and Smith 2016). The NLD and its allied ethnic parties, such as the Shan League for Democracy and Chin National League for Democracy, were among them. Additionally, in several states, ethnic parties merged in an effort to avoid vote-splitting, and the analysis also explores whether this strategy worked. Party mergers were relatively rare in 2015 compared to 2020. Only the Arakan National Party was a merger of the Rakhine Nationalities Development Party and the Arakan League for Democracy prior to the 2015 election. The number of parties merging in 2020 grew dramatically, including the Kachin State People's Party, the Mon Unity Party, and the Chin National Democratic Party, among others. We also include parties that represent small groups (such as farmers and sub-ethnic groups) and parties that have links with the military (such as the USDP) in the model. The number of seats at the regional/state level and the number of parties at the constituency level are also control variables. We analyze the data using an OLS or logistic regression model, depending on the structure of the dependent variables (seat share, votes, or winning a seat). Table 1 shows how voters' evaluations of the economy and preference for federalism influence their vote choices and vote intentions in the 2015 survey. Note that the survey was conducted before the 2015 election, and the ruling party of the time was the  USDP. Since the country is divided into seven regions and seven states plus the capital (Naypyidaw), where regions are mainly occupied by Bamars, but states have a very high proportion of ethnic minorities, these two groups may vote differently due to their geographic concentration, and, therefore, deserve separate attention. Thus, the Bamar majority and ethnic minorities are examined separately in the analyses. Furthermore, the models compare voters' choices between the NLD and the USDP, and between the NLD and ethnic parties for ethnic voters (Table D in the appendix compares the USDP to other options). Ethnic parties and other parties are combined into one category. Except for the 2010 election, in which the NLD boycotted and the splinter groups were counted as "other" parties that attracted pro-democracy voters, supporters of "other" parties were very few in the following elections. This category of parties is primarily supported by ethnic voters, as shown in Figure 1. As a result, the supporters of "other" parties do not significantly affect results.

Results and Discussion
In the Bamar electorate, economic evaluation has no effect on the probability of voting for the NLD as compared to the USDP (Models 1, 3, and 4 in Table 1). Due to concerns about collinearity, we examine the effects of partisanship and vote choice separately on voting intentions. Most of the variations are explained either by partisanship or the vote choices made in the last election. Voting for the NLD is more common among party identifiers of other parties (including those without a party ID). Minority voters, on the other hand, are less likely to report that they vote for ethnic parties over the USDP when they are happy about the current economic situation (Model 2 in Table 1). A good economic evaluation, however, increases the likelihood of a NLD vote for the 2015 election when controlling for vote choice (Model 4). When partisanship is taken into account, improving economic conditions reduce the likelihood of voting intention for the NLD (Model 5). Although the results appear contradictory, both indicate that the economy is a significant factor for ethnic voters. The USDP spearheaded political and economic reforms that ethnic voters seemed to associate with economic improvement. With a NLD vote, deeper reforms were expected to take place and the economy would continue to improve. Though economic development is poor throughout Myanmar, in states where ethnic voters concentrate, the economy is worse due to sporadic ethnic conflicts and the neglect of the military regime. It may explain why ethnic voters are more likely to cast economic votes compared to Bamars. When comparing the vote intentions of ethnic voters between ethnic parties and the NLD, only living in a state is statistically significant (Models 7 and 8). Ethnic voters in states are more likely to say they plan to vote for ethnic parties over the NLD in the 2015 election. Ethnic minorities' support for federalism, however, has no bearing on their vote choice or vote intention.
