Does Government Accounting Information Matter to Gain Votes? Evidence from Local Elections in Indonesia

: Voting decisions have become a strategic tool to analyze whether voters consider an incumbent's economic policies make it worthwhile re-electing the incumbent. If a strong correlation between the information on economic policies and the election results could be proved, evidence would thus be found that government accounting captures data on the attributes of government finances, consistent with the information incorporated by constituents in their voting decisions. This study investigates the association between government accounting information and local election outcomes and observes two regional government clusters that held local elections in 2017 and 2018, in which 198 incumbents ran for office again. By employing regression analysis, this study finds that several accounting information items impact the vote acquisition of the incumbents. The specific finding is that budget accounting information has a more decisive influence on votes for the incumbents than financial accounting information does. The result implies budget-ary accounting information that represents service performance is beneficial for gaining votes. The limitation that should be considered is the emphasis on the assumption of voter rationality, in which the voters accumulate performance information on the incumbents for their voting decisions. It is highly improbable that voters will allocate time to collecting and reading government financial statements on purpose.


Introduction
In the era of public sector reform, government accounting plays a strategic role in providing information to measure and evaluate financial performance and government service performance (Mir & Sutiyono, 2013;Brusca et al., 2019). In addition, it communicates the results of government policies to stakeholders (Cuadrado-Ballesteros et al., 2019). Consequently, external reporting is being transformed to make accounting information more consistent with accountability requirements (Rogošić, 2021;Steccolini, 2005). Furthermore, the role of government financial reports in making the government accountable is emphasized (De Vries & Giger, 2014;Ryan et al., 2002).
The political economy literature proposes that accountability is considered an essential element in democracy, for which accountability for an incumbent's performance is a central issue in voting decisions (Lee et al., 2016;Plescia & Kritzinger, 2017). The attention of all parties with political interests (including the voters) on the accountability of an incumbent has strengthened (De Vries & Giger, 2014;Rutherford, 1992). It is more likely that voters have a more profound concern for government accounting information during an election, even though they may obtain it from various sources (Feroz & Wilson, 1994;Ingram & Copeland, 1981). Voters are assumed to impound all the information about the incumbents, to decide whether to retain or replace them (A. M. Cunha et al., 2018). Thus, the relationship between government accounting information and election results is plausible at election times.
Empirical research has provided evidence that government accounting information is correlated with election outcomes (Ingram & Copeland, 1981;Feroz & Wilson, 1994;Brusca & Montesinos, 2006;Aidt et al., 2011;Fernandes et al., 2013;Maria Teresa Balaguer-Coll et al., 2015;Cunha et al., 2018;Cuadrado-Ballesteros et al., 2019). These findings also strengthen the argument that the accounting system can capture government financial attributes that are consistent with the voters' information in their voting decisions (Ingram & Copeland, 2012;Narayan et al., 2014;Pajala et al., 2018) However, the empirical evidence internationally contradicts the empirical evidence in the Indonesian context. For example, Yuhertiana et al. (2019) found that audit opinion mediated by online financial statements did not affect the electability of an incumbent.
The study concludes that voters do not rely on government financial information when choosing regional heads. Other studies (e.g., Yuliati et al., 2016;Aksa & Saputra, 2019;Kristanto, 2020) found that financial performance, accountability, and audit opinions have no impact on an incumbent's chances of being re-elected. However, those studies provide evidence that is far from being well defined as to whether voters hold well-performed incumbents with good accountable for their office.
The previous findings resonate with the pattern of the local political market in Indonesia, characterized by widespread corruption, domination of power by a specific elite group, the entrenchment of dynastic politics, money politics, and ethnoreligious polarization (Hadiz, 2004;Mietzner, 2012). When local politics is characterized by undesirable practices (e.g., corruption, money politics, domination of dynastic power, etc.), most politicians are perceived as dishonest; voters may conclude that all politicians are. Subsequently, voters assume that discriminating among candidates based on their performance is not worth the effort (Svolik, 2013). Moreover, the vote-buying practice creates a "bad culture" of political apathy, and accountability evaluations are less likely to be associated with voting decisions (Keefer & Khemani, 2005).
