Village capacity in maintaining infrastructure : evidence from rural Indonesia

Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia Arya Gaduh November 2010 THE WORLD BANK OFFICE JAKARTA Indonesia Stock Exchange Building, Tower II/12-13th Fl. Jl. Jend. Sudirman Kav. 52-53 Jakarta 12910 Tel: (6221) 5299-3000 Fax: (6221) 5299-3111 Printed in November 2010 Village Capacity in Maintaining Infrastructure: Evidence from Rural Indonesia is a product of staff of the World Bank. The findings, interpretation and conclusion expressed herein do not necessarily reflect the views of the Board of Executive Directors of the World Bank or the government they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denomination and other information shown on any map in this work do not imply any judgment on the part of the World Bank concerning the legal status of any territory or the endorsement of acceptance of such boundaries. Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia Arya Gaduh November 2010 Acknowledgements It took a team to complete this study. Daan Pattinasarany provided input into questionnaire design, while Ali Subandoro managed the early process necessary to get the study going. Data was collected by a team of trained enumerators, who went and interviewed 3840 households in 32 villages every three months for a whole year. Th e data were entered and cleaned by another team into an elaborate information system that was designed by Nur Cahyadi. Meanwhile, two individuals have been at the forefront of this study from the beginning. Yulia Herawati and Gregorius Kelik Endarso contributed to the design of the questionnaire, designed and implemented the training for enumerators, supervised survey implementations in the fi eld, as well as the data cleaning process. In addition, Gregorius Kelik Endarso provided the needed research assistance in the drafting of the fi nal report. It is safe to say that their eff orts are responsible for the high quality of the fi nal dataset on which this report is based. The author is also grateful to Richard Gnagey, Scott Guggenheim, Susanne Holste, Lina Marliani, Monica Singhal, and Susan Wong, as well as participants in the brown-bag lunch discussion held at Bappenas, which provided detailed and critical comments on the early draft of the report. Juliana Wilson did the fi nal editing of the report. Moreover, the author thanks Scott Guggenheim and Susan Wong in particular, whose support and assistance from the very beginning have made the study possible. Finally, we thank the Decentralization Support Facility (DSF) and the PNPM Support Facility (PSF) for their fi nancial support. iii Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia Table of Contents Acknowledgements iii Executive Summary 1 Chapter 1 Introduction 5 Chapter 2 General Methodology 7 2.1 Village and Household Sampling Strategy 8 2.2 Infrastructure Maintenance Costs: Definition and Sampling Strategy 9 Chapter 3 Sample Village Characteristics 13 3.1 The Broader Context: Where do Our Sample Villages Fit? 13 3.2 Village Characteristics 18 Chapter 4 Household Characteristics: Income, Expenditure, and Willingness to Pay 25 4.1 Income 26 4.2 Expenditures 31 4.3 Willingness to pay 36 Chapter 5 Village Resources 41 Chapter 6 Empirical Analysis 45 6.1 Paying for Maintenance: 6.2 Determinants of villagers’ willingness to pay 57 Chapter 7 Conclusion 63 Chapter 8 Policy Recommendations 65 References 68 List of Tables Table 1: Periodic (P) and routine (R) maintenance activities: Roads 9 Table 2: Periodic (P) and routine (R) maintenance activities: Bridges 10 Table 3: Periodic (P) and routine (R) maintenance activities: Piped water systems 10 Table 4: Sample kecamatan according to the 2004 Poverty Map 14 Table 5: Share of villages in kecamatan with different road types and conditions, 2008 15 Table 6: Share of villages in kecamatan with different types of water sources, 2008 17 Table 7: Sample villages by topography 19 Table 9: Number of villages with maintenance activities in the last 12 months by province 20 Table 8: Number of villages with maintenance activities and upgrades in the last 12­ months 20 Table 10: Implementer of maintenance activities and upgrades 21 Table 11: Sources of resources for infrastructure maintenance, by infrastructure type and province 22 iv Table 12: How villagers contribute by types of maintenance activities 24 Table 13: How villagers contribute by types of infrastructure 24 Table 14: Monthly per-capita income by province, Aug 2008-Jul 2009 27 Table 15: Average and CV of monthly per-capita income, by province and income source 29 Table 16: Number of households reporting experiencing shock by source and province, Aug 2008-Jul 2009 30 Table 17: Monthly nominal per-capita expenditure in sample villages, Aug 2008- Jul 2009 31 Table 18: Monthly nominal per-capita expenditure, Aug 2008-Jul 2009 33 Table 19: Monthly per-household local tax and consumption 34 Table 20: Monthly per-household water expenditure and consumption 35 Table 21: Villagers’ average willingness to pay by province 36 Table 22: Mean and median household willingness to pay for infrastructure maintenance (Rp/month) 37 Table 23: Distribution of willingness to pay per household per month in all villages, by province and infrastructure 39 Table 24: Willingness to pay per household per month for villages with bridges and water systems 40 Table 25: Annual income, consumption, and total maintenance cost 42 Table 26: Annual income, consumption, and maintenance cost excluding unskilled labor 43 Table 27: Maintenance costs and willingness to pay per-household: Roads 48 Table 28: Maintenance costs and willingness to pay per-household: Bridges 49 Table 29: Maintenance costs and willingness to pay per-household: Piped water 49 Table 30: Maintenance costs and willingness to pay: all infrastructure 50 Table 31: Resource collection under median and optimal user fees: Roads 53 Table 32: Resource collection under median and optimal user fees: Bridges 54 Table 33: Resource collection under median and optimal user fees: Piped water system 55 Table 34: Resource collection under median and optimal user fees: all infrastructure 56 Table 35: Interval regression on the willingness to pay for roads+ 60 Table 36: Complaints and satisfaction rates, by province 61 Table 37: Recipients of respondents’ complaints 62 List of Figures Figure 1: Monthly per-capita income by province, Aug 2008 - Jul 2009 27 Figure 2: Monthly per-capita income flow by income source and province, Aug 2008 – July 2009 28 Figure 3: Monthly nominal per-capita expenditure, Aug 2008-Jul 2009 32 v Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia vi Executive Summary Inadequate rural infrastructure creates constraints to economic growth in these areas by limiting productive growth and impeding the development of human capital. Recognition of the rural- poverty-alleviation role of infrastructure development motivated the central government, as well as donor agencies, to direct a significant amount of aid into supporting such projects.1 Naturally, the poverty-targeting criterion plays an important part in deciding on the location of project activities. As priorities for infrastructure projects are usually given to poorer regions and villages, a problem arises. Currently, infrastructure is often provided under the implicit assumption that local people are able to pay the full maintenance costs. However, there is no systematic evidence that this assumption is justified. 1 About 76% of almost US$ 700 million-worth of investments channeled through the World Bank’s Kecamatan Development Program (KDP) were for infrastructure projects, with 23% for economic activities, and 1% for education and health activities (ADB, 2005). Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia Research Questions and Design This study examines the assumption that villagers are able to finance the necessary maintenance of infrastructure on their own. More specifically, the study seeks answers to the following questions. i. ty in­ Do villagers in poor villages have the resources to maintain their priori­­ frastructure on their own? ii. ces for infrastructure If they do, to what extent are villagers willing to use their resour­ maintenance? iii. How do village characteristics affect resource availability and their­willingness to pay for infrastructure maintenance? To this end, we took a purposive sample of 32 relatively poor villages across five provinces in Java, Sumatra, Kalimantan, Sulawesi, and East Nusa Tenggara. To see how our sample fit relative to other villages in the country and in each province, we do some comparative analysis. We find that with some minor exceptions, our sample villages are mainly located in the poorest 40% of kecamatan (sub-districts) nationally and provincially. With regard to infrastructure quality, compared to their district averages, these kecamatan (sub-districts) have slightly greater shares of villages with a majority of asphalt roads, slightly lesser access to municipal waterworks (PAM) and ground water compared to their district averages. Once we selected our sample villages, in each village, our team of infrastructure engineers performed measurements of the maintenance costs necessary to ensure the long term sustainability of village infrastructure. We considered three types of infrastructure: roads, bridges, and piped water systems. To estimate the inflow of resources, we collected detailed information on household income and consumption. To account for possible income and consumption fluctuations between seasons, we collected data from four Consecutive Surveys, each three months apart. Using this data, we examine how much of an extra burden maintenance costs would be if they were fully born by these households. We also did a similar calculation assuming that villagers contribute fully the necessary unskilled labor. Using the contingent valuation method, we also asked households to state their Willingness To Pay (WTP) maintain infrastructure individually and as a bundle. We use these data to estimate resources that would be available in different ways to collect contributions from households. Household data were collected between July 2008 and August 2009. Results We find that, with respect to their current welfare, the cost of maintaining village infrastructure was significant in our sample villages. When distributed equally across households in each village, the total costs to maintain village infrastructure in this study—village roads, bridges, and piped water systems when they exist—amount to between 0.1% and 2.8% of households’ total consumption, with a median of 1.1% (see Table 25). If we assume that villagers can supply all of the necessary unskilled labor, maintenance costs amount to between 0.1% and 1.4% of their consumption, with a median of 0.5% (see Table 26). The bulk of these costs come from the cost to maintain rural roads. While these percentages might appear small, this “local tax” for infrastructure is likely to be burdensome, given the low income of many villagers. Moreover, since income and consumption 2  Executive Summary are not easily observable, it may be difficult to create with a mechanism to collect contributions based on household income or consumption share in these villages. If we use villagers’ stated willingness to pay instead, we find that the resources collected fell well below what would necessary to fully finance these costs. Table 30 Shows that between 21% (when villagers do not contribute any unskilled labor) and 63% (when villagers contribute all unskilled labor) of sample villages would be able to afford the maintenance of all three types of infrastructure. When looked at separately for each type of infrastructure, the heaviest burden comes from financing roads: only between 21% and 43% of villages would be able to afford road maintenance based on their stated willingness to pay for road maintenance (see Tables 27, 28, and 29 for comparisons). It is, however, reasonable to think that households may not contribute their full WTP. Our econometric analysis alludes to the presence of the temptation to free ride on other households’ contributions. When households did so, total contributions would fall. Under scenarios in which a single-priced monthly household user fee is established through a voting procedure, we find that the number of villages whose contributions would fall short of those necessary to finance proper infrastructure maintenance increases. Using their WTPs as the basis for calculations, we find that only between 10% and 20% of the sample villages’ resources collected through these user fees would be adequate to finance the maintenance of all three types of infrastructure (Table 34). Table 35 presents our empirical analysis of the determinants of household WTP to pay for road maintenance. It shows, as predicted by theory, that household income and opportunity costs from a damaged road were positively correlated with WTP. In addition, we also find that the increase in WTP due to increased opportunity costs only happened when households have suffered actual incidences of failed roads and the coefficient was only significant for monetary opportunity cost. We also find that institutional responsiveness is important: Satisfactory responses to complaints regarding road problems increased WTP. Perceived administrative quality of the various levels of governments did not seem to play a major role in influencing the WTPs. Interestingly enough, a higher perceived trust of fellow villagers and their helpfulness appears to reduce households WTPs. Policy Recommendations Based on our results, we recommend the following: Institutionalize infrastructure maintenance with clearly defined roles and responsibilities for the different levels of administration. Our study finds that there is a significant willingness on the part of villagers to contribute towards infrastructure maintenance. For most villages, there would be adequate resources to conduct routine maintenance. However, villagers may need a significant financial support to ensure that periodic maintenance is conducted properly. This suggests that villagers can take up the responsibility for routine maintenance. At the same time, district governments and outside agencies need also to step up their support of the types of maintenance activities villagers have difficulty affording. District governments need to gradually reallocate resources towards maintenance instead of upgrades. Table 11 suggests that district governments tend to use their resources to support upgrades and rehabilitation efforts instead of periodic maintenance. One study suggests that the cost ratio between upgrades and maintenance is about 3.5 to 1 (Dongges et. al., 2007). There is a strong case for a gradual shift from construction and rehabilitation or upgrades towards developing the necessary resources and institutions to undertake rural infrastructure maintenance activities. 3 Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia Newly constructed infrastructure needs to be accompanied with an explicit and detailed maintenance plan that states clearly the resources necessary to implement it. Our data shows that the maintenance costs vary much more than the villagers’ willingness to pay. These variations are driven, among other factors, by local conditions as well as the volume and design of the infrastructure. Estimating the maintenance costs long after the fact—which may be necessary in order to estimate the resource gap that will need to be plugged—can be cumbersome, costly, and may not be particularly accurate. On the other hand, these variations are likely to be better understood by the initial implementers of the infrastructure. It is therefore crucial that new projects be accompanied with plans for sustainable maintenance that can be used and understood by the various agencies that may need to be involved in plugging the resource gap. At the village level, there needs to be a designated institution responsible for maintenance. Our data suggest that villages do implement routine maintenance on their own. However maintenance can be implemented more efficiently if they are implemented at the correct time (Dongges et. al., 2007). Moreover, these activities will require villager contributions and we find that villagers’ willingness to contribute are positively and significantly correlated with the responsiveness of an institution in immediately addressing reported infrastructure problems. This designated institution or person can therefore act to coordinate maintenance efforts as well as respond to potential problems in a timely fashion. Having an institution responsible for maintenance—this could well come from an existing village institution—will be instrumental in ensuring the sustainability of the maintenance efforts. The assignment of maintenance activities to villagers needs to take into account the possibly unequal distribution of burdens towards poorer households. Our data show that maintenance costs can be reduced significantly when villagers are expected to contribute all of the unskilled labor. However having villages supply all unskilled labor may amount to a regressive “informal tax”, where poorer households “pay” more (in the form of labor) for public goods.2 It is important to address this potential issue in the process of institutionalizing maintenance activities at the village level. Studies need to be undertaken on effective and efficient means to collect and channel resources to ensure that infrastructure is well-maintained in the long term. This study provides insight into the resource gaps faced by villagers in fulfilling the maintenance requirements for their infrastructure. It also suggests a role for outside agencies to support maintenance efforts in villages, including, but not limited to, district governments. However, we still lack a good understanding on effective mechanisms to deliver resources in a manner that ensures that infrastructure is well-maintained in the long term or how effective these mechanisms are in different contexts. Moreover, we also need to understand better the effectiveness of different resource collection strategies for different types of infrastructure to improve the design of the village-level maintenance institutions. 2 For a discussion of this “informal tax”, see Olken and Singhal (2009). 4 Chapter 1 Introduction The lack of key infrastructure in rural areas is often considered an underlying cause of rural poverty in Indonesia. Inadequate rural infrastructure creates constraints to economic growth in rural areas by limiting productive growth and impeding the development of human capital. This recognition of the rural-poverty-alleviation role of infrastructure development motivated the central government, as well as donor agencies, to direct a significant amount of aid into supporting such projects.3 Naturally, the poverty-targeting criterion plays an important part in deciding on the location of project activities. As priorities for infrastructure projects are usually given to poorer regions and villages, a problem arises. Currently, infrastructure is provided under the implicit assumption that local people would be able to pay, either for partial contribution to the development or the upkeep of infrastructure. However, there is no systematic evidence that this assumption is justified. 3 About 76% of almost US$ 700 million-worth of investments channeled through the World Bank’s Kecamatan Development Program (KDP) were for infrastructure projects, with 23% for economic activities, and 1% for education and health activities (ADB, 2005). Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia This study examines the assumption that villagers would be able to finance the necessary maintenance of their infrastructure on their own. More specifically, the study seeks answers to the following questions: i. Do villagers in poor villages have the resources to maintain their priority infrastructure on their own? ii. If they do, to what extent are villagers willing to use their resources for infrastructure maintenance? iii. How do village characteristics affect resource availability and their willingness to pay for infrastructure? To this end, we took a purposive sample of 32 relatively poor villages across five provinces in Java, Sumatra, Kalimantan, Sulawesi, and East Nusa Tenggara. In each village, our team of infrastructure engineers performs measurements of the maintenance costs necessary to ensure the long term sustainability of village infrastructure. For this study, we consider three types of infrastructure: roads, bridges, and piped water systems. We then compare this with resources that would be available to villagers. More specifically, we use the income and consumption data to estimate the inflow of resources to households in these villages. Then, we examine how much of an extra burden would maintenance costs be if they were fully born by these households. Using the contingent valuation method, we also asked households to state their willingness to pay to maintain their infrastructure individually and as a group. We use these data to estimate resources that would be available under different ways to collect contributions from households. Our study finds that these maintenance costs would add a burden of between 0.1% and 4.3% of households’ total income, with a median of 1.4%. If we use consumption, which typically is better measured than income, these maintenance costs would amount to between 0.1% and 2.8% of these households’ consumption, with a median of 1.1%. Meanwhile, using the willingness to pay (WTP) data, we find that in 21% to 63% of our sample villages the maximum voluntary contributions from these households would be adequate to finance the three types of infrastructure considered in the study, depending on our assumption of the amount of unskilled labor that villagers contribute to infrastructure maintenance. These numbers fell to between 10% and 20% if we consider the more realistic mechanism of using user fees to collect these contributions. The report is organized as follows. In the next section, we describe the general methodology of the study. It is followed by a discussion on the characteristics of our sample villages in Section 3. Since we have a small sample, this section begins by giving a sense of where our sample villages fit within the context of the country. Section 4 describes the welfare characteristics and willingness to pay of the households in our sample, followed by a discussion of what that would entail for total available resources in the villages in Section 5. Section 6 discusses in detail some of our empirical analysis linking the total maintenance cost to the villagers’ willingness and ability to pay for them. More specifically, it discusses the extent to which village contributions can play a role in financing infrastructure maintenance under different scenarios. It also presents our empirical analysis of the determinants of villagers willingness for infrastructure, particularly for roads. Section 7 concludes. 6 Chapter 2 General Methodology We conducted our study in 32 villages in five provinces during the period of August 2008 and July 2009. At the beginning of the period, our team of engineers estimated the costs of infrastructure maintenance by sampling different points of three types of infrastructure, namely roads, bridges, and piped water systems, when they exist. Meanwhile, household data were collected every three months during the period. Within each village, we interviewed village officials as well as a sample of 120 households, which were followed by our panel survey throughout the year. We describe the sample and infrastructure selection strategy in more detail below. Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia 2.1 Village and Household Sampling Strategy In this study, we wanted to focus on villages that are representative of the poorer villages in different topographical locations across Indonesia. Hence, we sampled 32 villages in five island groups: Sumatra, Java, Kalimantan, Sulawesi, and East Nusa Tenggara. To find the poorer villages in these island groups, ideally we would select based on some direct measures of village-level poverty. However, since reliable poverty maps were only available at the kecamatan level or higher, we had to approximate. We first limited the sample villages to those located in kecamatan that are among the poorest 40% in each island group based on BPS’s 2004 kecamatan-level poverty map. To ensure the tractability of the survey, we then limited the sample to rural villages with between 300 and 600 households based on the Potensi Desa (Podes) 2005 data. Once we obtained a list of these villages, we categorized them in terms of their island groups and topography—whether each is located in a coastal, flatland, or hilly/mountainous area. Within each category, we selected 4-8 villages whose poverty characteristics (as provided by the Podes 2005 data) are around the median in each kecamatan.4 The final list of 32 villages was determined after we consulted with the field team regarding the feasibility of conducting the survey within the allotted time, i.e., one survey wave in each quarter. Within each village, the survey interviewed village officials and a random sample of households. To construct the sampling frame for households, the survey team conducted a mini-census of households. Based on the household consumption categories collected during the mini-census, 120 households were stratified-random sampled from each village. 4 Unlike the BPS’s poverty map which tries to estimate kecamatan-level poverty based on its households’ consumption level, Podes asks its village informant—typically the village head—to estimate how many households fall under the pra-sejahtera (or pre-welfare) level. When doing national level comparison, the former is likely to be more accurate; but we had to resort to the latter to choose the villages in the absence of a recent village-level poverty map. 8  General Methodology 2.2 Infrastructure Maintenance Costs: Definition and Sampling Strategy Defining Maintenance Our study focused on two types of infrastructure maintenance activities. The first, routine maintenance, refers to infrastructure maintenance work that needs to be performed at least once a year. This includes light maintenance work to ensure that the infrastructure is working properly, such as checking and cleaning the side-road drains after heavy rain. The second, periodic maintenance, needs to be done once every few years. The cost calculations in the study assume that periodic maintenance needs to be performed once every five years. Periodic maintenance refers to light or medium structural repairs to village infrastructure. Tables 1, 2, and 3 list the different maintenance activities and their classification as periodic or routine activities that we consider when we estimate the maintenance costs for different types of infrastructure. Table 1: Periodic (P) and routine (R) maintenance activities: Roads Work on the main road Slopes Replace damaged segment of concrete construction P Periodic maintenance of trees on the slopes P Repair damages to asphalt penetration P Periodic maintenance of weeds on the slopes P Resurface with asphalt and sands P Periodic repairs of damaged covering pairs P Resurface gravel roads P Periodic repairs of gabion P Re-gravel telford roads P Clean and maintain retaining wall, etc R Rearrange paving stone roads P Clearing of gabion R Routine maintenance of terrace, diversion Patch holes in telford roads R R channel, etc Patch holes or damaged segments in paving stone R roads Add sand and stones covering on telford stones R Drainage tunnel Resurface road tracks on telford roads R Repairs of the outside of the drainage tunnel P Periodic repairs of damaged pairs on tunnels P Periodic repairs of damaged concrete on the Shoulder lane P tunnels Repair drainage of the shoulder P Periodic repairs of damaged gabion on tunnels P Reshaping of the shoulder lane R Clearing of drainage tunnels R Repair the volume of eroded materials R Other buildings Side drainage Construction-protecting gabion P Periodic repairs of damaged pairs on other Reshape the soil channel and scour check P P buildings Periodic repairs of damaged concrete on other Repair damaged lining pairs P P buildings Maintain the side drainage P Periodic repairs of gabions on other buildings P Clean up and light repairs of the drainage R Repairs of the outside of the buildings P Clearing of areas around the buildings R 9 Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia Table 2: Periodic (P) and routine (R) maintenance activities: Bridges Bridge’s upper structure Road approach Periodic maintenance of the ramp covering Periodic maintenance of the wood, concrete planks P P pair Periodic maintenance of upper structure P Routine clearing of the road approach R Periodic maintenance of the metal construction P Painting of the metal, wood construction P River banks around the bridge General clean up of bridge, drainage channel, etc R Periodic maintenance of the bridge wing P Work on the bridge abutment and pillar Periodic maintenance of the concrete structure P Periodic maintenance of the abutment, pillar P Periodic maintenance of gabion P Maintenance of grass on cliff (replanting Periodic maintenance of the safety rock pairs P P volume) Periodic maintenance, abutment metal, wood, Clearing of the bridge wing pair (from plants, P R concrete etc) Periodic painting of wood part of abutment and Vegetation clearing around bridge R R pillar construction Periodic painting of the metal part of abutment and R pillar Clearing of water under the bridge R Maintenance of bridge ramp R Table 3: Periodic (P) and routine (R) maintenance activities: Piped water systems Main building for transmission pipes Piping Painting of walls for buildings P Replacement of pipes Maintenance/overhaul of the pump set P Pairs of stones to maintain Maintenance/overhaul of the hydrant set P Concrete constructions to maintain Repairs/replacement of metal parts on buildings Maintenance of stones, bricks, and concretes P etc Maintenance/replacement of building metals etc P Painting of metal construction Maintenance of building pipes P Checks on the pipelines Clean-up of the building R Clean-up of the building areas and drains R Distribution buildings Maintenance of public taps P Painting of walls, metal constructions P Maintenance of stones, bricks, and concretes P Periodic maintenance of metal constructions P Maintenance of pipes P Clean-up of the building areas and drains R 10  General Methodology There is a third type of maintenance activity, namely emergency maintenance, that we do not consider here. Emergency maintenance refers to activities to repair damages due to unforeseen circumstances (Dongges, et. al., 2007). Examples of such emergencies would be damage due to landslides, flood, or major accidents. Provision for this type of maintenance is obviously critical to ensuring that the road or other infrastructure can provide uninterrupted service to villagers. However, due to its unforeseen nature, we cannot incorporate the cost of emergency maintenance in this study. Infrastructure Sampling Strategy and Extrapolation to Village-Level Costs To establish maintenance costs, the team sampled different sections of the infrastructure that we will study to allow us to estimate the total cost of maintenance to the villages. For roads, we excluded roads whose maintenance is not the responsibility of the villagers, such as district or provincial roads. We then ranked the roads in terms of the number of beneficiaries and selected those with the largest number. We include only roads with a total length of at least 500 meters. For a homogeneous road, we sample one segment of the road to calculate the cost. Otherwise, we sampled segments that are representative of road types (in terms of its material), morphology (hilly or flat), and use (residential or non-residential). Once the costs for these segments were calculated, we inflated these costs using the total length of the existing village road divided by the segment length to establish a village-level maintenance cost. Similar to roads, for bridges we prioritized those with the largest beneficiaries. We only included bridges that were considered permanent, these can be one of three types: concrete, steel, or wood. Similarly, for the water system, we picked one piped water system with the largest number of beneficiaries. The cost is calculated for both transmission and distribution systems. Within the sampled water systems, we examine segments of the system with the highest number of users and figure out the costs for these segments. We use the costs for these segments as a basis for extrapolating the costs for the system covering the whole village. However, for water systems, we do not extrapolate based on the total length of the existing system. In our sample villages, existing systems only cover between 1.4% to 28.6% of villagers in each of villages with water systems. Instead, we inflate the cost using the total number of villagers divided by the number of users of our sampled segments in each of these villages using data on the number of beneficiaries collected during the infrastructure survey. Before we proceed, we need qualify what we mean by “village-level costs”. First, our calculations of “village-level costs” applies only to our sampled roads, bridges, and piped water systems. Although we sampled our infrastructure projects such that they are the most heavily used ones, they were often not the only ones that the villages had to maintain. We cannot generalize these costs to all infrastructure available in the villages. Henceforth, our calculated village-level costs should be interpreted as a lower bound to the cost of maintaining infrastructure in these villages. Second, we 11 Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia could only calculate costs based on maintenance of existing infrastructure. As the result, we do not have costs for bridges and piped water systems in almost half of our sample villages. Moreover, when calculating village-level costs, we assume that the maintenance costs are homogeneous within villages. Obviously, this is a strong assumption. Finally, all cost calculations include both material and labor necessary based on local wages and prices. 12 Chapter 3 Sample Village Characteristics 3.1 The Broader Context: Where do Our Sample Villages Fit? The above sample selection process gives us 32 villages located in 29 kecamatan in 21 districts across 5 provinces. The following provides a sense of how these sample villages fit within the broader context in terms of welfare and infrastructure. We first look at the welfare indicators. Using the 2004 kecamatan-level poverty map, we can examine where these kecamatan are in terms of poverty. Table 4 presents the kecamatan poverty rate as well as its normalized ranking nationally and within each province. The ranking ranged from 0 to 1, where 0 means it is the poorest, and 1, the richest. Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia Table 4: Sample kecamatan according to the 2004 Poverty Map % National Prov. Province District Kecamatan Poor rank Rank Central Java Blora Japah 30.75 0.18 0.2 Pati Margoyoso 21.58 0.38 0.51 Pemalang Warungpring 35.13 0.12 0.1 Rembang Kragan 28.96 0.21 0.24 Rembang Sluke 28.73 0.22 0.25 Tegal Bojong 30.53 0.19 0.21 Wonosobo Kaliwiro 35.1 0.12 0.1 West Kalimantan Kubu Raya Telok Pa’kedai 17.02 0.51 0.37 Landak Menyuke 20.66 0.40 0.25 Pontianak Toho 14.1 0.61 0.5 Singkawang Singkawang Timur 13.8 0.63 0.54 Singkawang Singkawang Utara 19.67 0.43 0.28 Sintang Kayan Hulu 21.06 0.39 0.23 Lampung Lampung Selatan Kalianda 30.86 0.18 0.21 Lampung Timur Way Bungur 39.35 0.08 0.09 Pesawaran Kedondong 26.48 0.26 0.38 Pesawaran Padang Cermin 26.25 0.27 0.4 Way Kanan Banjit 38.38 0.09 0.1 Way Kanan Baradatu 30.1 0.19 0.26 East Nusa Tenggara Kupang Amabi Oefeto - - - Kupang Amarasi 38.01 0.09 0.26 Kupang Kupang Timur 36.86 0.1 0.29 Timor Tengah Selatan Amanuban Selatan 37.43 0.09 0.27 Timor Tengah Utara Noemuti 41.23 0.06 0.18 South Sulawesi Bone Amali 21.47 0.38 0.27 Bone Lappariaja 34.8 0.12 0.06 Luwu Lamasi 23.33 0.33 0.24 Luwu Larompong Selatan 21.67 0.38 0.27 Tana Toraja Sopai - - - Source: BPS Poverty Map 2004 Nationally, four kecamatan in West Kalimantan were not within the 40% poorest in 2004. Of these, two (namely, Toho and East Singkawang) were within the 40% poorest kecamatan in the province. Kecamatan Pati in Java did not fall within the 40% poorest kecamatan in the province, but is within 40% poorest kecamatan nationally. These discrepancies were possibly due the application of the poverty criterion at the island group level, instead of the national or province level. Meanwhile, the 2004 poverty data for two kecamatan in East Nusa Tenggara and South Sulawesi (Amabi Oefeto and Sopai) are not available. Overall, except for the two kecamatan in West Kalimantan, the poverty rates are higher than the national average of 16.7% in 2004. 14  Sample Village Characteristics Table 5: Share of villages in kecamatan with different road types and conditions, 2008 Road type Accessible by 4+wh. Province District Kecamatan Ashpalt Hardened Earth Others vehicle all year? Dist. Kec Dist Kec Dist Kec Dist Kec Dist. Kec Lampung Lampung Selatan Kalianda 0.62 0.85 0.10 0.04 0.28 0.11 0.00 0.00 0.99 0.96 Lampung Timur Way Bungur 0.28 0.75 0.54 0.25 0.18 0.00 0.00 0.00 0.98 1.00 Way Kanan Banjit 0.17 0.30 0.60 0.65 0.23 0.05 0.00 0.00 0.96 1.00 Way Kanan Baradatu 0.17 0.59 0.60 0.27 0.23 0.14 0.00 0.00 0.96 0.91 Pesawaran Padang Cermin 0.62 0.73 0.23 0.27 0.15 0.00 0.00 0.00 0.97 0.86 Pesawaran Kedondong 0.62 0.67 0.23 0.14 0.15 0.19 0.00 0.00 0.97 0.95 Central Java Wonosobo Kaliwiro 0.63 0.24 0.36 0.76 0.00 0.00 0.00 0.00 1.00 1.00 Blora Japah 0.66 0.50 0.33 0.50 0.01 0.00 0.00 0.00 0.98 1.00 Rembang Kragan 0.94 1.00 0.06 0.00 0.00 0.00 0.00 0.00 1.00 1.00 Rembang Sluke 0.94 1.00 0.06 0.00 0.00 0.00 0.00 0.00 1.00 1.00 Pati Margoyoso 0.92 1.00 0.07 0.00 0.01 0.00 0.00 0.00 0.99 1.00 Pemalang Warungpring 0.86 0.50 0.11 0.33 0.03 0.17 0.00 0.00 0.96 1.00 Tegal Bojong 0.93 0.94 0.07 0.06 0.01 0.00 0.00 0.00 0.99 0.88 East Nusa Kupang Amarasi 0.18 0.33 0.45 0.56 0.37 0.11 0.00 0.00 0.90 1.00 Tenggara Kupang Kupang Timur 0.18 0.77 0.45 0.15 0.37 0.08 0.00 0.00 0.90 1.00 Kupang Amabi Oefeto 0.18 0.00 0.45 0.29 0.37 0.71 0.00 0.00 0.90 0.86 Timor Tengah Amanuban Selatan 0.30 0.20 0.37 0.70 0.34 0.10 0.00 0.00 0.91 1.00 Selatan Timor Tengah Utara Noemuti 0.30 0.00 0.38 0.82 0.31 0.18 0.00 0.00 0.96 0.91 West Landak Menyuke 0.38 0.31 0.13 0.13 0.50 0.56 0.00 0.00 0.65 0.56 Kalimantan Pontianak Toho 0.73 1.00 0.05 0.00 0.23 0.00 0.00 0.00 0.84 1.00 Sintang Kayan Hulu 0.19 0.14 0.11 0.21 0.69 0.64 0.01 0.00 0.59 0.43 Kubu Raya Telok Pa’kedai 0.48 0.00 0.02 0.00 0.51 1.00 0.00 0.00 0.42 0.15 Singkawang Singkawang Timur 1.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 1.00 1.00 Singkawang Singkawang Utara 1.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 1.00 1.00 South Sulawesi Bone Lappariaja 0.50 0.22 0.38 0.56 0.12 0.22 0.00 0.00 0.92 0.78 Bone Amali 0.50 0.80 0.38 0.20 0.12 0.00 0.00 0.00 0.92 1.00 Luwu Larompong Selatan 0.40 0.60 0.38 0.40 0.22 0.00 0.00 0.00 0.84 1.00 Luwu Lamasi 0.40 1.00 0.38 0.00 0.22 0.00 0.00 0.00 0.84 1.00 Tana Toraja Sopai 0.26 0.50 0.40 0.50 0.34 0.00 0.00 0.00 0.80 1.00 Source: Podes 2008 We next examine the relative quality of the infrastructure in the kecamatan where our sample villages reside using Podes 2008 (hereafter, Podes). Since our study focuses on roads and bridges, and water sources, we employ the Podes questions on road and water sources. Podes has questions on the main means of transportation to and from the village, and, if it is by land, the dominant road types in each respondent village. There are three main categories: asphalt, hardened, and earth road, plus an “other” option. In addition, Podes includes a question of whether these roads are accessible by 4-or-more wheeled vehicles all year long. For a sense of the relative quality of roads, we calculate the share of villages in each kecamatan with the different types of dominant road types 15 Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia and its year-long accessibility for cars and the like. We use district averages of these kecamatan-level shares as benchmark of the relative quality of the infrastructure in our sample. Table 5 presents these numbers. The relative qualities of infrastructure in between our sample kecamatan and their district counterparts differ by provinces. In Lampung, kecamatan in our sample tend to have better quality roads than their counterparts. Overall, sample kecamatan have a higher share of villages with asphalt roads, and a lower share with earth roads. However, in terms of accessibility to vehicles with 4 or more wheels, they are comparable to the district average, except for Kecamatan Padang Cermin, which has a lower share of year-long accessible villages than the district average. Meanwhile, in Central Java, about half of the kecamatan have a higher share of villages with asphalt, and the other half, a significantly lower share. Meanwhile, these kecamatan have comparable (and fairly small) shares of villages with earth roads to their district counterparts, except for Kecamatan Warungpring, with more of the earth roads. Across the province, almost all of the villages have roads that are accessible to four-or-more wheeled vehicles all year long. In East Nusa Tenggara, more than half of the kecamatan have a much smaller share of villages with a majority asphalt road than the district average. Sample kecamatan tend to have a higher share of hardened roads, and a lower share of earth roads, except for Kecamatan Amabi Oefeto, which has a larger share of villages with earth roads compared to the district average. Only about 90% of the districts where our sample villages reside have year-long accessibility to four-or-more-wheeled vehicles. Our sample kecamatan tend to have comparable or slightly higher accessibility, except for Kecamatan Amabi Oefeto, which as a much lower accessibility than its district average. Excluding kecamatan located in Singkawang—where all villages have a majority of asphalt roads— the road quality in our West Kalimantan sample kecamatan tended to be worse than the district average. There tend to be a lower share of villages with asphalt roads, and a higher share of those with earth roads. Our sample kecamatan also tended to have a much lower share of villages with year-long accessibility to four-or-more-wheeled vehicles. The exception in West Kalimantan is Kecamatan Toho, which tends to have a higher share of villages with asphalt road, a lower share with earth road, and a higher accessibility than its district average. In South Sulawesi, our sample kecamatan tend to have better road qualities than the district average, with the exception of Kecamatan Lappariaja. Except for one kecamatan, our sample kecamatan have shares of villages with asphalt roads that are higher than the district average, no villages with earth roads, and all villages with year-long access to four-or-more-wheeled vehicles. On water sources, Podes asks village informants for the main source of drinking and cooking water used in the village. Aside from the “other” option, it has six options: PAM (the state-owned water company), electric/manual pump, well, springs, river/lake, and rain. We re-categorized these options into PAM, manually extracted ground water (for the second and third options), and other natural sources (for the remaining three). We implemented the same calculations as before to find the share of villages in the sample kecamatan with different types of water sources and that have households that purchase water for drinking and cooking. The results are presented in Table 6. 