THE EFFICIENY OF VILLAGE GOVERNMENT SPENDING IN INDONESIA: A META-FRONTIER ANALYSIS

ARTICLE INFO Introduction: This study aims to measure the efficiency of village government spending and examine the effects of village transfers (Dana Desa) and institutional properties on its efficiency. Background Problems: The village law has endowed extra grants to village governments, which questions if the villages are prepared to adequately handle large increases in funding. Novelty: While previous studies address the misappropriation in spending within the municipal dimension, this study explores the spending efficiency of the autonomous sub-municipal governments and explains the impacts of both lump-sum grants and bureaucracy factors on spending efficiency. Research Methods: This study analyzes the Indonesian 2014 Village Governments’ dataset, using the meta-cost frontier in order to measure village spending efficiency, then it probes the causal impacts of endowed fiscal transfers and bureaucratic factors on the obtained efficiency. Findings/ Results: The results suggest that granting direct transfers would exacerbate the spending inefficiencies of autonomous villages. Administrative factors such as a large bureaucracy and a lack of bureaucratic capacity within the body of village governments positively affect spending inefficiencies. Conclusion: The result of this research reflects that there is a need to evaluate the village governance policy to increase spending efficiencies, specifically focusing on the adequacy of village institutions to handle village transfers. Article history: Received 1 April 2019 Received in revised form 5 September 2019 Accepted 23 January 2020


INTRODUCTION
Indonesian decentralization reform is gradually progressing, with the fragmentation of responsibilities within government authorities (Smoke and Lewis, 1996). This decentralization reform comes with various law packages, the most recent being Law 6/2014 on Village Government. Its implementation is progressing nationwide and has been implemented at the sub-municipal level. The law formalizes the creation of village governments, which are quasi-federal structures in the articulation of democratic legitimacy and public service provisions at a sub-municipal level (Hlepas, Kersting, Kuhlmann, Swianiewicz & Teles, 2018).
The village law greatly increases village budgets, injecting an additional 21 trillion rupiah (about US$ 1.5 billion) into village budgets in 2014 as top-down transfers (Aspinall & Rohman, 2017). These transfers account more than 70% of rural/village government revenues. Given the newness of such large endowments from central sources, there have been somewhat limited studies into village financial management systems that question their preparedness to adequately handle large increases in funding. The qualitative study from the Corruption Eradication Commission finds that unclear regulations and guidance about village transfer spending, and the lack of capacity of the executive body in ruralgovernments, have potentially caused government expenditure misallocations. 1 Antlöv, Wetterberg & Dharmawan (2016) argue that the village budget is usually overstated. From the experience of the National Program of Village Empowerment from 2006 to 2014, they found that there is a significant gap between a village's proposed budget and the budget's realization. Lewis (2015) provides empirical evidence of the flaw in village transfers, for example, poor villages get fewer transfers. However, there is no study to explore the potential abuse of the transfers from the side of the village governments' expenditure. The studies on addressing misappropriation in spending are abundant but limited to the dimension of the sub-national or municipal level (e.g. Akai, Sato, & Yamashita, 2001;Boetti, Piacenza, & Turati, 2012;De Borger & Kerstens, 1996).
Thus, this paper tries to fill the gap above by investigating the efficiency of village government spending post the implementation of Law 6/2014 on Village Government. By analyzing the parametric spending inefficiency from 80,179 Indonesian villages in 2014, the paper shows that autonomous villages have higher spending efficiencies than the villages with nonautonomous status. The autonomous villages are the villages with independent governance that have the rights to village transfers. Meanwhile, the non-autonomous villages (kelurahan) are sub-governments under the responsibility of the municipal governments (Martinez-Bravo, 2014). It appears that the independency to govern promotes the efficiency of public spending. Likewise, granting direct top-down transfers improves villages' government spending efficiency, albeit bureaucracy factors such as ballooning bureaucracy and the lack of the village officials' capacity worsen efficiency.
The first half of this article discusses the background and framework of the village decentralization framework and its transfers, together with theoretical support for the effects of bureaucracy on its efficiency. The second half of the paper provides empirical results.