Even though scholars have often determined the elections for regional heads as being low-quality democracy, Fossati (2016) finds Indonesian voters, in some instances, using their votes based on their assessments of an incumbent's performance. These findings confirm the conceptual framework of retrospective voting (see John Ferejohn, 1986;Healy & Malhotra, 2013), where accountability issues are the primary source of information for voters to evaluate elected politicians. Moreover, several political changes in Indonesia support the apparent connection between voting and accountability. First, there was a significant increase in infrastructure development from the central government during the Jokowi era, the elected president of Indonesia for the 2014 to 2018 period (Warburton, 2016;Salim & Negara, 2018). As previously documented, infrastructure spending is a critical winning strategy because it has high visibility, from the voters' perspectives, and ultimately increases the incumbents' electability (Veiga & Veiga, 2007). Second is the practice of the massive manipulation of discretionary spending by local governments ahead of local elections (Setyawan & Setyorini, 2018;Darmastuti & Setyaningrum, 2019;Rahmanto et al., 2021). The incumbents rely on their authority in designing discretionary budgets (such as social assistance and grant spending) as a vote-winning strategy. These policy practices are believed to be recorded in government accounting information; hence we investigate whether the information does increase the voter turnout at the local level.
This study observes the results of the direct elections for regional heads in Indonesia during 2017 and 2018, in which 270 local governments held elections. The election results show that 128 of the 198 (65%) incumbents won; the rest were defeated 1 . These results reinforce the idea that direct elections become a channel through which voters punish policymakers who are believed to be unaccountable for the provision of public goods and services (Barro, 1973;John Ferejohn, 1986).
The study classifies the regions that held local elections based on their sociodemographic characters. This grouping aims to understand how voters process accounting information in their voting decisions. The analysis of the suitability of the incumbents economic policies with the basic needs of the voters is shaped by the sociodemographic conditions (Feroz & Wilson, 1994;Ingram & Copeland, 1981;Tiebout, 1956). This voter grouping is supported by development disparities and high economic disparities between the regions in Indonesia (Akita, 1988;Suwanan, 2015).
This study finds a strong correlation between government accounting information and election results proxied by the votes for the incumbents. Hence, this study provides different findings of clusters characterized by their sociodemographics. For example, in groups with affluent characteristics, information on the increase in local incomes and asset balances positively affects the votes for the incumbents. On the other hand, income taxes and deficits prove to be unpopular policies in the polls. Meanwhile, in clusters with common welfare, the increase in asset balances, education spending, social assistance spending, and local income increase the votes for the incumbents. However, housing spending, local tax revenues, and deficits lower the number of ballots. In particular, this study also shows that local government's capital expenditure (infrastructure) does not secure re-election for an incumbent.
The remainder of this article is organized as follows. The following section explains the role of accounting information in voting decisions, which is equipped with a review of the results of previous studies. Then we describe the variables, data, and models used, followed by the empirical results. The final part is this study's discussion, conclusions, and limitations.

Literature Review
Voting decisions are adapted from the context of investment decisions detailed by the Financial Accounting Standard Board (FASB) (1980). The idea of usefulness departs from the context "to be relevant, accounting information must be able to make a difference in decision making, by helping users of the information predict the present and future events based on past outcomes or to confirm or correct expectations." However, public sector accounting information users' information needs differ from that of commercial reporting (Fernandes et al., 2013;Torres, 2004). Moreover, many parties doubt the users of government accounting information (P. S. Gomes et al., 2015;van Helden & Reichard, 2019). Since public sector accounting reform proponents do not specify how government accounting information is to be used, no decision process model underlies how government accounting, including the role commercial-like information plays, is presented (McLeod & Harun, 2014;Pajala et al., 2018;Zimmerman, 1977).
The field of voting depicts the relationship between politicians and citizens as voters. Most scholars articulate that voting decisions are a part of the agency theory's accountability concept (see John Ferejohn, 1986;Chan & Rubin, 1987). Accountability arises from delegating authority to elected officials to manage public resources. Based on this perspective, it is reasonable that what elected politicians do is linked to the voters via the responsibility hypothesis (Plescia & Kritzinger, 2017;van Helden, 2016;Nannestad & Paldam, 1994). This hypothesis describes that a voter most likely observes the economic improvement, assesses the performance, and then alters his/her vote accordingly.