16  Sample Village Characteristics Table 6: Share of villages in kecamatan with different types of water sources, 2008 Water source Province District Kecamatan PAM Ground Natural Others Dist Kec Dist Kec Dist Kec Dist Kec Lampung Lampung Selatan Kalianda 0.02 0.04 0.83 0.63 0.13 0.30 0.02 0.04 Lampung Timur Way Bungur 0.02 0.00 0.96 1.00 0.03 0.00 0.00 0.00 Way Kanan Banjit 0.00 0.00 0.93 0.90 0.07 0.10 0.00 0.00 Way Kanan Baradatu 0.00 0.00 0.93 1.00 0.07 0.00 0.00 0.00 Pesawaran Padang Cermin 0.02 0.14 0.83 0.55 0.13 0.32 0.02 0.00 Pesawaran Kedondong 0.02 0.00 0.83 0.71 0.13 0.19 0.02 0.10 Central Java Wonosobo Kaliwiro 0.28 0.38 0.00 0.00 0.72 0.62 0.00 0.00 Blora Japah 0.04 0.00 0.85 1.00 0.11 0.00 0.00 0.00 Rembang Kragan 0.12 0.04 0.63 0.63 0.25 0.33 0.00 0.00 Rembang Sluke 0.12 0.00 0.63 0.50 0.25 0.50 0.00 0.00 Pati Margoyoso 0.12 0.00 0.72 0.95 0.15 0.05 0.00 0.00 Pemalang Warungpring 0.05 0.00 0.69 0.83 0.26 0.17 0.00 0.00 Tegal Bojong 0.05 0.06 0.81 0.18 0.14 0.76 0.00 0.00 East Nusa Kupang Amarasi 0.03 0.00 0.63 0.56 0.33 0.44 0.00 0.00 Tenggara Kupang Kupang Timur 0.03 0.00 0.63 0.85 0.33 0.15 0.00 0.00 Kupang Amabi Oefeto 0.03 0.00 0.63 0.57 0.33 0.29 0.00 0.14 Timor Tengah Selatan Amanuban Selatan 0.00 0.00 0.13 0.40 0.86 0.60 0.01 0.00 Timor Tengah Utara Noemuti 0.11 0.00 0.28 0.45 0.59 0.55 0.02 0.00 West Landak Menyuke 0.00 0.00 0.08 0.06 0.90 0.94 0.01 0.00 Kalimantan Pontianak Toho 0.00 0.00 0.19 0.63 0.81 0.38 0.00 0.00 Sintang Kayan Hulu 0.04 0.00 0.31 0.00 0.65 1.00 0.00 0.00 Kubu Raya Telok Pa’kedai 0.00 0.00 0.01 0.07 0.98 0.93 0.01 0.00 Singkawang Singkawang Timur 0.32 0.60 0.14 0.20 0.54 0.20 0.00 0.00 Singkawang Singkawang Utara 0.32 0.00 0.14 0.00 0.54 1.00 0.00 0.00 South Sulawesi Bone Lappariaja 0.09 0.00 0.66 0.89 0.25 0.11 0.00 0.00 Bone Amali 0.09 0.00 0.66 0.33 0.25 0.67 0.00 0.00 Luwu Larompong Selatan 0.03 0.00 0.72 0.80 0.25 0.20 0.00 0.00 Luwu Lamasi 0.03 0.00 0.72 1.00 0.25 0.00 0.00 0.00 Tana Toraja Sopai 0.09 0.00 0.27 0.00 0.61 1.00 0.02 0.00 Source: Podes 2008 We can immediately observe that across the five provinces, PAM coverage was very low, even in Java. Except in Wonosobo, Rembang, and Pati in Central Java, and Singkawang in West Kalimantan, the district average for the share of villages with PAM connections in the kecamatan was less than 10%. These shares were generally even lower for our sample kecamatan: out of 29 sample kecamatan, they were only higher in five: Kalianda and Padang Cermin in Lampung, Kaliwiro, and Bojong (barely) in Central Java, and Singkawang Timur in West Kalimantan. In the rest, the coverage is equal to—i.e. at zero—or lower than the district average. Most villagers in our sample kecamatan rely on either ground water or other natural sources of water for drinking and cooking. 17 Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia In Lampung, ground water, accessed through a well or pump, is generally the main source of water. The district average of the share of villages in each kecamatan using ground water is between 83% and 96%. Some of our sample kecamatan, however, have less reliance on ground water and more on other natural sources than their district averages. Kecamatan Kalianda, Padang Cermin, and Kedondong reported shares of 63%, 55%, and 71% respectively, lower than their district averages of 83%. Meanwhile in Java, though in most districts villages still mainly use ground water as their source for cooking and drinking, there is a larger share of villages that use other natural sources than in Lampung. The exception is in Wonosobo district, where villages do not use ground water for cooking and drinking, and instead either use PAM or other natural sources of water. In about half of our sample, kecamatan have a higher share of villages using ground water than the district average. In the other half, more villages in our sample kecamatan use other natural sources than ground water compared to the district average. In East Nusa Tenggara, outside of Kupang district, villagers rely much more on other natural sources of water. Our sample kecamatan appear to have greater access to ground water than their district averages. However, even there, less than half can rely on ground water. In Kupang, about two-thirds of villages rely on ground water, and only about one-third have to rely on the other natural sources. In West Kalimantan, the majority of villages have to rely on natural sources of water other than manually-extracted ground water. Except in Toho and Singkawang Timur, our sample kecamatan rely very heavily on natural sources of water. In Toho, a majority of villages make use of ground water, while in the more urban Singkawang Timur, 60% of villages have access to PAM sources. Meanwhile, half of our sample kecamatan in South Sulawesi have more access to manually-extracted ground water than the district average. Except in Tana Toraja, villages in districts where our samples are located mainly use ground water for cooking and drinking. In our sample kecamatan Sopai in Tana Toraja, all villages have to rely on other natural sources for water. In sum, with some minor exceptions, our sample villages are mainly located in the poorest 40% kecamatan nationally and provincially. With regards to infrastructure quality, compared to their district averages, these kecamatan have slightly greater shares of villages with a majority of asphalt roads, slightly lesser access to PAM and ground water compared to their district averages. 3.2 Village Characteristics In this study, we considered topographical conditions as one of the criteria for village selection. Within each province, we selected more or less equal numbers of villages that are located either in the coastal, flatland, or hills/mountainous areas. Table 7 below summarizes the sample villages by topography. Overall, the final sample slightly oversampled villages located in flatland areas. The final sample under-sampled villages in coastal area in South Sulawesi, flatland area in West Kalimantan, and hills/mountainous area in Lampung and South Sulawesi. 18  Sample Village Characteristics Table 7: Sample villages by topography Province Coast Flatland Hills/Mountns Total Lampung 2 3 1 6 Central Java 3 2 3 8 East Nusa Tenggara 2 2 2 6 West Kalimantan 2 1 3 6 South Sulawesi 1 4 1 6 Total 10 12 10 32 Source: Podes 2008, own calculations The sample villages have sizes between 230 and 669 households, with almost half (15 villages) with more than 500 households. While the mean number of households in the villages (459.5 households) is much lower than the national average based on Podes 2008 data (801.9), the median of the number of households in the sample (479.5 households) is comparable to that at the national level (472 households).5 We also compare village density, which we calculated as the number of households per hectare. The average village density in our sample (0.87 households per hectare ) is much lower than at the national level (3.21 households per hectare). The median density in our sample (0.65 households per hectare) is slightly lower than that nationally (0.76 households per hectare). Village Maintenance Practices Next, we examined the maintenance practices of these villages. To this end, we asked village informants (typically the village head) whether there had been any infrastructure activity in the village in the past 12 months. For each type of infrastructure, we then asked what kind of maintenance activities were done and the locations (for up to three locations for each type of infrastructure). We then classified these answers into the different types of maintenance activities. Table 8 summarizes maintenance activities for different types of infrastructure across our sample. It also categorizes the different types of activities as routine maintenance (e.g., light clean-up of roads, removal of weeds, etc.) or periodic maintenance (e.g., asphalting, filling in potholes, replacing water pipes). Otherwise, the maintenance activities might be considered a sustainable upgrade (i.e., a minor or major upgrade of the infrastructure, e.g., concrete reinforcement). Routine maintenance tends to be the least costly, and the upgrades tend to be the most costly. Table 8 shows that many villages perform their own maintenance activities. As shown above, out of 32 villages, 22 villages performed some activities to maintain their roads. These activities are not limited to the more routine (and least costly) type of maintenance, namely routine maintenance. Three villages performed routine maintenance, 8 performed periodic maintenance and 8 performed sustainable upgrades. In 4 villages, the informant did not elaborate on the types of maintenance 5 Before we implemented the first wave of interviews, we conducted a census of the sample villages. The numbers of households in these villages based on our census are comparable to those of Podes 2008. 19 Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia activities performed, in part because some of these works were either performed by private or some government contractors. Note here that since we asked for three locations for each infrastructure, a village could have performed more than one type of maintenance activities. Table 8: Number of villages with maintenance activities and upgrades in the last 12­ months Total w/ Maintenance Activities Type Upgrade Periodic Routine infrastructure* (last 12 mo) Roads 32 22 8 8 3 Bridges 21 7 3 3 0 Water systems** 13 7 1 5 0 Source: VRRI Village data; * VRRI infrastructure survey. **The total number in this row (based on the VRRI Infrastructure Survey) only accounts for villages with piped water system. Table 9: Number of villages with maintenance activities in the last 12 months by province Maintenance Total w/ Activities Upgrade Periodic Routine infrastructure* (last 12 mo) Roads Lampung 6 5 2 0 2 Central Java 8 6 0 4 2 East Nusa Tenggara 6 2 0 1 1 West Kalimantan 6 3 1 0 2 South Sulawesi 6 6 0 3 1 Bridges Lampung 4 1 0 1 0 Central Java 5 2 0 1 1 East Nusa Tenggara 2 2 0 0 2 West Kalimantan 6 1 0 1 0 South Sulawesi 4 1 0 0 0 Water systems** Lampung 2 1 0 1 0 Central Java 5 2 0 2 0 East Nusa Tenggara 3 2 0 1 1 West Kalimantan 1 1 0 1 0 South Sulawesi 2 1 0 0 0 Source: VRRI Village data; * VRRI infrastructure survey. Note: **The total number in this row (based on the VRRI Infrastructure Survey) only accounts for villages with piped water system. 20  Sample Village Characteristics Meanwhile, out of 21 villages with bridges, only a third (or seven villages) performed some kind of maintenance in the last 12 months. Three villages performed activities that would fall under upgrades (strengthening the concrete), while the other three performed periodic maintenance (e.g., replacement of the bridge planks). For water systems, 7 villages out of 13 with piped systems performed mainly periodic maintenance of pipe replacements. Table 9 disaggregates the data by province. Most of our sample villages in Lampung, Central Java, and South Sulawesi performed some kind of road maintenance. About half of the sample villages in West Kalimantan did so, and only a third in East Nusa Tenggara. Meanwhile, in the past 12 months, bridges did not get maintained very often—except in East Nusa Tenggara, where two out of two villages with a bridge perform some kind of maintenance, albeit only routine. Table 10: Implementer of maintenance activities and upgrades Upgrade Periodic Routine Sub-total All Villagers only 19 22 9 50 Villagers and government 1 1 0 2 Villagers and an infrastructure organization 1 1 0 2 Infrastructure organization only 0 3 0 3 Government 0 0 0 0 Others 1 3 0 4 Subtotal 22 30 9 61 By Province: Lampung Villagers only 4 2 6 12 Villagers and an infrastructure organization 1 0 0 1 Infrastructure organization only 0 2 0 2 Central Java Villagers only 5 12 0 17 Infrastructure organization only 0 1 0 1 East Nusa Tenggara Villagers only 5 1 0 6 Villagers and an infrastructure organization 0 1 0 1 Others 1 0 0 1 West Kalimantan Villagers only 5 1 3 9 Government 0 0 0 0 Others 0 3 0 3 South Sulawesi Villagers only 0 6 0 6 Villagers and government 1 1 0 2 Total 22 30 9 61 Source: VRRI village data 21 Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia Next, we examine who performed these maintenance activities as well as the source of resources to perform these activities. Table 10 and Table 11 summarize, respectively, the answers of the village informants regarding the implementers of and the source of resources for these activities. We counted all maintenance activities that are listed by the village informants and that included explanations of maintenance performed. A glance at Panel A of Table 10 shows that villagers were overwhelmingly responsible for these maintenance activities. In 82% of all maintenance activities, villagers were solely responsible for performing maintenance. In an additional 6.7% of these activities, villagers work together with other institutions to perform maintenance. The pattern is similar across provinces. It is also true for different types of maintenance activities—with 100% of routine maintenance being done by villagers only. The picture is very different for maintenance resources. Table 11 summarizes the source of maintenance resources, grouped by infrastructure type and provinces. The table suggests that villagers are still the most significant contributors to these activities. For instance, let us look at Table 11: Sources of resources for infrastructure maintenance, by infrastructure type and province Upgrade Periodic Routine Total Upgrade Periodic Routine Total By Infrastructure By Province Roads Lampung Villagers only 10 8 6 24 Villagers only 5 2 3 10 Villagers & district gov. 0 1 0 1 Villagers & district Villagers & private 0 1 0 1 0 0 3 3 government sector Villagers & private Villagers & others 1 1 0 2 0 0 3 3 sector Central government 1 0 0 1 Infrastructure District government 3 1 0 4 0 1 0 1 organization Donor organization 3 3 0 6 Central Java Others 0 2 0 2 Villagers only 0 7 0 7 Subtotal 18 16 9 43 Villagers & district 0 1 0 1 government Bridges Central government 1 0 0 1 Villagers only 0 4 0 4 District government 3 0 0 3 District government 1 0 0 1 Subtotal 3 4 0 7 Donor organization 3 3 0 6 Others 0 2 0 2 Water East Nusa Tenggara Villagers only 0 4 0 4 Villagers only 3 0 0 3 Villagers & district gov. 0 4 0 4 District government 3 1 0 4 Infrastructure 0 1 0 1 Donor organization 0 1 0 1 organization West Kalimantan District government 1 0 0 1 Donor organization 0 1 0 1 Villagers only 2 1 3 6 Subtotal 1 10 0 11 Villagers & district 0 3 0 3 government Source: VRRI village data District government 3 0 0 3 South Sulawesi Villagers only 0 6 0 6 Villagers & others 1 1 0 2 Total all provinces 22 30 9 61 22  Sample Village Characteristics the example of road maintenance on Panel A. In 55.8% of these maintenance activities villagers themselves provide the resources. In an additional 14% of the cases villagers bear part of the cost, while either government or private agencies provide the rest. However, in about 30% of the cases, these activities are funded fully by outside sources. As expected, activities that were fully resourced by outside institutions are sustainable and periodic maintenance activities, which can become too costly for these villagers (see Table 27). About 39% of all sustainable maintenance activities and 37.5% of periodic maintenance are fully resourced from outside these villages. Meanwhile, all routine maintenance activities are fully resourced by villagers. What about the role of district governments? To what extent do they participate in supporting village infrastructure-maintenance efforts, particularly the more expensive periodic type? Table 11 suggests that district governments tend to be more actively involved in supporting upgrades rather than maintenance. Out of 22 upgrade activities, district governments fully financed 7 of them, while it only fully financed 1 out of 30 periodic maintenance activities and provided partial resources for 5 periodic activities. Given that, in the long term, investment in periodic maintenance tends to bring a higher return than upgrades, there appears to be a need for district governments to reallocate some of its resources towards helping villagers with periodic maintenance. For the maintenance of bridges, all sustainable maintenance are fully resourced from outside, while periodic maintenance—which, in this case, refers only to replacing bridge planks—are all fully resourced by villagers. Meanwhile, 27% of maintenance work on water facilities were fully resourced by outside sources. Meanwhile, Panel B of Table 11 shows how these capabilities to provide for maintenance activities vary across province. There were comparable number of maintenance activities in our sample villages in both Lampung and Central Java; however, those in Lampung were more likely to resource their maintenance activities on their own. These can reflect either financial capabilities, access to outside resources (which are more likely to be available in Central Java), or both. Meanwhile, maintenance activities in South Sulawesi were either fully or partly resourced by villagers. In West Kalimantan, apart from routine resources, about a third of the maintenance activities were fully resourced by villagers, a third were partly resourced by them, and the remaining third were fully funded by outside resources. Our sample villages in East Nusa Tenggara appeared to be the least capable; they have the least number of maintenance activities and these were mostly resourced by outside sources. Table 12 tabulates the number of villages by the types of contributions villagers make for each of the maintenance activities summarized in the previous two tables. Panel A shows the frequencies of combinations of the different forms of villager contributions. Meanwhile Panel B uses information in Panel A to examine the extent to which villagers provide the different forms of contributions individually. Overall, in 81% of the cases, villagers contribute some labor; in some 43% of the cases, they contribute some money. For sustainable maintenance activities, labor is the main form of contribution from these villagers. Labor is still the most important form of contributions for periodic maintenance, followed by labor and material. In the case of routine maintenance, villagers contribute both money and labor in all cases. Table 13 breaks down the types of contributions by the types of infrastructure projects. 23 Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia Finally, we also inquired whether villages have community organizations to manage the different types of infrastructure. Only a small number of villages have any kind of infrastructure-management organization: of 32 villages, 4 villages had organizations for water, 1 village for roads, and none for bridges. Table 12: How villagers contribute by types of maintenance activities Contributions Upgrade Periodic Routine Unclassified Total Money only 0 4 0 0 4 Money and labor 1 9 6 0 16 Money, labor, material, and snacks 1 2 0 1 4 Money, labor, and snacks 2 0 3 0 5 Money and snacks 0 1 0 0 1 Labor only 13 5 0 0 18 Labor and material 1 3 0 0 4 Labor, material and snacks 0 5 0 0 5 Labor and snacks 3 0 0 2 5 Not applicable 1 1 0 6 8 Subtotal 22 30 9 9 70 Villagers contribute some: Money 4 16 9 1 30 Labor 21 24 9 2 56 Material 2 10 0 2 14 Snacks 6 8 3 3 20 Source: VRRI village data Table 13: How villagers contribute by types of infrastructure Contributions Road Bridge Water System Total Money only 0 0 4 4 Money and labor 10 3 3 16 Money, labor, material, and snacks 3 0 1 4 Money, labor, and snacks 5 0 0 5 Money and snacks 0 1 0 1 Labor only 17 1 0 18 Labor and material 2 0 2 4 Labor, material and snacks 5 0 0 5 Labor and snacks 3 1 1 5 Not applicable 5 2 1 8 Subtotal 50 8 12 70 Villagers contribute some: Money 18 4 8 30 Labor 45 5 7 57 Material 10 0 3 13 Snacks 16 2 2 20 Source: VRRI village data 24 Chapter 4 Household Characteristics: Income, Expenditure, and Willingness to Pay In this section, we examine the three characteristics of the households that are relevant to answering our research questions: income, expenditure, and willingness to pay. We use income as a rough measure of resource inflows into households in these villages. However, in rural areas, fluctuation tends to be high. As such, measured income will include the transitory components, and hence, may not accurately reflect household welfare. We use expenditure as a measure of overall welfare of the households. In addition, we also include a discussion of households’ responses to the willingness to pay (WTP) questions at the end of the section. Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia 4.1 Income Before we begin with our discussion of income, we need to start with a caveat with regard to income measurement. Deaton (1997) has argued that income measurements are rife with problems. In general, recall bias, seasonality, imputations when the prices of goods and services are not readily available, and biases coming from questionnaire design—problems that are associated with expenditure questionnaire—tend to affect the precision and accuracy of income questions more than they do expenditure. In addition, income is a more sensitive subject, and households may have an incentive to bias their answers downward. Also, accurate income measurements require knowledge of the returns of different assets; again, a sensitive subject about which respondents may not have an incentive to answer truthfully. Furthermore, measurements of self-employment incomes are complicated since households do not necessarily account money inflows and outflows, as well as inventory, carefully enough to allow the researchers to measure the true income precisely. Measurement errors are likely to be significant. The fact that this survey is a quarterly survey may address some of the issues related to seasonality and recall bias, albeit imperfectly. However, there are still a myriad of other issues in measuring income that could not be addressed satisfactorily.6 Hence, we consider the income measure as a rough estimate that will allow us to study the inflow and fluctuations of resources into these households over the year and across various seasons. This survey considers four types of income: non-business income (i.e., incomes from assets), salary, agriculture self-employment income, and non-agriculture self-employment income. For both agriculture and non-agriculture self-employment income, at every survey wave, we asked about their revenue, costs, as well as existing inventory. However, because we are interested mainly in the inflow of resources, we excluded business inventories in our income calculations in this section. Table 14 shows our calculations of monthly per-capita income of households aggregated at the province level for the sample villages. The average income in our sample was highest in Lampung, followed by Central Java (except in the first survey wave), West Kalimantan, and South Sulawesi. In addition, Figure 1 plots income across the different waves in these provinces. A casual observation of Figure 1 suggests that income fluctuation was strongest in Lampung, the province with the highest income, followed by Central Java. In these provinces, there did not seem to be a clear trend of income. Meanwhile, there were very mild income fluctuations among the remaining three provinces. However, in both West Kalimantan and South Sulawesi, income appeared to trend downward slightly across the four quarters between August 2008 and July 2009. Fluctuations were relatively mild in East Nusa Tenggara, where the average income was lowest among the provinces. 6 A comparison between the (typically more reliable) measurements of per-capita expenditure (Table 17) with income (Table 14) suggests that our income measures may have been somewhat of an underestimate. However, in comparing the two, note that our income measures in Table 14 did not include agriculture and non-agriculture inventories at the end of the fourth wave, net transfers, or loans. 26  Household Characteristics: Income, Expenditure, and Willingness to Pay Table 14: Monthly per-capita income by province, Aug 2008-Jul 2009 Wave Province 1 2 3 4 Lampung 363,503 762,912 385,928 530,932 Central Java 156,837 355,085 266,402 335,364 East Nusa Tenggara 148,253 165,488 155,376 157,351 West Kalimantan 291,134 266,955 239,099 245,592 South Sulawesi 225,091 198,827 161,845 159,307 Figure 1: Monthly per-capita income by province, To examine the source of Aug 2008 - Jul 2009 income fluctuations, we decompose incomes in these 700,000 provinces according to their 600,000 sources—namely, whether 500,000 Lampung Central Java they come from non-business Monthly PCI 400,000 East Nusa Tenggara income, salary, agriculture, West Kalimantan 300,000 South Sulawesi or non-agriculture self 200,000 employment. Panels A to E of 100,000 Figure 2 show the contribution of different 1 2 3 4 Wave sources on income at the province. In general, we observe that salary and non-business income do not significantly contribute to income fluctuations in these provinces; instead, self-employment income appears to be the more important source of fluctuations over the year. In Panel A, we see that non-agriculture income was the main source of income fluctuations in the province. In particular, there was an upward spike of non-agriculture income in the second wave of the survey, corresponding to the beginning of the Idul Fitri holiday, where many traders typically enjoy income boosts, before it stabilized in the following waves. At the same time, agriculture income appeared to have fallen significantly in waves 2 and 3 before picking up in wave 4. We observe from Panel B above a somewhat similar pattern in Central Java, where non-agriculture income was up between waves 1 and 2, although it tended to be more stable across the last three waves compared to that in Lampung. Non-agriculture self-employment income fell slightly between waves 2 and 3, but picked up in the following wave. However, unlike in Lampung, self- employment agriculture income in Central Java tends to be more stable across the four waves. Meanwhile, Panel D shows that in West Kalimantan, non-agriculture self-employment income did not play as significant role as the source of income compared to that in Central Java and Lampung. Instead, on average, salary was the dominant source of income in the province. As we have discussed above, there appeared to be a slight downward trend in income in West Kalimantan. This trend came mainly from falling agriculture self-employment income. The trend began in wave 2 and continued in wave 3, and stabilized at that level in wave 4. As we discuss below, a negative shock to commodity prices that began in wave 2 was responsible for the fall in agriculture income in the 27 Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia Figure 2: Monthly per-capita income flow by income source and province, Aug 2008 – July 2009 A. Lampung B. Central Java 800000 400000 700000 350000 600000 300000 500000 250000 Monthly PCI Monthly PCI 400000 200000 300000 150000 200000 100000 100000 50000 0 0 1 2 3 4 1 2 3 4 Wave Wave 1. Non-business income 3. Agriculture 1. Non-business income 3. Agriculture 2. Salary 4. Non-agriculture 2. Salary 4. Non-agriculture C. East Nusa Tenggara D. West Kalimantan 400000 400000 350000 350000 300000 300000 250000 Monthly PCI 250000 Monthly PCI 200000 200000 150000 150000 100000 100000 50000 50000 0 0 1 2 3 4 1 2 34 Wave 1. Non-business income Wave 3. Agriculture 2. Salary 4. Non-agriculture 1. Non-business income 3. Agriculture 2. Salary 4. Non-agriculture E. South Sulawesi 400000 350000 300000 250000 Monthly PCI 200000 150000 100000 50000 0 1 2 3 4 Wave 1. Non-business income 3. Agriculture 2. Salary 4. Non-agriculture Source: VRRI Household data province. In Panel E, we observe that non-agriculture income was the largest source of income, followed by salary and agricultural income in South Sulawesi. As in West Kalimantan, there was a slight downward trend in income, which mostly came from fluctuations in agriculture income. In Panel C, we observe that there was little fluctuation of overall income in East Nusa Tenggara. Of the various sources of income, salary was the most important. Agriculture income was the second most significant source of income in the first two waves, while non-agriculture income became important in the last two waves, compensating for the fall in the agriculture income during the last two waves. 28  Household Characteristics: Income, Expenditure, and Willingness to Pay Table 15: Average and CV of monthly per-capita income, by province and income source Wave CV 1 2 3 4 Lampung Non-business income 69,998 98,739 75,505 87,189 0.15 Salary 81,319 82,810 81,120 88,818 0.04 Agriculture 119,784 58,019 22,635 109,295 0.59 Non-agriculture 92,403 523,344 206,668 245,631 0.68 Central Java Non-business income 55,586 104,823 76,548 88,677 0.25 Salary 70,253 81,247 71,560 89,934 0.12 Agriculture 27,110 44,334 29,773 40,684 0.23 Non-agriculture 3,888 124,682 88,521 116,068 0.66 East Nusa Tenggara Non-business income 26,180 31,639 39,413 43,156 0.22 Salary 51,820 48,334 49,822 51,303 0.03 Agriculture 42,964 57,613 31,024 26,629 0.35 Non-agriculture 27,289 27,902 35,118 36,263 0.15 West Kalimantan Non-business income 29,141 21,554 26,947 38,661 0.25 Salary 105,673 112,161 117,516 101,514 0.06 Agriculture 103,703 78,498 50,197 56,318 0.34 Non-agriculture 52,617 54,742 44,438 49,098 0.09 South Sulawesi Non-business income 76,166 78,284 58,050 41,498 0.27 Salary 163,975 139,102 139,292 130,193 0.1 Agriculture 192,648 119,488 86,311 148,366 0.33 Non-agriculture 242,485 259,608 201,881 157,865 0.21 Source: VRRI Household data, author’s own calculations To examine further the role of each of the income sources to total income fluctuation in each of these provinces, Table 15 presents the average and coefficient of variation of these averages across the four interview waves, disaggregated by the different income sources. When we examine the coefficient of variation across these interview waves, we observe that the mean non-agriculture self-employment income has the highest variation in Lampung and Central Java. In the remaining three provinces, agriculture income has the highest variation. Among the provinces, agriculture income fluctuation was highest in Lampung and lowest in Central Java. 29 Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia Table 16: Number of households reporting experiencing shock by source and province, Aug 2008-Jul 2009 Wave Wave 1 2 3 4 1 2 3 4 Lampung (n = 670 households) West Kalimantan (n = 696 households) Death in the household 6 0 0 0 Death in the household 9 8 6 3 Illness in the household 21 11 12 8 Illness in the household 40 21 15 28 Loss of employment of Loss of employment of 12 7 7 4 22 15 13 10 HH member(s) HH member(s) Natural disasters 0 0 1 5 Natural disasters 1 11 17 5 Crop failure 114 78 33 33 Crop failure 10 40 55 143 Product/commodity price Product/commodity price 13 42 30 7 5 345 356 54 shocks shocks Climate-related shocks 49 71 25 3 Climate-related shocks 142 372 243 115 Others 4 2 3 6 Others 22 5 5 12 Central Java (n = 939 households) South Sulawesi (n = 718 households) Death in the household 9 4 4 4 Death in the household 16 1 4 2 Illness in the household 31 22 12 21 Illness in the household 45 16 4 10 Loss of employment of Loss of employment of HH 19 4 3 8 15 4 1 1 HH member(s) member(s) Natural disasters 2 0 6 3 Crop failure 91 41 96 29 Product/commodity price Crop failure 71 31 51 82 14 2 4 4 shocks Product/commodity price 5 2 6 1 Climate-related shocks 37 13 35 22 shocks Climate-related shocks 20 12 14 10 Others 16 10 5 9 Others 19 15 23 45 East Nusa Tenggara (n = 698 households) ALL (n = 3721 households) Death in the household 5 1 3 4 Death in the household 45 14 17 13 Illness in the household 5 3 4 2 Illness in the household 121 62 35 61 Loss of employment of HH Loss of employment of HH 0 2 0 1 68 32 24 24 member(s) member(s) Natural disasters 1 1 0 0 Natural disasters 4 12 24 13 Crop failure 7 9 25 30 Crop failure 293 199 260 317 Product/commodity price Climate-related shocks 0 1 11 0 37 391 396 66 shocks Others 1 0 0 0 Climate-related shocks 248 469 328 150 Others 62 32 36 72 Source: VRRI Household data, author’s own calculations. What accounted for income fluctuations in these provinces throughout the year? Table 16 presents parts of the answer. The table summarizes responses to a question that asked respondents whether their households experienced some kind of negative economic shocks during the three months 30  Household Characteristics: Income, Expenditure, and Willingness to Pay prior to the interview. The most significant negative economic shocks overall are agriculture in nature—which explained how falling agriculture income appeared to contribute to falling income in several of these provinces, mainly in the second and third waves. In the first wave, crop failure and climate-related shocks were the factors most often cited by respondents. Crop failure was cited across provinces, with the highest incidence reported in Lampung, followed by South Sulawesi, and Central Java. Meanwhile, the highest incidence of climate-related shocks were reported by West Kalimantan, followed by Lampung in a distant second. However, in the second and third wave, there were significant jumps in the incidences of product/ commodity price shocks and climate-related shocks. A closer look suggests that a disproportionate share of the former incidences were reported by West Kalimantan households. Enumerators in the province reported that falling world commodity prices, particularly rubber, due to the economic crisis had impacted many households there. A similar plight was experienced by producers of cacao and rubber in Lampung. However, there appeared to be more diversity among agricultural producers in Lampung compared to West Kalimantan. Crop failures played an important role in creating income fluctuations all across the provinces, although their contributions, while not unimportant, were not as dramatic as those of falling commodity prices in West Kalimantan. 4.2 Expenditures General expenditures We next examine the average economic welfare of villagers in our sample villages. For our measure of welfare, we use households’ monthly per-capita expenditures. Table 17 below summarizes the mean nominal monthly per-capital expenditures in our sample villages across the four survey waves. The data show the different levels of development (and prosperity) between the different regions in our sample, despite the fact that they are mostly located in the 40% poorest kecamatan in their respective provinces. Lampung and Central Java were the two most prosperous regions, followed by West Kalimantan. South Sulawesi and East Nusa Tenggara were at the bottom. The ranking was consistent with our income ranking, although when we compare the monthly per- capita expenditure with income, it appears that our measures underestimate the true income in these villages. Table 17: Monthly nominal per-capita expenditure in sample villages, Aug 2008- Jul 2009 Wave 1 Wave 2 Wave 3 Wave 4 Lampung 438,853 552,507 418,449 459,663 Central Java 433,802 486,662 393,640 430,832 East Nusa Tenggara 220,280 235,037 242,016 217,383 West Kalimantan 398,863 357,731 282,099 396,957 South Sulawesi 311,596 307,098 237,887 226,432 Source: VRRI household data, own calculations 31 Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia Figure 3: Monthly nominal per-capita expenditure, Throughout the year, the Aug 2008-Jul 2009 various regions experienced 600,000 different cycles that affected their consumption. Figure 3 500,000 shows how monthly per- capita expenditures fluctuate 400,000 Lampung over the year. In Lampung 300,000 Central Java and Central Java, household East Nusa Tenggara West Kalimantan per-capita expenditure went 200,000 South Sulawesi up in the second wave of the 100,000 survey, before falling back to around the level found in the 0 first wave. The second wave Wave 1 Wave 2 Wave 3 Wave 4 of the survey was collected in the three months beginning in November 2008, and since the expenditure questionnaire asked about expenses in the previous month, for households that were interviewed in November, the second wave of survey therefore captured the expenditures of some of the households during the Idul Fitri holiday, which fell on October 1, 2008 However, we did not find a similar pattern in West Kalimantan where per-capita expenditure fell during waves 2 and 3, and in South Sulawesi where per capita consumption fell in waves 2, 3, and 4. As we have mentioned in our discussion on income, the sharp fall in the price of rubber may have significantly affected purchasing power in these villages. Meanwhile, no significant fluctuation was found among the villages in East Nusa Tenggara. The sharpest drop between the first and fourth wave was observed in South Sulawesi. Table 18 provides a more detailed breakdown of per-capita expenditures in each of the sample villages. Public goods and infrastructure expenditures In a later section, we will explore the extent to which infrastructure maintenance costs may act as an additional tax burden to villagers. However, before we consider how those costs become an additional burden, it may be useful to look at the extent to which villagers are paying local taxes. We ask respondents about their expenditures on local public goods, such as trash, road, bridge, and water system maintenance, and neighborhood collections (iuran RT). Table 19 summarizes the average monthly expenditure on these local taxes, as well as their share as a percentage of consumption. Overall, they are insignificant, amounting to less than 0.1% of our sample households’ consumption in all bar one village (Village 24). 32  Household Characteristics: Income, Expenditure, and Willingness to Pay Table 18: Monthly nominal per-capita expenditure, Aug 2008-Jul 2009 Province/Village ID Wave 1 Wave 2 Wave 3 Wave 4 Lampung 1 481,274 478,804 401,906 443,012 2 510,060 730,220 574,854 536,276 3 386,450 470,135 340,528 334,671 4 443,642 469,835 472,248 743,339 5 368,266 595,676 311,696 339,855 6 457,056 579,847 415,409 360,824 Central Java 7 344,925 384,245 315,429 337,970 8 381,873 386,746 353,016 322,305 9 363,906 380,551 337,895 295,574 10 468,015 550,348 403,622 509,002 11 483,269 629,412 478,694 583,137 12 463,749 427,612 358,757 463,171 13 534,568 767,181 528,376 549,621 14 431,972 373,111 373,031 385,880 East Nusa Tenggara 15 225,036 235,357 265,048 204,870 16 212,897 262,554 235,085 235,938 17 212,708 210,642 233,546 194,411 18 206,097 226,708 239,968 215,299 19 195,340 222,129 215,554 197,687 20 269,812 253,096 262,864 256,093 West Kalimantan 21 393,647 224,177 198,791 319,073 22 458,663 455,969 323,629 496,006 23 313,270 282,856 272,574 331,748 24 439,124 536,371 348,736 445,609 25 398,677 278,674 239,005 383,879 26 392,620 374,200 311,288 405,429 South Sulawesi 27 311,888 233,508 230,946 216,611 28 319,889 423,963 254,243 273,565 29 257,508 416,826 276,342 253,050 30 297,559 349,692 254,605 258,345 31 321,000 218,069 211,780 178,087 32 361,802 200,888 199,541 178,932 Source: VRRI data, author’s own calculations 33 Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia Table 19: Monthly per-household local tax and consumption Province vid HH consumption Local tax % of HH Cons. Lampung 1 1,915,696 7 0.00% 2 2,670,972 820 0.03% 3 1,700,188 166 0.01% 4 2,507,650 460 0.02% 5 1,437,218 14 0.00% 6 1,677,492 1,492 0.09% Central Java 7 1,338,415 222 0.02% 8 1,297,386 1,122 0.09% 9 1,168,903 796 0.07% 10 1,568,555 468 0.03% 11 1,946,179 13 0.00% 12 1,465,667 219 0.01% 13 1,979,672 506 0.03% 14 1,314,921 608 0.05% East Nusa Tenggara 15 887,094 19 0.00% 16 1,185,005 0 0.00% 17 1,007,547 8 0.00% 18 883,098 0 0.00% 19 992,353 7 0.00% 20 1,099,967 66 0.01% West Kalimantan 21 1,246,733 0 0.00% 22 1,975,047 0 0.00% 23 1,289,781 58 0.00% 24 1,829,268 2,858 0.16% 25 1,559,003 114 0.01% 26 1,486,823 275 0.02% South Sulawesi 27 1,045,858 0 0.00% 28 1,365,275 0 0.00% 29 1,206,702 160 0.01% 30 1,272,316 0 0.00% 31 1,005,602 0 0.00% 32 1,019,723 0 0.00% Source: VRRI household data, all waves. Another expenditure that is relevant to the infrastructure studied here is households’ current water consumption. Table 20 looks at the monthly per-household water consumption. There is a wide variation in the amount that households spent for water. On average, households spent about Rp2158 to obtain water, with a median of Rp594. In a quarter of the villages, households had no expenditure to obtain water. On average, water expenditure accounts for 0.14% of total households expenditure, with the median at 0.04%. 