JBICI - JICA · generate the distribution of public good investment. Proposition 1: (i) Investment...
Transcript of JBICI - JICA · generate the distribution of public good investment. Proposition 1: (i) Investment...
ISSN 1347-9156
Governance Decentralization and
Infrastructure Provision in Indonesia
Shamal Chowdhury, Futoshi Yamauchi, Reno Dewina
JBICI
Working Paper
NO. 25
October 2007
JBIC Institute (JBICI)
The JBICI Working Papers are based on the research done by staffs of Japan
Bank for International Cooperation (JBIC) and published by the JBIC Institute. The views expressed in this paper are those of the author and do not necessarily represent the official position of JBIC.
1
Governance Decentralization and Infrastructure Provision in Indonesia1
Shyamal Chowdhury2 University of Sydney
Futoshi Yamauchi3
International Food Policy Research Institute
Reno Dewina3 International Food Policy Research Institute
Abstract: This paper examines the recent decentralization of governance in Indonesia and its impact on local infrastructure provision. The decentralization of decision-making power to local authorities in Indonesia has created the dependency of local public infrastructures on local resources. However, due to the difference in initial endowments, this may result in the divergence in local public infrastructures. The divergence may also happen due to differences in preference within communities. Using a village-level survey conducted in 1996, 2000, and 2006, this paper finds that (i) though local public infrastructure depends on local resources and preferences, (ii) decentralization has improved the availability of local public infrastructures, and (iii) communities are converging to a similar level of local public infrastructures. 1 We would like to thank Japan Bank for International Cooperation for financial support and the Indonesian National Statistic Office for discussion on PODES village data. This paper is part of our background research on decentralization, infrastructure and poverty dynamics in Indonesia. Remaining errors are ours. 2 Agricultural and Resource Economics, A04 RD Watt Building, Science Road, University of Sydney, NSW 2006, Australia. Tel: +2-93512786, Email: [email protected] 3 International Food Policy Research Institute, Washington, DC, USA.
2
1. Introduction Decentralization refers to the devolution of political decision-making power to elected local-level, small-scale entities (Bardhan 2002). With the fall of planned economies and centralized states, the decentralization of governance and the delivery of public services by local authorities have taken the centre stage of policies favored by development agencies and donor communities (World Bank 1999). Like many other countries, Indonesia also has embarked on the path to decentralization in the late 1990s following the economic and financial crisis. Decentralized services can be a way to achieve allocative efficiency since they should reflect the preference of the actual users of the services and the cost of the service should be reflected in the user fee for such services. Decentralization can also reduce information asymmetry that prevails otherwise between the central planner and the actual users of services. Decentralization however also makes local public services dependent on local resource availability leading to the possibility of divergence in the amount of local public good available between rich and poor communities. Since decisions are taken at local level, difference in preference within communities over different public goods may also result in the divergence in the type of public goods available. In this paper, the objective is to answer two questions: first, does decentralization lead to a divergence in the amount of local infrastructure availability between rich and poor communities? Since due to decentralization investment on local public infrastructure depends largely on local resources, differences in local resource endowment across communities may result in the differences in the amount of investment on public infrastructure. Second, does preference heterogeneity within a community lead to the divergence in the type of local infrastructure provision? Since the type of public infrastructure chosen depends on preferences at local level, differences in the preference within a community may result in the difference in the type of local public infrastructure provision. Three distinct characteristics made Indonesia an attractive test-case to answer the two research questions at hand: first, the relative exogeneity of the decentralization experiment; second, large internal variations in the availability of resources among local government jurisdictions; and third, the availability of both pre- and in-decentralization data on infrastructure at local level, which is discussed in more details in Section 4. In Indonesia, the decision to decentralize was exogenous, at least to the local communities, and the implementation of decentralization was relatively quick –as opposed to gradualism a big-bang approach was followed. Starting from the Dutch colonial role, for most of its modern history, Indonesia was governed by a centralized system where the local governments mostly functioned as implementing agencies of policies and programs designed by the central government. The reform agenda formulated by the multilateral donors and development agencies following the Asian financial crisis paved the devolvement of fiscal and political authorities from the central
3
to the regional governments. Through the enactment of Law 22/1999 on regional governance, responsibility for much of the government expenditures was decentralized to local (district) governments. The sweeping legislative and administrative changes in local governments introduced in 1999 have brought momentous changes on what local public infrastructures/goods will be financed and how they will be financed. Though all major tax bases, e.g., value added taxes, personal and corporate income tax, are still controlled by the central government, the local governments receive a united transfer over which they have full discretion. Barring five functions – security and defense, law and judiciary, monetary and macroeconomic policies, religious affairs, and foreign relations – local governments are responsible for all other functions including the financing and management of local public infrastructures, health and educational services. In the fiscal year 2002, the local government financed 44.3% of transportation development, 21% of health and social services, and 16% of education development (Eckardt and Shah 2006). Needless to say that transportation infrastructure, health and education constitute the major expenditure outlay for the local governments. Large variations in local resource availability and resource potentials among local government jurisdictions make Indonesia an attractive test case for the current research. In 2000 prior to the implementation of decentralization, the per capita income in the top 20 percent of the districts was three times more than in the bottom 20 percent of the districts. Similarly the poverty rates also differed significantly among local jurisdictions (poverty figures for top and bottom quartiles of the districts in 1999 were 8.53% and 43.07%, respectively).1 These reflect the uneven distribution of economic activities as well as natural resource endowments among local government jurisdictions. Though a conditional intergovernmental transfer program based on the fiscal needs addresses the heterogeneity of local governments, fiscal disparities still persist. Even after equalization transfers, for instance, the richest district has roughly 70 times as much as per capita revenue as the poorest district (Eckardt and Shah 2006). Our work complements the recently developed theoretical literature that focuses on the link between decentralization and welfare (see for instance, Bardhan and Mookherjee 2006, Besley and Coate 2003 and the reference therein). It also complements the cross-country empirical literature on the link between democratization and economic growth (Barro 1996 and the literature that follows it). While the theoretical literature in the past emphasized on the trade-of between the uniformity of service delivery under centralization, and uneconomic scale and cross-regional externalities under decentralization, the recent literature emphasizes on the limited accountability of local government officials and capture by local elites and interest groups. The current debate on the relative responsiveness of centralized versus decentralized delivery of public goods needs to be tested empirically, and the current work is an attempt in this direction.2
1 Authors’ calculation based on BPS data 2 The issue of spatial inequality and externalities in the context of Indonesia are discussed in another paper of ours (Yamauchi, Chowdhury and Dewina, 2007)
4
The remainder of this paper proceeds as follows: Section 2 builds a model that links decentralization of decision making to local resources and preferences. The main implications are that without progressive subsidy from the central government, decentralization may increase cross-community inequality in public goods provision, and heterogeneous preferences may increase investment favorable to large groups within a community. Section 3 describes the empirical approach followed in this paper that maps the initial local endowments and preferences to subsequent allocation on local public goods during pre- and in-decentralization periods. The unit of analysis here is village instead of district. This is due the availability of disaggregated data at village level. We, however, control of district fixed effects in the estimation process. The parametric approach followed here should be capable of distinguishing the relative responsiveness of public infrastructures provision during pre- and in-decentralization periods. Section 4 describes the data used to construct the village-level panels for the two periods – pre- and in-decentralization, and gives the summary statistics of the variables used. Section 5 describes the empirical results on the impact of local income and preferences on public goods provision in pre-decentralization and in-decentralization periods. Section 6 concludes the paper with some possible policy implications.
2. Model 2.1. Basic Setting This section sets up the environment in which the community decides on public good investment and allocation. Assume for simplicity that there is one public good. Also assume that the community allocates its resource into consumption and investment on public good, and receives returns (services) from such public good investment. For simplicity, assume that there is no subsidy from the central government. The community taxes the residents to raise funds for building the public good. The tax rate is assumed to be unique within the community. Therefore, community residents have to decide the rate following some mechanism. Before discussing the mechanism, let us describe individual decision problem.
Assume a degenerate income distribution in the community. Income iy is given to individual i . With tax rate t , the per-capita public good q is formed through ( )q tE y= if q q≥ − (or ( )
qE yt −≥ ) where ( )E y is the average community income, which is
exogenous to each individual in the community, and q− is the minimum sufficient level of the public good. Assume that 0q = if ( )tE y < q− . Meanwhile, we assume that this condition holds.
We write the utility for household as ( ) ((1 ) ) ( )i iV t y u t y f qβ; = − +
5
where ( )u . and ( )f . are utility functions for the consumption and public good service respectively, and β is the discount factor ( (0 1)β ∈ , ). Assume that both utility functions are strictly increasing and concave. Since the problem is single-peaked, the first order condition for agent i , ( ) ( ) ( )i iu c y E y f qβ′ ′= provides the most preferred tax rate for i . In other words, the most-preferred tax rate is determined by
( ) ( )( )
( )i
ii
u c E yMRS yf q yβ
′
′= = .
The relative income position affects the individually most-preferred tax rate.
Let ( )it y denote the tax rate, i.e., ( ) argmax ( )i it y V t y≡ ; . Given the non-degenerate income distribution, we also have a non-degenerate distribution of the most preferred tax rate. The community uses majority voting to decide the community’s tax rate. By median voter theorem, the tax rate t∗ is equal to
( ) argmax ( )m mt y V t y= ;
where my is the median income in the community. In this autarky case, since community income level (median, taking into account the distribution) affects investment in public good. The distribution of the community median income (across communities) will generate the distribution of public good investment. Proposition 1: (i) Investment in public good depends on community income level. (ii) Cross-community income distribution generates the distribution (inequality) of
public good investment across communities. In short, a higher average (median) community income makes it possible to collect higher local tax to build more local public goods. Therefore, without progressive subsidy from the central government, cross-community inequality in public goods may increase through decentralization. 2.2. Two goods, two groups: (1,2) - potential conflict between groups Suppose there are two groups in the community. Let φ denote the fraction of group 1 (so (1 φ− ) is the fraction of group 2). Public goods are 1 2( )q q, Let α denote investment share for the public good 1, given the total investment is q ,
6
1
2 (1 )q qq q
α
α
=
= −
The utility from services or next period returns is 1 2( )f q q, Therefore, each individual attempts to maximizes his/her value, as in the previous section 1 2( ) ((1 ) ) ( )i iV t y u t y f q qβ; = − + ,
Here the community can have a potential conflict between the two groups in determining the investment share kα
∗ . Each group tries to equalize marginal products from the two public goods, i.e., 1 2f f=
The most-preferred tax t for individual i with income iy is given by, given α 1 2( ) (1 )i iu c y f fβµ α α′ ⎡ ⎤
⎢ ⎥⎣ ⎦= + −
Here we assume α is determined through bargaining between the two groups, influenced by group proportions in the community population. 1 2 1 2(1 )α φα φ α α α∗ ∗ ∗ ∗⎛ ⎞
⎜ ⎟⎝ ⎠
= + − ∈ ,
Note that it is single peaked, so we obtain ( )i it y∗ . By median voter theorem, the median income determines tax rate. Then, community composition determines resource allocation. Proposition 2: Greater proportion of one group increases investment share favorable to the group. For example, gender of village head, male/female proportions among voters, and particular occupation or social group may matter in the allocation of public good investments.
3. Empirical Approach We have followed a parametric approach that focuses on the relationship between initial village resource and preference and subsequent stock of local public infrastructures. This
7
approach can help to map the changes in stock of local public infrastructures – an increase, no change, and deterioration – with the local resource availability and preferences. The main specification is:
iiiii ZINFINCINF εδϕβα ++•+•+=∆ ',0,0 (1)
The change in infrastructure stock between two period in village i, iINF∆ , is defined as a categorical variable and is our primary variable of interest. Initial measures of local resources, iINC ,0 , is one of the primary explanatory variables, where “0” denotes the initial period resource availability for village i. In addition, village characteristics such as population, existence of port/ and topography are included in iZ as further controls. The first measure of local public infrastructure considered here is the village road – the road that connects a village with the regional road network. As discussed, after the devolution of power, financing and maintenance of this type of roads falls under the purview of local authorities. Intuitively, villages with more resources are likely to invest more and maintain their roads better. However, investment in roads could be endogenously determined since the initial stock of village roads may affect the current income and subsequent investments in roads. This is accounted for by incorporating iINF ,0 that indicates infrastructure stock for village i in period 0. The second measure of local public infrastructure considered here is village streetlight. A specification similar to road is followed here, and intuitively, it should follow the same rationale. The third measure of local public goods considered here is the presence of school (or absence of it) – both primary and junior high school in the village. And the fourth measure of local public goods considered here is the healthcare infrastructure, namely if a village possesses polyclinic and puskesmas. These are the two primary healthcare giving facilities in Indonesia. Four public infrastructures considered here – road, streetlight, health and education – constitute, on average, the major expenditure outlay for the local governments in Indonesia as discussed in Section 1. The changes between pre- and in-decentralization periods should be a benchmark for measuring the relationship between decentralization and local infrastructure provision in Indonesia. In estimating the equation (1), we have some concerns. First, decentralization was implemented at the district level, so that observed and unobserved investment behavior should differ across districts. We compare results with and without district fixed effects to see how district-level unobservables affect public good investment behavior in both periods. Since we think that investment behavior is more homogeneous within a district than across districts, it is reasonable to control district fixed effects. However, we are not sure of possible direction of bias if the fixed effects are not controlled.