A voting decision is a form of decision taken by the voter to maximize his/her interests in the future (De Vries & Giger, 2014). Voters are assumed not to have complete information about the consequences of future decisions. Therefore, the correspondence between their votes and preferences over possible electoral outcomes will be less precise (Lupia, 1994). To be able to make decisions, voters need relevant information. Government accounting provides voters with useful information to assess an incumbent's policies and performance because it reflects the economic consequences of an incumbent's political choices (Chan & Rubin, 1987;Rogošić, 2021). An incumbent's policies impact assets, debt, income, and spending. These accounting numbers surrogate that incumbent's priorities in fulfilling the needs of his/her constituents (Ingram & Copeland, 1981;James & John, 2007;Ryan et al., 2002).
In the past, the voting literature has shed light on retrospective voting. The electorate has a strategic role as an appraiser of past policies, actions, and the government's performance ( (Tilley & Hobolt, 2011). Since then, a bulk of empirical research has amassed a tremendous amount of information on retrospective voting. The direction moves toward what an incumbent is doing to improve citizens' welfare. Theoretically, voters are expected to identify the conditions in which they believe the incumbents are accountable for past performance in those areas of public life that the incumbents oversee (Cutler, 2004;Di Pietro, 2019;John Ferejohn, 1986).
The earliest studies provide mixed evidence regarding the correlation between accounting figures and election results. Ingram et al. (1981) prove a strong correlation between accounting ratios and the odds of winning a major mayor in the United States.
They include sociodemographic factors that underlie the accounting ratios' abilities to discriminate the election results. They find the accounting ratios representing the service level and the municipal debt-related could differentiate the election result. The subsequent empirical findings (Bradbury & Scott, 2015;Brusca et al., 2019;A. Cunha et al., 2016;D. Gomes & Sargiacomo, 2013;James & John, 2007) reinforce Feroz et al. (1994) andIngram et al. (1981), who prove that voters have concerns about the level of services for health, welfare, and debt. Next, Fernandes et al. (2013) carried out a new approach incorporating an index of accounting information's quality and financial accounting ratios as explanatory variables. The study classified a sample of Portuguese municipalities of small cities and large cities, where the dividing indicator was 20,000 inhabitants. Fernandes et al. (2013) found that voters were more concerned about providing goods and services to the communities in the smaller towns. Meanwhile, voters in the larger cities were more concerned with their quality of life and the politicians' fiscal policies. The first two studies employed a survey instrument as a citizens' opinion poll to measure the incumbents' electability. There was little doubt as to whether the polls would be the same as the final election results. This study prefers to use the final official election results to provide more reliable findings. The two studies did not correlate voter satisfaction with the incumbents' performances, but instead linked it to the independent data in the accounting figures. Nonetheless, Ingram & Copeland (1981) and Feroz & Wilson (1994)  While municipal capital expenditures are increased/reduced in the pre-election period, these costs will likely go unnoticed by voters (Veiga & Veiga, 2007).
An analysis of the taxation factor shows inconclusive findings. Cunha et al. (2018) found that the ratio of tax revenue divided by total income positively affects a mayor's reelection chances. Likewise, the findings of Maria Teresa Balaguer-Coll et al. (2015) show that tax revenue positively influences the re-election of mayors. However, María Teresa Balaguer-Coll & Brun-Martos (2013) prove that tax revenue does not affect a mayor's re-election. Veiga & Veiga (2007) show that mayors visibly lowered city taxes in the preelection period. Tax increases were also associated with a decrease in the chances of a successful re-election bid (Lowry et al., 1998). These findings confirm that voters have different reactions to the tax policies of city governments.