34  Household Characteristics: Income, Expenditure, and Willingness to Pay Table 20: Monthly per-household water expenditure and consumption Province vid HH consumption Water exp. % of hh cons Lampung 1 1,915,696 4,592 0.24% 2 2,670,972 2,271 0.09% 3 1,700,188 4,696 0.28% 4 2,507,650 4,779 0.19% 5 1,437,218 0 0.00% 6 1,677,492 0 0.00% Central Java 7 1,338,415 288 0.02% 8 1,297,386 2,529 0.19% 9 1,168,903 5,500 0.47% 10 1,568,555 458 0.03% 11 1,946,179 10,408 0.53% 12 1,465,667 3,546 0.24% 13 1,979,672 5,788 0.29% 14 1,314,921 729 0.06% East Nusa Tenggara 15 887,094 46 0.01% 16 1,185,005 83 0.01% 17 1,007,547 4,785 0.47% 18 883,098 58 0.01% 19 992,353 0 0.00% 20 1,099,967 0 0.00% West Kalimantan 21 1,246,733 2,208 0.18% 22 1,975,047 71 0.00% 23 1,289,781 0 0.00% 24 1,829,268 2,583 0.14% 25 1,559,003 0 0.00% 26 1,486,823 8 0.00% South Sulawesi 27 1,045,858 0 0.00% 28 1,365,275 0 0.00% 29 1,206,702 7,696 0.64% 30 1,272,316 3,717 0.29% 31 1,005,602 1,917 0.19% 32 1,019,723 292 0.03% Source: VRRI household data, all waves. 35 Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia 4.3 Willingness to pay Above, we have described the income and expenditure, which together reflect resources that are available to households in these villages. However, in order to answer the question of whether villagers can pay for maintaining the infrastructure in their villages, we need to measure not only resources that are available to households, but also whether villagers would be willing to contribute some of these resources for maintaining their infrastructure. Our study directly addresses this question by asking our survey respondents their willingness to pay for maintaining their infrastructure. Using information elicited by this question, we come up with a profile of villagers’ willingness—and, implicit in that, ability—to pay for maintaining the roads, bridges, and piped water systems in their respective village. To elicit respondents’ willingness to pay, we used an iterative bidding procedure with a randomized starting value. The interviewer begins by describing the quality of a village infrastructure to the respondent. Following a description of the quality of the infrastructure, the interviewer described a hypothetical village meeting in which a decision about contributions to maintain the infrastructure is to be decided. In the case of roads and bridges, villagers may still be asked to provide some labor contributions. Meanwhile, in the question for water system, villagers will not have to contribute additional labor. The interviewer then asks the respondent whether the respondent will be willing to pay a certain amount to maintain the infrastructure. If the respondent answers “Yes”, the interviewer asks the same question using a higher value X + e; otherwise, the interviewer uses a lower value X - e. If the highest value provided in the survey is reached, the interviewer asks the respondent to name a value that he or she is willing to pay to maintain the infrastructure. We used this approach to minimize anchoring bias. Anchoring is the tendency of respondents to answer close to the first value introduced at the start of the question. To minimize this, we randomized the starting value, X, across the respondents. By randomizing the starting points, we expect that the anchoring bias will cancel each other out when we calculate the village averages. We use a value of Rp.1,000 for the increment e, and Rp.15,000 for the maximum value before the interviewer asks the respondent to name his or her willingness to pay. Table 21: Villagers’ average willingness to pay by province Province Road Bridge Water All Lampung 6,061 4,894 5,587 9,797 Central Java 5,497 3,705 4,546 8,379 East Nusa Tenggara 4,558 3,945 5,333 10,319 West Kalimantan 6,399 4,499 9,483 12,911 South Sulawesi 3,562 2,300 4,240 4,794 Source: VRRI data (Wave 2), own calculations 36  Household Characteristics: Income, Expenditure, and Willingness to Pay Table 22: Mean and median household willingness to pay for infrastructure maintenance (Rp/month) Road Bridge Water All Mean Median Mean Median Mean Median Mean Median Lampung 1 5,892 5,000 5,058 3,000 5,617 5,000 9,415 6,000 2 7,420 5,000 5,790 5,000 6,458 5,000 12,283 10,000 3 5,992 5,000 5,083 5,000 6,250 5,000 9,208 8,000 4 5,782 5,000 5,000 3,000 5,800 5,000 10,350 7,000 5 4,714 4,000 3,642 2,000 3,580 3,000 7,538 5,000 6 6,567 5,000 4,800 5,000 5,815 5,000 9,966 8,000 Central Java 7 3,842 3,000 2,408 2,000 3,533 3,000 4,683 4,000 8 5,667 5,000 4,458 3,000 4,600 4,000 7,975 5,000 9 4,217 4,000 2,958 2,000 3,975 3,000 7,483 5,000 10 4,667 3,000 3,383 2,000 3,908 3,000 7,575 5,000 11 6,000 5,000 3,613 2,000 4,286 3,000 8,102 5,000 12 5,758 5,000 3,567 2,000 4,842 4,000 9,025 5,000 13 7,492 5,000 4,933 3,500 6,233 5,000 11,782 10,000 14 6,342 5,000 4,317 3,000 4,983 5,000 10,433 10,000 East Nusa Tenggara 15 5,361 5,000 4,708 5,000 6,125 5,000 10,800 10,000 16 4,895 5,000 3,957 2,000 5,474 5,000 11,145 5,500 17 4,717 2,500 3,975 2,000 4,658 3,000 9,375 5,000 18 4,133 2,000 3,158 2,000 6,117 3,000 12,345 5,000 19 3,975 2,000 3,723 2,000 4,840 2,000 8,699 5,000 20 4,283 2,000 4,150 2,000 4,783 2,000 9,542 5,000 West Kalimantan 21 3,942 3,000 2,202 2,000 7,417 5,000 9,242 7,000 22 8,158 5,000 3,429 3,000 10,575 10,000 14,267 10,000 23 3,042 2,000 2,583 2,000 7,025 5,000 8,658 5,000 24 4,825 3,000 2,750 2,000 10,508 10,000 13,250 11,000 25 10,092 10,000 8,433 7,000 10,563 10,000 14,850 13,500 26 8,333 7,000 7,567 5,000 10,839 10,000 17,235 15,000 South Sulawesi 27 3,333 2,500 2,342 1,000 3,825 3,000 4,175 3,000 28 5,706 5,000 3,292 1,000 5,521 3,000 7,261 5,000 29 2,383 1,000 1,592 1,000 3,899 3,000 4,185 3,000 30 3,500 2,000 2,479 1,000 4,367 3,000 4,733 4,000 31 3,067 2,000 1,625 1,000 3,445 3,000 4,033 3,000 32 3,400 2,000 2,475 1,000 4,383 3,000 4,392 3,000 37 Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia Table 21 summarizes the average (or mean) households’ WTP to maintain different types of infrastructure, broken down by province. For respondents who chose values above Rp.15.000, we included their answer if they answered the open question following the maximum value in our iterative bidding procedure, or Rp.15,000 if the answer was missing. This procedure has the unfortunate effect of underestimating the mean WTPs, and therefore throughout the analysis, we sometimes make use of the median instead of the mean WTPs. The WTP questions were asked to villagers irrespective of the availability of the infrastructure in their village. The first three columns summarize the WTP for each individual infrastructure. Meanwhile, the fourth column lists the WTP if they were to pay a single fee to maintain all kinds of infrastructure. Different provinces exhibit different priorities. In Lampung and Central Java, villagers are most willing to pay for road maintenance followed by water systems. In comparison, in East Nusa Tenggara, West Kalimantan and South Sulawesi, the WTPs were higher for water than for roads. The willingness to pay for maintaining a good water system was very high in West Kalimantan compared to the rest of the provinces, even though its per-capita expenditure was third among the provinces. Our data cannot explain why there was a higher relative WTP for maintaining roads relative to piped water system in Lampung and Central Java, but not in the other three provinces. However, since these two provinces are the ones more integrated to the wider market, one plausible explanation is that the economic returns from roads in these provinces are higher than in the rest of provinces. Generally, villagers are least willing to pay for bridge maintenance—perhaps because in our sample there were more villagers benefiting from roads and water systems. In addition, one thing that was true across the provinces is that the average WTP to finance maintenance of all three maintenance is much lower than the sum of the averages for individual infrastructure. A more detailed breakdown that included both the mean and median household WTP by village is provided in Table 22. The previous tables show us the centered values of the distribution of the households’ WTP for infrastructure maintenance. However, we may want to see the distribution of the WTPs across different values. Panels A, B, C, and D of Table 23 present these values grouped in Rp 5,000 increments. For roads, we can observe in Panel A that overall 10.3% of households had zero WTP. The share of respondents with zero WTP was highest in South Sulawesi and lowest in Lampung. It is interesting to note that even though East Nusa Tenggara is the province with the lowest per-capita expenditure in our sample, the share of households with zero WTP was the second lowest after Lampung. On the other hand, households in Central Java, which is the province with the second highest per-capita expenditure, have a second highest share of households with zero willingness to pay for road maintenance. About 63.1% of the respondents are willing to pay between Rp.1,000-5,000, with Rp.5000 as the mode of the distribution of the WTPs (i.e., the value with the highest number of responses). Hence, almost three-fourths of the respondents overall were not be willing to pay more than Rp.5,000 for road maintenance. The share of households willing to pay up to Rp.5,000 only was highest in South Sulawesi, followed by East Nusa Tenggara, Lampung, Central Java, and West Kalimantan. Meanwhile, about 3.5% of the overall households were willing to pay more than Rp.15,000 a month for road maintenance. Lampung, which is the province with the highest per- capita expenditure in our sample, led with 5.7%, followed, surprisingly, by East Nusa Tenggara. South Sulawesi has the smallest share of households willing to pay more than Rp.15,000. 38  Household Characteristics: Income, Expenditure, and Willingness to Pay Table 23: Distribution of willingness to pay per household per month in all villages, by province and infrastructure East Nusa West South Roads Lampung Central Java All Tenggara Kalimantan Sulawesi Nothing 19 2.6% 119 12.4% 68 9.5% 82 11.4% 108 15.0% 396 10.3% Rp.1000-5000 485 67.4% 547 57.0% 512 71.2% 377 52.4% 499 69.4% 2420 63.1% Rp.6000- 142 19.7% 194 20.2% 82 11.4% 152 21.1% 79 11.0% 649 16.9% 10000 Rp.11000- 33 4.6% 69 7.2% 25 3.5% 85 11.8% 26 3.6% 238 6.2% 15000 >15000 41 5.7% 31 3.2% 32 4.5% 24 3.3% 7 1.0% 135 3.5% Total 720 100.0% 960 100.0% 719 100.0% 720 100.0% 719 100.0% 3838 100.0% East Nusa West South Bridges Lampung Central Java All Tenggara Kalimantan Sulawesi Nothing 66 9.2% 278 29.0% 84 11.7% 148 20.6% 179 24.9% 755 19.7% Rp.1000-5000 505 70.2% 503 52.4% 527 73.2% 410 57.1% 488 67.9% 2433 63.4% Rp.6000- 91 12.7% 128 13.3% 64 8.9% 113 15.7% 38 5.3% 434 11.3% 10000 Rp.11000- 33 4.6% 32 3.3% 23 3.2% 36 5.0% 11 1.5% 135 3.5% 15000 >15000 24 3.3% 19 2.0% 22 3.1% 11 1.5% 3 0.4% 79 2.1% Total 719 100.0% 960 100.0% 720 100.0% 718 100.0% 719 100.0% 3836 100.0% East Nusa West Piped Water Lampung Central Java South Sulawesi All Tenggara Kalimantan Nothing 77 10.7% 257 26.8% 47 6.5% 7 1.0% 91 12.7% 479 12.5% Rp.1000-5000 439 61.0% 454 47.3% 501 69.6% 245 34.1% 489 68.2% 2128 55.5% Rp.6000- 124 17.2% 160 16.7% 100 13.9% 264 36.7% 89 12.4% 737 19.2% 10000 Rp.11000- 42 5.8% 60 6.3% 36 5.0% 164 22.8% 34 4.7% 336 8.8% 15000 >15000 38 5.3% 29 3.0% 36 5.0% 39 5.4% 14 2.0% 156 4.1% Total 720 100.0% 960 100.0% 720 100.0% 719 100.0% 717 100.0% 3836 100.0% All East Nusa West Lampung Central Java South Sulawesi All Infrastructure Tenggara Kalimantan Nothing 19 2.6% 93 9.7% 40 5.6% 7 1.0% 77 10.7% 236 6.2% Rp.1000-5000 285 39.6% 422 44.0% 352 48.9% 181 25.2% 486 67.7% 1726 45.0% Rp.6000- 234 32.5% 233 24.3% 161 22.4% 215 29.9% 102 14.2% 945 24.6% 10000 Rp.11000- 78 10.8% 97 10.1% 57 7.9% 158 22.0% 31 4.3% 421 11.0% 15000 >15000 104 14.4% 115 12.0% 110 15.3% 158 22.0% 22 3.1% 509 13.3% Total 720 100.0% 960 100.0% 720 100.0% 719 100.0% 718 100.0% 3837 100.0% Panel B shows the distribution of households’ WTP for bridges. Close to one-fifth of the households have zero WTPs—almost double the number with zero WTPs for roads. This is not surprising, since for most respondents, roads are likely to be more relevant to their livelihoods than bridges. About 63.4% were willing to pay between Rp.1,000 and Rp.5,000 for bridge maintenance. Overall, less than 20% of households would be willing to pay more than Rp.5,000 to maintain bridges in their villages. 39 Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia As mentioned above, the survey asked the WTP question irrespective of whether the infrastructure referred to in the questions were present in these villages. While this presents no problem with regards to the WTP measures for roads (which are present in all of the villages), it may skew these measures for bridges and water systems if the distribution of responses of households in villages without these types of infrastructure are systematically different from those with the infrastructure. Since our interest is in studying the willingness to pay for maintenance, we are more interested in the latter type of villages. Table 24 presents the distribution of responses from villages with the infrastructure for bridges and water systems. For bridges, there was a higher share of households in villages with these infrastructure who have zero WTP, although there was a slightly lower share with WTP up to Rp.5,000. For water systems, there was a higher share at the lower ends of WTP distribution overall: there was a larger share of households with zero WTPs and with WTPs between Rp.1,000 and Rp.5,000. This seems to suggest that households in villages with the infrastructure tend to state a lower willingness to pay compared to those who were asked to state their WTP based on hypothetical infrastructure. However, this casual observation does not control for the different characteristics between villages with and without bridges or water systems. These characteristics may have been responsible for the difference in the willingness to pay for maintenance between the two groups of villages. Table 24: Willingness to pay per household per month for villages with bridges and water systems E. Nusa West Bridges Lampung Central Java S. Sulawesi All Tenggara Kalimantan Nothing 35 7.3% 176 29.3% 30 12.5% 148 20.6% 153 25.5% 542 20.5% Rp.1000-5000 342 71.3% 310 51.7% 170 70.8% 410 56.9% 409 68.2% 1641 62.2% Rp.6000-10000 65 13.5% 81 13.5% 26 10.8% 113 15.7% 28 4.7% 313 11.9% Rp.11000-15000 21 4.4% 18 3.0% 5 2.1% 36 5.0% 8 1.3% 88 3.3% >15000 17 3.5% 15 2.5% 9 3.8% 13 1.8% 2 0.3% 56 2.1% Total 480 100.0% 600 100.0% 240 100.0% 720 100.0% 600 100.0% 2640 100.0% E. Nusa West Piped Water Lampung Central Java S. Sulawesi All Tenggara Kalimantan Nothing 34 9.4% 126 21.0% 28 7.8% 2 1.7% 31 13.0% 221 13.2% Rp.1000-5000 214 59.4% 334 55.7% 264 73.3% 26 21.7% 171 71.6% 1,009 60.1% Rp.6000-10000 67 18.6% 81 13.5% 36 10.0% 49 40.8% 27 11.3% 260 15.5% Rp.11000-15000 24 6.7% 42 7.0% 17 4.7% 35 29.2% 7 2.9% 125 7.4% >15000 21 5.8% 17 2.8% 15 4.2% 8 6.7% 3 1.3% 64 3.8% Total 360 100.0% 600 100.0% 360 100.0% 120 100.0% 239 100.0% 1,679 100.0% 40 Chapter 5 Village Resources In this section, we try to answer the first of our research questions, namely whether villagers have the resources to finance their own infrastructure. One way to approach this question is to examine how much of an additional burden would transferring the full cost of infrastructure maintenance be to these villagers in terms of the income that they generated. The first and second columns of Table 25 present the annual income and consumption from our household data for these villages. Meanwhile, Panel C of the table presents the annualized periodic and routine costs that were necessary to maintain these villages’ infrastructure properly. Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia Table 25: Annual income, consumption, and total maintenance cost C. Annual periodic and routine cost B. Province vid A. Income % of Consumption Rupiah % of cons income Lampung 1 6,694,273,024 5,080,425,472 60,211,418 0.90% 1.19% 2 14,465,620,992 12,179,630,080 143,965,178 1.00% 1.18% 3 5,315,733,504 6,814,352,896 120,697,529 2.27% 1.77% 4 4,259,000,832 7,402,582,016 49,806,373 1.17% 0.67% 5 7,615,383,552 7,640,249,856 108,729,166 1.43% 1.42% 6 11,645,253,632 5,354,554,368 32,233,364 0.28% 0.60% Central Java 7 2,391,934,976 5,476,795,392 36,307,074 1.52% 0.66% 8 3,029,536,768 6,943,611,392 91,190,771 3.01% 1.31% 9 6,436,700,160 6,802,330,112 113,896,246 1.77% 1.67% 10 5,337,646,080 7,472,593,920 78,715,472 1.47% 1.05% 11 7,118,802,944 9,365,013,504 171,380,245 2.41% 1.83% 12 5,672,835,584 7,158,316,544 60,141,497 1.06% 0.84% 13 11,614,415,872 11,355,395,072 39,645,959 0.34% 0.35% 14 1,612,817,920 2,713,996,800 51,522,781 3.19% 1.90% East Nusa Tenggara 15 3,663,342,592 4,886,112,256 26,060,619 0.71% 0.53% 16 6,076,235,776 6,285,267,968 38,904,618 0.64% 0.62% 17 3,740,581,376 5,211,032,576 74,525,178 1.99% 1.43% 18 2,222,495,744 2,458,543,616 2,517,202 0.11% 0.10% 19 2,346,012,416 2,536,454,656 24,110,973 1.03% 0.95% 20 4,879,995,904 5,148,011,520 59,649,605 1.22% 1.16% West Kalimantan 21 3,382,002,432 4,592,963,072 42,841,475 1.27% 0.93% 22 7,094,712,320 9,314,318,336 111,422,409 1.57% 1.20% 23 5,741,086,208 6,825,518,592 144,811,262 2.52% 2.12% 24 8,371,660,288 8,034,143,744 71,338,695 0.85% 0.89% 25 5,302,328,832 6,790,970,368 26,142,682 0.49% 0.38% 26 2,602,045,440 3,318,588,416 34,300,763 1.32% 1.03% South Sulawesi 27 2,062,812,672 3,175,225,856 88,218,479 4.28% 2.78% 28 8,451,226,624 7,454,402,048 31,067,303 0.37% 0.42% 29 7,706,818,560 7,225,728,512 44,131,120 0.57% 0.61% 30 5,869,850,624 8,687,370,240 22,778,165 0.39% 0.26% 31 3,414,336,000 5,695,731,200 63,002,107 1.85% 1.11% 32 4,482,786,304 5,249,534,464 126,302,889 2.82% 2.41% Source: VRRI data, own calculations. 42  Village Resources Table 26: Annual income, consumption, and maintenance cost excluding unskilled labor C. Annual periodic and routine cost B. Province vid A. Income % of % of Consumption Rupiah income cons Lampung 1 6,694,273,024 5,080,425,472 41,323,041 0.62% 0.81% 2 14,465,620,992 12,179,630,080 82,940,534 0.57% 0.68% 3 5,315,733,504 6,814,352,896 75,281,538 1.42% 1.10% 4 4,259,000,832 7,402,582,016 36,184,727 0.85% 0.49% 5 7,615,383,552 7,640,249,856 24,683,489 0.32% 0.32% 6 11,645,253,632 5,354,554,368 10,431,949 0.09% 0.19% Central Java 7 2,391,934,976 5,476,795,392 26,833,625 1.12% 0.49% 8 3,029,536,768 6,943,611,392 67,339,257 2.22% 0.97% 9* 6,436,700,160 6,802,330,112 10 5,337,646,080 7,472,593,920 62,245,934 1.17% 0.83% 11* 7,118,802,944 9,365,013,504 12 5,672,835,584 7,158,316,544 34,898,724 0.62% 0.49% 13 11,614,415,872 11,355,395,072 29,111,188 0.25% 0.26% 14 1,612,817,920 2,713,996,800 19,616,898 1.22% 0.72% East Nusa Tenggara 15 3,663,342,592 4,886,112,256 12,062,782 0.33% 0.25% 16 6,076,235,776 6,285,267,968 24,392,409 0.40% 0.39% 17 3,740,581,376 5,211,032,576 38,579,555 1.03% 0.74% 18 2,222,495,744 2,458,543,616 1,639,789 0.07% 0.07% 19 2,346,012,416 2,536,454,656 2,117,122 0.09% 0.08% 20 4,879,995,904 5,148,011,520 15,326,924 0.31% 0.30% West Kalimantan 21 3,382,002,432 4,592,963,072 21,698,687 0.64% 0.47% 22 7,094,712,320 9,314,318,336 46,463,799 0.65% 0.50% 23 5,741,086,208 6,825,518,592 98,339,135 1.71% 1.44% 24 8,371,660,288 8,034,143,744 47,479,863 0.57% 0.59% 25 5,302,328,832 6,790,970,368 11,663,339 0.22% 0.17% 26 2,602,045,440 3,318,588,416 21,882,611 0.84% 0.66% South Sulawesi 27 2,062,812,672 3,175,225,856 43,240,819 2.10% 1.36% 28 8,451,226,624 7,454,402,048 16,021,501 0.19% 0.21% 29 7,706,818,560 7,225,728,512 15,592,006 0.20% 0.22% 30 5,869,850,624 8,687,370,240 14,803,946 0.25% 0.17% 31 3,414,336,000 5,695,731,200 33,013,723 0.97% 0.58% 32 4,482,786,304 5,249,534,464 74,036,783 1.65% 1.41% Source: VRRI data, own calculations. * Labor cost data were missing for these villages. 43 Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia Table 25 shows that, if we distribute the cost of maintaining infrastructure equally across households in these villages, it would amount to 0.1% (Village 18) to 4.3% (Village 27) of extra burden—or a “tax” if you will—to these villagers relative to their income, with a median of 1.4%. However, as discussed above in the introduction of the income section, there are problems with the measurement of income. In particular, income reporting tends to be systematically downward biased.7 Since consumption tends to be measured with less bias, we may want to consider examining the additional tax from paying for maintenance cost using our consumption measures. Based on the consumption measures, paying for the full cost of periodic and routine maintenance amounts to a tax of between 0.1% (Village 18) to 2.8% (Village 14) relative to their consumption, with a median of 1.1%. The maintenance costs presented above included costs for both labor and material. One could argue that this would overestimate the true maintenance costs, since as we observe before, in many of the villages with maintenance activities, their main contributions are in the form of labor, especially for routine maintenance activities. Table 26 addresses this issue by excluding unskilled-labor costs from our calculations of the total maintenance costs. If we assume that all unskilled labor are provided by villagers, we find that the cost of infrastructure maintenance will create an additional tax burden of 0.1% (0.1%) to 2.2%(1.4%) with respect to their income (consumption). The median tax from maintenance with respect to their income (consumption) will be 0.6%(0.5%). 