Indonesian Village Decentralization Framework
Indonesia's decentralization policy began after the collapse of the authoritarian regime of the Soeharto era in 1998. The enacted decentralization law in 1999enforced the full transfer of authority from central to local governments. There are two significant amendments to the 1999 decentralization law; Law 32/2004 on Local Government and Law 23/2014 on Local Governance. Despite the progressive stage of decentralization, the central and local governments are still at the stage of finding the proper fragmentation for the balance of authority (Aspinall, 2003). The granting of autonomy was supposed to reduce development inequality between the central and local governments, and between the rich and poor regions (Roudo & Chalil, 2016). However, the rising inequality between rural and urban areas propels the public demands for expanding decentralization to the village level (Antlöv, 2003).
As a hurried response, the government introduced Law 6/2014 on Village Government. The law acknowledges that village governments (pemerintahan desa) have the right to govern and to provide public services based on their own needs and circumstances, within their administrative boundaries.
Village governments have responsibilities to provide certain public services for their "local" indigenous people. Government Regulation 43/2014 provides the details of the village governments' authority, which are: (1) basic public services (health and education), for instance, cram schools (taman bacaan), maternal health posts, emergency units (posyandu), or others; (2) village-scale infrastructure and transportation, for instance, agro-irrigation, local markets, village roads, and others. Following these authorities and functions, the central government decided to transfer the funds using an intergovernmental transfer scheme.
Law 6/2014 obliges the central government to share at least 10% of the total fiscal transfers in the national budget with the village governments. 2 The use of village transfers is not Chalil explicit; however, 30% of these transfers should go to village government operations and 70% for capital development spending. 3 Lewis (2015) shows the calculation for village transfers, which can be perceived as a general formula with several demographic and geographic variables. Based on the calculation process, the nature of village transfers is close to lump-sum grants since they do not represent the real needs for the villages' development. Furthermore, Lewis (2015) suggests that village transfers are unequally distributed. He finds that villages with high levels of poverty receive less funding than villages with greater access to funding, especially oil and gas transfers. In his remarks, the villages' service responsibilities are unclearly defined and the villages' public financial systems are inadequately prepared to handle the transfers.
On the other hand, Maharjan (2014) suggests that the inadequate capacity within the village governments is a result of the lack of resources and budgets to train the village apparatuses. Despite the limitation of resources and capacity, village communities give above average scores for the performance of their villages' apparatuses for capturing the villages' needs, providing services, and accountability. He remarks that future village transfers should be increased to overcome the local problems.
The above literature shows a problem with the village transfers. First, the village transfers potentially exacerbate the efficiency of public services due to the inadequate capacity of the villages' governments. Second, the village transfers are essential to source funding for the villages' development. The following section explains the theoretical underpinning of the impact of village transfers and village capacity's relationship with spending efficiency. enforced the stipulation of Law 6/2014 that mandated the village transfers. 3 It is mandated on Government Regulation 43/2014.

Theoretical Underpinning: Maximizing Bureaucracy Model
The model for maximizing bureaucracy (Niskanen, 1968) is appropriate to explain the impacts of fiscal transfers and institutions on spending efficiency. The model starts with the assumption that the bureaucracy maximizes the total budget given the utility demands and costs, subject to production constraints. The total budget is derived from the marginal values of taxpayers with the decreasing marginal utility of public goods' consumption Q.
Then the bureaucrats who provide public goods face a constraint, in that the total cost of providing public goods must not exceed the total budget (TB), the "no fat" bureaucracy indicates that the total budget should cover the minimum total costs (Niskanen, 1968). Assuming that the aggregate cost function has a linear production 4 :

=
(4) In Figure 2, the equilibrium Q is located at = , and the maximized budget leads to an equilibrium = . Moesen & Van Cauwenberge (2000) extend this model by adding a fiscal residuum as a difference between the budget and the costs incurred. They explore the later version of Niskanen's model, where bureaucrats are inefficient producers, generating a utility for bureaucrats that take the form of, for example, coffee breaks, political appointments, and complicated procedures. The utility curve of bureaucrats intersects the fiscal residuum curve (in the lower panel, Figure 2), and produces public goods at level Q*, which is lower than the social optimum level ( = ).