There is less evidence that government accounting information influences election results in the Indonesian context. Previous studies analyzed accountability measures using audit opinions and local performance indices, but these studies found inconsistent results. Yuhertiana et al. (2019) examined the role of internet financial reporting and audit opinions on regional elections in Indonesia, in 2018. Their findings are that internet financial reporting and audit opinions have no impact on an incumbent's re-election, but audit opinions affect internet financial reporting. They conclude that voters do not base their choices for regional heads on the financial reporting on the internet. Other findings documented that the dimensions of government accountability and audit opinion are not the determining factors for an incumbent's victory in several of the elections for the regional heads in Indonesia (Aksa & Saputra, 2019;Kristanto, 2020;Yuliati et al., 2016).
Financial reporting is based on the fundamental assumption that interested parties are rational, evaluative, and maximize their interests (Zimmerman, 1977). Therefore, the ruling party has discretion in financial management; increasing their perquisites also increases their re-election chances. This proposition is proven by empirical evidence that the incumbents use discretionary spending to win elections (Duchin et al., 2019;Stratmann et al., 2016). The behavior is identical to the political phenomenon, which is often referred to as the pork barrel. It means the efforts made by the incumbent include allocating funds and getting closer to the constituents in their area to give them an incentive to re-elect the incumbent (John Ferejohn, 1986).
In short, this study takes advantage of local elections in Indonesia to prove the possibility of the usability of accounting information's influence on incumbents' votes.
This study proposes the following hypothesis: Government accounting information will likely attract voters to vote for the incumbent.

Data and Samples
This study utilized local governments in Indonesia that held elections for the regional heads in 2017 and 2018. The purpose of this study was to examine the influence of local government accounting information on votes for the incumbents. The data used to achieve the objectives were from incumbents who ran again in the 2017 and 2018 elections. There were 272 regional governments that held local elections in 2017 and 2018.
In these local elections, 198 of the candidates were the incumbents. However, the data that were used in this study came from 193 local governments, because of incomplete financial data from five of the local governments.

Variables and Research Model
The research model depicted retrospective voting. The concept was based on the idea that voters use elections to judge whether the government is accountable, based on past performance (Cutler, 2004;Healy & Malhotra, 2013). The electoral selection allows voters to maintain an incumbent whose characteristics align with their interests-the incumbent's factors include financial accountability that surrogate represent in government accounting information (Ingram & Copeland, 1981). The model was developed based on previous studies (Brusca et al., 2019;Feroz & Wilson, 1994). The research model constructed the effect of accounting information on the number of votes for the incumbents seeking reelection.
Government accounting information was not the sole factor affecting the number of votes for the incumbents. Therefore, we included other variables in the model, including governance factors (Bastida & Benito, 2007;Boyne et al., 2010;Di Pietro, 2019;Ferraz & Finan, 2008;McCabe et al., 2008) and political factors (Aldrich, 1993;Martins & Veiga, 2013;Sakurai & Menezes-Filho, 2008). Consequently, this study built a model that connected government accounting information, governance, and political factors to elections for the regional heads in Indonesia. The proposed model was as follows: Commission. The independent variable consisted of accounting information, which was classified into financial accounting (Financial) and budget accounting (Budgetary) (Fernandes et al., 2013;Ingram & Copeland, 1981). The Financial variable consisted of accounts in the balance sheet, namely assets, investment, debt, and leverage. Meanwhile, Budgetary was an item from budget accounting, which consisted of the spending and revenue group. The Financial and Budgetary variables used budget changes starting from when the incumbents were first elected until the last budget before the election.
The differences were divided by the account balance in the first year of each incumbent's tenure, to avoid nominal deflation. 2 In addition to the accounting information variable, we included local governance and political factors as control variables. Governance consisted of the audit opinion, budget transparency, and quality of government (Bastida & Benito, 2007;Di Pietro, 2019;Ferraz & Finan, 2008). Meanwhile, the political aspects (Pol) consisted of political competition and alignment with the national rules (Martins & Veiga, 2013;Sakurai & Menezes-Filho, 2008;Veiga & Veiga, 2007). A summary of the explanations of the research variables is presented in Appendix 1.