7 This problem is clearly illustrated by Table 25, where consumption exceeds income in 25 out of 32 villages by between 0.3% and 129%, with a median of 31.3%. 44 Chapter 6 Empirical Analysis 6.1 Paying for Maintenance: Assessing Different Scenarios With data on maintenance costs and villagers’ willingness to pay, we can now begin to answer the extent to which these villages can maintain their own infrastructure. The answer to this question varies depending on what we assume about the mechanism that we will use to collect the contributions that these villagers claim they are willing to pay. We shall consider two scenarios. First, we shall consider the simplest case where villagers can somehow be induced to voluntarily contribute the full value of their stated willingness to pay. Note, however, that theoretically we do not have a credible resource collection mechanism to allow Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia us to induce households to contribute the full value of their willingness to pay. Therefore, this should be considered the best-case (and hypothetical) scenario amongst voluntary mechanisms. We will examine, under this scenario, the extent to which households are able to pay for maintaining infrastructure. Second, we shall consider the more realistic mechanism of charging a single user fees to all households. Because the survey only collected monthly per-household WTP, our user fees scenario would be based on a monthly collection of user fees for the right to use the infrastructure. We assume that households that do not pay this fee would be cut off from access to the infrastructure. Obviously, this second scenario is a mere analytical tool to get at a more realistic amount that can be collected from villagers. We do not, in any way, suggest that this would be the proper user-fee policy to implement for the various types of infrastructure considered here. For instance, user fees for piped water may be better collected based on usage (instead of a flat fee) depending on the local context. At any rate, we use this scenario to examine two questions. First, how much financial resources would be generated by this scenario? Second, how many households in each village would be excluded from the use of the different types of infrastructure? Scenario 1: Villagers Pay Their Full Willingness to Pay Suppose that villagers pay their stated willingness to pay. How much resources would be available to finance the maintenance of infrastructure? Would these resources be adequate to fund the necessary maintenance to ensure the sustainability of their infrastructure? To answer this question, we use our estimates of the periodic and routine maintenance costs for each village to find the amount that each household would need to contribute on average to maintain the quality of their infrastructure. We then compare this number with the average that each head of household stated he or she was willing to contribute for the infrastructure. Since our sample is representative of the village population, if this willingness to pay is more than the average maintenance cost, it implies that contributions of villagers would be adequate to finance the maintenance costs. Tables 27, 28, and 29 present the average costs that households need to contribute for maintenance and households’ willingness to pay for roads, bridges, and piped water, respectively, across all villages. We consider both the costs that include and exclude those of unskilled labors. We begin with comparing the costs with the separate WTPs for each infrastructure, starting with roads. Based on total maintenance costs, Table 27 shows that the maximum voluntary contributions of villagers would cover costs in only 7 out of the 32 villages surveyed—less then 25% of the villages sampled. However, since we ask respondents their WTPs assuming that they may still be asked to contribute some (unskilled) labor, we need to compare their WTPs with the total cost excluding unskilled labor. In this case, 13 out of 30 villages—or 43% of our sample villages—would be able to finance the material, equipment rental, and skilled labor cost necessary to maintain their roads. Before we proceed, it is important to clarify how we would interpret the results here. Since our WTP question did not include a question on their willingness to supply unskilled labor, it would seem 46  Empirical Analysis that the cost estimates that exclude unskilled-labor cost would be the more accurate ones. However, this may not be true, since for some types of roads, such as telford or macadam roads, periodic maintenance is a very labor-intensive process. For villages with these types of roads, excluding the cost of unskilled labor may reduce the maintenance costs significantly. However, at the same time, excluding all of the unskilled-labor cost, we assume that villagers would be able to supply all of the unskilled labor necessary for this purpose. Our data, however, cannot say whether this assumption is justified. As such, we take the number of villages that are able to self-finance when we compare villager WTPs with the total costs as the lower bound, while that compared to the costs that exclude unskilled-labor costs as the upper bound. In other words, our results show that between about a quarter and slightly more than two-fifths of our sample would be able to finance road maintenance if each household contributes their maximum willingness to pay for road maintenance. Looking across different provinces, we find that when we consider total costs, none of the villages in Lampung, one in eight villages in Central Java and two in six villages each in East Nusa Tenggara, West Kalimantan, and South Sulawesi would have adequate resources to finance road maintenance from their own contributions. However, when we exclude unskilled labor from the total cost, five out of six villages in East Nusa Tenggara would now be able to finance the maintenance of their road from contributions alone. From the data, we find that the majority of villages in East Nusa Tenggara are able to finance road maintenance through the villagers’ own contributions because of the generally lower cost of road maintenance there, due to a combination of lower volume, local prices, and need for a given road type. In addition, two villages in Lampung, and an additional village in South Sulawesi are now able to finance their road maintenance from contributions alone. In the above analysis, we use the average (or mean) of willingness to pay (WTP). Instead of using the mean, we can use the median, which gives us central values that are less sensitive to the effects of outliers. Except for Village 9, the median WTP is lower than its mean. When we use the median, only 4 out of the 32 villages—a mere 13%—would be able to afford the total costs of maintenance, and only 9 out of 30 villages—or 30% of the villages—would be able to afford the cost of maintenance, even if they supplied all of the unskilled labor themselves. Table 28 below presents our results for bridges. In our sample of 32 villages, only 21 villages have permanent bridges that are maintained by the village. Total maintenance costs per households for the bridges in our sample were relatively low compared to roads. As such, using the average WTPs, in 17 out of 21 villages, the voluntary contributions from villagers would be adequate to finance their maintenance. The number of villages able to self-finance their bridge maintenance does not change when we exclude the cost of unskilled labor. Meanwhile, as shown in Table 29, we find that in 11 out of 14 villages with piped water systems, in all bar one villages, voluntary contributions based on the mean WTP would be adequate to finance the total cost of maintenance. When we exclude unskilled-labor costs, in all villages bar one (Village 27), the maximum possible contributions based on the WTPs would be adequate to finance the maintenance of the water systems. 47 Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia Table 27: Maintenance costs and willingness to pay per-household: Roads Costs excluding Willingness All costs unskilled labor to pay** Province vid Periodic Periodic Periodic Routine Periodic Routine Mean Median +Routine +Routine Lampung 1 14,922 1,297 16,220 12,960 344 13,305 5,892 5,000 2 16,641 1,400 18,041 13,934 108 14,042 7,420 5,000 3 20,300 4,640 24,940 14,483 1,160 15,644 5,992 5,000 4 7,346 6,432 13,778 6,525 3,036 9,561 5,782 5,000 5 10,584 9,869 20,453 2,158 2,485 4,643 4,714 4,000 6 5,418 3,658 9,076 1,673 919 2,592 6,567 5,000 Central Java 7 6,856 518 7,374 5,953 50 6,003 3,842 3,000 8 13,968 2,436 16,405 11,685 472 12,158 5,667 5,000 9* 10,157 6,827 16,984 4,217 4,000 10 15,126 1,117 16,243 12,857 97 12,954 4,667 3,000 11* 34,338 1,277 35,615 6,000 5,000 12 6,148 4,589 10,737 4,992 1,320 6,312 5,758 5,000 13 5,683 512 6,195 4,708 41 4,749 7,492 5,000 14 15,606 9,357 24,963 7,136 2,369 9,504 6,342 5,000 East Nusa 15 1,847 2,884 4,731 1,469 721 2,190 5,361 5,000 Tenggara 16 3,890 2,804 6,693 3,445 762 4,207 4,895 5,000 17 6,948 4,448 11,396 5,156 1,097 6,253 4,717 2,500 18 709 195 904 569 20 589 4,133 2,000 19 351 5,717 6,068 155 20 175 3,975 2,000 20 186 9,503 9,689 117 2,119 2,236 4,283 2,000 West 21 1,321 1,305 2,627 822 62 884 3,942 3,000 Kalimantan 22 16,357 6,937 23,293 8,793 802 9,595 8,158 5,000 23 21,603 4,820 26,423 17,030 760 17,791 3,042 2,000 24 7,366 4,323 11,689 6,446 993 7,439 4,825 3,000 25 2,520 2,472 4,992 1,803 177 1,981 10,092 10000 26 8,146 6,817 14,963 4,494 5,064 9,557 8,333 7,000 South 27 5,096 11,532 16,628 3,871 2,770 6,641 3,333 2,500 Sulawesi 28 3,387 2,303 5,690 2,049 885 2,934 5,706 5,000 29 1,879 2,111 3,990 603 935 1,539 2,383 1,000 30 1,894 1,442 3,336 1,463 706 2,168 3,500 2,000 31 7,437 2,661 10,098 4,894 291 5,184 3,067 2,000 32 11,059 4,029 15,088 7,403 462 7,865 3,400 2,000 Source: VRRI data, own calculations. Note: * Labor cost data missing for these villages. **Based on wave 2 data. 48  Empirical Analysis Table 28: Maintenance costs and willingness to pay per-household: Bridges Willingness to All costs Costs excluding unskilled labor pay** Province vid Periodic Periodic Periodic Routine Periodic Routine Mean Median +Routine +Routine Lampung 1 155 133 288 97 114 211 5,058 3,000 3 2,413 153 2,567 2,288 28 2,316 5,083 5,000 4 471 2,623 3,094 342 2,355 2,697 5,000 3,000 6 316 707 1,023 185 491 677 4,800 5,000 Central Java 8 163 415 578 115 284 399 4,458 3,000 9* 243 271 515 2,958 2,000 10 177 103 280 100 12 112 3,383 2,000 12 83 71 155 51 38 89 3,567 2,000 13 137 73 210 77 8 85 4,933 3,500 East Nusa 16 300 342 642 162 231 392 3,957 2,000 Tenggara 17 780 372 1,152 514 183 696 3,975 2,000 West 21 4,388 4,615 9,003 4,205 802 5,006 2,202 2,000 Kalimantan 22 169 164 333 163 94 258 3,429 3,000 23 644 297 942 623 169 792 2,583 2,000 24 734 2,050 2,784 711 1,630 2,341 2,750 2,000 25 663 347 1,009 636 61 697 8,433 7,000 26 173 232 405 165 82 246 7,567 5,000 South Sulawesi 27 457 4,039 4,497 254 3,031 3,285 2,342 1,000 29 243 21 264 80 3 83 1,592 1,000 31 409 617 1,026 269 375 644 1,625 1,000 32 78 9,369 9,446 55 6,462 6,517 2,475 1,000 Source: VRRI data, own calculations. Note: * Labor cost data missing for these villages **Based on wave 2 data. Table 29: Maintenance costs and willingness to pay per-household: Piped water Willingness to All costs Costs excluding unskilled labor pay** Province vid Periodic Periodic Periodic Routine Periodic Routine Mean Median +Routine +Routine Lampung 1 1,698 4,500 6,197 1,398 669 2,067 5,617 5,000 2 3,876 9,654 13,530 3,056 1,091 4,147 6,458 5,000 3 675 1,933 2,608 571 253 824 6,250 5,000 Central Java 7 514 985 1,499 432 123 554 3,533 3,000 8 30 26 56 24 2 26 4,600 4,000 9* 1,054 1,017 2,071 3,975 3,000 12 702 720 1,422 626 118 744 4,842 4,000 13 285 222 507 217 24 242 6,233 5,000 East Nusa 17 396 1,465 1,861 309 201 510 4,658 3,000 Tenggara 19 367 2,998 3,365 230 424 653 4,840 2,000 20 737 2,255 2,992 593 429 1,023 4,783 2,000 West Kalimantan 24 1,172 598 1,770 925 105 1,030 10,508 10,000 South Sulawesi 27 4,327 3,606 7,932 3,851 465 4,316 3,825 3,000 29 928 2,188 3,115 709 273 982 3,899 3,000 Source: VRRI data, own calculations. Note: * Labor cost data missing for these villages **Based on wave 2 data. 49 Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia Finally, Table 30 shows comparisons between the per-household annualized maintenance costs for all of the infrastructure in each village and the average and median household WTP for the maintenance of all infrastructure. Mean WTPs would cover the total costs of maintenance in only 7 out of 32 villages, while they would cover the costs excluding the unskilled-labor costs in 19 out of 30 villages—or between 22% to 63% of the sample villages. Meanwhile, if villagers are to focus on fully financing routine maintenance, 25 out of 32 villages would be able to do so if they all contribute their full WTP. Table 30: Maintenance costs and willingness to pay: all infrastructure P+R costs Willingness to pay** Province vid Periodic Routine P+R costs w/o unskilled Mean Median Lampung 1 16,774 5,930 22,704 15,582 9,252 9,000 2 20,517 11,054 31,571 18,189 10,713 10,000 3 23,388 6,726 30,114 18,783 8,942 8,000 4 7,817 9,055 16,872 12,258 9,271 7,000 5 10,584 9,869 20,453 4,643 7,456 5,000 6 5,734 4,365 10,098 3,268 9,515 10,000 Central Java 7 7,370 1,503 8,873 6,558 4,996 4,000 8 14,161 2,877 17,039 12,582 7,171 5,000 9 11,454 8,115 19,570 6,833 5,000 10 15,303 1,220 16,523 13,066 6,938 5,000 11 34,338 1,277 35,615 7,824 5,000 12 6,933 5,381 12,314 7,145 8,563 5,500 13 6,105 806 6,912 5,075 10,238 10,000 14 15,606 9,357 24,963 9,504 9,913 10,000 East Nusa Tenggara 15 1,847 2,884 4,731 2,190 8,550 5,000 16 4,190 3,145 7,335 4,599 9,048 5,000 17 8,125 6,285 14,409 7,459 7,471 5,000 18 709 195 904 589 9,310 5,000 19 719 8,714 9,433 828 6,910 5,000 20 923 11,758 12,681 3,258 8,092 5,000 West Kalimantan 21 5,709 5,920 11,629 5,890 8,911 7,000 22 16,526 7,100 23,626 9,852 11,924 10,000 23 22,247 5,117 27,364 18,583 8,158 5,000 24 9,272 6,971 16,243 10,810 10,924 10,000 25 3,182 2,819 6,002 2,678 11,912 10,000 26 8,319 7,049 15,368 9,804 14,167 15,000 South Sulawesi 27 9,880 19,177 29,057 14,243 4,067 3,000 28 3,387 2,303 5,690 2,934 6,858 5,000 29 3,050 4,320 7,370 2,604 4,757 5,000 30 1,894 1,442 3,336 2,168 5,158 5,000 31 7,846 3,278 11,123 5,829 3,892 3,000 32 11,137 13,398 24,534 14,382 4,347 3,000 Source: VRRI data, own calculations. Note: * Labor cost data missing for these villages **Based on wave 2 data. 50  Empirical Analysis Scenario 2: User fees As we have argued before, it may not be realistic to consider the total WTPs as the resources available for infrastructure maintenance. When households are asked to voluntarily contribute to maintain infrastructure there is the temptation to free-ride on the contributions of others and contribute much less that their WTP values. We do not have a mechanism to collect each households’ true WTPs. As such, we should consider these WTP values as the maximum that villagers are willing to contribute. A more realistic approach would be to consider common mechanisms of resource collections, and consider the amount of resources that would be available for maintenance given the distribution of households’ WTPs in each village. One such mechanism is user fees. Given how the WTP values were collected, we consider a scenario where households would be required to pay a flat monthly fee that will grant them access the infrastructure. Hence, this calculation may not be applicable to other, more sophisticated, forms of user fees, such as those based on usage. Furthermore, we only focus the analysis on single-price user fees—namely that all households face the same monthly fee. That is, we do not consider the possibility of price discrimination among different households because it is not clear how such a price discrimination would be implemented in villages. As we discussed in the beginning of this section, this user fee scenario is a mere analytical tool to add a bit of realism into the amount of resources that can be collected voluntarily by and from villagers. We calculate the available resources under the assumption that households will refuse to pay user fees above their WTPs. We consider two different user fee scenarios: the resource-maximizing (or “optimal”) user fee, or the “politically feasible” (or “median”) user fee. The former refers to the level of single-price user fee that would maximize the total amount of resources collected in the village given the distribution of WTPs in the village. An increase in user fees will increase the revenue collected per household, but will decrease the number of households willing to pay this fee. Using our discrete WTP data, we can therefore find the optimal level of user fee that maximizes the total resources collected. While this level of user fee maximizes the total resources collected, in a democratic setting where the user fee is decided through a one-household one-vote mechanism, the optimal user fee may not be politically feasible if it is rejected by the majority of the households in the village. Alternatively, we consider a scenario if which each household representative will vote for a particular level of user fee insofar that it is below his or her WTPs. Under a majority rule decision rule, this implies that the “politically feasible” user fee will be somewhere around or below the median WTP.8 Under this decision rule, the optimal user fee will be politically feasible if and only if it is less or equal to this “median” user fee. Tables 31, 32, and 33 present the optimal and median user fee for each of the village and summarize the amount of resources that would be available under the median and optimal fees for roads, bridges, and piped water systems respectively. They also calculated the financing gaps 8 More specifically, suppose we have a village with 100 households and we sort their WTPs from the lowest to the highest. Under this decision rule, the fee that is politically feasible will be one less or equal to the WTP of the 49th household from the bottom. 