For respondents who did not indicate their vote choice or vote intention, age and living in a state are the factors determining their option (no vote) over the USDP (Table D in the appendix). Younger voters are more likely than older voters to report a no vote than voting for the USDP, so are minority voters living in states. Additionally, those who identified with other/ethnic parties or voted for other/ethnic parties in the 2010 election are also more likely to report a no vote or no voting intention than voting for the USDP. As a whole, young voters and ethnic minorities are less likely to disclose their political orientation than to report a vote for the USDP. Since these voters are more likely to vote for the NLD than the USDP, the findings suggest that a substantial number of respondents may vote for the NLD despite not revealing their choices in the survey. Table 2 shows the factors influencing vote choices in the 2015 election and vote intentions in the upcoming 2020 election (Table E in the appendix compares the USDP to other options). There is still an important role for party identification in determining vote choices for the NLD over the USDP among both Bamar majority and ethnic minorities (Models 1 and 2). Only satisfaction with the NLD government is a statistically significant factor affecting ethnic voters' choice between ethnic parties and the NLD (Model 3). Among the Bamar voters, in addition to party identification and vote choice, higher income and living in urban areas are associated with a higher likelihood of voting intention for the NLD than the USDP in the 2020 election (Models 4 and 5). For ethnic voters, those who felt the economic situation was better than before would tend to vote for the NLD rather than the USDP (Model 7). Voting intentions between ethnic parties and the NLD, the significant factors are satisfaction with the NLD government and being an urbanite (Models 8 and 9). All other factors are too small to have any statistically significant impact. Minorities living in states do not favor ethnic parties over the NLD, contrary to the findings in Table 1, indicating the NLD's growing popularity among ethnic voters living in states. Among those who did not reveal their vote choices and intentions (Table E in the appendix), the factors determining their decision (no vote) over the USDP are similar to those found in Table D. However, improving economic conditions lead ethnic voters to vote for ethnic parties or report a no vote than for the USDP, re-confirming that economic conditions are a major consideration for minorities when casting a ballot for ethnic parties or (more likely) the NLD.
Based on Tables 1 and 2, Hypothesis 1 is largely supported by the results. Bamar voters' vote choices or vote intentions between the two Bamar major parties are not influenced by economic evaluation, whether current or retrospective. Neither income, education nor being an urban dweller show consistent effects, indicating that a sociological model of voting behavior has yet emerged in Myanmar (Lazarsfeld, Berelson, and Gaudet 1968). Most of the variations are explained by partisanship or vote choice in the last election. Even people without a party ID or those who decline to reveal their vote choices are more likely to vote for the NLD than the USDP in the upcoming election. A comparison of distributions of predicted probabilities for voting the NLD (vote intention) in the upcoming 2015 and 2020 elections is shown in Figure 3. Although the likelihood of voting for the NLD decreases among Bamar voters identifying with the NLD in the 2020 election, those without a party ID are more likely to vote for the NLD than they were in 2015. This applies to ethnic voters as well. Minority voters are more likely to vote for the NLD even if they identify with the USDP. According to the findings, a majority of voters prefer the NLD to the USDP regardless which party they identify with. Because of the military's long-term poor record of managing the economy and protecting human rights, the results are not surprising. Minority groups, on the other hand, are particularly affected by economic factors. They are more likely than Bamar voters to cast economic votes. After so many years of ethnic conflict, minorities are likely to crave normalcy. In this case, voting may not be based on considerations of political rights, but rather, based on economic concerns. However, the NLD is also popular among many ethnic voters, Note: Standard errors are in the parentheses. Significance level *p < 0.05; **p < 0.01; ***p < 0.001.
making the majority of explanatory variables ineffective predictors to distinguish vote choice between ethnic parties and the NLD. Although the NLD is popular with both Bamar and ethnic voters, some ethnic parties still managed to win seats. The results in Table 3 then provide insight into which type of party is more electorally viable. Based on the information available for the 2020 election, the dependent variable is the seat share of winning parties at the regional/state level. Results indicate that ethnic parties and parties with military ties tend to win fewer seats (Model 1). A limited number of seats can be won by ethnic parties since they tend to compete in constituencies with a high proportion of co-ethnic voters. Despite the USDP's participation in every constituency, it won a much smaller proportion of seats at the state and regional levels than the NLD. When focusing on ethnic parties, it appears that only merger parties can win more seats at the state level (Model 4). Rebel groups-turned-parties, by contrast, have no impact on seat share because these parties usually compete in a small number of districts. This suggests that party mergers make viable parties stronger since ethnic parties can gain more seats through this electoral coordination. However, the results do not indicate that parties that won no seats in the previous election will be able to win seats when they merge to avoid vote splitting. We need constituency-level data to examine whether party mergers increase electability. Table 4 analyzes the 2015 electoral data in which information is available down to the constituency level to understand whether the characteristics of parties can increase electability. Regressing on the vote counts and whether each party won the only constituency seat, the results show rebel groups that turned into parties not only garner more votes but also have a higher chance of winning a seat. Furthermore, parties that participated in the 1990 election and merged parties are also more likely to win votes and seats. As a result, two types of ethnic parties are viable, those that merge and those with rebel roots. However, a merger of parties may pose a problem.