Cluster Analysis
Cluster analysis groups individuals or objects into clusters so that things in the same cluster are more similar than objects in other clusters (Hair et al., 2014). However, we realized that the composition of the constituents had heterogeneity in their sociodemographic characteristics that reflect their basic needs. Therefore, the cluster analysis's output was the group (a cluster). Table 1 presents the sociodemographic characteristics used as the basis for the grouping. The ratio of workers in the agricultural sector-2017 (%) Nonfarming Worker The ratio of workers in the non-agricultural sector-2017 (%) Formal Worker The ratio of workers in the formal sector-2017 (%) Informal Worker The ratio of workers in the informal sector-2017 (%) Source: Sociodemographic data from official data published by the Central Bureau of Statistics on the page www.bps.go.id.
This sociodemographic characteristics referred to previous research (McGee & Fabricant, 1953). The selected sociodemographic characteristics indicated the basic needs of the voter population. Thus, the conditions' profile correlation was carried out so it would be closer to the voters' preferences for the incumbents' economic policies.

Multiple linear regression analysis
This study employed regression analysis to analyze the relationship between vote acquisition (as the dependent variable) and the incumbents' economic performance, as represented by financial accounting information, budgeting, governance, and political factors. In addition, multiple linear regression analysis was performed to find out the direction and how much influence the independent variables had on the dependent variable (Ghozali, 2018). Thus, the result produced explanatory factors determining the incumbents' vote acquisition in each local election.

Cluster Analysis
The cluster analysis used the K-means procedure to maximize each cluster's homogeneity while maximizing the heterogeneity between the groups. The resulting groups should exhibit high internal homogeneity and high external heterogeneity. The

Descriptive Analysis
The descriptive statistical analysis aimed to explain the mean and standard deviations in the data. Changes in the use of accounting information, and its balance, occurred at the beginning of the incumbents' tenures (2012) and at the end of their term in office (2017). The balance change was intended for voters to collect information, not only on single one data date. Still, voters can respond more quickly to the level and quality of services provided during the incumbents' tenures. Table 3 contains a descriptive analysis of the data. between the independent variables as they had a multicollinearity level of more than 10 (VIF score > 10).
In general, the change in total expenditure for all the samples was 0.9408, while Cluster 1 was 0.8974, and Cluster 2 was 0.9729. These three figures indicated that the average total spending was close to the same model. However, each type of expenditure showed various changes for the group All Samples, Cluster 1, and Cluster 2. Interestingly, social assistance spending had the highest increase compared to expenditure items in the three groups. The All Samples, Cluster 1 and Cluster 2, were 2.8186, 1.3090, and 3.9338 respectively. Social assistance spending in Cluster 2 experienced a massive increase. Cluster 1 had a budgeting policy of increasing the economy's budget items, including security and order, and environmental protection.
A descriptive analysis of the character of local governments' governance showed Cluster 1 had better governance than Cluster 2. The governance component consisted of the audit opinion, governance performance index, and budget transparency, where Cluster 1 had three indexes which were greater than those in Cluster 2. However, for political competitions, which stated the number of candidates competing in local elections, the two clusters had similar averages of 2.9756 and 2.9640 for Cluster 1 and Cluster 2 respectively.
In addition, in Cluster 1, more incumbents were allied with the national ruling party than there were in Cluster 2.