51 Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia in these villages in both cases where villagers do and do not fully contribute their unskilled labor for maintenance. In general, with the user fees, naturally the number of villages that are able to self-finance their maintenance falls. For roads, Table 31 below shows that in 18 out of 32 villages, we have optimal fee levels that are be different from the median level. In one of these 18 villages, Village 26, the optimal user fee is lower than the median user fee, while it is higher in the remaining 17 villages. Implementing the optimal user fee in these 17 villages would exclude between 50.9% to 79.2% of households from access to the village roads (see column D). Meanwhile, implementing the median user fee will exclude between 5.3% to 49.6% of the households in these villages. With the median user fees, the number of villages that would be able to collect enough resources to finance the total cost of road maintenance falls from 7 villages above to a mere 2 out of 32 villages. If villagers fully contribute the necessary unskilled labor, 4 out of 30 villages would have adequate resources to finance road maintenance. The financing gaps among the villages that are unable to finance the total costs of road maintenance range from Rp4.0 million to Rp156.1 million annually. The financing gaps if villagers provide all of the unskilled labor ranges from Rp.0.7million to Rp88.1 million annually. For the most part, the median fee varies between Rp.2000 and Rp.5,000 (with two exceptions of Rp.7,000 in Village 26 and Rp.10,000 in Village 25). Meanwhile, under the optimal user fees, only 2 out of 32 villages will be able to finance the total cost of maintenance. If villagers contribute all of the necessary unskilled labor, 6 out of 30 villages would have the resources to finance road maintenance. The financing gap ranges from Rp3.7 million to Rp156.1 million annually, with a median of Rp37.4 million. Meanwhile, the financing gap when villagers contribute all of the necessary unskilled labor ranges from Rp0.7 million to Rp87.0 million, with a median of Rp13.6 million. For almost all villages, the optimal fee for road maintenance is Rp.5,000. Table 32 shows that for bridges, 14 out of 21 villages would be able to self-finance the total cost of bridge maintenance in their villages with the median user fees. If villagers are to supply all of the unskilled labor, 15 out of 20 villages would be able to self-finance the cost to maintain bridge. Hence, between two-thirds and three-forth of the villages would be able to afford the cost of maintenance with the median user fee scenario. The financing gaps among the seven villages that are unable to finance the total cost range from Rp0.1 million (Village 17) to Rp44.8 million (Village 32) annually. Meanwhile, the financing gaps when villages contribute all of the unskilled labor range from Rp2.6 million (Village 4) to Rp29.7 million (Village 32) annually. The median user fees mainly vary between Rp.1,000 and Rp.3,000 per household. With the median user fee, between 10.3% to 48.3% of households in these villages would be unwilling to pay the fee and therefore not have access to them. Meanwhile, with the optimal user fees, villagers in 16 out of 21 villages would be able to finance the total cost of bridge maintenance. When we exclude the unskilled-labor cost, 15 in 20 villages would be able to do so—hence, roughly, three-fourths of the sample with bridges would have adequate resources with the optimal user fees. When the cost includes the unskilled-labor cost, the financing gaps range from Rp2 million (Village 4) to Rp44.8 million (Village 32), while without it, the financing gaps range from Rp.0.8 million (Village 4) to Rp29.7 million (Village 32). In more than 75% of these villages, the optimal fee for bridges is Rp.5,000. With the optimal fee, between 3.4% to 85% of households in these villages would be would be unwilling to pay the fee and therefore not have access to them. 52  Table 31: Resource collection under median and optimal user fees: Roads C. Ann. resource D. % HHs E. Resources – F. Resources – Maintenance costs B. User fees** collected excluded Cost w/ median fee Cost w/ optimal fee Province Vid w/o Cost w/o Cost w/o All Median Optimal Median Optimal Median Optimal Total cost Total unskilled unskilled unskilled Lampung 1 43,014,214 35,283,548 5,000 5,000 7,577,143 7,577,143 42.9% 42.9% -35,437,071 -27,706,405 -35,437,071 -27,706,405 2 82,268,483 64,031,887 5,000 5,000 17,040,000 17,040,000 25.3% 25.3% -65,228,483 -46,991,887 -65,228,483 -46,991,887 3 99,957,775 62,699,539 5,000 5,000 14,045,128 14,045,128 29.9% 29.9% -85,912,647 -48,654,411 -85,912,647 -48,654,411 4 40,673,164 28,224,642 5,000 5,000 8,270,690 8,270,690 44.0% 44.0% -32,402,474 -19,953,952 -32,402,474 -19,953,952 5 108,729,166 24,683,489 4,000 5,000 11,191,579 13,056,842 47.4% 50.9% -97,537,587 -13,491,910 -95,672,324 -11,626,647 6 28,969,415 8,272,205 5,000 5,000 10,731,724 10,731,724 32.8% 32.8% -18,237,691 2,459,519 -18,237,691 2,459,519 Central Java 7 30,174,640 24,565,370 3,000 5,000 6,866,238 7,629,153 44.1% 62.7% -23,308,402 -17,699,132 -22,545,487 -16,936,217 8 87,798,031 65,066,944 5,000 5,000 15,194,237 15,194,237 43.2% 43.2% -72,603,794 -49,872,707 -72,603,794 -49,872,707 9* 98,848,786 4,000 5,000 11,938,462 12,933,333 48.7% 55.6% -86,910,324 -85,915,453 10 77,381,689 61,712,314 3,000 5,000 8,994,103 11,499,310 37.1% 51.7% -68,387,586 -52,718,211 -65,882,379 -50,213,004 11* 171,380,245 5,000 5,000 15,272,870 15,272,870 36.5% 36.5% -156,107,375 -156,107,375 12 52,440,777 30,826,435 5,000 5,000 14,980,336 14,980,336 38.7% 38.7% -37,460,441 -15,846,099 -37,460,441 -15,846,099 13 35,536,104 27,239,893 5,000 9,000 17,158,974 17,208,000 40.2% 66.7% -18,377,130 -10,080,919 -18,328,104 -10,031,893 14 51,522,781 19,616,898 5,000 5,000 6,996,610 6,996,610 32.2% 32.2% -44,526,171 -12,620,288 -44,526,171 -12,620,288 East Nusa 15 26,060,619 12,062,782 5,000 5,000 15,461,053 15,461,053 43.9% 43.9% -10,599,566 3,398,271 -10,599,566 3,398,271 Tenggara 16 35,501,028 22,312,209 5,000 5,000 14,383,729 14,383,729 45.8% 45.8% -21,117,299 -7,928,480 -21,117,299 -7,928,480 17 58,940,495 32,341,722 2,000 5,000 6,277,128 10,609,231 39.3% 59.0% -52,663,367 -26,064,594 -48,331,264 -21,732,491 18 2,517,202 1,639,789 2,000 5,000 3,602,824 4,094,118 35.3% 70.6% 1,085,622 1,963,035 1,576,916 2,454,329 19 15,510,262 446,809 2,000 5,000 3,228,632 3,699,474 36.8% 71.1% -12,281,630 2,781,823 -11,810,788 3,252,665 20 45,575,470 10,517,004 2,000 5,000 5,758,345 8,313,104 38.8% 64.7% -39,817,125 -4,758,659 -37,262,366 -2,203,900 West 21 9,675,877 3,256,550 3,000 5,000 5,719,895 5,978,421 48.3% 67.5% -3,955,982 2,463,345 -3,697,456 2,721,871 Kalimantan 22 109,850,652 45,249,266 5,000 5,000 22,338,948 22,338,948 5.3% 5.3% -87,511,704 -22,910,318 -87,511,704 -22,910,318 23 139,828,492 94,147,722 2,000 5,000 6,009,560 7,175,593 43.2% 72.9% -133,818,932 -88,138,162 -132,652,899 -86,972,129 24 51,339,082 32,674,086 3,000 5,000 8,478,470 10,120,696 35.7% 53.9% -42,860,612 -24,195,616 -41,218,386 -22,553,390 25 21,745,462 8,627,638 10,000 10,000 24,364,068 24,364,068 44.1% 44.1% 2,618,606 15,736,430 2,618,606 15,736,430 26 33,397,583 21,332,516 7,000 5,000 8,145,846 10,396,923 47.9% 6.8% -25,251,737 -13,186,670 -23,000,660 -10,935,593 South 27 50,484,078 20,163,436 3,000 5,000 4,592,269 5,485,210 49.6% 63.9% -45,891,809 -15,571,167 -44,998,868 -14,678,226 Sulawesi 28 31,067,303 16,021,501 5,000 5,000 15,370,588 15,370,588 43.7% 43.7% -15,696,715 -650,913 -15,696,715 -650,913 29 23,892,967 9,214,475 1,000 5,000 4,391,200 6,237,500 26.7% 79.2% -19,501,767 -4,823,275 -17,655,467 -2,976,975 30 22,778,165 14,803,946 2,000 5,000 7,510,800 12,802,500 45.0% 62.5% -15,267,365 -7,293,146 -9,975,665 -2,001,446 31 57,193,445 29,363,870 2,000 5,000 6,136,000 8,260,000 45.8% 70.8% -51,057,445 -23,227,870 -48,933,445 -21,103,870 32 77,672,236 40,487,708 2,000 5,000 5,748,600 7,936,500 44.2% 69.2% -71,923,636 -34,739,108 -69,735,736 -32,551,208 Source: VRRI data, own calculations. Note: *Labor cost missing **Based on wave 2 data. 53 Empirical Analysis 54 Table 32: Resource collection under median and optimal user fees: Bridges C. Ann. resource D. % HHs E. Resources – F. Resources – Maintenance costs B. User fees collected excluded Cost w/ median fee Cost w/ optimal fee Evidence from Rural Indonesia Province vid w/o Cost w/o Cost w/o All Median Optimal Median Optimal Median Optimal Total cost Total cost unskilled unskilled unskilled Lampung 1 762,623 558,846 3,000 5,000 4,546,286 6,156,429 42.9% 53.6% 3,783,663 3,987,440 5,393,806 5,597,583 Village Capacity in Maintaining Infrastructure 3 10,286,561 9,280,504 5,000 5,000 11,647,179 11,647,179 41.9% 41.9% 1,360,618 2,366,675 1,360,618 2,366,675 4 9,133,209 7,960,085 3,000 5,000 5,344,138 7,125,517 39.7% 51.7% -3,789,071 -2,615,947 -2,007,692 -834,568 6 3,263,949 2,159,744 5,000 5,000 8,530,345 8,530,345 46.6% 46.6% 5,266,396 6,370,601 5,266,396 6,370,601 Central Java 8 3,092,691 2,135,423 3,000 5,000 9,660,814 12,019,322 39.8% 55.1% 6,568,123 7,525,391 8,926,631 9,883,899 9* 2,995,143 2,000 5,000 6,964,103 8,705,128 40.2% 70.1% 3,968,960 5,709,985 10 1,333,783 533,620 2,000 5,000 5,667,517 8,419,138 40.5% 64.7% 4,333,734 5,133,897 7,085,355 7,885,518 12 756,232 436,225 2,000 5,000 5,663,799 8,824,034 42.0% 63.9% 4,907,567 5,227,574 8,067,802 8,387,809 13 1,202,016 485,511 3,000 5,000 9,118,769 13,482,051 47.0% 53.0% 7,916,753 8,633,258 12,280,035 12,996,540 East Nusa 16 3,403,591 2,080,199 2,000 5,000 7,101,966 12,136,271 33.1% 54.2% 3,698,375 5,021,767 8,732,680 10,056,072 Tenggara 17 5,958,757 3,601,623 2,000 5,000 5,835,077 9,946,154 43.6% 61.5% -123,680 2,233,454 3,987,397 6,344,531 West 21 33,165,598 18,442,137 2,000 2,000 4,588,842 4,588,842 37.7% 37.7% -28,576,756 -13,853,295 -28,576,756 -13,853,295 Kalimantan 22 1,571,756 1,214,534 3,000 5,000 7,694,527 11,376,316 45.6% 51.8% 6,122,771 6,479,993 9,804,560 10,161,782 23 4,982,770 4,191,413 2,000 3,000 5,830,170 6,323,492 44.9% 60.2% 847,400 1,638,757 1,340,722 2,132,079 24 12,226,590 10,282,872 2,000 3,000 5,499,548 5,499,548 37.4% 58.3% -6,727,042 -4,783,324 -6,727,042 -4,783,324 25 4,397,220 3,035,701 7,000 5,000 15,762,814 21,041,694 48.3% 3.4% 11,365,594 12,727,113 16,644,474 18,005,993 26 903,180 550,095 5,000 5,000 10,015,385 10,015,385 10.3% 10.3% 9,112,205 9,465,290 9,112,205 9,465,290 South Sulawesi 27 13,651,783 9,974,084 1,000 3,000 2,500,235 2,908,437 17.6% 68.1% -11,151,548 -7,473,849 -10,743,346 -7,065,647 29 1,583,364 497,536 1,000 5,000 3,642,700 4,491,000 39.2% 85.0% 2,059,336 3,145,164 2,907,636 3,993,464 31 5,808,662 3,649,852 1,000 5,000 5,519,300 8,250,500 19.2% 75.8% -289,362 1,869,448 2,441,838 4,600,648 32 48,630,654 33,549,076 1,000 1,000 3,823,200 3,823,200 32.5% 32.5% -44,807,454 -29,725,876 -44,807,454 -29,725,876 Source: VRRI data, own calculations. Note: *Labor cost missing **Based on wave 2 data.  Table 33: Resource collection under median and optimal user fees: Piped water system C. Ann. resource D. % HHs E. Resources – F. Resources – Maintenance costs B. User fees collected excluded Cost w/ median fee Cost w/ optimal fee Province vid Cost w/o Cost w/o All w/o unskilled Median Optimal Median Optimal Median Optimal Total cost Total cost unskilled unskilled Lampung 1 16,434,581 5,480,647 5,000 5,000 7,695,536 7,695,536 42.0% 42.0% -8,739,045 2,214,889 -8,739,045 2,214,889 2 61,696,695 18,908,646 5,000 5,000 13,920,000 13,920,000 38.9% 38.9% -47,776,695 -4,988,646 -47,776,695 -4,988,646 3 10,453,193 3,301,495 5,000 5,000 13,017,436 13,017,436 35.0% 35.0% 2,564,243 9,715,941 2,564,243 9,715,941 Central Java 7 6,132,434 2,268,254 3,000 3,000 6,554,136 6,554,136 46.6% 46.6% 421,702 4,285,882 421,702 4,285,882 8 300,050 136,890 4,000 5,000 10,885,424 12,699,661 49.2% 52.5% 10,585,374 10,748,534 12,399,611 12,562,771 9* 12,052,317 3,000 5,000 9,849,231 12,435,897 43.6% 57.3% -2,203,086 9,849,231 383,580 12,435,897 12 6,944,488 3,636,064 4,000 5,000 10,014,252 12,107,395 48.7% 50.4% 3,069,764 6,378,188 5,162,907 8,471,331 13 2,907,839 1,385,785 5,000 10,000 17,158,974 20,590,770 40.2% 64.1% 14,251,135 15,773,189 17,682,931 19,204,985 East Nusa 17 9,625,926 2,636,209 3,000 5,000 8,089,539 10,609,231 47.9% 59.0% -1,536,387 5,453,330 983,305 7,973,022 Tenggara 19 8,600,711 1,670,313 2,000 5,000 3,587,369 4,596,316 29.8% 64.0% -5,013,342 1,917,056 -4,004,395 2,926,003 20 14,074,135 4,809,920 2,000 5,000 6,569,380 9,326,897 30.2% 60.3% -7,504,755 1,759,460 -4,747,238 4,516,977 West 24 7,773,023 4,522,905 10,000 10,000 29,025,392 29,025,392 33.9% 33.9% 21,252,369 24,502,487 21,252,369 24,502,487 Kalimantan South Sulawesi 27 24,082,617 13,103,299 3,000 5,000 4,898,420 6,378,152 46.2% 58.0% -19,184,197 -8,204,879 -17,704,465 -6,725,147 29 18,654,788 5,879,996 3,000 5,000 9,281,400 13,223,500 48.3% 55.8% -9,373,388 3,401,404 -5,431,288 7,343,504 Source: VRRI data, own calculations. Note: *Labor cost missing **Based on wave 2 data. 55 Empirical Analysis 56 Table 34: Resource collection under median and optimal user fees: all infrastructure C. Ann. resource D. % HHs E. Resources – F. Resources – Maintenance costs B. User fees** collected excluded Cost w/ median vote Cost w/ optimal fee Province vid w/o Cost w/o Cost w/o All Median Optimal Median Optimal Median Optimal Total Total unskilled unskilled unskilled Lampung 1 60,211,418 41,323,041 6,000 10,000 8,666,357 11,839,286 46.0% 55.0% -51,545,061 -32,656,684 -48,372,132 -29,483,755 Evidence from Rural Indonesia 2 143,965,178 82,940,534 10,000 10,000 23,040,000 23,040,000 49.0% 49.0% -120,925,178 -59,900,534 -120,925,178 -59,900,534 3 120,697,529 75,281,538 8,000 10,000 16,443,077 18,155,898 49.0% 55.0% -104,254,452 -58,838,461 -102,541,631 -57,125,640 4 49,806,373 36,184,727 7,000 10,000 10,866,414 12,215,172 47.0% 59.0% -38,939,959 -25,318,313 -37,591,201 -23,969,555 5 108,729,166 24,683,489 5,000 5,000 17,953,158 17,953,158 32.0% 32.0% -90,776,008 -6,730,331 -90,776,008 -6,730,331 6 32,233,364 10,431,949 8,500 10,000 12,988,138 15,409,655 49.0% 52.0% -19,245,226 2,556,189 -16,823,709 4,977,706 Village Capacity in Maintaining Infrastructure Central Java 7 36,307,074 26,833,625 4,000 5,000 8,738,847 9,883,220 47.0% 52.0% -27,568,227 -18,094,778 -26,423,854 -16,950,405 8 91,190,771 67,339,257 5,000 5,000 18,369,152 18,369,152 31.0% 31.0% -72,821,619 -48,970,105 -72,821,619 -48,970,105 9* 113,896,246 5,000 10,000 16,912,820 18,405,128 42.0% 68.0% -96,983,426 -95,491,118 10 78,715,472 62,245,934 5,000 10,000 13,758,103 16,016,897 42.0% 66.0% -64,957,369 -48,487,831 -62,698,575 -46,229,037 11* 171,380,245 5,000 10,000 17,155,826 17,155,826 29.0% 64.0% -154,224,419 -154,224,419 12 60,141,497 34,898,724 5,000 10,000 18,468,908 19,289,748 24.0% 61.0% -41,672,589 -16,429,816 -40,851,749 -15,608,976 13 39,645,959 29,111,188 10,000 10,000 33,337,436 33,337,436 42.0% 42.0% -6,308,523 4,226,248 -6,308,523 4,226,248 14 51,522,781 19,616,898 10,000 10,000 11,019,661 11,019,661 47.0% 47.0% -40,503,120 -8,597,237 -40,503,120 -8,597,237 East Nusa 15 26,060,619 12,062,782 10,000 10,000 29,472,632 29,472,632 46.0% 46.0% 3,412,013 17,409,850 3,412,013 17,409,850 Tenggara 16 38,904,618 24,392,409 6,500 10,000 17,530,170 23,373,560 45.0% 56.0% -21,374,448 -6,862,239 -15,531,058 -1,018,849 17 74,525,178 38,579,555 5,000 10,000 15,250,769 16,797,948 41.0% 68.0% -59,274,409 -23,328,786 -57,727,230 -21,781,607 18 2,517,202 1,639,789 5,000 10,000 9,825,882 10,059,832 29.0% 64.0% 7,308,680 8,186,093 7,542,630 8,420,043 19 24,110,973 2,117,122 5,000 5,000 6,838,421 6,838,421 46.0% 46.0% -17,272,552 4,721,299 -17,272,552 4,721,299 20 59,649,605 15,326,924 5,000 10,000 13,584,828 13,787,586 42.0% 71.0% -46,064,777 -1,742,096 -45,862,019 -1,539,338 West 21 42,841,475 21,698,687 6,500 5,000 12,021,474 15,350,000 46.0% 17.0% -30,820,001 -9,677,213 -27,491,475 -6,348,687 Kalimantan 22 111,422,409 46,463,799 10,000 10,000 35,990,528 35,990,528 24.0% 24.0% -75,431,881 -10,473,271 -75,431,881 -10,473,271 23 144,811,262 98,339,135 5,000 10,000 19,732,882 22,423,728 25.0% 58.0% -125,078,380 -78,606,253 -122,387,534 -75,915,407 24 71,338,695 47,479,863 12,000 10,000 26,581,148 30,171,130 50.0% 31.0% -44,757,547 -20,898,715 -41,167,565 -17,308,733 25 26,142,682 11,663,339 13,500 10,000 28,314,000 37,284,408 50.0% 14.0% 2,171,318 16,650,661 11,141,726 25,621,069 26 34,300,763 21,882,611 15,000 15,000 19,458,462 19,458,462 42.0% 42.0% -14,842,301 -2,424,149 -14,842,301 -2,424,149 South Sulawesi 27 88,218,479 43,240,819 3,000 5,000 5,357,647 6,760,841 41.0% 55.0% -82,860,832 -37,883,172 -81,457,638 -36,479,978 28 31,067,303 16,021,501 5,000 10,000 15,600,000 18,811,764 43.0% 66.0% -15,467,303 -421,501 -12,255,539 2,790,263 29 44,131,120 15,592,006 3,000 5,000 9,580,800 13,473,000 47.0% 55.0% -34,550,320 -6,011,206 -30,658,120 -2,119,006 30 22,778,165 14,803,946 4,000 5,000 13,883,600 16,785,500 49.0% 51.0% -8,894,565 -920,346 -5,992,665 1,981,554 31 63,002,107 33,013,723 3,000 5,000 9,204,000 12,272,000 46.0% 57.0% -53,798,107 -23,809,723 -50,730,107 -20,741,723 32 126,302,889 74,036,783 3,000 5,000 9,266,400 10,725,000 40.0% 58.0% -117,036,489 -64,770,383 -115,577,889 -63,311,783  Empirical Analysis For piped water systems, Table 33 shows that under the median user fee, 6 out of 14 villages would be able to afford the total cost of maintenance. If villagers contribute all unskilled labor, the number increases to 12 villages. The median fee was more varied between Rp.2,000-5,000 with Rp.3,000 being the most frequent median fee, found in 5 out of the 14 villages. Meanwhile, under the optimal fee, 8 out of 14 villages would be afford the total cost of maintenance, while 12 would be able to afford the cost if all unskilled labor are provided by villagers. The optimal fee was Rp.5,000 in 11 villages, Rp.10,000 in 2 villages, and Rp.3,000 in one village. With the median user fee, between 29.8% and 49.2% of households in these villages would not be willing to pay the fee and not have access to the water system. Meanwhile, with the optimal fee, between 33.9% and 64.1% of households would not pay the fee and lose access to the water system. Table 34 considers these fee scenarios when the WTPs are asked for all infrastructure as a bundle. As we have discussed above, the WTPs for the overall infrastructure tends to be lower than the sum of the WTPs for individual infrastructure. Under the median user fee, only 3 out of 32 villages would have adequate resources to finance the total costs for all three types of infrastructure. If villagers contribute all of the necessary unskilled labor, the number increases to 6 (out of 30 villages). Hence, roughly 10-20% of our sample villages would be able to afford the cost of maintenance under the median user fees. Under the optimal user fee, 3 out of 32 villages could afford the total costs of maintenance for all infrastructure. If we assume that villagers provide the unskilled labor, 8 out of 30 villages would be able to afford the full maintenance under the optimal fee scenario. Therefore, under the optimal user fee scenario, only about 10–25% of villages would be able to afford the cost of maintaining these three types of infrastructure in their villages. 6.2 Determinants of villagers’ willingness to pay We next use a simple econometric model to explore the question of the determinants of respondents’ willingness to pay. In theory, we expect that respondents’ income and the opportunity cost from poorly maintained roads to be important determinants of the willingness to pay. However, we are also interested in looking at the role of institutional quality, as well as villagers’ knowledge about village administration and perception of social trust in the village in determining their willingness to pay. As explained before, the iterative bidding procedure used to elicit respondents’ WTP gives us WTP responses that are in the form of discrete values from Rp.0 up to Rp.15,000. If the respondent was willing to pay up to Rp.15,000, he or she would be asked to state the maximum value that she would be willing to pay. However, many respondents did not state specific maximum values of their willingness to pay, creating missing observations at values greater than Rp.15,000. In additions, respondents are, naturally, not allowed to state negative WTPs, although this is theoretically possible.9 Because of the discrete nature of the data, and the fact that WTP values above Rp.15,000 and below Rp.0 are “missing”, we specified the regression using an interval model that are censored from above and below. 9 An example of a negative WTP would be when a high-traffic use of the well-maintained infrastructure creates negative costs to the respondents beyond the benefits that they obtain from good quality infrastructure. 57 Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia We estimate the following simple model: WTP i= . STARTBID i. INC i. OC i . INST i. OTH . D p  i where WTP is the respondents’ willingness to pay, INC is the log of respondent income, OC is the opportunity costs (in time and monetary terms) of a well-maintained infrastructure, INST is respondent perception of the quality of the institutions (including that of social trust), OTH is other control variables, such as the respondent’s mode of transport, and Dp is the province dummy to control for province-level effects. In addition, we also include STARTBID, which is the random starting bid for the contingent valuation questions, to control for the anchoring bias. The subscript i refers to respondent i. To estimate this model, we use data from Wave 4 of the VRRI survey, which contains information on all of the relevant variables above. Before we proceed with the estimation, we need to be clear about what we mean by “opportunity cost” (i.e., the variable OC). The WTP question elicits the maximum value that respondents are willing to pay to ensure that they have a smoothly working village infrastructure, be that a road, bridge, or piped water system. The opportunity cost of having improper maintenance would, therefore, be the cost from having village infrastructure that would occasionally fail to work smoothly. We measure this by taking the difference in the costs—in terms of both monetary and time values—of using the (main) village infrastructure and an alternative infrastructure when the main infrastructure failed. As an example, for roads, this would be the difference in the costs to reach the place of their main employment using the main road versus the alternative road that the respondents would use when the main road became damaged. In the past two years, some respondents in our sample experienced a breakdown in the main village infrastructure that they use, and therefore, could state the real cost of having to use an alternate. However, not all respondents had ever experienced such breakdowns. For these respondents, we hence asked a “what-if ” question, i.e., what would the costs of using an alternative infrastructure be if the main infrastructure had broken down. The costs of an alternate we obtained from these respondents are, therefore, hypothetical. As we shall see, whether respondents’ answer is based on real experiences or not changes the results in important ways. Meanwhile, for INST, we made use of several variables. We used the respondent satisfaction level with regard to the quality of the village administration. We also asked the question: “What do you think about the statement that villagers trust their […] officials?”, where the blank was filled with either “village” and “sub-district” and the respondent was asked whether he or she strongly disagrees, disagrees, agrees, or strongly agrees. To gauge the effect of social trust on WTP, we use two questions, namely, “What do you think of the statement that villagers here are trustworthy?” and “What do you think of the statement that villagers here are helpful to each other?” where the respondent was again asked to express their level of agreement. We recode the answers such that a low number implies a low level of trust. In addition, we also asked if the respondent knows about the use of the village development funds as a measure of respondent knowledge of village finances. 