As the parties in Rakhine State demonstrate, merging parties are easily disintegrated. The Arakan National Party (ANP) was formed by the merger of the Rakhine Nationalities Development Party (RNDP) and the Arakan League for Democracy (ALD). A split occurred in the party in 2017, and the party's leaders re-registered for the 2020 elections. Election results in 2020 showed the ANP and the Arakan Front Party splitting the seats. In contrast, rebel groups that have turned into parties, such as the Pa-O National Organization, tend to have stable support within the constituencies they serve. It can be argued that these parties are only participating in districts with a large co-ethnic population. However, there are many other ethnic parties that also focus on co-ethnic populations, but most often, they fail to gain seats. Therefore, striving to prove their credibility, in this case by fighting for their ethnic groups, can help them gain stable support from their voters. The result is in line with Hypothesis 2. As a result of the availability of electoral data, we can only analyze whether the characteristics of the party have any effect on electability at the district level for the 2015 election. With the findings of the regional/state level, the conclusion seems to be that merged parties are not only capable of winning more seats, but also likely to win a seat at the district level. It is unclear, however, whether the strategy of party merger increases the likelihood of winning in the 2020 election without information about the district level. As mentioned earlier, this strategy may help ethnic parties gain more seats, but such parties are also prone to disintegration, which is inimical to establishing social roots and institutionalizing the party. In contrast, rebel groups-turned-parties enjoy more stable support, but their bases tend to be small, and they may only be concerned with the welfare of one ethnic group. Therefore, it remains to be seen whether such stable support can contribute to overall party institutionalization among ethnic parties.
Our findings support the arguments that military dominance strengthens prodemocracy forces that benefit the NLD, and ethnic conflicts give ethnic parties with rebel backgrounds the social roots to maintain stable electoral support. Yet, there are alternative explanations for the NLD's victory in 2020 that need further discussion. One alternative explanation involves vote splitting among ethnic parties and the cancellation of elections in several state constituencies, which might have benefited the NLD. In 2010, there were 15 parties representing their constituencies in the lower house. In 2015 and 2020, the number of parties in the lower house decreased to 12. Meanwhile, there were 91 parties registered to run in 2015 and 94 in 2020. The fact that more contesting parties don't necessarily translate into more parties entering parliament seems to support the argument of vote splitting. However, there would be fewer parties in parliament if vote splitting continued to plague ethnic parties. Apparently, party mergers prevented the negative effect of vote splitting, while parties with rebel roots continued retaining their support.
In 2020, elections were canceled in 15 constituencies, all in states where ethnic parties might have an advantage. The NLD's seat share in 2020 would have increased as a result of the cancellation (the denominator becomes smaller). However, the NLD gained 255 seats in 2015 and 258 in 2020 in the lower house. As the number of seats won by the NLD increases in 2020, and since those districts with canceled elections in theory should vote for ethnic parties, it suggests that the NLD's number of seats is not affected by those canceled elections.
A second alternative explanation is that the NLD benefitted from the FPTP electoral rule when small party supporters voted strategically. This explanation seems to be supported by Figure 3, in which identifiers of ethnic parties were more likely to vote for the NLD in 2020 than in 2015. However, Duverger's Law also suggests that FPTP with single-member districts would tend to form a two-party system. In Myanmar, the effective number of parties (ENP) 3 based on the number of seats in the lower house was 1.56 in 2010, 1.57 in 2015, and 1.47 in 2020. If there are other parties in parliament than the two major parties, the ENP is likely to be more than two in most democracies using FPTP. It is, therefore, rare for a particular party to hold a large number of seats in a stable democracy, even in the case of FPTP. A lopsided distribution of seats toward the NLD in 2015 and 2020 is not simply due to strategic voting because FPTP did not benefit the USDP at all even though in theory this major party should also get an advantage. In other words, there are other reasons for the NLD victory than the electoral rule bias. The bipolar political climate is what persists to allow the NLD to win more seats in 2020, as this article argues.