The regression results
The study used the votes for the incumbents and the outcome of the local elections as the dependent variable. The independent variables were financial accounting information, budget accounting information, local governance, and political factors. The results were predicted to explain how these independent variables influenced the acquisition of votes for the incumbents. This research employed cross-sectional data because we only observe the local election in the period of 2017-2018. The model met the goodness of fit, including normality, multicollinearity, and heteroscedasticity. Based on a normality analysis, the model's residuals met the normality assumption. The Kolmogorov-Smirnov (K-S) test showed evidence of this, with a significance of 0.257 (> 0.05). Then, the multicollinearity analysis used the variance inflation factor (VIF) with an indicator below 10, except for the variables of assets, investments, and liabilities. Therefore, these three variables were excluded from the model. The model was freed from heteroscedasticity by regressing it, where the dependent variable used the residuals from the linear regression. The heteroscedastic test's results showed a significance (p-value) of 1.00 (> 0.05). Since the study used cross-section data instead of intertemporal data, it did not make confounding errors for the current period (t) with the previous period (t-1). Thus, the autocorrelation test was not required (Ghozali, 2018). Table 4 shows the regression's result. The independent variable represented a set of financial accounting and budget accounting information items. The use of multiple items of information aimed to capture the outcomes of the incumbents' policies that coincided with voter preferences, as worked by Brusca & Montesinos (2006), A. Cunha et al. (2016), Feroz & Wilson (1994) and Ingram & Copeland (1981). Economic policies implemented by local governments can shape various social and political outcomes, ranging from public services such as health, education, infrastructure, housing, social protection, environmental issues, and many more (Fossati, 2016). The output of the analysis found that financial accounting information and budget accounting affected the votes for the incumbents in different ways. For example, information items such as assets and local revenue had a positive impact, while local taxes and deficits reduced the number of votes for the incumbents. As controls, governance and political factors also impacted the incumbents' successful re-election. Whether using All data, Cluster 1, or Cluster 2, the results showed that only part of the accounting information significantly affected the votes for the incumbents. The analysis results using all the samples showed the changes in assets, social aid, local taxes, local income, and the incumbents' vote acquisition deficits. Also, the findings in the two clusters revealed differences. The grouping of the voters based on sociodemographic characteristics led to an in-depth exploration of voting decisions. The hypothesis was partially accepted based on these results, because only part of government's accounting information increased the votes for the incumbents.
For governance factors, the audit opinion and budget transparency impacted voters' alignments with the incumbents. Audit opinion increased the vote for the incumbents, whereas local accountability decreased it in Cluster 1. In Cluster 2, local accountability and budget transparency affected the vote acquisition. If local accountability was positive (up), budget transparency went in the opposite direction. From the descriptive data, budget transparency in Cluster 2 was lower than in Cluster 1. Political factors such as political competition and alliances with ruling national parties significantly affected the votes for the incumbents. These two political variables substantially impacted all our tests, Cluster 1 and Cluster 2. Thus, it showed that political factors dominate vote acquisition in Indonesia.
The model was tested using a logit regression (presented in Appendix 2) as a robustness test. The dependent variable used a binary variable; if the incumbents win it was 1, otherwise 0. The independent variable used the variable items in Table 4. Some budget accounting information items impacted the incumbents' chances of winning, along with governance and political factors. As with the results in Table 4, policies on education spending benefitted incumbents in Cluster 2. Similarly, local incomes had the same impact on the incumbents winning. In summary, the results from the logistic regression supported the main findings.

Discussion
This study examines the influence of government accounting information on voting for the incumbents as a voting decision. The local elections in Indonesia allow for investigating the correlation; moreover, not all incumbents won in the local elections. This study supports Ingram & Copeland (1981), who said sociodemographic characteristics influence local financing patterns. Comparing the descriptive data between Cluster 1 and Cluster 2 shows that the incumbents' economic policy packages differ. Cluster 1 has higher social and economic characteristics and a fiscal policy focusing on the economy, security and order, and the environment. While Cluster 2 has a lower socioeconomic characteristic, the incumbents' policies prioritize public services, education, social protection, and housing. Also, the incumbents in Cluster 2 tripled social assistance spending in the leadup to the local elections. The local tax revenue and income increases were higher than in Cluster 1. This finding supports Däubler et al. (2021) and Lewis-beck & Stegmaier (2000), by emphasizing that different cities have different economic policy packages.
The differences in the fiscal policies of the incumbents cause variations in the government's accounting information. They lead to large variations in the findings when analyzed using three approaches. Using all the samples (All Data), the first approach shows four accounting information items (assets, local taxes, local income, and deficits) that impact the votes for the incumbents. The second approach uses a sample of local governments in Cluster 1, and the analysis proves that economic expenditure and local revenue affect the votes for the incumbents. In Cluster 1, the economic spending policy disappointed constituents, thus reducing the number of votes for the incumbents. This is understandable because the expenditure has not resolved the unemployment problems (previously illustrated that Cluster 1 has a higher unemployment rate than Cluster 2). On the other hand, the ability of local governments to raise local revenues benefits the voters.