8
Second, we are also aware of possible correlation between the initial conditions and fixed factors, and unobserved shocks. For instance, well-endowed villages are likely to have good infrastructure to start with, and such villages also have high propensity to invest more on infrastructure or improve infrastructures further. This positive correlation may create upward bias in the estimated effect of initial infrastructure condition on the infrastructure investment (or improvement). Third, resource constraint and demand for services may differ across different types of infrastructure and villages. These factors change parameters of our interest. For example, demand for roads depends on economic activities in the village, which is correlated with the initial income level (and also road conditions). This makes it difficult to identify resource constraint against demand heterogeneity. However, if economic conditions are stable between pre-and in-decentralization periods, the comparison of parameters between the periods enables us to identify the role of resource constraint. For the preference heterogeneity within a village (PRE), the gender of the village head, his/her education, age and duration are considered. The rationale is similar to Chattopadhyay, Duflo (2004) where they found the evidence of gender differentiated preference for infrastructures. This gives us the following specification:
iiiiii ZINFINCPREINF εδϕβγα ++•+•+•+=∆ ',0,0,0 (1′)
While the first specification (1) should allow us to see the impact of local resource availability on local public goods, the second specification (1′) should allow for the preference heterogeneity within a local area controlling for resource constraints. We examine the roles of village head characteristics (age, gender, education and tenure) and, in some year, gender composition of voters in the village. The latter measure could be correlated with economic activities in the village, since migration opportunities change the composition. In the equation (1’), controlling the initial income and infrastructure condition also mitigates this potential bias. Besides the points we discussed above, we are concerned with exogeneity of village preference measure since it can be correlated with unobserved component of infrastructure condition. However, this possibility is mitigated in the above equation as we examine changes in infrastructure over time, not at a given point in time. One issue not discussed so far is the macro shock of the economic crisis that had taken place during the first period of our analysis. Being a major shock, it had affected all villages and districts irrespective of rich or poor. It is not unrealistic to assume that the effects of the shock were randomly distributed and there is no reason to believe a priori that there was any systematic relation between the shock and the resource availability between the rich and the poor villages. As a result, we do not need to control for the shock. Even if the shock had different effects on different districts, it should not matter for our estimation as long as it is not systematically correlated with the initial infrastructure conditions of the centralized regime. Besides, the impact of financial crisis
9
on infrastructure investment can be assumed homogeneous across districts in the centralized regime.
4. Data The data for this study has come from the Village Potential Statistics (PODES) community (desa) survey conducted periodically by Indonesia’s Central Bureau of Statistics (BPS). The PODES survey contains detailed information on the stock of public infrastructures at village in addition to the characteristics of villages, their geography and topography, the natural disaster that they faced in the past years. We have used three round of PODES survey – 1996, 2000, and 2006, and were able to match (part of the) communities across survey rounds for the PODES for the pre- and in- decentralization periods. Since the decentralization started in 2001, we have used the first two rounds, 1996 and 2000, for the pre-decentralization period, and compared 2000 with 2006 to examine changes due to decentralization (before and after analysis). We have matched the 1996 and 2000 rounds by village (i) recovering province, district and sub-district from codes that have changed from round to round and (ii) constructing unique codes for these administrative units, taking into account their mergers and splits. In this process, we have kept (recorded) village names to match the data, so some observations were missed if they changed village names (along with their splits). In this process, nearly 84 percent of the 1996 villages were matched between 1996 and 2000. It is about 81 percent of the 2000 villages between 2000 and 2006. Public infrastructures that are available and chosen for the current study are village road, village streetlight, schools and health care facilities. Among them, the first two, village road and streetlight, are non-excludable as well as non-rival, and externalities are limited within a village. The PODES survey includes information on the type of village road (soil, hardened, paved etc), and state of street lighting [Table 1]. Since we were comparing two periods (1996 and 2000, and 2000 and 2006), for road there could be 16 different states, and for street lighting there could be four different states. In stead of all these combinations, we have used three states for both road and lighting – improvement, no change, and deterioration [Table 2]. In addition, for school and health infrastructures, the presence (or absence of it) of primary school and junior high school, and polyclinic and puskesmas are considered, respectively. For the estimation purpose, for the local road and streetlight, three states – deterioration (e.g., from hardened road to soil road or from having streetlight to no light), no change, and improvement (e.g., from soil road to paved road) are constructed as a categorical variable (1, 2, 3, respectively). Similarly changes in education3 and health infrastructures
3 Note that there was a large decline in the number of schools (Table 1), both primary and junior high schools, between 1996 and 2000. Though there might be underreporting by villagers, this may not explain the total decline. Though it remains to be explored, most appropriately by using school census data, most
10
between two periods are categorized as one and three depending on if the changes were positive or negative, and no change is categorized as two. This is conditioned on the initial state at (t-1) period, which appears as a lagged dependent categorical variable. For schools, in addition, the mean distance of schools in district in initial period is also interacted with the initial village condition. For the estimation purpose both ordinary least squares (OLS) as well as ordered probit are used. Turning to the explanatory variables, for the availability of local resource at village level, the proportion of households that live in poverty (pre-prosperous households and prosperous households level I4) are taken as the indicator of income assuming that the higher the proportion of households in poverty, the lower the availability of resources for public goods. Though transfer from the central government via districts may reduce it, the cross-community differences in resource availability still prevails and it does not necessarily eliminate the resource constraint as discussed in Section 1. Note that due to the possible endogeniety between income and public goods, the income taken is a one period lag or income at the beginning of the period. The second explanatory variable of interest is the preference. Here a set of characteristics of the head of the village such as gender, age, education and duration, and the gender composition of the voters are considered. Note that the proportion of villages with female heads is very small though the total number is not due to the large sample size of PODES data. Table 2 provide summary statistics of all the above and other variables utilized in the estimation.
5. Empirical Results
Local income and changes in local public goods Table 3 to Table 6 show the estimation results where the availability of local public goods –roads, streetlight, schools, and health facilities - is regressed on local income controlling for other factors described in equation (1) including the initial stock of public goods. For all tables, a negative coefficient for village income indicator shows the dependence of local public goods on local income. Estimated coefficients are derived from ordinary least square (OLS), OLS with district fixed effects (FE), and ordered probit with district fixed effects (FE). Results are shown for both pre- (1996-2000) and in-
plausible causes are: many privately-run schools might have been shutdown, and many community-run schools might have become unsustainable. 4 The Family Planning Coordinating Board (BKKBN) of Indonesia classifies all households into five welfare status: (i) Pre-Prosperous Households (Keluarga Pra Sejahtera or KPS), (ii) Prosperous Households Level I (Keluarga Sejahtera I or KS I), (iii) KS II, (iv) KS III, and (v) KS III+. Poor households are often equated as the KPS households, and they are often defined to include the KPS and KS I households. See BKKBN (1994) for further details.
11
decentralization (2000-2006) periods. To ensure comparability between these two periods, parameters for the later period have been adjusted so that they now reflect the equal length of time (4 years). As can be seen in the table, the results show a significant dependence of public goods – both public infrastructures and public schools and health care facilities – on local income/resource. For the local roads, villages with more poor households have experienced either deterioration or less improvement in road condition compared to villages that have few poor households. Note that the reverse causality of road/public goods to income has been controlled for in the estimation by taking the income at the beginning of the period. It is therefore the local income, in addition to other factors, that determines the local public goods here. Results similar to the local roads are found for all public goods considered here. Local government authorities in Indonesia are not financially self-sufficient and their revenue-raising capabilities vary widely depending on local economic activities and resources. This inequality among local jurisdictions prevails even after financial transfers from the national government, and for the poor jurisdictions, local revenues fall well short of their expenditure responsibilities forcing them to have less public goods. This is true for all the public goods discussed here. Though one should expect such dependency of local public goods on local income not to exist in pre-decentralization period, the empirical findings show otherwise. How the decentralization has changed this dependence? A comparison between the estimated coefficients ( β ) of income between the two periods – pre- and in- decentralization suggests that the dependency of local public infrastructures on local income has substantially declined following the decentralization. While it should rather increase in in-decentralization period compared to pre-decentralization period, to our surprise, it has in fact decreased. This finding remains equally valid across public goods and across estimation methods. Though the addition of district fixed effects reduces the size of the income coefficient, it does not change the pre- and in-decentralization differences. This finding is counter-intuitive and goes against our initial prediction. It seems that the decentralization has eased the local resource constraints. Among the plausible explanations, competition among local government jurisdictions may have enhanced efficiency of local governments (Tiebout 1956, Oates 1972). Alternatively, this may have happened due to linking of local benefits with local costs, and local governments may now tailor the public goods and services to local circumstances (Musgrave 1959). Decentralization might have changed the priority among competing public goods that are locally financed, and the local public goods considered here might get more priority of local citizens compared to other alternative use of resources available to the local government jurisdictions. This issue of priority/preference is discussed further in the next section.
12
It seems that decentralization has led to an improvement in the local public goods availability. While the divergence in local public goods still exists to a large extent, the dependency on local income, which was initially thought to be the main impediment against convergence in in-decentralization period, seems not to exist. Horizontal inequalities among local jurisdictions are not uncommon in other countries and vertical transfers may not be sufficient to ensure equality. Nonetheless local jurisdictions provide public goods according to local demand and capabilities including fiscal and financial capabilities. The Indonesian experience also supports to this direction.
Local preference and changes in local public goods As discussed in the previous section, the preference of different groups within a village may vary between alternative public goods and the possibility of rank ordering of these alternatives may enable them to choose one public good over others. Table 7 to Table 10 present the estimation results for the impact of local preference on changes in local public goods availability over time where the availability of local public goods –roads, streetlight, schools, and health facilities - is regressed on a set of preference indicators controlling for other factors described in equation (1′) including the initial stock of public goods and local resource availability. Estimated coefficients are derived from the ordinary least square (OLS), OLS with district fixed effects (FE), and ordered probit with district fixed effects (FE). Results are shown for both pre- (1996-2000) and in-decentralization (2000-2006) periods. Two explanatory variables that are of most interest for the preference proxies are gender of the head of the village, and voter composition of the village. For the local road, gender of the village head does not have any significant impact. While voter composition does not have any significant impact in pre-decentralization period, it has become significant in in-decentralization period implying that more women voters results in better local roads. One plausible explanation is that the benefits of local roads are widely shared and known to the voters, and the road ranks very high at the preference list of women voters. This is not surprising given the findings of other studies (Jacoby 2000, Gibson and Rozelle 2003) on the benefits of rural roads for all households, particularly to the poorer ones, and women headed households (Chowdhury et al 2005). Unlike roads, the streetlight is significantly affected by the gender of the village head being the case that female village heads prefer streetlight more compared to male village heads. In addition, years of schooling of village head have a positive impact while the village head’s duration has a negative impact. Note that the size of the effect of female heads has in fact increased during the decentralized period when the devolution of power to local authority has made it possible to reflect their preferences. Similarly, higher the percentage of women voters in a village, higher is the likelihood of having streetlight in the village. This finding of gender-differentiated preference in public good allocation is similar to that found in Chattopadhyay and Duflo (2004). One plausible reason for this is the security and mobility of women at nighttime that may improve with village streetlight.
13
In the case of educational infrastructure, female village head has a positive impact on the supply of junior high school. However, effect becomes insignificant once controlled for district fixed effects. Similarly years of schooling have a positive impact but becomes insignificant once controlled for district fixed effects. Note that Indonesia has a quite diverse educational system with general education, technical education, religious education, public and private institutions - all co-existing together. Therefore allocating public funds at local government jurisdictions could be quite complex and not all of them are considered here. In the case of health infrastructure, the impact of gender differentiated preference indicators is either mixed or insignificant during in-decentralization period. Two plausible explanations are: first, unlike road and streetlight, the health (and educational) infrastructures are more complex to build and require resources beyond building physical structures such as doctors and nurses (teachers). Attracting and retaining such resources require policies and plans not considered here. Second, the typical free-riding and coordination problem of public goods due to externalities may arise that go beyond local resources and preferences. While externalities for local roads and road lights are easy to capture within a local community, health (and education) infrastructures are not easily excludable5. Why duration of elected local government officials is related negatively with local public goods provision? One plausible explanation is that elected representatives in Indonesia have a short time horizon that discounts the long-term prospects in favor of more immediate payoffs. If reelection prospect diminishes with duration in the office, the goal of public officials might systematically diverge from maximizing citizen welfare, especially in their second terms (Brennan and Buchanan 1980). Since the introduction of decentralization, Indonesian local government selection process, in addition, has also experienced some experimentation, which may have created some uncertainty about the future on public officials.
6. Conclusions In this paper, we have examined the impact of recent decentralization of governance in Indonesia on local infrastructure provision where unobserved heterogeneity and initial conditions are appropriately accounted for. The decentralization of decision-making power to local authorities in Indonesia has created the dependency of local public infrastructures on local government jurisdictions. Horizontal inequality that pre-exists among local government jurisdictions allows the possible divergence in local public good provision. The decentralization, in addition, has also allowed local preferences to be accounted for. If, for instance, women as a group prefer a particular public good to other alternatives, it may get financed in in-decentralization period.
5 Of course mechanisms, such as discriminatory pricings, can be designed for the out of community beneficiaries.
14
To model the dynamics of public good provision, we have introduced state dependence of public goods by conditioning the supply of public good in period t on the lagged public good stock in period (t-1). The availability of pre- and in-decentralization period data at the level of local government jurisdictions has facilitated this empirical exercise. In addition, the information on income indicators at village level, village administration and voters’ composition has allowed us to account for local level resources and preferences. The empirical findings of this paper support the notion that the decentralization has linked local public goods to local government endowments; the local public goods such as local road, streetlight, schools, and health facilities at village level depend on local resources. Despite the transfer of resources from the central government to the district government following the decentralization, the local public goods still depend on local resource availability, and poorer communities have less public goods than richer communities. This finding remains valid even after controlling for district fixed effects. However, contrary to our predictions, decentralization in Indonesia that has linked local public goods provision to local resources has also improved the availability of local public goods across communities. Though cross-community difference exists, they are converging to a situation with similar access to local public goods. This may happen due to the inter-jurisdictional competition and efficiency gain that economists have long been argued in decentralized states. Given the resource availability, the supply of some of the public goods considered here also depends on local preferences. It is found that women village heads and women voters have a strong preference for local roads improvement and village streetlight. However, for educational and health infrastructures, preferences are not explained well, and there might be coordination and free-riding problems due to externalities that are not well accounted here. Based on Indonesian experiences, the current decentralization and democratization trend observed in developing countries deserves cautious optimism. The devolution of decision-making power to local government jurisdictions in developing countries and the delivery of local public goods by them can enhance efficiency and reflect local preferences better than a centralized system. However, for some public goods, such as education and health, local authorities might not have the sufficient capacity, or may require inter-jurisdictional coordination to capture the externalities, that currently lacks.
15
References Bardhan, Pranab, 2002, “Decentralization of governance and development,” Journal of
Economic Perspectives, 16: 185-205. Bardhan, P. and Mookherjee, D. 2006. Decentralization and Accountability in
Infrastructure Delivery in Developing Countries. Economic Journal, 116(January): 101-127.