Conclusion
Previous studies of Myanmar's elections examined the results from the perspective of normal politics, thus contributing NLD victory to electoral rule bias, economic performance, and party strategy. In this article, the role of the military is placed at the center of the analysis. Due to the military's dominance in politics, a bipolar environment persists between the pro-democracy and pro-military camps. A contest of this sort forms a centripetal force in favor of democracy. From survey results, it appears that the majority of Bamar voters did not consider economic performance when deciding on their support for the NLD. By contrast, minorities who longed for normalcy would vote for parties capable of improving their economic well-being. However, as a result of the long-term conflicts between the military and ethnic communities, ethnic voters tended to place their hopes in the NLD. People who believed ethnic voters were more concerned with their political rights might be surprised to learn that a preference for federalism is not related to their vote choice.
While most ethnic parties in Myanmar suffered from electoral system bias, some ethnic parties had nevertheless managed to secure stable electoral support despite this disadvantage. While ethnic parties that had engaged in party mergers had increased their chances of winning, some without such a strategy still enjoyed support from their core constituencies. Certain parties were able to cultivate their social roots, and this article argues that having a rebel heritage gave them credibility. Unless there is a long-term ethnic conflict in the country, not many countries would have parties relying on rebel backgrounds for credibility. When analyzing ethnic parties in Myanmar, this factor should be taken into consideration.
The findings of this article have theoretical implications. For countries in transition, like Myanmar, these few elections are the prototype of a party system that will develop later, but it is based on the assumption that politics will gradually become normal. In such a scenario, incumbents' performance, and specifically the economy in Myanmar's case, will be key factors in voting decisions. Yet, Myanmar's politics would never be normal, due to the 2008 constitution, which gave tremendous power to the military. In realizing that people would continue to vote for the NLD despite its poor performance, the military ended the democratic experiment abruptly. However, democratic elections would not co-exist with military dominance. It is, therefore, a futile effort for the military to seek legitimacy through electoral mandate in the first place.
Political instability has been caused by ethnic conflicts in independent Myanmar. Redressing ethnic tensions can be achieved through ethnic representation in state institutions. The effectiveness of formal rules, such as the federal system and lenient electoral rules, depends on the assumption that ethnic voters will vote for parties representing them (Lijphart 2007). Ethnic voting is not a default choice for ethnic minorities, however; it depends on how ethnic parties establish themselves as credible advocates for the community's interests. When ethnic communities lack access to such parties, they will vote for the majority party they believe can provide a better standard of living for minorities, as was the case in Myanmar. As well as being explained by the strategic behavior incentivized by the FPTP rule, the result could also be derived from the mundane hope of anyone living through prolonged conflict.
During the short life of Myanmar's democracy, military dominance and ethnic conflicts had shaped the electoral results as well as preventing Myanmar from developing a normal political system. Even though the military has no plans to hold another election in the near future, several things may remain intact if elections are allowed again. One thing the coup will do is strengthen the credibility of ethnic parties still fighting the military. Rebel groups-turned-parties will remain popular in their constituencies. The small size of the strongholds of these ethnic parties likely means that the military will allow them to win seats to showcase the representative nature of the parliament. Given that the biggest challenge is the NLD, the military may do what the Thai military did to disintegrate the pro-democracy forces: postpone the election as long as possible until the charisma of the opposition leader completely fades. In light of the age of Aung San Suu Kyi and several top officials of the NLD, this seems likely (it also points to the leadership crisis within the NLD). Another lesson can be learned from Cambodia, where opposition parties were banned altogether and the ruling party competed with other government-sponsored parties in elections to demonstrate competitiveness. Either way, the military wins elections, but never legitimacy.