Finally, a third regression analysis approach with samples belonging to Cluster 2 shows more attractive evidence. The findings in Cluster 2 noted more significant accounting information items. These items include assets, housing spending, education, social assistance, local taxes, local income, and deficits. Particularly for education expenditure and social aid, the two accounting information influences significantly prove the voters' alignment with the incumbents prioritizing voter needs in Cluster 2. As explained in the previous section, Cluster 2 consists of districts with less well educated people, lower per capita incomes, and HDI; the poverty rate is also higher than in Cluster 1.
The findings in Cluster 2 show that voters are satisfied with the incumbents' policies that increase the education budget. Cluster 2 is characterized by people with low levels of education, low HDI, and lower per capita incomes. Therefore, it is reasonable if Cluster 2 chooses the incumbents that increase the education budget. The result implies that voters in Cluster 2 expect policies oriented toward improving education because of the conditions. The increase in education spending helps voters obtain an education at a low cost, or perhaps for free. It seems that the incumbents in Cluster 2 take advantage of the voters' poor conditions. Thus, the offer of higher education spending suits the primary needs of the voters in Cluster 2. Statistically, the increase in education spending in Cluster 2 in the period from 2012 to 2017 was higher than in Cluster 1. This suitability between the incumbents and the voters' response to the fiscal policy designs is political equilibrium (Drazen & Eslava, 2006). The findings prove that there has been a political balance in which rational voters favor incumbents who target them with prior spending, even though they know it may be electorally motivated. It implies that accounting information captures the incumbents' strategies that deliberately design spending according to the expectations of the targeted voters.
The increase in local assets significantly affects the votes for the incumbents.
The increase in total assets represents increased facilities, infrastructure, and the local governments' provision of goods and services to serve their citizens. With the availability of such resources, voters consider the regional governments' ability to serve the residents.
Thus, voters are willing to vote for the incumbent governments, trusting them to continue providing reasonable goods and services. Based on these findings, information on the increase in assets leads to local governments in Indonesia appearing confident about their ability to provide services to the community. Therefore, it results in respect for the voters, who then side with the local governments.
The other finding on the positive effects of social aid spending is providing social assistance as a grand strategy to gain the incumbents more votes. The impact of social aid spending in Cluster 2 is more substantial than in Cluster 1. This finding is reasonable because the sociodemographic conditions in Cluster 2 support the effectiveness of a social aid spending policy as a winning strategy for the incumbents. Several studies that use the Indonesian context prove that the practice of pork barrel spending on social assistance occurs ahead of any elections, especially in local governments where the incumbents re-run in the next election (Nugraheni & Subaweh, 2008). It is widely known that the incumbents use their discretionary budgets, especially to attract swing voters, and increase their chances of being re-elected, (Duchin et al., 2019). This finding confirms Stratmann et al. (2016), who found that legislators who were elected based on geographic constituent representation zones use their discretionary budgets to please their constituents and fulfill previous campaign promises, and to secure their local constituency's loyalty. Regarding the voter's reaction to social aid, this phenomenon shows the opportunistic behavior of the voters.  (2013) and Drazen & Eslava (2006). Increasing local incomes attests to the visual performance of an incumbent, from a voter's perspective. Furthermore, the increase in local revenues improves the local governments' capacity to provide public services beyond the minimum service standards financed by the transfer financing from the central government. We also acknowledge that local governments in Indonesia that have the highest local income can provide better public services, particularly those that fulfill the poor or the lower-income people's needs (Narayan et al., 2014).
The study also proves that raising local taxes is not beneficial for winning in Cluster 2. This finding confirms that raising taxes is not popular from the voters' perspective. An increase in tax rates is identical to a decrease in the voters' disposable income. Moreover, local tax increases' negative impact suggests that constituents resist the burden placed on them to fund local public services. These findings implicitly support Veiga & Veiga (2007), who said that the move to lower city taxes in the pre-election period does not affect the mayor's victory. Other researchers, such as Maria Teresa Balaguer-Coll et al. (2015) and Lowry et al. (1998), conclude that increasing taxes also decreases an incumbent's re-election chances. Another item of budget accounting information that reduces the votes for the incumbent is a budget deficit policy. The larger the deficit, the more voters interpret it as increasing the local government's financial risk. Local voters are fiscally conservative. Consequently, it can be accepted that policies to raise local taxes and budget deficits reduce the votes for the ones proposing them.