Barro, Robert J, 1996. Democracy and Growth. Journal of Economic Growth, 1(1): 1-27. Besley, T. and Coate, S. 2003. Centralized versus Decentralized Provision of Local
Public Goods: A Political Economy Analysis. Journal of Public Economics, 87(12): 2611-37.
BKKBN (1994), Pembangunan Keluarga Sejahtera di Indonesia Berdasarkan UU No. 10 Tahun 1992 dan GBHN 1993 [Family Welfare Development in Indonesia Based on Law No. 10/1992 and State Guidelines 1993], Kantor Menteri Negara Kependudukan/Badan Koordinasi Keluarga Berencana Nasional, Jakarta.
Brennan, Geoffrey and Buchanan, James. 1980. The Power to Tax. New York: Cambridge University Press.
Chattopadhyay, R., and E. Duflo. 2004. Women as Policy Makers: Evidence from a Randomized Policy Experiment in India. Econometrica, 72 (5): 1409-1443.
Chowdhury, S., M. Torero, Ali, A M M S., Rana, M., Garrett, J., Jahan, I., Islam, M. Sherman, T.2005. Urban-Rural Linkages in Bangladesh: Impact of infrastructure and the food value chain on the livelihoods and migration of landless households, women and girls in the Northwestern Region. Manuscript. IFPRI.
Eckardt, S., and A. Shah. 2006. “Local Government Organization and Finance: Indonesia.” In Local Governance in Developing Countries, ed. A. Shah, 233-274. Washington, DC: The World Bank.
Gibson, John and Scott Rozelle. 2003. Poverty and Access to Roads in Papua New Guinea. Economic Development and Cultural Change, 52(1) October 2003, 159-185.
Jacoby, Hanan. 2000. Access to Markets and the Benefits of Rural Roads. Economic Journal, 110 (465) July: 713-737.
Musgrave, Richard. 1959. Public Finance. New York: McGraw Hill. Oates, Wallace. 1972. Fiscal Federalism. New York: Harcourt Brace Jovanovich. Tiebout, Charles. 1956. A Pure Theory of Local Expenditures. Journal of Political
Economy, 64: 416-24. World Bank. 1999. World Development Report: Entering the 21st Century. Oxford
University Press. Yamauchi, Futoshi, Shyamal Chowdhury, and Reno Dewina, 2007, “Spatial Coordination,
Decentralization and Public Good Allocation: Nonparametric Evidence from Indonesia,” Manuscript, International Food Policy Research Institute, Washington D.C.
16
Tables and Figures Table 1: Road, Road Lightening, Schools and Health Facilities, and their Changes: 1996 – 2000, and 2000 – 2006
1996 2000 2006 1996-2000 2000-2006
Road (in %)Paved 57.51 55.42 58.03 - -Hardened 25.02 28.88 25.77 - -Soil 17.12 15.26 15.77 - -Other 0.35 0.45 0.54 - -Improved - - - 13.73 19.57No change - - - 74.07 68.66Deteriorated - - - 12.2 11.77
Village Road Lightening (in %)Yes 42.9 40.55 57.24No 57.1 59.55 42.76Improved - - - 12.39 23.23No change - - - 73.88 72.75Deteriorated - - - 13.73 4.02
Schools (in %)Primary school (yes-1) 92.42 73.98 91.3Improved 0.59 18.28No change 83.95 80.76Deteriorated 15.47 0.96
Junior high school (yes-1) 41.46 24.93 44.42Improved 2.87 21.64No change 78.93 76.22Deteriorated 18.2 2.14
Health Facilities (in %)Puskesmas (yes-1) 4.3 5.6 12.1
Improved 2.34 7.66No change 95.98 89.98Deteriorated 1.68 2.36
Polyclinic (yes-1) 10.9 11.6 10.7Improved 3.13 2.33No change 95.08 95.91Deteriorated 1.79 1.76
17
Tab
le 2
: Sum
mar
y St
atis
tics
Var
iabl
e N
ame
Var
iabl
e D
escr
iptio
n19
96-2
000
2000
-200
6
Mea
nS
td. D
ev.
Mea
nS
td. D
ev.
Roa
d_C
hang
eC
hang
e in
villa
ge ro
ad b
etw
een
(t-1)
and
t pe
riods
. Thr
ee p
ossi
ble
stat
es a
re: d
eter
iora
tion-
1, n
o ch
ange
-2, a
nd im
prov
emen
t-3.
2.01
40.
509
2.06
70.
562
Roa
d_ha
rdIf
type
of v
illage
road
was
har
dene
d in
(t-1
) per
iod
equa
ls 1
, els
e 0
0.25
00.
433
0.28
90.
453
Roa
d_pa
ved
If ty
pe o
f villa
ge ro
ad w
as p
aved
in (t
-1) p
erio
d eq
uals
1, e
lse
00.
575
0.49
40.
554
0.49
7Li
ght_
Cha
nge
Cha
nge
in v
illage
ligh
teni
ng b
etw
een
(t-1)
and
t pe
riods
. Thr
ee p
ossi
ble
stat
es a
re: d
eter
iora
tion-
1, n
o ch
ange
-2, a
nd im
prov
emen
t-3.
1.98
60.
448
2.18
10.
484
Ligh
t(t-1
)If
villa
ge h
ad ro
ad li
ghte
ning
in (t
-1) e
qual
s 1,
els
e 0
0.42
90.
495
0.40
40.
491
Prim
ary
scho
ol c
hang
eC
hang
e in
prim
ary
scho
ol b
etw
een
(t-1)
and
t pe
riods
. Thr
ee p
ossi
ble
stat
es a
re: d
eter
iora
tion-
1, n
o ch
ange
-2, a
nd im
prov
emen
t-3.
1.85
10.
372
2.17
30.
403
Prim
ary
scho
ol (t
-1)
If vi
llage
had
prim
ary
scho
ol in
(t-1
) per
iod
equa
ls 1
, els
e 0
0.92
40.
265
0.74
00.
439
Mea
n di
stan
ce p
rimar
y (t-
1)A
vera
ge d
ista
nce
of p
rimar
y sc
hool
s in
dis
trict
(in
Km
) in
(t-1)
per
iod
0.44
99.
849
1.28
032
.540
Juni
or h
igh
scho
ol c
hang
eC
hang
e in
juni
or h
igh
scho
ol b
etw
een
(t-1)
and
t pe
riods
. Thr
ee p
ossi
ble
stat
es a
re: d
eter
iora
tion-
1,
no c
hang
e-2,
and
impr
ovem
ent-3
.1.
847
0.43
32.
195
0.44
7Ju
nior
hig
h sc
hool
(t-1
)If
villa
ge h
ad ju
nior
hig
h sc
hool
in (t
-1) p
erio
d eq
uals
1, e
lse
00.
415
0.49
30.
249
0.43
3M
ean
dist
ance
juni
or (t
-1)
Ave
rage
dis
tanc
e of
juni
or h
igh
scho
ol in
dis
trict
(in
Km) i
n (t-
1) p
erio
d7.
084
33.1
227.
930
36.5
70P
olyc
linic
_Cha
nge
Cha
nge
in p
olyc
linic
bet
wee
n (t-
1) a
nd t
perio
ds. T
hree
pos
sibl
e st
ates
are
: det
erio
ratio
n-1,
no
chan
ge-
2, a
nd im
prov
emen
t-3.
2.01
30.
221
2.05
30.
312
Pol
yclin
ic(t-
1)If
villa
ge h
ad a
pol
yclin
ic in
(t-1
) equ
als
1, e
lse
00.
042
0.20
00.
055
0.22
7P
uske
sma_
Cha
nge
Cha
nge
in p
uske
smas
bet
wee
n (t-
1) a
nd t
perio
ds. T
hree
pos
sibl
e st
ates
are
: det
erio
ratio
n-1,
no
chan
ge-2
, and
impr
ovem
ent-3
.2.
007
0.20
12.
006
0.20
2
Pus
kesm
as(t-
1)If
villa
ge h
ad a
pus
kesm
as in
(t-1
) equ
als
1, e
lse
00.
106
0.30
80.
111
0.31
4
Inco
me(
t-1)
The
avai
labi
lity
of lo
cal r
esou
rce
at v
illage
leve
l in
the
(t-1)
per
iod.
If a
villa
ge h
ad 5
0% o
r mor
e fa
milie
s in
wel
fare
equ
als
1, e
lse
0.0.
606
0.48
90.
495
0.50
0
Hea
d’s
Gen
der
Gen
der o
f villa
ge h
ead
(mal
e –1
, fem
ale-
0)0.
980
0.13
90.
977
0.14
9A
ge o
f villa
ge h
ead
Age
of v
illage
hea
d in
yea
rs44
.655
8.68
344
.424
8.92
6E
duca
tion
of v
illage
hea
dN
ot c
ompl
eted
–1,
els
e 0
0.02
60.
160
0.04
00.
197
Edu
catio
n of
villa
ge h
ead
Prim
ary
scho
ol –
1, e
lse
00.
259
0.43
80.
171
0.37
7E
duca
tion
of v
illage
hea
dJu
nior
hig
h sc
hool
–1,
els
e 0
0.27
60.
447
0.28
10.
449
Edu
catio
n of
villa
ge h
ead
Hig
h sc
hool
–1,
els
e 0
0.35
00.
477
0.39
20.
488
Edu
catio
n of
villa
ge h
ead
Aca
dem
y –1
, els
e 0
0.03
90.
193
0.03
50.
184
Edu
catio
n of
villa
ge h
ead
Uni
vers
ity –
1, e
lse
00.
046
0.21
00.
075
0.26
4D
urat
ion
of v
illage
hea
dD
urat
ion
of v
illage
hea
d in
yea
rs5.
692
5.05
44.
637
5.11
3%
of w
omen
vot
ers
% o
f wom
en v
oter
s in
villa
ge a
mon
g to
tal v
oter
s49
.067
8.86
950
.169
4.45
3
Pop
ulat
ion(
t-1)
The
size
of v
illage
pop
ulat
ion
in (t
-1) p
erio
d29
85.0
0085
74.0
0028
71.0
0037
62.0
00S
tatio
n_P
ort(t
-1)
If th
e vi
llage
had
any
sta
tion/
term
inal
/airp
ort/s
eapo
rt in
per
iod
(t-1)
per
iods
equ
als
1, e
lse
00.
057
0.23
10.
068
0.25
2D
isas
ter(
t-1)
Any
dis
aste
r in
the
last
thre
e ye
ars:
yes
1, n
o-2
0.53
40.
499
0.39
40.
489
Topo
grap
hyV
illage
topo
grap
hy0.
715
0.45
10.
715
0.45
1
18
Tabl
e 3:
Loc
al In
com
e an
d C
hang
e in
Loc
al P
ublic
Goo
ds -
Villa
ge R
oad
Dep
ende
nt V
aria
ble-
Cha
nge
in v
illag
e ro
ad: d
eter
iora
tion-
1, n
o ch
ange
-2, i
mpr
ovem
ent -
3
1996
-200
020
00-2
006
OLS
OLS
Ord
ered
Pro
bit
OLS
OLS
Ord
ered
Pro
bit
with
Dis
trict
FE
with
Dis
trict
FE
with
Dis
trict
FE
with
Dis
trict
FE
Vill
ages
with
50%
or m
ore
fam
ilies
in p
re-w
elfa
re a
nd w
elfa
re 1
in
(t-1)
per
iod:
yes
-1, n
o-0
-0.0
835
-0.0
645
-0.1
949
-0.0
347
-0.0
347
-0.0
950
(0.0
105)
***
(0.0
070)
***
(0.0
212)
***
(0.0
086)
***
(0.0
062)
***
(0.0
170)
***
Roa
d_ha
rden
ded
in (t
-1) p
erio
d eq
uals
1, e
lse
0-0
.287
6-0
.303
6-0
.887
8-0
.145
4-0
.224
9-0
.633
1(0
.022
3)**
*(0
.021
4)**
*(0
.055
5)**
*(0
.018
4)**
*(0
.014
5)**
*(0
.035
9)**
*
Roa
d_pa
ved
in (t
-1) p
erio
d eq
uals
1, e
lse
0-0
.667
9-0
.743
7-2
.328
-0.4
825
-0.5
951
-1.6
632
(0.0
217)
***
(0.0
215)
***
(0.0
676)
***
(0.0
167)
***
(0.0
145)
***
(0.0
471)
***
Pop
ulat
ion
of v
illag
e in
(t-1
) per
iodX
10^4
0.01
330.
0075
10.
0234
0.08
070.
0330
0.10
00(0
.004
9)**
*(0
.003
5)**
*(0
.012
)***
(0.0
113)
***
(0.0
067)
***
(0.0
193)
***
If th
e vi
llage
had
any
term
inal
/sta
tion/
port
in (t
-1) p
erio
d: y
es-1
, no-
00.
0566
0.06
490.
2059
0.03
180.
0488
0.13
79(0
.009
9)**
*(0
.009
5)**
*(0
.029
8)**
*(0
.007
9)**
*(0
.006
2)**
*(0
.017
3)**
*
Vill
age
topo
grap
hy (h
ill a
rea
-0, f
latla
nd-1
)0.
0503
0.05
640.
1702
0.05
110.
0501
0.13
75(0
.010
9)**
*(0
.010
1)**
*(0
.029
8)**
*(0
.009
5)**
*(0
.007
1)**
*(0
.018
8)**
*
Any
dis
aste
r in
last
thre
e ye
ars:
y-1
, no-
2-0
.014
8-0
.032
5-0
.102
6-0
.023
9-0
.007
9-0
.023
1(0
.010
8)(0
.010
3)**
*(0
.030
3)**
*(0
.007
5)**
*(0
.006
1)(0
.016
7)C
onst
ant
2.49
972.
7056
1.65
401.
9084
(0.0
253)
***
(0.0
244)
***
(0.0
195)
***
(0.0
128)
***
Obs
erva
tions
4916
849
168
4916
851
995
5199
551
995
R-s
quar
ed0.
230.
270.
250.