Effec
In The effect of increasing the capacity of the bureaucrats will increase the total budget, since: Hence, the performance of the bureaucrats is a monotonic increase, so ( , ) is positive.
Therefore, TB increases, leading to a new equilibrium Q" where the provided service goods are larger than Q in Niskanen's model. Now we set the hypothesis as described below.
H2: Bureaucracy's capacity improves government efficiency In contrast, bureaucracy's size has a diminishing return on bureaucratic performance, which changes the sign.
The new equilibrium exists at the point c, the lower panel of Figure 2. Increasingthe bureaucracy's size decreases the total budget curve to TB". By keeping the utility of the bureaucrats unchanged, the provided public goods falls to Q", which is lower than the initial Niskanen's model, Q * . Now we set the hypothesis as described below.
H3: Increasing the bureaucracy's size reduces government efficiency The theoretical framework above demonstrates that extra budget funds increase the social optimal of providing public goods; conversely, the bureaucracy has two effects: increasing the size of the administration negatively affects the social optimal of delivering public goods; but bureaucratic quality creates a better condition. In the next step, these arguments are tested by using empirical data from Indonesian village governments.

Measuring the Cost Efficiency of Rural Government Expenditures
The studies on measuring the efficiency of government are abundant (e.g., see: Akai, Sato, & Yamashita, 2001;Battese & Coelli, 1992;and Boetti, et al., 2012). For each i producers, total expenditure (E), the quantity of produced outputs, and the input price; the cost frontier is expressed as: Where = = ∑ is expenditure incurred by the producer i, i = 1,2,..,I ; y i is a vector of N outputs produced by producer i, ( , ; ) is the cost frontier, w is the input price, which is assumed to be monetized, and β is the vector parameters to be estimated; exp(v i ) is the random stochastic shocks that varies for each producer. CE i is the cost efficiency of producer i, which is evaluated as follows: Where cost efficiency is a ratio of the minimum cost attainable in an environment characterized by exp(v i ). Assumingthe equality of revenue and expenditure, procured expenditure is equal to minimum cost if CE i = 1. Ashortfall of observed expenditure to minimum feasible cost is shown by 0≤CE<1 (Kumbhakar & Lovell, 2000).

Model Specification
The model specification follows Battesse & Coelli (1992): Where C i is the total revenues of village i, as a proxy of total expenditure, in this condition, procured expenditures should meet total revenues. x i represents output factors that are provided by ruralgovernments. µ is random noise, and δ is the inefficiency terms, both are i.i.d. The function is return to scale and follows the Cobb-Douglas rule. In the next step, the inefficiency terms are products of corruption, monitoring, size of the government (in matrix Z). For efficiency's measurement, the statisticprogramming measures cost inefficiency as: Where the estimated cost inefficiency has an exponential distribution, which varies from 1 (minimum cost) to infinity (Battesse & Coelli, 1992). Following the formula of Khumbakar & Lovell (2000), cost efficiency takes the form of the inverse of the estimated cost inefficiency: Therefore, cost efficiency varies from 0 (perfectly inefficient) to 1 (perfectly efficient).

Meta-Cost Frontier Approach
The meta-cost frontier approach allows us to estimate the cost efficiency in different given groups within the data. The method of the metacost frontier is similar to the cost frontier's estimation but the meta-cost frontier estimates the efficiency by groups. Battese & Rao's (2002) article presents the detailed methods. For j groups, the cost frontier becomes: Parameters ( ) ; ( ) are governed under the meta-cost frontier ; with the following restriction: Therefore, the evaluated cost efficiency (withingroup) becomes: The following equation represents the evaluated estimated efficiency within j groups: Village governments can choose their administrative status, whether they want to be autonomous (village government) or not autonomous (kelurahan/urban communities). 5 Since the village dataset contains all the sub-municipal governments, I divided the data into village and non-village groups. As the first step for each group, the calculation lists the sub-municipal governments which are the most inefficient, and the most efficient, in spending. The second step models the effect of bureaucracy on the inefficiency of ruralgovernments.