Regarding the increase in capital expenditure and housing expenditure, the incumbents who designed the rises in infrastructure and housing spending did not receive the voters' appreciation. Analyzing the three sample groups provides evidence that supports the idea that infrastructure development does not encourage voters to re-elect the incumbents. This finding contradicts Veiga & Veiga (2007), who state that increased capital expenditures positively affect the likelihood of re-election, in the pre-election period.
However, Brusca et al. (2019) prove that governments with higher capital investment do not have an advantage in elections. In other words, infrastructure development policies are not always popular with voters. This argument supports Banerjee et al. (2010), who state that performance information only impacts voter behavior if it is directly related to the voters' welfare, and primarily related to local policies' performance at solving daily problems.
The audit opinion is the governing factor influencing an incumbent's victory. By obtaining an unqualified audit opinion, the local government has an image of itself as a financially accountable institution. Besides, the audit opinion is also a measure of the incumbent's achievement in financial management, and the public perceives it as clean government (Bradbury & Scott, 2015). Thus, the audit opinion can raise a positive image in the eyes of voters. In Cluster 2, the test results show budget transparency harms votes for the incumbents. This study uses a checklist of the budget information's content presented on an official local website. The descriptive statistics show that transparency is 24.46%; meaning local budget transparency is still low. Thus, the lack of budget information from government websites decreases the votes for the incumbents. The quality factor of the local governments' administration reduces the incumbents' chances of winning in Cluster 2. The appropriate explanation for this finding is that the performance index in Cluster 2 is lower than that in Cluster 1.
Political factors, namely political competition and national party alliances, significantly affect the incumbents' electability. Political competition and the incumbents' party alliances affect the votes for the incumbents in Cluster 1 and Cluster 2. The effect of political competition shows the same results in both clusters, lowering the chance of an incumbent's victory. The more competitors there are, the more the votes are divided.
On the other hand, an alliance with the national ruling party shows significant positive results.
Meanwhile, with political party alliances, although the results in both clusters have a positive effect, the level of significance in Cluster 2 is more vital. This finding proves that perceptions of "national party brands" are more substantial in Cluster 2. Thus, in the 2017/2018 local elections, an incumbent's victory was primarily determined by the incumbent's alliance with the national ruling party. Local elections are an extension of national political contests, so a partnership with the ruling national parties attracts local voters (Reif & Schmitt, 1980). In voting decisions, voters simply elect the incumbent based on use a method shortcut with the majority party's identity, in which guides voters to choose candidates, with less knowledge. In general, voters rely on the name of their political party (Aldrich, 1993) This study also provides evidence of voter behavior that maximizes their interests based on these findings. A voter would choose an incumbent to increase local assets, education spending, and incomes. These items represent the incumbent government's ability to provide better public services. Conversely, voters penalize incumbents who raise local taxes and increase the deficit, because these two items might reduce their disposable income. Especially for an excess budget deficit, voters reject the additional financial risk, which will eventually be added to their burden.

Conclusion and Limitations
This study explores the results of elections for the regional heads in Indonesia during 2017 and 2018, in which 198 incumbents ran for re-election. This study identifies the correlation between local government accounting information, local governance, and political factors with voters' information needs during local elections. The result concludes that voters are still concerned with budget accounting information that is relevant to their needs when making decisions.
This study has several limitations. First, this study assumes that voters are rational individuals who consider all the relevant information in their voting decisions. Even though it is highly improbable that voters would look for this information, they are more likely to behave with rational ignorance. Therefore, the conclusions made in this study still depend on these assumptions. Secondly, the budget classifications are limited to the categories based on the local government's function. These classifications are global and do not show the expenditures that are genuine solutions to people's problems. Future studies should clarify the details, such as spending on poverty, food, and clean water. These three aspects are directly related to the daily needs of the voters, so the accountability measures can be more accurate in assessing the voter against the incumbent. In addition, we recommend that further studies focus on the readiness of an incumbent to handle critical situations such as the COVID-19 outbreak and extreme climate change. We also recommend using a composite index to simplify items of government accounting information, so that the model is more concise.