34
Rob
ust s
tand
ard
erro
rs in
par
enth
eses
* si
gnifi
cant
at 1
0%; *
* si
gnifi
cant
at 5
%; *
** s
igni
fican
t at 1
%D
istrc
it sp
ecifi
c co
nsta
nts
are
not r
epor
ted
in th
e ta
ble
19
Tabl
e 4:
Loc
al In
com
e an
d C
hang
e in
Loc
al P
ublic
Goo
ds -
Roa
dlig
htD
ep. V
aria
ble:
Cha
nge
in v
ill ro
adlig
ht: d
eter
iorta
tion-
1, n
ocha
nge-
2, im
prov
ed-3
1996
-200
020
00-2
006
OLS
OLS
with
Ord
ered
pro
bit
OLS
OLS
with
Ord
ered
pro
bit
Dis
trict
FE
with
dis
trict
FE
Dis
trict
FE
with
dis
trict
FE
Vill
ages
with
50%
or m
ore
fam
ilies
in p
re-w
elfa
re a
nd w
elfa
re 1
in
(t-1
) per
iod:
yes
-1, n
o-0
-0.0
856
-0.0
751
-0.3
076
-0.0
115
-0.0
305
-0.1
352
(0.0
115)
***
(0.0
074)
***
(0.0
009)
***
(0.0
090)
(0.0
041)
***
(0.0
006)
***
Vill
age
road
ligh
teni
ng in
(t-1
) per
iod:
yes
-1, n
o-0
-0.4
654
-0.6
373
-12.
3672
-0.3
711
-0.5
409
-12.
0419
(0.0
143)
***
(0.0
117)
***
(0.0
052)
***
(0.0
113)
***
(0.0
079)
***
(0.0
051)
***
Pop
ulat
ion
of v
illag
e in
(t-1
) per
iod
X 1
0^4
0.02
630.
0263
0.04
390.
1152
0.06
790.
6860
(0.0
089)
***
(0.0
089)
***
(0.0
0037
)***
(0.0
186)
***
(0.0
100)
***
(0.0
023)
***
If th
e vi
llage
had
any
term
inal
/sta
tion/
port
in (t
-1) p
erio
d: y
es-1
, no
-00.
0525
0.09
80.
4305
-0.0
174
0.04
870.
2196
(0.0
125)
***
(0.0
097)
***
(0.0
017)
***
(0.0
082)
**(0
.005
9)**
*(0
.001
5)**
*
Vill
age
topo
grap
hy (h
ill a
rea
-0, f
latla
nd-1
)0.
0997
0.09
890.
408
0.06
380.
0496
0.22
20(0
.012
2)**
*(0
.009
2)**
*(0
.001
3)**
*(0
.012
5)**
*(0
.006
0)**
*(0
.001
0)**
*
Any
dis
aste
r in
last
thre
e ye
ars:
y-1
, no-
2-0
.044
9-0
.001
9-0
.004
1-0
.006
40.
0106
0.04
25(0
.013
1)**
*(0
.008
7)(0
.000
8)**
*(0
.007
9)(0
.004
1)**
*(0
.000
7)**
*
Con
stan
t2.
1898
2.32
271.
5438
1.41
57(0
.015
8)**
*(0
.010
8)**
*(0
.015
1)**
*(0
.006
9)**
*
Obs
erva
tions
5212
752
127
5212
755
129
5512
955
129
R-s
quar
ed0.
230.
350.
280.
5R
obus
t sta
ndar
d er
rors
in p
aren
thes
es*
sign
ifica
nt a
t 10%
; **
sign
ifica
nt a
t 5%
; ***
sig
nific
ant a
t 1%
Dis
trcit
spec
ific
cons
tant
s ar
e no
t rep
orte
d in
the
tabl
e
20
Tabl
e 5a
: Loc
al In
com
e an
d C
hang
e in
Loc
al P
ublic
Goo
ds -
Prim
ary
Scho
ols
Dep
. Var
iabl
e: C
hang
e in
the
avai
labi
lity
of p
rimar
y sc
hool
s in
vill
age
1996
-200
020
00-2
006
OLS
OLS
Ord
ered
Pro
bit
OLS
OLS
Ord
ered
Pro
bit
with
Dis
trict
FE
with
Dis
trict
FE
with
Dis
trict
FE
with
Dis
trict
FE
Vill
ages
with
50%
or m
ore
fam
ilies
in p
re-w
elfa
re a
nd w
elfa
re 1
in (t
-1)
perio
d: y
es -1
, no-
0-0
.004
1-0
.003
3-0
.048
50.
0040
0.00
010.
0335
(0.0
016)
**(0
.001
5)**
(0.0
400)
(0.0
018)
**(0
.001
3)(0
.024
7)If
the
villa
ge h
ad a
prim
ary
scho
ol in
(t-1
) per
iod:
yes
-1, n
o -0
-0.1
028
-0.1
507
-3.1
620
-0.1
283
-0.1
913
-1.6
525
(0.0
108)
***
(0.0
120)
***
(0.0
780)
***
(0.0
120)
****
(0.0
131)
***
(0.0
440)
***
Mea
n di
stan
ce in
dis
trict
in (t
-1) p
erio
d X
intit
ial c
ondi
tionX
10^4
14.2
300
14.3
600
59.6
000
0.04
000.
1667
2.05
31(5
.750
0)**
(6.1
600)
**(2
1.00
00)
***
(0.2
066)
(0.2
533)
(0.9
999)
**
Pop
ulat
ion
of v
illag
e in
(t-1
) per
iod
X10
^40.
0027
0.00
101.
4000
0.01
880.
0095
1.20
65(0
.001
0)**
*(0
.000
4)**
*(0
.370
0)**
*(0
.003
5)**
*(0
.001
9)**
*(0
.166
7)**
*
If th
e vi
llage
had
any
term
inal
/sta
tion/
port
in (t
-1) p
erio
d: y
es-1
, no-
00.
0027
-0.0
004
-0.1
220
0.00
13-0
.002
3-0
.062
3(0
.002
2)(0
.002
0)(0
.095
0)(0
.001
9)(0
.002
0)(0
.050
7)V
illag
e to
pogr
aphy
(hill
are
a -0
, fla
tland
-1)
-0.0
016
-0.0
005
-0.0
467
-0.0
026
0.00
060.
0023
(0.0
021)
(0.0
014)
(0.0
460)
(0.0
026)
(0.0
015)
(0.0
267)
Any
dis
aste
r in
last
thre
e ye
ars:
y-1
, no-
2-0
.000
9-0
.001
7-0
.066
6-0
.007
20.
0014
0.03
19(0
.001
8)(0
.001
6)(0
.052
0)(0
.003
4)**
(0.0
009)
(0.0
180)
*
Con
stan
t2.
0970
2.11
561.
4472
1.50
44(0
.011
3)**
*(0
.008
3)**
*(0
.012
9)**
*(0
.010
2)**
*
Obs
erva
tions
5212
752
127
5512
955
129
5512
9R
-squ
ared
0.05
0.08
0.1
0.17
Rob
ust s
tand
ard
erro
rs in
par
enth
eses
* si
gnifi
cant
at 1
0%; *
* si
gnifi
cant
at 5
%; *
** s
igni
fican
t at 1
%D
istrc
it sp
ecifi
c co
nsta
nts
are
not r
epor
ted
in th
e ta
ble
21
Tabl
e 5b
: Loc
al In
com
e an
d C
hang
e in
Loc
al P
ublic
Goo
ds -
Juni
or H
ight
Sch
ools
Dep
. Var
iabl
e: C
hang
e in
the
avai
labi
lity
of ju
nior
hig
h sc
hool
s in
vill
age
1996
-200
020
00-2
006
OLS
OLS
Ord
ered
Pro
bit
OLS
OLS
Ord
ered
Pro
bit
with
Dis
trict
FE
with
Dis
trict
FE
with
Dis
trict
FE
with
Dis
trict
FE
Vill
ages
with
50%
or m
ore
fam
ilies
in p
re-w
elfa
re a
nd w
elfa
re 1
in (t
-1)
perio
d: y
es -1
, no-
0-0
.015
6-0
.014
0-0
.110
3-0
.000
6-0
.003
3-0
.015
5(0
.003
4)**
*(0
.002
9)**
*(0
.000
6)**
*(0
.002
3)(0
.002
0)*
(0.0
120)
If th
e vi
llage
had
a ju
nior
hig
h sc
hool
in (t
-1) p
erio
d: y
es-1
, no
-0-0
.156
6-0
.170
0-1
0.73
98-0
.149
7-0
.157
3-1
.650
2(0
.004
6)**
*(0
.005
1)**
*(0
.006
6)**
*(0
.004
3)**
*(0
.004
7)**
*(0
.022
9)**
*
Mea
n di
stan
ce in
dis
trict
in (t
-1) p
erio
d m
ultip
lied
by in
titia
l con
ditio
n X
10
^4-0
.010
00.
0900
0.55
00-0
.033
30.
0133
0.10
13(0
.040
0)(0
.050
0)(0
.030
0)**
*(0
.026
7)(0
.033
3)(0
.261
3)P
opul
atio
n of
vill
age
in (t
-1) p
erio
d X
10^
40.
0126
0.00
630.
0362
0.06
000.
0533
0.54
73(0
.004
4)**
*(0
.002
5)**
(0.0
160)
**(0
.006
7)**
*(0
.006
7)**
*(0
.057
3)**
*
If th
e vi
llage
had
any
term
inal
/sta
tion/
port
in (t
-1) p
erio
d: y
es-1
, no-
00.
0273
0.02
560.
1897
0.02
280.
0153
0.08
00(0
.005
3)**
*(0
.005
0)**
*(0
.001
9)**
*(0
.004
0)**
*(0
.003
8)**
*(0
.022
7)**
*
Vill
age
topo
grap
hy (h
ill a
rea
-0, f
latla
nd-1
)-0
.000
40.
0002
0.00
230.
0006
0.00
280.
0100
(0.0
035)
(0.0
032)
(0.0
011)
**(0
.003
0)(0
.002
5)(0
.015
3)A
ny d
isas
ter i
n la
st th
ree
year
s: y
-1, n
o-2
-0.0
077
0.00
030.
0067
0.00
190.
0046
0.02
79(0
.003
2)**
(0.0
030)
(0.0
008)
***
(0.0
026)
(0.0
019)
**(0
.011
3)**
Con
stan
t2.
0708
1.37
971.
4428
(0.0
052)
***
(0.0
041)
***
(0.0
036)
***
Obs
erva
tions
5212
755
129
5512
955
129
R-s
quar
ed0.
070.
10.
12R
obus
t sta
ndar
d er
rors
in p
aren
thes
es*
sign
ifica
nt a
t 10%
; **
sign
ifica
nt a
t 5%
; ***
sig
nific
ant a
t 1%
Dis
trcit
spec
ific
cons
tant
s ar
e no
t rep
orte
d in
the
tabl
e
22
Tabl
e 6a
: Loc
al In
com
e an
d C
hang
e in
Loc
al P
ublic
Goo
ds -
Hea
lthca
re in
fras
truc
ture
Dep
. Var
iabl
e: C
hang
e in
the
avai
labi
lity
of p
olyc
linic
in v
illag
e
1996
-200
020
00-2
006
OLS
OLS
Ord
ered
OLS
OLS
Ord
ered
Pro
bit
with
Dis
trict
FE
with
Dis
trict
FE
with
Dis
trict
FE
with
Dis
trict
FE
Vill
ages
with
50%
or m
ore
fam
ilies
in p
re-w
elfa
re a
nd w
elfa
re 1
in (t
-1) p
erio
d:
yes
-1, n
o-0
-0.0
298
-0.0
243
-0.3
411
-0.0
167
-0.0
151
-0.1
335
(0.0
039)
***
(0.0
032)
***
(0.0
011)
***
(0.0
039)
***
(0.0
023)
***
(0.0
005)
***
If th
e vi
llage
had
a p
olyc
linic
in (t
-1) p
erio
d: y
es-1
, no
-0-0
.509
2-0
.589
7-1
3.83
08-0
.454
9-0
.480
9-9
.316
2(0
.023
0)**
*(0
.022
5)**
*(0
.015
1)**
*(0
.013
3)**
*(0
.010
9)**
*(0
.009
3)**
*
If th
e vi
llage
had
a p
ushk
esm
as in
(t-1
) per
iod:
yes
-1, n
o-0
0.08
280.
073
0.61
320.
0341
0.03
960.
2372
(0.0
074)
***
(0.0
057)
***
(0.0
022)
***
(0.0
051)
***
(0.0
044)
***
(0.0
022)
***
Pop
ulat
ion
of v
illag
e in
(t-1
) per
iod
X10
^40.
0227
0.01
10.
0478
0.14
190.
1012
0.45
62(0
.007
7)**
*(0
.004
1)**
*(0
.000
43)*
**(0
.010
7)**
*(0
.011
3)**
*(0
.001
7)**
*
If th
e vi
llage
had
any
term
inal
/sta
tion/
port
in (t
-1) p
erio
d: y
es-1
, no-
00.
0536
0.05
40.
4112
0.02
520.
0324
0.18
66(0
.007
2)**
*(0
.006
8)**
*(0
.002
7)**
*(0
.006
3)**
*(0
.004
9)**
*(0
.002
3)**
*
Vill
age
topo
grap
hy (h
ill a
rea
-0, f
latla
nd-1
)0.
0119
0.01
010.
1801
0.00
830.
0137
0.12
09(0
.002
9)**
*(0
.002
6)**
*(0
.001
6)**
*(0
.004
3)*
(0.0
030)
***
(0.0
011)
***
Any
dis
aste
r in
last
thre
e ye
ars:
y-1
, no-
2-0
.004
0.00
290.
04-0
.001
60.
0000
0.00
49(0
.003
4)(0
.002
7)(0
.001
1)**
*(0
.002
8)(0
.002
2)(0
.000
9)**
*
Con
stan
t2.
0264
2.00
821.
3465
1.31
39(0
.004
8)**
*(0
.003
4)**
*(0
.005
1)**
*(0
.003
7)**
*
Obs
erva
tions
5212
752
127
5212
755
129
5512
955
129
R-s
quar
ed0.
200.
250.
220.