Data Description
The Indonesian Central Bureau of Statistics (BPS) provides an enormous dataset for the 2014 Village Development Survey (Potensi Desa 2014) covering all the village administrations in Indonesia (80,179 observations). The dataset is divided into two groups, the autonomous village (desa) group and the non-autonomous village (kelurahan) group. The autonomous villages account for 88.55% of the total observations and the non-autonomous villages account for 11.45%. The employed variables are organized into three categories: expenditure (E i ), output (y i ), and efficiency variables (Z).

Expenditure:
Holding the equality, the total revenue of the village government is a proxy of its total expenditure. It is a sum of own-source revenue, grants/contributions from higher tiers of government, and village transfers.
Output: There are two approaches to output, first is the physical output approach and second is the expenditure approach. The variables are as follows:

RESULTS AND DISCUSSION
At first, the meta-cost frontier analysis estimates the cost efficiency of village governments' expenditure. The observations are classified into two groups; the first one is the (autonomous) village government and the second one is the urban community (kelurahan-non-autonomous). Each group has different decision-making processes since urban communities do not have autonomy for discretionary spending. After estimating the cost frontier curve for each group, a meta-cost frontier curve is constructed. There are two model approaches. Model (1) is the output-approach and model (2) is the expen-diture-approach, for the robustness check. 6 Table  3 and Table 4 present the results respectively. The produced cost efficiency under exponential distribution has the maximum value of one for perfect efficiency. 7 6 Input price is indexed, since for the output approach mainly physical outputs are used as output variables. In Indonesia, physical outputs (construction, buildings, roads, etc.) have been standardized through government regulation. On the other hand, the expenditure approach does not necessarily impose an input price. 7 Programming on SFA produced a different scale of cost efficiency, which follows the estimation technique by Battese and Coelli (1992) where CE={1,Infinity} where perfect efficiency takes the value as one. I scale the efficiency measurement by taking the inverse of CE, to be consistent with CE under Kumbhakar and Lovell's (2003) approach, where CE = {0.1}. Figure 3 plots the CE for village, non-village and all observations using the output approach. Figure 4 shows a similar presentation using the expenditure approach. Table 5 presents the statistical summaries of cost efficiency. From the mean efficiency scores, an autonomous village is 22% more efficient than a non-autonomous village (kelurahan). This means that the autonomy status of the government can lead to efficiency. On the other hand, urban community (kelurahan) status is non-autonomous and entirely controlled by the upper tier of government; resulting in mean-efficiency scoresthat are lower than those of the autonomous village. However, a non-autonomous unit is 6% more efficient than an autonomous unit in its expenditure approach, which implies a smaller deficit/surplus and more care taken to balance expenditure. Through the lens of public spending, it is efficient, but it is worth noting that this approach does not represent the effectiveness of the expenditure.  Notes: Cost frontier model is carried using log transformation. All inputs and outputs are indexed by population. Technical efficiency is scaled as exponential distributions. Standard errors in brackets; *** denotes significance at 1% level, **at 5% and * at 10%. Source: Data analysis   result implies that a percent of increase of village transfer shares cut around 2,67%-2,77% of inefficiency scores. By comparing the magnitude of the impacts of the village transfers, the village group has a lower magnitude than the non-village one, which implies village transfers positively affect government spendings' efficiency, especially for the non-village governments. Similar effects are observed with the expenditure approach. Figure 5 and Figure 6 show the relationship.

Chalil
The existence of monitoring and representatives of the community on the villages' governing bodies should reduce inefficiency. The likelihood estimations do not report any significance. It seems that the presence of an adhoc representative does not have any effect. Figures 7 and 8 show that, for the distribution of CE by the availability of representatives, the result seems ambiguous since, from the graph, there is no significant difference in CE's distribution between an available representative government and the not-available one. The plausible explanation is that enlarging the participation of the community in the governmental process may increase the efficiency of spending. However, if the community's members are not well educated, the effect may be the opposite. Better-educated people may be also more altruistic as leaders, and thus they are less likely to engage in resources misappropriation (Mansuri & Rao, 2012). Vol. 35, No. 1, 2020 13 Source: Data analysis Relation Between Efficiency Score and Portion of Village Transfer by Village Status The validation of the second hypothesis is about the correlation between expenditure efficiency and the size of the bureaucracy. Adding a unit of government officers per family increases output inefficiency by 20.57 units and expenditure inefficiency by 5.22 units, which implies that increasing the size of the rural government's administration distorts expenditure efficiency. The magnitude is greater for nonvillage governments where the difference of the coefficients is 20 points for the output approach and 26 points for the expenditure approach. This finding shows that increasing the numbers of apparatus leads to less efficient government spending, and the effect is greater for the nonautonomous government. The results confirm the second hypothesis, where an increase in the size of the bureaucracy leads to spending inefficiency. Kau & Rubin (1981) and Treisman (2000) confirmed similar findings. Figure 9 and Figure  10 present the two-way plot showing a negative relation between the bureaucracy's size and its efficiency.