58  Empirical Analysis The survey also asked whether the respondent has ever lodged a complaint regarding an infrastructure. Having lodged a complaint can be construed as placing a high value on the particular infrastructure. Conditional on having lodged a complaint, the survey followed up by asking the respondent whether the complaint was responded to satisfactorily. We use both of these two variables to see how satisfactory responses to complaints may have affected WTP. We also included dummy variables for the different economic sectors in which family members were employed or have businesses. Finally, we included the village dummy variable to control for fixed effects at the village level and a dummy variable indicating whether a village received the village allocation fund (or alokasi dana desa, ADD) during the survey period. In principle, we would have liked to do this estimation on all three types of infrastructure—the road, bridge, and piped water systems. However, since only a subset of the whole sample were actually users of the bridge and piped water system, the sample sizes available to do estimations for the bridges and piped water systems were much lower than that for roads. The low sample sizes for actual users, combined with the relatively large number of variables that we are interested in, resulted in the low power of the estimates. On the other hand, we can try running the estimates on the overall sample, irrespective of whether they used the infrastructure in question—i.e., estimating the effects of these variables on the WTP for hypothetical infrastructure. However, as we will show below, this could be misleading. As such, we decide to focus on estimating the model for roads only. Table 35 presents the interval regression of the various indicators on the WTP for roads. Meanwhile, the signs for the main variables are as expected from theory: a higher income is positively associated with a higher WTP, suggesting that road maintenance is a normal good that rises with income. Statistically, the estimate is highly significant. However, the coefficient for income is miniscule. Meanwhile, we find that there were differential impacts between the “hypothetical” versus real opportunity costs of a damaged road on WTP. For those who stated hypothetical costs of using the alternate road because they had never experienced damages to their main road to work in the past two years, the increase in the relative time cost has no significant effect on the WTP, and the increase in the relative monetary cost has a negative, although not statistically significant, effect on the WTP. On the other hand, respondents who were impacted by road damages in the past two years, the increase in the time and money opportunity costs from actually having to go through an alternative road are positive, although the coefficient for the time opportunity cost is not statistically significant. For these respondents, there is almost a one-to-one increase in WTP with respect to the increase in the additional transportation cost due to the breakdown in roads. We also examine the role of the respondents’ mode of transportation on their WTPs. For each respondent, we asked them of their mode of transport to go to work—whether they walk, use a bicycle, motorcycle, car, or other. All things equal, it appears that respondents using motorcycles as their main mode of transport have the highest willingness to pay for well-maintained road compared to those using other modes of transportation. Respondents who use cars also have a slightly higher WTP when compared to the benchmark “Other” category, although the difference is not statistically significant. 59 Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia Table 35: Interval regression on the willingness to pay for roads+ Coefficients Coefficients Variables Variables (std dev) (std dev) Starting bid for CV questionnaire 0.2425*** Complained about poor road quality -173.6632 (0.03) (355.26) Monthly income 0.0001** Is satisfied with response to complaints 1466.6899*** (0.00) (549.57) Opportunity cost of poor maintenance Knowledge of use of village funds 1118.4932*** Resp. did not experience road failure (413.27) Time (minute) 14.8441 Village dummy variables: (15.63) Receives village allocation fund (ADD) 4095.4534*** Money (Rp.) -0.6943 (1210.44) (0.48) Topography: Mountains/hills 594.1091 Resp. experienced road failure (1020.95) Time (minute) experienced damage 53.9326 Topography: Flatland 2442.6942** (37.97) (1019.56) Money if experienced damage (Rp.) 1.0034* Province: Central Java - (0.55) Mode of transport Province: East Nusa Tenggara - Foot 319.1193 (469.89) Province: West Kalimantan - Bike -149.3376 (703.63) Province: South Sulawesi - Motorbike 1434.5177*** (486.52) Other dummy variables not shown Car 670.0737 Sectors family members work in Included (564.09) Sectors family members own business in Included Satisfaction with regards to Village ID dummies Included The quality of the village administration 101.4066 (160.22) Constant -1305.2213 The quality of the road -241.1105 (2380.67) (264.40) Regression statistics: Trust of lnsigma Fellow villagers -956.8074*** _cons 8.4408*** (293.63) (0.02) Helpfulness of villagers -11.1474 (342.83) sigma 4632.26 Village officials -131.6200 ll -4594.64 (402.94) chi2 464.10 Sub-district officials 650.5856 (456.77) McKelvey & Zavoina’s R2: 0.163 Number of observations 1859 60  Empirical Analysis In terms of institutional quality, we did not find an effect of respondents’ perception of the quality and trustworthiness of the various levels of government and government officials on their WTP. For village level governments, the quality of the village administration and the trustworthiness of village officials do not significantly affect respondents WTP. Meanwhile, the coefficient on the trustworthiness of sub-district officials was positive, although this is not significant at the 10 percent level. With regards to social trust we find, somewhat surprisingly, negative coefficients on both trust variables towards fellow villagers. All else the same, the more the respondent agrees to the statement that “villagers are trustworthy” and “villagers are helpful to each other”, the lower is the respondent’s WTP. The coefficient is only significant at the 10% level for the former.10 One plausible interpretation is to see this as evidence for strategic free-riding among villagers; the belief that one’s fellow villagers are trustworthy means that one does not need to contribute as much. If this interpretation is correct, it highlights the difficulties of relying solely on voluntary contributions to finance these types of public goods. However, more evidence is needed to see the extent to which this interpretation is correct. Meanwhile, three other variables have significant and positive correlation with WTP in this sample. The first variable is having made a complaint about road quality, and having that complaint resolved satisfactorily. Those who make complaints are likely to be individuals who value the benefit of road use more than the cost of making the complaint. It is, hence, plausible to think that those who complained would state a higher WTP. However, our analysis shows that those who complain translates to a higher stated WTPs only when their complaints are satisfactorily resolved. All else the same, having had a complaint satisfactorily resolved increased WTP by around Rp.1,450. Otherwise, having made a complaint had a zero effect on the WTP. Table 36: Complaints and satisfaction rates, by province Observations % of observations Satisfaction Total Complained Satisfied Complained Satisfied rate Lampung 392 54 27 13.8% 6.9% 50.0% Central Java 450 64 18 14.2% 4.0% 28.1% East Nusa Tenggara 456 90 12 19.7% 2.6% 13.3% West Kalimantan 328 83 34 25.3% 10.4% 41.0% South Sulawesi 234 78 33 33.3% 14.1% 42.3% Table 36 presents the breakdown of where these results come from. Overall, we found that about 19.8% of respondents included in our regressions have lodged a complaint about roads. About a third of them received satisfactory responses to their complaints. The province with highest share of respondents complaining is South Sulawesi, while the lowest is Lampung. In West Kalimantan, South Sulawesi and Lampung, about 40-50% of those complaints were responded satisfactorily. The lowest satisfaction rate is found in East Nusa Tenggara, where only 13.3% of complaints were responded satisfactorily. 10 The coefficient for the response to “villagers are helpful to each other” becomes significant (and remains negative) when we remove the variable containing the responses for the “villagers are trustworthy” question. 61 Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia We also asked each respondent: “To whom do you report your infrastructure problem?” and the respondent can list more than one recipients of their complaints. Table 37 lists the different parties that receive respondents’ complaints regarding the roads problems. The top four recipients of respondents’ complaints are village officials, followed by the village heads, hamlet head, and head of the neighborhood administration (RT). Among these four, the satisfaction rate was highest for those reporting to the hamlet head, followed by the village head. However, when we see beyond the top four, the highest satisfaction rate was obtained by respondents who lodged their complaints to the infrastructure management group. Since such groups are rarely present in these villages, and the sample size of those reporting to them are admittedly very small. Table 37: Recipients of respondents’ complaints Satisfaction Complained Satisfied Rate Village officials 139 43 30.9% Village head 137 49 35.8% RW head 95 38 40.0% RT head 85 20 23.5% Family and relatives 16 2 12.5% Gov admin other than vil admin 13 5 38.5% Project management (e.g., KDP) 10 4 40.0% Infrastructure management group 7 4 57.1% Community group/NGO 6 2 33.3% Regional parliament 3 0 0.0% Others 3 1 33.3% The second variable that is associated with a higher WTP is knowledge of how the village allocation fund is used. Those who claim to know how village funds are used, all else the same, have WTPs that are on average around Rp.1,100 higher than those who do not. This finding needs to be interpreted carefully. On the one hand, this may be seen as evidence for the value of greater transparency in increasing people’s WTP for public goods. On the other hand, it is also plausible that people who know how village funds are used may have been individuals who have a higher propensity to care about public welfare, and therefore, are more likely to have a higher WTP for public goods irrespective of the level of transparency of village funds. Finally, the dummy variable indicating that a village receives the village allocation fund (ADD) is also positive and significant. Having received the village allocation fund does not reduce one’s willingness to contribute for maintenance. At the same time, one needs to be careful not to interpret this as saying that higher village allocation fund levels cause higher WTP. As before, it may well be that factors associated with higher WTPs (such as a higher political awareness) may also be conducive to their villages’ receipt of village allocation funds. 62 Chapter 7 Conclusion Our study provides an insight into the implicit assumption that villagers are able to finance proper maintenance of rural infrastructure. We found that, with respect to their current welfare, the costs to maintain their infrastructure are significant in our sample villages. When distributed equally across households in each village, the total costs to maintain the village infrastructure in this study—to wit, village roads, bridges, and piped water systems when they exist—amount to between 0.1% and 2.8% of households’ total consumption, with a median of 1.1%. If we assume that villagers can supply all of the necessary unskilled labor, maintenance costs amount to between 0.1% and 1.4% of their consumption, with a median of 0.5%. The bulk of these costs come from the cost to maintain rural roads. Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia While these percentages might appear small, this “local tax” for infrastructure is likely to be burdensome, given the low income of many villagers. Moreover, since income and consumption are not easily observable, it may be difficult to create a mechanism to collect contributions based on household income or consumption share in the village. If we use villagers’ stated willingness to pay instead, we find that the resources collected fell well below what would necessary to fully finance these costs. Between 21% (when villagers do not contribute any unskilled labor) and 63% (when villagers contribute all unskilled labor) of sample villages would be able to afford the maintenance of all three types of infrastructure. When looked at separately for each infrastructure, the heaviest burden comes from financing roads. Only 21% to 43% of villages would be able to afford road maintenance based on their stated willingness to pay for it. It is, however, reasonable to think that households may not contribute their full WTP. Our econometric analysis alludes to the presence of the temptation to free ride on other households’ contributions. As households did so, total contributions would fall. Under scenarios in which a single-priced monthly per-household user fee are established through a voting procedure, we find the number of villages whose contributions would fell short of those necessary to finance proper infrastructure maintenance increases. Using their WTPs as the basis for calculations, we find that only between 10% and 20% of the sample villages resources collected through these user fees would be adequate to finance the maintenance of all three infrastructures. Meanwhile, our examination of the determinants of households’ WTP to pay for road maintenance shows, as predicted by theory, household income and opportunity costs from a damaged road were positively correlated with WTP. In addition, we also find that the increase in WTP due to increased opportunity costs only happened when households have suffered actual incidences of failed roads and the coefficient was only significant for monetary opportunity cost. We also find that satisfactory responses to complaints regarding road problems increased WTP. Perceived administrative quality of the various levels of governments did not seem to play a major role in influencing WTPs. Interestingly enough, a higher perceived trust of fellow villagers and their helpfulness appears to reduce households WTPs. 64 Chapter 8 Policy Recommendations Based on these results, we have the following recommendations. Institutionalize infrastructure maintenance with clearly defined roles and responsibilities for the different levels of administration. Our study finds that there is a significant willingness on the part of villagers to contribute towards infrastructure maintenance, even though their contributions are not adequate to fully finance the all of the necessary maintenance for all of the infrastructure in their villages. For most of the villages, there would be adequate resources to conduct routine maintenance. However, villagers may need significant financial support to ensure that periodic maintenance is conducted properly. This suggests that villagers can take up the responsibility for routine maintenance, especially in the presence of an institution responsible for maintenance (see below). At the same time, district governments and outside agencies need also to step up their support of the types of maintenance activities that villages are unlikely to be able to afford. Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia District governments need to gradually reallocate resources towards maintenance instead of upgrades. Table 11 suggests that district governments tend to use their resources to support upgrades and rehabilitation efforts instead of periodic maintenance. However, one study suggests that the cost ratio between upgrades and maintenance is about 3.5 to 1 (Dongges et. al., 2007). There is a strong case for a gradual shift from construction and rehabilitation or upgrades towards developing the necessary resources and institutions to undertake rural infrastructure maintenance activities. To quote Dongges et. al. in their discussion of rural roads, policymakers should consider maintenance as “an essential part of the investments in the transport sector, and not as an afterthought only considered when the demands for trunk roads have been taken care of ”. The same can be said of rural infrastructure generally. Newly constructed infrastructure needs to be accompanied with a clear maintenance plan, which states clearly the resources necessary to implement it. Our data show that the maintenance costs vary much more than the villagers’ willingness to pay. These variations are driven, among others, by local conditions as well as the volume and design of the infrastructure. Accounting for the maintenance costs long after the fact, which may be necessary to estimate the resource gap that needs to be plugged, can be cumbersome, costly, and may not be particularly accurate. On the other hand, these variations are likely to be better understood by the initial implementers of the infrastructure. It is therefore crucial that new projects be accompanied with plans for sustainable maintenance that can be used and understood by the various agencies that may want to be involved in plugging the resource gap. At the village level, there needs to be a designated institution responsible for maintenance. Our data suggest that villages do implement routine maintenance on their own. However, as suggested by Dongges et. al. (2007, p.19), these routine maintenance can be implemented more efficiently if they are implemented at the correct time. This is likely to apply in the case of periodic maintenance. Moreover, these activities will require villager contributions—which are likely to be forthcoming only if these villagers recognize the value of participating in maintenance activities. Our study finds villagers’ willingness to contribute is positively and significantly correlated with the responsiveness of an institution in immediately addressing reported infrastructure problems. This designated institution (or person) can therefore act to coordinate maintenance efforts as well as respond to potential problems. Having an institution responsible for maintenance—which may well come from an existing village institution—will be instrumental in ensuring the sustainability of the maintenance efforts. The assignment of maintenance activities to villagers needs to take into account the possibly unequal distribution of burdens towards poorer households. Our data show that maintenance costs can be reduced significantly when villagers are expected to contribute all of the unskilled labor. However having villages supply all unskilled labor may amount to a regressive “informal tax”, where poorer households “pay” more (in the form of labor) for public goods.11 It is important to address this potential issue in the process of institutionalizing maintenance activities at the village level. 11 For a discussion of this “informal tax”, see Olken and Singhal (2009). 66  Policy Recomendations There needs to be further study of effective and efficient means to collect and channel resources to ensure that village infrastructure is well-maintained in the long term. This study provides an insight into the resource gaps faced by villagers in fulfilling the maintenance requirements for their infrastructure. It also suggests a role for outside agencies, including but not limited to district governments, to support maintenance efforts in villages. However, we still lack a good understanding on effective mechanisms to deliver resources in a manner that ensures that infrastructure is well-maintained in the long term or whether different mechanisms work better for certain types of infrastructure and communities, but not on others. Moreover, we also need to understand more the effectiveness of different resource collection strategies for different types of infrastructure to improve the design of the village-level maintenance institutions. 67 Village Capacity in Maintaining Infrastructure Evidence from Rural Indonesia References Asian Development Bank (ADB). 2005. Proposed Loan, Republic of Indonesia: Rural Infrastructure Support Project. Deaton, A. 1997. The Analysis of Household Surveys: A microeconometric approach to development policy. Baltimore, Maryland: John Hopkins University Press. Dongges, Ch., G. Edmonds, and B. Johannessen. 2007. Rural road maintenance: Sustaining the benefits of improved access. Bangkok: International Labour Office. Olken, B. and M. Singhal. 2009. “Informal taxation”, NBER Working Paper 15221, accessible from http://papers.nber.org/papers/w15221 World Bank. 2005. Global Economic Prospects 2005: Economic Implications of Remittances and Migration. Washington DC: World Bank. 68 70