28
Rob
ust s
tand
ard
erro
rs in
par
enth
eses
* si
gnifi
cant
at 1
0%; *
* si
gnifi
cant
at 5
%; *
** s
igni
fican
t at 1
%D
istrc
it sp
ecifi
c co
nsta
nts
are
not r
epor
ted
in th
e ta
ble
23
Tabl
e 6b
: Loc
al In
com
e an
d C
hang
e in
Loc
al P
ublic
Goo
ds -
Hea
lthca
re in
fras
truc
ture
Dep
. Var
iabl
e: C
hang
e in
the
avai
labi
lity
of p
uske
smas
in v
illag
e19
96-2
000
2000
-200
6O
LSO
LSO
rder
ed P
robi
tO
LSO
LSO
rder
ed P
robi
tw
ith D
istri
ct F
Ew
ith D
istri
ct F
Ew
ith D
istri
ct F
Ew
ith D
istri
ct F
EV
illag
es w
ith 5
0% o
r mor
e fa
mili
es in
pre
-wel
fare
and
wel
fare
1 in
(t-1
) per
iod:
ye
s -1
, no-
0-0
.009
9-0
.011
-0.1
427
-0.0
033
-0.0
025
-0.0
404
(0.0
021)
***
(0.0
022)
***
(0.0
008)
***
(0.0
013)
**(0
.001
2)**
(0.0
162)
**
If th
e vi
llage
had
a p
ushk
esm
as in
(t-1
) per
iod:
yes
-1, n
o-0
-0.1
948
-0.1
942
-11.
0575
-0.1
373
-0.1
375
-8.6
149
(0.0
119)
***
(0.0
102)
***
(0.0
123)
***
(0.0
067)
***
(0.0
063)
***
(0.0
322)
***
If th
e vi
llage
had
a p
olyc
linic
in (t
-1) p
erio
d: y
es-1
, no
-00.
0462
0.03
60.
2865
0.01
140.
0101
0.09
46(0
.008
2)**
*(0
.008
4)**
*(0
.003
1)**
*(0
.009
5)**
*(0
.008
3)**
*(0
.004
2)**
*
Pop
ulat
ion
of v
illag
e in
(t-1
) per
iod
0.00
816
0.00
575
0.03
420.
0314
0.02
780.
2165
(0.0
029)
***
(0.0
024)
***
(0.0
004)
***
(0.0
051)
***
(0.0
056)
***
(0.0
020)
***
If th
e vi
llage
had
any
term
inal
/sta
tion/
port
in (t
-1) p
erio
d: y
es-1
, no-
00.
0429
0.04
40.
3638
0.02
280.
0202
0.16
18(0
.005
3)**
*(0
.005
8)**
*(0
.002
4)**
*(0
.001
3)**
(0.0
012)
**(0
.016
2)**
*
Vill
age
topo
grap
hy (h
ill a
rea
-0, f
latla
nd-1
)0.
0033
0.00
380.
0525
0.00
110.
0041
0.05
81(0
.002
0)(0
.002
0)*
(0.0
014)
***
(0.0
017)
(0.0
018)
**(0
.024
2)**
Any
dis
aste
r in
last
thre
e ye
ars:
y-1
, no-
20.
0015
0.00
10.
0191
0.00
26-0
.000
6-0
.013
7(0
.001
8)(0
.002
1)(0
.000
9)**
*(0
.001
4)*
(0.0
013)
(0.0
170)
Con
stan
t2.
0232
2.02
361.
3397
1.33
60(0
.002
6)**
*(0
.003
1)**
*(0
.002
0)**
*(0
.002
3)**
*
Obs
erva
tions
5212
752
127
5212
755
129
5508
255
082
R-s
quar
ed0.
090.
100.
09R
obus
t sta
ndar
d er
rors
in p
aren
thes
es*
sign
ifica
nt a
t 10%
; **
sign
ifica
nt a
t 5%
; ***
sig
nific
ant a
t 1%
Dis
trcit
spec
ific
cons
tant
s ar
e no
t rep
orte
d in
the
tabl
e
24
Tabl
e 7:
Loc
al P
refe
renc
e an
d C
hang
e in
Loc
al P
ublic
Goo
ds -
Villa
ge R
oad
Dep
ende
nt V
aria
ble-
Cha
nge
in v
illag
e ro
ad: d
eter
iora
tion-
1, n
o ch
ange
-2, i
mpr
ovem
ent -
3
1996
-200
020
00-2
006
OLS
OLS
OLS
Ord
ered
Pro
bit
Ord
ered
Pro
bit
OLS
OLS
OLS
Ord
ered
Pro
bit
Ord
ered
Pro
bit
with
Dis
trict
FE
with
Dis
trict
FE
with
Dis
trict
FE
with
Dis
trict
FE
with
Dis
trict
FE
with
Dis
trict
FE
with
Dis
trict
FE
with
Dis
trict
FE
Vill
age
head
's g
ende
r in
t-1 p
erio
d (m
ale-
1, fe
mal
e-0)
-0.0
004
-0.0
045
-0.0
034
-0.0
093
-0.0
06-0
.021
0-0
.007
3-0
.007
2-0
.023
0-0
.022
8(0
.014
7)(0
.013
2)(0
.013
2)(0
.041
2)(0
.041
4)(0
.010
2)**
(0.0
091)
(0.0
091)
(0.0
260)
(0.0
260)
% o
f wom
en v
oter
in v
illag
e-0
.000
1-0
.000
40.
0025
0.00
69(0
.000
4)(0
.001
4)(0
.000
7)**
*(0
.002
1)**
*V
illag
e he
ad's
age
in (t
-1) p
erio
d 0.
0022
0.00
160.
0015
0.00
480.
0045
0.00
180.
0013
0.00
130.
0036
0.00
36(0
.000
4)**
*(0
.000
3)**
*(0
.000
3)**
*(0
.001
0)**
*(0
.001
0)**
*(0
.000
3)**
*(0
.000
2)**
*(0
.000
2)**
*(0
.000
5)**
*(0
.000
5)**
*V
illag
e he
ad's
edu
catio
n in
(t-1
) per
iod
Not
com
plet
ed0.
0074
-0.0
309
-0.0
331
-0.1
067
-0.1
104
0.02
04-0
.020
6-0
.021
1-0
.051
7-0
.053
2(0
.056
2)(0
.048
3)(0
.051
8)(0
.147
5)(0
.159
2)(0
.036
0)(0
.032
1)(0
.031
9)(0
.089
4)(0
.088
6)P
rimar
y sc
hool
0.08
40.
0196
0.02
80.
0417
0.06
640.
0139
-0.0
371
-0.0
375
-0.0
971
-0.0
983
(0.0
562)
(0.0
45)
(0.0
459)
(0.1
387)
(0.1
425)
(0.0
377)
(0.0
340)
(0.0
339)
(0.0
945)
(0.0
939)
Juni
or h
igh
scho
ol0.
1012
0.04
490.
0519
0.12
20.
1425
0.04
17-0
.019
7-0
.020
1-0
.049
5-0
.050
7(0
.056
9)*
(0.0
459)
(0.0
47)
(0.1
409)
(0.1
457)
(0.0
384)
(0.0
336)
(0.0
334)
(0.0
935)
(0.0
928)
Hig
h sc
hool
0.13
070.
0721
0.07
920.
2019
0.22
30.
0819
0.00
890.
0085
0.03
040.
0296
(0.0
573)
**-0
.046
(0.0
470)
*(0
.141
6)(0
.145
8)(0
.038
3)**
(0.0
335)
(0.0
333)
(0.0
933)
(0.0
927)
Aca
dem
y0.
1791
0.11
010.
1176
0.32
160.
3441
0.10
310.
0233
0.02
270.
0693
0.06
77(0
.057
5)**
*(0
.047
2)**
(0.0
475)
**(0
.145
4)**
(0.1
475)
**(0
.039
1)**
*(0
.034
3)(0
.034
2)(0
.095
9)(0
.095
4)U
nive
rsity
0.14
460.
0775
0.08
520.
2131
0.23
50.
1033
0.02
050.
0201
0.06
230.
0612
(0.0
569)
**(0
.046
0)*
(0.0
472)
*(0
.141
5)(0
.146
8)(0
.038
8)**
*(0
.033
7)(0
.033
5)(0
.094
1)(0
.093
5)V
illag
e he
ad's
dur
atio
n in
yea
rs-0
.001
6-0
.000
3-0
.000
2-0
.001
-0.0
007
-0.0
027
-0.0
002
-0.0
002
-0.0
007
-0.0
007
(0.0
006)
***
(0.0
005)
(0.0
005)
(0.0
015)
(0.0
016)
(0.0
005)
***
(0.0
004)
(0.0
004)
(0.0
010)
(0.0
010)
Vill
ages
with
50%
or m
ore
fam
ilies
in p
re-w
elfa
re a
nd w
elfa
re 1
in (t
-1)
perio
d: y
es -1
, no-
0-0
.075
2-0
.059
9-0
.058
7-0
.181
6-0
.179
5-0
.030
0-0
.031
8-0
.031
7-0
.087
3-0
.087
0(0
.010
4)**
*(0
.007
1)**
*(0
.007
1)**
*(0
.021
2)**
*(0
.021
7)**
*(0
.008
3)**
*(0
.006
1)**
*(0
.006
1)**
*(0
.016
7)**
*(0
.016
7)**
*R
oad
hard
ende
d in
(t-1
) per
iod
equa
ls 1
, els
e 0
-0.2
925
-0.3
062
-0.2
998
-0.8
979
-0.8
829
-0.1
529
-0.2
261
-0.2
267
-0.6
381
-0.6
402
(0.0
218)
***
(0.0
213)
***
(0.0
219)
***
(0.0
553)
***
(0.0
569)
***
(0.0
175)
***
(0.0
145)
***
(0.0
145)
***
(0.0
358)
***
(0.0
357)
***
Roa
d_pa
ved
in (t
-1) p
erio
d eq
uals
1, e
lse
0-0
.679
5-0
.751
1-0
.748
2-2
.354
8-2
.361
1-0
.494
7-0
.599
2-0
.599
9-1
.678
6-1
.681
6(0
.021
4)**
*(0
.021
5)**
*(0
.022
0)**
*(0
.067
6)**
*(0
.069
5)**
*(0
.015
8)**
*(0
.014
3)**
*(0
.014
3)**
*(0
.047
0)**
*(0
.046
9)**
*P
opul
atio
n of
vill
age
in (t
-1) p
erio
d X
10^
40.
0108
0.00
665
0.00
650.
021
0.02
080.
0629
0.02
730.
0273
0.08
400.
0840
(0.0
042)
**(0
.003
3)**
(0.0
032)
**(0
.011
)*(0
.011
)*(0
.010
0)**
*(0
.006
4)**
*(0
.006
4)**
*(0
.018
7)**
*(0
.018
7)**
*If
the
villa
ge h
ad a
ny te
rmin
al/s
tatio
n/po
rt in
(t-1
) per
iod:
yes
-1, n
o-0
0.05
070.
0609
0.06
060.
1947
0.19
50.
0289
0.04
600.
0459
0.13
070.
1307
(0.0
097)
***
(0.0
095)
***
(0.0
094)
***
(0.0
296)
***
(0.0
295)
***
(0.0
076)
***
(0.0
062)
***
(0.0
062)
***
(0.0
172)
***
(0.0
173)
***
Vill
age
topo
grap
hy (h
ill a
rea
-0, f
latla
nd-1
)0.
0472
0.05
20.
051
0.15
710.
1543
0.04
650.
0471
0.04
690.
1295
0.12
91(0
.010
7)**
*(0
.010
1)**
*(0
.010
2)**
*(0
.029
7)**
*(0
.030
2)**
*(0
.009
1)**
*(0
.007
1)**
*(0
.007
1)**
*(0
.018
7)**
*(0
.018
7)**
*A
ny d
isas
ter i
n la
st th
ree
year
s: y
-1, n
o-2
-0.0
123
-0.0
318
-0.0
346
-0.1
007
-0.1
088
-0.0
221
-0.0
076
-0.0
077
-0.0
221
-0.0
223
(0.0
107)
(0.0
102)
***
(0.0
103)
***
(0.0
302)
***
(0.0
305)
***
(0.0
073)
***
(0.0
060)
(0.0
060)
(0.0
166)
(0.0
165)
Con
stan
t2.
3061
2.61
572.
6129
1.56
241.
8763
1.75
28(0
.067
5)**
*(0
.054
4)**
*(0
.058
2)**
*(0
.049
7)**
*(0
.035
9)**
*(0
.051
8)**
*O
bser
vatio
ns49
168
4916
846
954
4916
846
954
5199
551
995
5199
551
995
5199
5R
-squ
ared
0.23
0.28
0.28
0.26
0.34
0.34
Rob
ust s
tand
ard
erro
rs in
par
enth
eses
*
sign
ifica
nt a
t 10%
; ** s
igni
fican
t at 5
%; *
** s
igni
fican
t at 1
%D
istrc
it sp
ecifi
c co
nsta
nts
are
not r
epor
ted
in th
e ta
ble
25
Tabl
e 8:
Loc
al P
refe
renc
e an
d C
hang
e in
Loc
al P
ublic
Goo
ds -
Villa
ge R
oadl
ight
Dep
ende
nt V
aria
ble-
Cha
nge
in v
illag
e ro
ad li
ght:
dete
riora
tion-
1, n
o ch
ange
-2, i
mpr
ovem
ent -
319
96-2
000
2000
-200
6O
LSO
LSO
LSO
rder
ed P
robi
tO
rder
ed P
robi
tO
LSO
LSO
LSO
rder
ed P
robi
tO
rder
ed P
robi
tw
ith D
istri
ct F
Ew
ith D
istri
ct F
Ew
ith D
istri
ct F
Ew
ith D
istri
ct F
Ew
ith D
istri
ct F
Ew
ith D
istri
ct F
Ew
ith D
istri
ct F
Ew
ith D
istri
ct F
EV
illag
e he
ad's
gen
der i
n t-1
per
iod
(mal
e-1,
fem
ale-
0)-0
.036
1-0
.018
3-0
.018
-0.0
733
-0.0
715
-0.0
337
-0.0
128
-0.0
128
-0.0
547
-0.0
551
(0.0
140)
**(0
.011
9)(0
.011
9)(0
.000
9)**
*(0
.000
8)**
*(0
.008
3)**
*(0
.006
9)*
(0.0
069)
*(0
.000
7)**
*(0
.000
7)**
*
% o
f wom
en v
oter
in v
illag
e X
10^
20.
0098
90.
0274
0.08
860.
3130
(0.0
42)
(0.0
016)
***
(0.0
533)
*(0
.001
3)**
*
Vill
age
head
's a
ge in
(t-1
) per
iod
X 1
0^2
0.07
130.
0713
0.12
10.