Journal of Indonesian Economy and Business,
Finally, the third hypothesis, which seeks correlations between bureaucracy's capacity and expenditure efficiency. In Table 3, the education of the village leader and his/her deputy has no significant effect on production efficiency. In Table 4, the education of the leader, as a proxy of bureaucratic capacity, influences the inefficiency negatively. The impact of education only matters for the (autonomous) village government. These findings are aligned with Aspinall & Rohman's (2017) recommendation, where the capacity of the bureaucracy improves the spending efficiency, especially when locals can elect their leaders according to their ability. Note that only autonomous village governments can hold local elections, non-autonomous villages cannot. For a non-autonomous village, the upper tier government directly appoints the village's leader.

CONCLUSION
Decentralization reforms in Indonesia reach a new level by acknowledging the autonomy of the village government, the smallest unit in the government structure. Following this reform, the central government allocates top-down transfers to the villages, yet these transfers are prone to corruption and inefficiencies.
The findings show that the villages could efficiently spend their expenditure, with effi-ciency scores above 68%. The village governments perform better if they have autonomous status and have full control over their village transfers. Several factors may explain this efficiency. Increasing the number of bureaucrats diminishes village spending efficiency, as expected. In contrast, increasing the bureaucratic capacity appears to improve the spending efficiency, even though the statistical results are only convincing for the group of autonomous villages.
This study's contribution is twofold. As a pioneering move, Indonesia expanded its decentralization policy by granting autonomy to the sub-municipal administrations. Thus, given the experience and policy practice in Indonesia, this study is the first to explore the efficiency of spending by the autonomous sub-municipal governments. Second, the paper contributes by explaining the impacts of both the lump-sum grants and the bureaucracy on determining expenditure efficiency. This paper complements the works by Boetti et al (2012) andDe Borger &Kerstens (1996). 10 Source: Data analysis Technical Efficiency of output approach (Y-axis) versus Size of Bureaucracy (X-axis) 10  Kerstens (1996) explore the efficiency of municipal/city governments by employing data envelopment analysis (DEA) and stochastic frontier analysis (SFA). These techniques produce the parametric and non-parametric measurement of inefficiency terms, albeit assuming there is a technology homogeneity among the observations. This paper uses a meta-frontier approach that allows for technology heterogeneity among the observations.

STUDY LIMITATION AND POLICY IMPLICATION
This study has three major limitations. First, this study employed village survey data from the National Statistics Agency. The budget data contains self-reporting assessments from the village governments, which arguably need to be validated by the upper tier governments. Therefore, the problem of measurement errors may arise during the empirical analysis. The solution to correct this problem is to combine the maximum likelihood estimators in the first step to estimate the efficiency scores, and use instrumental variable analysis in the second step to evaluate the efficiency determinants. Second, this study limits the measurement of government efficiency to the view of parametric cost efficiency, while assuming the price of providing public goods is indexed. However, it does not apply in real practice since the government's spending efficiency may come from the result of effective management, organization, and cheaper costs for delivering public services. Third, the paper uses the village leaders' and deputy leaders' education as the representative variable for the villages' bureaucratic capacity, which is challenging. The bureaucracy's capacity is a general concept, which needs a comprehensive measurement. A detailed study to measure the broad meaning of bureaucratic capacity should take place as a milestone in future research agendas, which later could be associated with the village government's efficiency.
The results suggest several policy implications, for instance creating a lean government structure for village governments and focusing on capital expenditure and physical infrastructure. Extending decentralization to the village level and increasing village transfers are appropriate policies; nevertheless, strengthening the current regulations, capacity and institutions are indispensable.