535
0.52
40.
1372
0.14
390.
1432
0.63
740.
6360
(0.0
42)*
(0.0
42)*
**(0
.027
)***
(0.0
02)*
**(0
.001
8)**
*(0
.033
3)**
*(0
.016
7)**
*(0
.016
7)**
*(0
.001
7)**
*(0
.001
5)**
*
Vill
age
head
's e
duca
tion
in (t
-1) p
erio
d N
ot c
ompl
eted
0.03
020.
024
0.04
270.
1929
0.31
950.
0375
-0.0
071
-0.0
075
-0.0
202
-0.0
208
(0.0
302)
(0.0
263)
(0.0
259)
(0.0
014)
***
(0.0
014)
***
(0.0
241)
(0.0
135)
(0.0
135)
(0.0
010)
***
(0.0
009)
***
Prim
ary
scho
ol0.
0562
0.02
670.
0454
0.20
710.
3301
0.06
12-0
.002
1-0
.002
40.
0027
0.00
21(0
.030
6)*
(0.0
257)
(0.0
254)
*(0
.001
7)**
*(0
.001
7)**
*(0
.024
8)**
(0.0
127)
(0.0
127)
(0.0
012)
**(0
.001
1)*
Juni
or h
igh
scho
ol0.
0792
0.04
870.
0676
0.29
810.
4208
0.10
280.
0161
0.01
580.
0672
0.06
68(0
.030
4)**
*(0
.025
6)*
(0.0
250)
***
(0.0
015)
***
(0.0
015)
***
(0.0
253)
***
(0.0
128)
(0.0
128)
(0.0
012)
***
(0.0
012)
***
Hig
h sc
hool
0.14
010.
0812
0.10
010.
4292
0.55
10.
1445
0.03
870.
0385
0.16
140.
1609
(0.0
311)
***
(0.0
259)
***
(0.0
252)
***
(0.0
018)
***
(0.0
018)
***
(0.0
258)
***
(0.0
131)
***
(0.0
130)
***
(0.0
013)
***
(0.0
013)
***
Aca
dem
y0.
2059
0.13
260.
1513
0.65
330.
7724
0.16
870.
0598
0.05
940.
2921
0.29
13(0
.032
2)**
*(0
.027
8)**
*(0
.026
8)**
*(0
.002
2)**
*(0
.002
1)**
*(0
.026
6)**
*(0
.014
3)**
*(0
.014
3)**
*(0
.001
3)**
*(0
.001
3)**
*
Uni
vers
ity0.
1939
0.10
440.
1216
0.54
620.
6576
0.16
360.
0465
0.04
620.
2151
0.21
44(0
.031
9)**
*(0
.026
9)**
*(0
.026
6)**
*(0
.002
0)**
*(0
.001
9)**
*(0
.026
8)**
*(0
.013
8)**
*(0
.013
7)**
*(0
.001
4)**
*(0
.001
4)**
*
Vill
age
head
's d
urat
ion
in y
ears
X 1
0^2
0.19
700.
1970
-0.0
021
-0.0
033
-0.0
021
-0.2
910
-0.0
746
-0.0
739
-0.2
837
-0.2
824
(0.0
72)*
**(0
.072
)**
(0.0
430)
(0.0
073)
(0.0
069)
(0.0
579)
***
(0.0
306)
**(0
.030
6)**
(0.0
057)
***
(0.0
053)
***
Vill
ages
with
50%
or m
ore
fam
ilies
in p
re-w
elfa
re a
nd w
elfa
re 1
in (t
-1)
perio
d: y
es -1
, no-
0-0
.076
5-0
.070
1-0
.072
2-0
.287
9-0
.295
4-0
.006
5-0
.027
7-0
.027
7-0
.125
7-0
.125
6(0
.011
3)**
*(0
.007
2)**
*(0
.007
5)**
*(0
.000
6)**
*(0
.000
5)**
*(0
.008
5)(0
.004
0)**
*(0
.004
0)**
*(0
.000
5)**
*(0
.000
4)**
*
Vill
age
road
ligh
teni
ng in
(t-1
) per
iod:
yes
-1, n
o-0
-0.4
857
-0.6
441
-0.6
525
-12.
4429
-12.
4711
-0.3
872
-0.5
447
-0.5
448
-11.
7584
-11.
7485
(0.0
140)
***
(0.0
117)
***
(0.0
106)
***
(0.0
054)
***
(0.0
051)
***
(0.0
111)
***
(0.0
077)
***
(0.0
077)
***
(0.0
051)
***
(0.0
051)
***
Pop
ulat
ion
of v
illag
e in
(t-1
) per
iod
X 1
0^4
0.02
220.
0222
0.01
150.
0405
0.03
950.
0960
0.06
230.
0623
0.63
930.
6399
(0.0
077)
***
(0.0
077)
***
(0.0
048)
***
(0.0
0032
)***
(0.0
0029
)***
(0.0
173)
***
(0.0
093)
(0.0
093)
(0.0
019)
(0.0
018)
***
If th
e vi
llage
had
any
term
inal
/sta
tion/
port
in (t
-1) p
erio
d: y
es-1
, no-
00.
0477
0.09
480.
096
0.41
720.
42-0
.017
70.
0463
0.04
630.
2092
0.20
96(0
.012
0)**
*(0
.009
4)**
*(0
.009
7)**
*(0
.001
5)**
*(0
.001
5)**
*(0
.007
8)**
(0.0
058)
***
(0.0
059)
***
(0.0
013)
***
(0.0
013)
***
Vill
age
topo
grap
hy (h
ill a
rea
-0, f
latla
nd-1
)0.
0958
0.09
510.
0968
0.39
340.
3973
0.06
080.
0471
0.04
700.
2140
0.21
36(0
.011
8)**
*(0
.009
1)**
*(0
.009
3)**
*(0
.000
9)**
*(0
.000
9)**
*(0
.012
2)**
*(0
.005
9)**
*(0
.005
9)**
*(0
.000
7)**
*(0
.000
7)**
*
Any
dis
aste
r in
last
thre
e ye
ars:
y-1
, no-
2-0
.039
1-0
.001
20.
0006
-0.0
011
0.00
53-0
.004
30.
0110
0.01
100.
0441
0.04
41(0
.013
0)**
*(0
.008
7)(0
.008
1)(0
.000
6)**
(0.0
006)
***
(0.0
080)
(0.0
041)
***
(0.0
041)
***
(0.0
005)
***
(0.0
004)
***
Con
stan
t2.
0822
2.27
112.
2258
1.42
391.
3509
1.30
70(0
.038
2)**
*(0
.021
0)**
*(0
.034
7)**
*(0
.033
7)**
*(0
.018
3)**
*(0
.034
2)**
*
Obs
erva
tions
5212
752
127
4985
052
127
4985
055
129
5512
955
129
5512
955
129
R-s
quar
ed0.
240.
350.
350.
30.
510.
51R
obus
t sta
ndar
d er
rors
in p
aren
thes
es
* si
gnifi
cant
at 1
0%; *
* si
gnifi
cant
at 5
%; *
** s
igni
fican
t at 1
%
Dis
trcit
spec
ific
cons
tant
s ar
e no
t rep
orte
d in
the
tabl
e
26
Tabl
e 9a
: Loc
al P
refe
renc
e an
d C
hang
e in
Loc
al P
ublic
Goo
ds -
Prim
ary
Scho
ols
Dep
. Var
iabl
e: C
hang
e in
the
avai
labi
lity
of p
rimar
y sc
hool
s in
vill
age
1996
-200
020
00-2
006
OLS
OLS
OLS
Ord
ered
Pro
bit
Ord
ered
Pro
bit
OLS
OLS
OLS
Ord
ered
Pro
bit
Ord
ered
Pro
bit
with
Dis
trict
FE
with
Dis
trict
FE
with
Dis
trict
FE
with
Dis
trict
FE
with
Dis
trict
FE
with
Dis
trict
FE
with
Dis
trict
FE
with
Dis
trict
FE
Vill
age
head
's g
ende
r in
t-1 p
erio
d (m
ale-
1, fe
mal
e-0)
-0.0
006
0.00
390.
0043
0.17
890.
1795
-0.0
051
-0.0
013
-0.0
013
-0.0
163
-0.0
169
(0.0
036)
(0.0
035)
(0.0
035)
(0.1
269)
(0.1
286)
(0.0
026)
**(0
.002
3)(0
.002
3)(0
.059
3)(0
.059
3)%
of w
omen
vot
er in
vill
age
X10
^20.
0008
0.08
240.
0458
0.43
86(0
.005
7)(0
.286
6)(0
.052
0)(0
.566
6)V
illag
e he
ad's
age
in (t
-1) p
erio
d X
10^2
0.02
220.
0157
0.00
880.
2422
-0.0
012
-0.0
034
0.00
600.
0059
0.04
020.
0398
(0.0
092)
**(0
.007
4)**
(0.0
001)
(0.0
023)
(0.2
148)
(0.0
120)
(0.0
073)
(0.0
073)
(0.1
467)
(0.1
467)
Vill
age
head
's e
duca
tion
in (t
-1) p
erio
d N
ot c
ompl
eted
0.01
420.
0110
0.01
690.
1944
0.29
630.
0063
0.00
580.
0056
0.03
490.
0322
(0.0
240)
(0.0
206)
(0.0
202)
(0.2
944)
(0.2
934)
(0.0
200)
(0.0
200)
(0.0
200)
(0.1
533)
(0.1
533)
Prim
ary
scho
ol0.
0177
0.01
010.
0154
0.22
660.
3313
0.01
170.
0077
0.00
750.
0658
0.06
43(0
.021
0)(0
.017
0)(0
.017
5)(0
.237
3)(0
.244
3)(0
.019
3)(0
.019
3)(0
.019
3)(0
.146
7)(0
.146
7)Ju
nior
hig
h sc
hool
0.02
140.
0120
0.01
670.
2494
0.34
210.
0121
0.00
770.
0075
0.04
740.
0459
(0.0
210)
(0.0
171)
(0.0
175)
(0.2
410)
(0.2
471)
(0.0
193)
(0.0
193)
(0.0
193)
(0.1
400)
(0.1
400)
Hig
h sc
hool
0.02
460.
0127
0.01
680.
2515
0.32
630.
0177
0.01
120.
0111
0.11
200.
1107
(0.0
212)
(0.0
173)
(0.0
177)
(0.2
474)
(0.2
522)
(0.0
193)
(0.0
193)
(0.0
193)
(0.1
400)
(0.1
400)
Aca
dem
y0.
0253
0.01
260.
0156
0.16
020.
1600
0.01
400.
0081
0.00
79-0
.004
2-0
.005
9(0
.020
9)(0
.017
0)(0
.017
7)(0
.255
8)(0
.272
3)(0
.019
3)(0
.019
3)(0
.019
3)(0
.153
3)(0
.153
3)U
nive
rsity
0.02
680.
0140
0.01
590.
2686
0.22
070.
0159
0.01
050.
0103
0.06
020.
0573
(0.0
212)
(0.0
175)
(0.0
181)
(0.2
712)
(0.2
807)
(0.0
193)
(0.0
193)
(0.0
193)
(0.1
467)
(0.1
467)
Vill
age
head
's d
urat
ion
in y
ears
x 1
0^2
0.01
86-0
.001
70.
0006
-0.0
154
0.10
51-0
.036
50.
0032
0.00
340.
1647
0.16
53(0
.015
3)(0
.011
7)(0
.010
6)(0
.342
9)(0
.342
1)(0
.014
7)**
(0.0
113)
(0.0
113)
(0.2
066)
(0.2
066)
Vill
ages
with
50%
or m
ore
fam
ilies
in p
re-w
elfa
re a
nd w
elfa
re 1
in (t
-1)
perio
d: y
es -1
, no-
0-0
.003
1-0
.003
1-0
.003
1-0
.045
9-0
.044
60.
0042
0.00
030.
0004
0.03
450.
0354
(0.0
015)
**(0
.001
4)**
(0.0
015)
**(0
.039
8)(0
.043
7)(0
.001
6)**
(0.0
013)
(0.0
013)
(0.0
253)
(0.0
253)
If th
e vi
llage
had
a p
rimar
y sc
hool
in (t
-1) p
erio
d: y
es-1
, no
-0-0
.104
5-0
.151
1-0
.162
1-3
.158
8-3
.138
0-0
.130
0-0
.191
3-0
.191
3-1
.651
8-1
.653
2(0
.010
6)**
*(0
.012
0)**
*(0
.011
9)**
*(0
.078
4)**
*(0
.088
0)**
*(0
.012
0)**
*(0
.013
3)**
*(0
.013
3)**
*(0
.044
0)**
*(0
.042
7)**
*
Mea
n di
stan
ce in
dis
trict
in (t
-1) p
erio
d X
intit
ial c
ondi
tion
X10
^20.
1458
0.14
350.
1530
0.59
700.
6288
0.00
080.
0017
0.00
182.
1131
2.15
31(0
.057
2)(0
.061
1)(0
.062
9)**
(0.2
069)
***
(0.2
208)
***
(0.0
021)
(0.0
026)
(0.0
026)
(0.0
107)
**-0
.010
7**
Pop
ulat
ion
of v
illag
e in
(t-1
) per
iod
X10^
40.
0024
0.00
100.
0010
1.39
001.
7600
0.01
730.
0091
0.00
911.
2065
1.20
65(0
.000
9)**
*(0
.000
4)**
*(0
.000
4)**
*(0
.370
0)**
*(0
.310
0)**
*(0
.003
5)**
*(0
.001
8)**
*(0
.001
8)**
*(0
.166
7)**
*(0
.166
7)**
*
If th
e vi
llage
had
any
term
inal
/sta
tion/
port
in (t
-1) p
erio
d: y
es-1
, no-
00.
0023
-0.0
006
-0.0
010
-0.1
200
-0.1
659
0.00
14-0
.002
5-0
.002
5-0
.066
7-0
.066
2(0
.002
1)(0
.002
0)(0
.002
0)(0
.094
4)(0
.094
8)*
(0.0
019)
(0.0
020)
(0.0
020)
(0.0
507)
(0.0
500)
Vill
age
topo
grap
hy (h
ill a
rea
-0, f
latla
nd-1
)-0
.002
1-0
.000
7-0
.000
8-0
.051
2-0
.060
6-0
.003
10.
0004
0.00
040.
0008
-0.0
002
(0.0
020)
(0.0
014)
(0.0
015)
(0.0
458)
(0.0
479)
(0.0
026)
(0.0
015)
(0.0
015)
(0.0
273)
(0.0
267)
Any
dis
aste
r in
last
thre
e ye
ars:
y-1
, no-
2-0
.000
4-0
.001
7-0
.001
7-0
.066
7-0
.070
2-0
.006
90.
0014
0.00
140.
0328
0.03
30(0
.001
9)(0
.001
6)(0
.001
7)(0
.052
0)(0
.060
9)(0
.003
3)**
(0.0
009)
(0.0
009)
(0.0
180)
*(0
.018
0)*
Con
stan
t2.
0659
2.09
412.
0988
1.44
251.
4952
1.47
19(0
.025
8)**
*(0
.021
0)**
*(0
.020
9)**
*(0
.022
7)**
*(0
.021
3)**
*(0
.031
3)**
*
Obs
erva
tions
5212
752
127
4985
055
129
5512
955
129
R-s
quar
ed0.
0500
0.08
000.
0900
0.1
0.17
0.17
Rob
ust s
tand
ard
erro
rs in
par
enth
eses
*
sign
ifica
nt a
t 10%
; **
sign
ifica
nt a
t 5%
; ***
sig
nific
ant a
t 1%
Dis
trcit
spec
ific
cons
tant
s ar
e no
t rep
orte
d in
the
tabl
e
27
Tabl
e 9b
: Loc
al P
refe
renc
e an
d C
hang
e in
Loc
al P
ublic
Goo
ds -
Juni
or H
igh
Scho
ols
Dep
. Var
iabl
e: C
hang
e in
the
avai
labi
lity
of ju
nior
hig
h sc
hool
s in
vill
age
1996
-200
020
00-2
006
OLS
OLS
with
O
LS w
ith
Ord
ered
Pro
bit
Ord
ered
Pro
bit
OLS
OLS
with
O
LS w
ith
Ord
ered
Pro
bit
Ord
ered
Pro
bit
Dis
trict
FE
Dis
trict
FE
Dis
trict
FE
Dis
trict
FE
Dis
trict
FE
Dis
trict
FE
Dis
trict
FE
Dis
trict
FE
Vill
age
head
's g
ende
r in
t-1 p
erio
d (m
ale-
1, fe
mal
e-0)
-0.0
093
-0.0
076
-0.0
075
-0.0
468
-0.0
466
0.00
170.
0000
50.
0000
5-0
.001
6-0
.001
6(0
.008
1)(0
.008
1)(0
.008
2)(0
.058
8)(0
.059
6)(0
.005
3)(0
.005
2)(0
.005
2)(0
.033
3)(0
.033
3)%
of w
omen
vot
er in
vill
age
X 1
0^2
0.01
110.
0786
0.02
390.
1453
(0.0
231)
(0.1
645)
(0.0
313)
(0.2
000)
Vill
age
head
's a
ge in
(t-1
) per
iod
X 10
^20.
0740
0.07
430.
0718
0.60
600.
5726
0.04
630.
0451
0.04
500.
2673
0.26
66(0
.016
6)**
*(0
.016
1)**
*(0
.016
4)**
*(0
.126
2)**
*(0
.126
2)**
*(0
.012
7)**
*(0
.010
7)**
*(0
.010
7)**
*(0
.066
7)**
*(0
.066
7)**
*
Vill
age
head
's e
duca
tion
in (t
-1) p
erio
d N
ot c
ompl
eted
-0.0
126
-0.0
103
-0.0
030
-0.1
054
-0.0
302
0.01
870.
0109
0.01
080.
0767
0.07
53(0
.018
9)(0
.020
0)(0
.021
8)(0
.182
1)(0
.195
9)(0
.012
0)(0
.012
0)(0
.012
0)(0
.100
0)(0
.100
0)P
rimar
y sc
hool
0.00
450.
0012
0.00
810.
0180
0.07
900.
0187
0.00
810.
0081
0.06
220.
0613
(0.0
186)
(0.0
194)
(0.0
209)
(0.1
749)
(0.1
876)
(0.0
120)
(0.0
120)
(0.0
120)
(0.0
933)
(0.1
000)
Juni
or h
igh
scho
ol0.
0201
0.01
410.
0209
0.13
240.
1900
0.02
270.
0123
0.01
230.
0880
0.08
73(0
.019
0)(0
.020
0)(0
.021
6)(0
.179
4)(0
.192
4)(0
.012
0)*
(0.0
127)
(0.0
127)
(0.1
000)
(0.1
000)
Hig
h sc
hool
0.02
250.
0148
0.02
140.
1371
0.19
130.
0281
0.01
970.
0196
0.13
070.
1293
(0.0
187)
(0.0
196)
(0.0
211)
(0.1
762)
(0.1
889)
(0.0
120)
**(0
.012
0)(0
.012
0)(0
.093
3)(0
.093
3)A
cade
my
0.04
170.
0328
0.03
870.
2760
0.32
440.
0272
0.02
300.
0229
0.14
400.
1433
(0.0
191)
**(0
.020
0)(0
.021
5)*
(0.1
782)
(0.1
904)
*(0
.012
7)**
(0.0
133)
*(0
.013
3)*
(0.1
000)
(0.1
000)
Uni
vers
ity0.
0338
0.02
040.
0268
0.18
630.
2395
0.02
650.
0225
0.02
250.
1440
0.14
27(0
.019
6)*
(0.0
202)
(0.0
212)
(0.1
801)
(0.1
886)
(0.0
127)
**(0
.012
7)*
(0.0
127)
*(0
.100
0)(0
.100
0)V
illag
e he
ad's
dur
atio
n in
yea
rs X
10^
2-0
.008
5-0
.018
4-0
.019
9-0
.106
1-1
.255
3-0
.018
3-0
.017
9-0
.017
7-0
.115
3-0
.115
3(0
.024
7)(0
.025
9)(0
.026
6)(0
.202
0)(2
.165
4)(0
.021
3)(0
.019
3)(0
.019
3)(0
.120
0)(0
.120
0)V
illag
es w
ith 5
0% o
r mor
e fa
mili
es in
pre
-wel
fare
and
wel
fare
1 in
(t-1
) pe
riod:
yes
-1, n
o-0
-0.0
127
-0.0
124
-0.0
117
-0.0
986
-0.0
919
0.00
03-0
.002
5-0
.002
5-0
.011
4-0
.011
4(0
.003
4)**
*(0
.002
9)**
*(0
.002
9)**
*(0
.021
8)**
*(0
.021
7)**
*(0
.002
3)(0
.001
9)(0
.001
9)(0
.012
0)(0
.012
0)If
the
villa
ge h
ad a
juni
or h
igh
scho
ol in
(t-1
) per
iod:
yes
-1, n
o-0
-0.1
598
-0.1
717
-0.1
711
-2.8
449
-2.8
446
-0.1
507
-0.1
580
-0.1
580
-1.6
472
-1.6
472
(0.0
046)
***
(0.0
051)
***
(0.0
052)
***
(0.0
313)
***
(0.0
313)
***
(0.0
043)
***
(0.0
047)
***
(0.0
047)
***
(0.0
227)
***
(0.0
227)
***
Mea
n di
stan
ce o
f jun
ior h
igh
scho
ol in
(t-1
) per
iod
X in
itial
vill
age
cond
ition
X
10^2
0.00
050.
0011
0.00
120.
0076
0.00
77-0
.000
20.
0002
0.00
020.
0016
0.00
15(0
.000
4)(0
.000
6)**
(0.0
006)
**(0
.004
2)*
(0.0
045)
*(0
.000
3)(0
.000
3)(0
.000
3)(0
.002
7)(0
.002
7)P
opul
atio
n of
vill
age
in (t
-1) p
erio
d X
10^4
0.01
180.
0061
0.00
600.
0348
0.03
420.
0576
0.05
510.
0551
0.53
390.
5346
(0.0
041)
***
(0.0
024)
**(0
.002
4)**
(0.0
150)
**(0
.015
0)**
(0.0
073)
***
(0.0
067)
***
(0.0
067)
***
(0.0
567)
***
(0.0
567)
***
If th
e vi
llage
had
any
term
inal
/sta
tion/
port
in (t
-1) p
erio
d: y
es-1
, no-
00.
0263
0.02
480.
0245
0.18
270.
1824
0.02
250.
0147
0.01
470.
0767
0.07
67(0
.005
4)**
*(0
.005
0)**
*(0
.005
1)**
*(0
.036
6)**
*(0
.037
3)**
*(0
.004
1)**
*(0
.003
7)**
*(0
.003
7)**
*(0
.022
7)**
*(0
.022
7)**
*
Vill
age
topo
grap
hy (h
ill a
rea
-0, f
latla
nd-1
)-0
.001
6-0
.001
2-0
.001
3-0
.009
4-0
.009
70.
0003
0.00
200.
0019
0.00
590.
0057
(0.0
035)
(0.0
032)
(0.0
033)
(0.0
245)
(0.0
248)
(0.0
029)
(0.0
025)
(0.0
025)
(0.0
160)
(0.0
160)
Any
dis
aste
r in
last
thre
e ye
ars:
y-1
, no-
2-0
.006
40.
0004
0.00
010.
0079
0.00
530.
0019
0.00
470.
0047
0.02
890.
0289
(0.0
032)
**(0
.003
0)(0
.003
1)(0
.023
6)(0
.024
0)(0
.002
5)(0
.001
9)**
(0.0
019)
**(0
.011
3)**
(0.0
113)
**
Con
stan
t2.
0288
2.00
731.
9955
1.33
451.
4105
1.39
85(0
.022
0)**
*(0
.022
4)**
*(0
.027
5)**
*(0
.014
7)**
*(0
.014
7)**
*(0
.020
7)**
*
Obs
erva
tions
5212
752
127
4985
055
129
5512
955
129
R-s
quar
ed0.
070.
090.
090.
10.
120.
13R
obus
t sta
ndar
d er
rors
in p
aren
thes
es*
sign
ifica
nt a
t 10%
; **
sign
ifica
nt a
t 5%
; ***
sig
nific
ant a
t 1%
Dis
trcit
spec
ific
cons
tant
s ar
e no
t rep
orte
d in
the
tabl
e
28
Tabl
e 10
a: L
ocal
Pre
fere
nce
and
Cha
nge
in L
ocal
Pub
lic G
oods
- H
ealth
care
Infr
astr
uctu
reD
ep. V
aria
ble:
Cha
nge
in th
e av
aila
bilit
y of
pol
yclin
ic in
vill
age
1996
-200
020
00-2
006
OLS
OLS
OLS
Ord
ered
Pro
bit
Ord
ered
Pro
bit
OLS
OLS
OLS
Ord
ered
Pro
bit
Ord
ered
Pro
bit
with
Dis
trict
FE
with
Dis
trict
FE
with
Dis
trict
FE
with
Dis
trict
FE
with
Dis
trict
FE
with
Dis
trict
FE
with
Dis
trict
FE
with
Dis
trict
FE
Vill
age
head
's g
ende
r in
t-1 p
erio
d (m
ale-
1, fe
mal
e-0)
-0.0
066
-0.0
091
-0.0
093
-0.1
136
-0.1
17-0
.006
9-0
.006
3-0
.006
3-0
.046
3-0
.046
3(0
.007
2)(0
.007
4)(0
.007
3)(0
.001
3)**
*(0
.001
1)**
*(0
.005
4)(0
.005
3)(0
.005
3)(0
.000
9)**
*(0
.000
7)**
*
% o
f wom
en v
oter
in v
illag
e X1
0^2
-0.0
17-0
.157
-0.0
045
-0.0
004
(0.0
160)
(0.0
023)
***
(0.0
015)
***
(0.0
207)
***
Vill
age
head
's a
ge in
(t-1
) per
iod
X 1
0^2
0.04
190.
0361
0.63
90.
0359
0.62
20.
0457
0.04
370.
0437
0.42
530.
4253
(0.0
140)
***
(0.0
110)
***
(0.0
031)
***
(0.0
110)
***
(0.0
028)
***
(0.0
227)
**(0
.011
3)**
*(0
.011
3)**
*(0
.001
9)**
*(0
.001
7)**
*
Vill
age
head
's e
duca
tion
in (t
-1) p
erio
d N
ot c
ompl
eted
-0.0
702
-0.0
543
-0.0
698
-0.5
2-0
.603
7-0
.030
6-0
.006
2-0
.006
2-0
.095
3-0
.095
3(0
.041
4)*
(0.0
365)
(0.0
396)
*(0
.002
3)**
*(0
.002
1)**
*(0
.020
0)(0
.016
0)(0
.016
0)(0
.001
1)**
*(0
.001
0)**
*
Prim
ary
scho
ol-0
.070
5-0
.049
1-0
.065
7-0
.460
8-0
.573
1-0
.021
80.
0016
0.00
160.
0129
0.01
29(0
.041
6)*
(0.0
364)
(0.0
391)
*(0
.002
4)**
*(0
.002
4)**
*(0
.019
3)(0
.015
3)(0
.015
3)(0
.001
1)**
*(0
.001
1)**
*
Juni
or h
igh
scho
ol-0
.065
6-0
.045
2-0
.061
2-0
.359
9-0
.467
7-0
.023
30.
0019
0.00
190.
0359
0.03
59(0
.041
9)(0
.036
6)(0
.039
3)(0
.002
4)**
*(0
.002
4)**
*(0
.020
0)(0
.015
3)(0
.015
3)(0
.001
3)**
*(0
.001
3)**
*
Hig
h sc
hool
-0.0
581
-0.0
396
-0.0
556
-0.2
56-0
.367
6-0
.018
50.
0055
0.00
550.
0720
0.07
20(0
.041
7)(0
.036
2)(0
.039
0)(0
.002
6)**
*(0
.002
5)**
*(0
.020
0)(0
.015
3)(0
.015
3)(0
.001
4)**
*(0
.001
3)**
*
Aca
dem
y-0
.027
4-0
.011
9-0
.027
1-0
.012
5-0
.116
7-0
.004
50.
0197
0.01
970.
1660
0.16
60(0
.042
5)(0
.037
0)(0
.039
7)(0
.002
5)**
*(0
.002
3)**
*(0
.021
3)(0
.016
0)(0
.016
0)(0
.001
5)**
*(0
.001
4)**
*
Uni
vers
ity-0
.032
-0.0
284
-0.0
452
-0.1
365
-0.2
519
-0.0
081
0.01
630.
0163
0.13
870.
1387
(0.0
426)
(0.0
368)
(0.0
390)
(0.0
021)
***
(0.0
020)
***
(0.0
207)
(0.0
160)
(0.0
160)
(0.0
018)
***
(0.0
017)
***
Vill
age
head
's d
urat
ion
in y
ears
X10
^2-0
.018
6-0
.002
96-0
.032
60.
0015
2-0
.025
8-0
.008
4-0
.003
8-0
.003
8-0
.142
0-0
.142
0(0
.025
0)(0
.022
0)(0
.009
9)**
*(0
.023
0)(0
.008
9)**
*(0
.032
0)(0
.019
3)(0
.019
3)(0
.006
1)**
*(0
.005
7)**
*
Vill
ages
with
50%
or m
ore
fam
ilies
in p
re-w
elfa
re a
nd w
elfa
re 1
in (t
-1) p
erio
d:
yes
-1, n
o-0
-0.0
268
-0.0
228
-0.0
223
-0.3
172
-0.3
046
-0.0
154
-0.0
143
-0.0
143
-0.1
267
-0.1
267
(0.0
038)
***
(0.0
032)
***
(0.0
032)
***
(0.0
012)
***
(0.0
011)
***
(0.0
038)
***
(0.0
023)
***
(0.0
023)
***
(0.0
005)
***
(0.0
005)
***
If th
e vi
llage
had
a p
olyc
linic
in (t
-1) p
erio
d: y
es-1
, no-
0-0
.513
3-0
.591
2-0
.595
2-1
3.87
51-1
3.82
17-0
.456
0-0
.482
0-0
.482
0-9
.219
1-9
.219
1(0
.023
2)**
*(0
.022
5)**
*(0
.023
1)**
*(0
.015
1)**
*(0
.015
5)**
*(0
.013
3)**
*(0
.010
7)**
*(0
.010
7)**
*(0
.009
3)**
*(0
.009
3)**
*
If th
e vi
llage
had
a p
uske
smas
in (t
-1) p
erio
d: y
es-1
, no-
00.
0801
0.07
160.
0722
0.60
090.
6037
0.03
350.
0390
0.03
900.
2320
0.23
20(0
.007
3)**
*(0
.005
7)**
*(0
.005
7)**
*(0
.001
9)**
*(0
.001
9)**
*(0
.005
1)**
*(0
.004
4)**
*(0
.004
4)**
*(0
.002
1)**
*(0
.002
1)**
*
Pop
ulat
ion
of v
illag
e in
(t-1
) per
iod
X 10
^40.
0217
0.01
070.
0462
0.01
050.
0458
0.13
870.
1000
0.10
000.
4440
0.44
40(0
.007
4)**
*(0
.004
1)**
*(0
.000
4)**
*(0
.004
0)**
*(0
.000
4)**
*(0
.010
7)**
*(0
.011
3)**
*(0
.011
3)**
*(0
.001
7)**
*(0
.001
6)**
*
If th
e vi
llage
had
any
term
inal
/sta
tion/
port
in (t
-1) p
erio
d: y
es-1
, no-
00.
0526
0.05
310.
0535
0.39
750.
4022
0.02
470.
0319
0.03
190.
1820
0.18
20(0
.007
1)**
*(0
.006
7)**
*(0
.006
7)**
*(0
.002
5)**
*(0
.002
4)**
*(0
.006
4)**
*(0
.004
9)**
*(0
.004
9)**
*(0
.002
1)**
*(0
.002
1)**
*
Vill
age
topo
grap
hy (h
ill a
rea
-0, f
latla
nd-1
)0.
0103
0.00
890.
009
0.16
160.
1658
0.00
760.
0129
0.01
290.
1140
0.11
40(0
.002
8)**
*(0
.002
6)**
*(0
.002
6)**
*(0
.001
1)**
*(0
.001
0)**
*(0
.004
2)*
(0.0
030)
***
(0.0
030)
***
(0.0
009)
***
(0.0
007)
**
Any
dis
aste
r in
last
thre
e ye
ars:
y-1
, no-
2-0
.002
60.
0032
0.00
330.
0481
0.04
76-0
.001
60.
0001
0.00
010.
0064
0.00
64(0
.003
3)(0
.002
7)(0
.002
8)(0
.000
7)**
*(0
.000
7)**
*(0
.002
7)(0
.002
2)(0
.002
2)(0
.000
6)**
*(0
.000
6)**
Con
stan
t2.
0754
2.04
772.
0723
1.35
521.
2985
1.29
92(0
.043
3)**
*(0
.037
8)**
*(0
.041
0)**
*(0
.024
0)**
*(0
.016
0)**
*(0
.019
3)**
*
Obs
erva
tions
5212
752
127
4985
052
127
4985
055
129
5512
955
129
5512
955
129
R-s
quar
ed0.
210.
250.
260.
220.
280.
07R
obus
t sta
ndar
d er
rors
in p
aren
thes
es
* sig
nific
ant a
t 10%
; ** s
igni
fican
t at 5
%; *
** s
igni
fican
t at 1
%D
istrc
it sp
ecifi
c co
nsta
nts
are
not r
epor
ted
in th
e ta
ble
29
Tabl
e 10
b: L
ocal
Pre
fere
nce
and
Cha
nge
in L
ocal
Pub
lic G
oods
- H
ealth
care
Infr
astr
uctu
reD
ep. V
aria
ble:
Cha
nge
in th
e av
aila
bilit
y of
pus
kesm
as in
vill
age
1996
-200
020
00-2
006
OLS
OLS
with
O
LS w
ith
Prob
it w
ith
Pro
bit w
ith
OLS
OLS
with
O
LS w
ith
Prob
it w
ith
Pro
bit w
ith
Dis
trict
FE
Dis
trict
FE
Dis
trict
FE
Dis
trict
FE
Dis
trict
FE
Dis
trict
FE
Dis
trict
FE
Dis
trict
FE
Vill
age
head
's g
ende
r in
t-1 p
erio
d (m
ale-
1, fe
mal
e-0)
-0.0
051
-0.0
069
-0.0
071
-0.0
791
-0.0
820
-0.0
038
-0.0
042
-0.0
042
-0.0
480
-0.0
484
(0.0
054)
(0.0
055)
(0.0
056)
(0.0
011)
***
(0.0
010)
***
(0.0
036)
(0.0
036)
(0.0
036)
(0.0
087)
***
(0.0
087)
***
% o
f wom
en v
oter
in v
illag
e X
10^2
0.01
410.
1590
-0.0
085
-0.4
213
(0.0
130)
(0.0
240)
***
(0.0
153)
(0.0
180)
***
Vill
age
head
's a
ge in
(t-1
) per
iod
X 1
0^2
0.03
690.
0408
0.03
950.
5250
0.52
700.
0219
0.02
670.
0267
0.31
400.
3180
(0.0
110)
***
(0.0
110)
***
(0.0
110)
***
(0.0
260)
***
(0.0
270)
***
(0.0
073)
***
(0.0
073)
***
(0.0
073)
***
(0.0
187)
***
(0.0
193)
***
Vill
age
head
's e
duca
tion
in (t
-1) p
erio
d N
ot c
ompl
eted
0.02
150.
0219
0.01
520.
3603
0.27
56-0
.000
70.
0014
0.00
140.
0074
0.01
01(0
.012
4)*
(0.0
121)
*(0
.011
7)(0
.001
5)**
*(0
.001
5)**
*(0
.008
0)(0
.008
0)(0
.008
0)(0
.004
9)(0
.005
0)**
Prim
ary
scho
ol0.
0236
0.02
840.
0212
0.45
240.
3620
0.00
150.
0045
0.00
450.
0364
0.03
93(0
.011
9)**
(0.0
116)
**(0
.010
8)**
(0.0
020)
***
(0.0
020)
***
(0.0
073)
(0.0
073)
(0.0
073)
(0.0
059)
***
(0.0
061)
***
Juni
or h
igh
scho
ol0.
0248
0.03
190.
0254
0.50
100.
4192
0.00
050.
0069
0.00
690.
0542
0.05
83(0
.012
0)**
(0.0
117)
***
(0.0
107)
**(0
.002
2)**
*(0
.002
3)**
*(0
.007
3)(0
.007
3)(0
.007
3)(0
.006
1)**
*(0
.006
3)**
*
Hig
h sc
hool
0.03
100.
0386
0.03
180.
5950
0.50
820.
0040
0.01
030.
0103
0.09
330.
0980
(0.0
120)
**(0
.011
8)**
*(0
.010
8)**
*(0
.002
3)**
*(0
.002
3)**
*(0
.007
3)(0
.007
3)(0
.007
3)(0
.005
9)**
*(0
.006
1)**
*
Acad
emy
0.04
720.
0536
0.04
600.
7508
0.65
790.
0063
0.01
330.
0133
0.11
930.
1253
(0.0
128)
***
(0.0
125)
***
(0.0
113)
***
(0.0
021)
***
(0.0
020)
***
(0.0
080)
(0.0
080)
(0.0
080)
(0.0
051)
***
(0.0
051)
***
Uni
vers
ity0.
0309
0.03
770.
0317
0.59
820.
5219
0.00
710.
0145
0.01
450.
1247
0.13
07(0
.012
3)**
(0.0
122)
***
(0.0
112)
***
(0.0
026)
***
(0.0
025)
***
(0.0
073)
(0.0
080)
*(0
.008
0)*
(0.0
059)
***
(0.0
060)
***
Vill
age
head
's d
urat
ion
in y
ears
X 1
0^2
-0.0
004
0.00
700.
0112
0.01
640.
0516
0.00
19-0
.015
0-0
.015
0-0
.029
1-0
.032
1(0
.017
0)(0
.017
0)(0
.018
0)(0
.095
0)**
*(0
.098
0)**
*(0
.013
3)(0
.012
7)(0
.012
7)(0
.066
7)(0
.066
7)V
illag
es w
ith 5
0% o
r mor
e fa
mili
es in
pre
-wel
fare
and
wel
fare
1 in
(t-1
) per
iod:
ye
s -1
, no-
0-0
.008
3-0
.009
8-0
.009
8-0
.128
0-0
.127
0-0
.002
6-0
.001
9-0
.001
9-0
.046
5-0
.046
2(0
.002
2)**
*(0
.002
2)**
*(0
.002
2)**
*(0
.000
7)**
*(0
.000
7)**
*(0
.001
3)**
(0.0
012)
(0.0
012)
(0.0
063)
***
(0.0
064)
***
If th
e vi
llage
had
a p
uske
smas
in (t
-1) p
erio
d: y
es-1
, no-
0-0
.196
2-0
.195
3-0
.195
7-1
1.03
82-1
1.05
78-0
.138
0-0
.138
0-0
.138
0-4
.269
6-4
.271
6(0
.011
9)**
*(0
.010
3)**
*(0
.010
4)**
*(0
.012
5)**
*(0
.012
4)**
*(0
.006
7)**
*(0
.006
3)**
*(0
.006
3)**
*(0
.011
3)**
*(0
.011
3)**
*
If th
e vi
llage
had
a p
olyc
linic
in (t
-1) p
erio
d: y
es-1
, no-
00.
0447
0.03
490.
0362
0.27
280.
2839
0.01
110.
0098
0.00
970.
0893
0.08
87(0
.008
1)**
*(0
.008
4)**
*(0
.008
4)**
*(0
.002
9)**
*(0
.002
7)**
*(0
.009
3)(0
.008
7)(0
.008
7)(0
.006
7)**
*(0
.006
7)**
*
Pop
ulat
ion
of v
illag
e in
(t-1
) per
iod
X 10
^40.
0078
0.00
550.
0054
0.04
090.
0396
0.03
040.
0269
0.02
690.
2313
0.23
06(0
.002
8)**
*(0
.002
4)**
(0.0
023)
**(0
.001
4)**
*(0
.001
4)**
*(0
.005
1)**
*(0
.005
6)(0
.005
6)**
*(0
.008
0)**
*(0
.008
7)**
*
If th
e vi
llage
had
any
term
inal
/sta
tion/
port
in (t
-1) p
erio
d: y
es-1
, no-
00.
0422
0.04
330.
0448
0.35
350.
3673
0.02
260.
0199
0.01
990.
1700
0.16
80(0
.005
3)**
*(0
.005
7)**
*(0
.005
4)**
*(0
.002
2)**
*(0
.002
1)**
*(0
.004
0)**
*(0
.003
9)**
*(0
.003
9)**
*(0
.006
3)**
*(0
.006
3)**
*
Vill
age
topo
grap
hy (h
ill a
rea
-0, f
latla
nd-1
)0.
0027
0.00
280.
0031
0.03
940.
0442
0.00
080.
0036
0.00
360.
0240
0.02
49(0
.002
0)(0
.002
0)(0
.002
1)(0
.001
0)**
*(0
.000
9)**
*(0
.001
7)(0
.001
8)**
(0.0
018)
**(0
.007
3)**
*(0
.007
3)**
*
Any
dis
aste
r in
last
thre
e ye
ars:
y-1
, no-
20.
0020
0.00
110.
0000
0.02
100.
0061
0.00
26-0
.000
6-0
.000
60.
0298
0.02
96(0
.001
8)(0
.002
1)(0
.002
1)(0
.000
7)**
*(0
.000
6)**
*(0
.001
4)*
(0.0
013)
(0.0
013)
(0.0
062)
***
(0.0
063)
***
Con
stan
t1.
9835
1.98
121.
9818
1.33
251.
3232
1.32
79(0
.013
5)**
*(0
.013
6)**
*(0
.013
8)**
*(0
.008
7)**
*(0
.008
7)**
*(0
.012
0)**
*
Obs
erva
tions
5212
752
127
4985
052
127
4985
055
129
5512
955
129
5512
955
082
R-s
quar
ed0.
090.
10.
10.
090.
110.
11R
obus
t sta
ndar
d er
rors
in p
aren
thes
es* s
igni
fican
t at 1
0%; *
* sig
nific
ant a
t 5%
; ***
sig
nific
ant a
t 1%
Dis
trcit
spec
ific
cons
tant
s ar
e no
t rep
orte
d in
the
tabl
e