The Political Economy of Gram Panchayats

in South India: Results and Policy Conclusions From a Research Project

The World Bank July 2005

ii

ABBREVIATIONS AND ACRONYMS

AP Andhra Pradesh BP Block Panchayat BPL Below Poverty Line CEO Chief Executive Officer CFC Center Finance Commission CSS Centrally Sponsored Schemes DDP Desert Development Program DEA Department of Economic Affairs DPC District Planing Committee DRDA District Rural Development Agency EAS Employment Assurance System EGS Education Guarantee Scheme EO Executive Officer GOI Government of India GP Gram Panchayat GS Gram Sabha IAS Indian Administrative Service IRDP Integrated Rural Development

Program JRY Jawahar Rozgar Yojana JSGY Jawahar Gram Samridhi Yojana KA Karnataka

KE Kerala MLA Member of Legislative Assembly MLC Member of Legislative Council MP Member of Parliament; NGO Non-governmental Organization OBC Other Backward Caste PRI Panchayat Raj Institution PS Panchayat Samitis SAS State Administrative Service SC/ST Scheduled Caste/Scheduled Tribe SFC State Finance Commission SGSY Swarnjanyanti Gram Swarozgar

Yojana TAD Tribal Area Development TN Tamil Nadu UNDP United Nations Development

Program VEC Village Education Committee VTC Voluntary Technical Experts and

Core ZP Zilla Parishad

Vice President : Praful C. Patel Country Director : Michael Carter Sector Director : Connie Bernard Sector Manager : Adolfo Brizzi Task Manager : Vijayendra Rao

ii

GRAM PANCHAYATS IN SOUTH INDIA: A Report on a Research Project

TABLE OF CONTENTS

Acknowledgements…………………………………………..……………………………….....iv Executive Summary………………………………………………………………………..…....v I. Introduction……………………………………………………………………...…….....1 II. Panchayats and Resource Allocation: A Comparison of the Indian States……….…3

Tables and Maps for Section 2..................................................................................6 Map 1………………………………………………………………………………….6

Table 2.1: Political Participation......................................................………………….7 Table 2.2: Gram Sabha Participation………………………………………………..8 Table 2.3: Public Goods Levels……………………......................………………….9 Table 2.4: GP Activity, from PRA…………………………………………………10 Table 2.5: Private Benefits….………………………………………………………11 Table 2.6: Village Level Participation………………………………………………12 Table 2.7: Household Willingness to Pay…………………………………………...13 Table 2.8: Inequality and Caste Domination……………………………………...14

III. Caste Reservations and the Politics of Public Good Provision……………..….15 Household Level Evidence…………....……………………………………………..16 Village Level Evidence…………………….…………………….………………….16 Tables for Section 3....................................................................................................18 Table 3.1: Summary Statistics......................................................…………………..18 Table 3.2: Effect of SC/ST Reservation on Resource Allocation…………….……..19

IV. Gram Sabhas and Political Participation…………………………………..……20 Determinants of holding a Gram Sabha and who attends……………………….21 Does Participation Matter?..........................................................................................22 Tables for Section 4....................................................................................................24 Table 4.1: Descriptive Statistics......................................................………………...24 Table 4.2: Gram Sabha: Occurrence and Attendance…………...………….……..25 Table 4.3: Gram Sabha Occurrence and Beneficiary Selection...…………….……..26

V. Political Selection and the Quality of Government…………………………………..27 Political Selection…………………..…….………………………………………….28 Policy Effects………………………………..……………….………………………29 Summarizing the results on Political Selection……………….……………….……30 Tables for Section 5....................................................................................................31 Table 5.1: Descriptive Statistics......................................................……...…………31 Table 5.2: Individual Characteristics and Politician Selection………..…………….32 Table 5.3: Village Characteristics and Politician Selection........... ...……………….33

iii

Table 5.4: Politician Characteristics and Beneficiary Selection…….………………34 Table 5.5: Village Characteristics and Beneficiary Selection for BPL cards……….35

VI. Policy Implications…………...……………………………….…………………………..36

References……………………………………………………………………..…………...……38 Annex A: Panchayats and Resource AllocationAnnex B: The Politics of Public Good ProvisionAnnex C: Participatory Democracy in ActionAnnex D: Political Selection and the Quality of Government

iv

ACKNOWLEDGMENTS

This report was jointly authored by Timothy Besley of the London School of Economics, Rohini Pande of Yale University and Vijayendra Rao of the Development Economics Research Group at the World Bank. Radu Ban and Jillian Waid provided excellent research assistance. It was supervised by the South Asia Rural Development Department of the World Bank under themanagement of Adolfo Brizzi. The research underlying the report was co-funded by the South Asia Rural Development Department, the Development Research Group of the World Bank, and the Department for International Development (DFID) of the United Kingdom. Valuable comments were provided by peer reviewers - Ruth Alsop, Rob Chase and Brian Levy, and by Adolfo Brizzi, Stephen Howes and Dina Umali Dieninger. The project benefited greatly from Luis Constantino’sguidance and support.

v

EXECUTIVE SUMMARY

1. Our aim in this report is to summarize the results from a research project on Panchayat Decentralization, and draw some policy implications. The project is an effort to understand the political economy and institutional context of village government in India with a focus on the South Indian states of Andhra Pradesh, Karnataka, Kerala and Tamil Nadu. 2. We use a unique sampling design constructed to control for differences in institutional history and cultural differences by comparing villages on either side of the border of these states which belonged to the same political entity prior to 1956 (when the states were reorganized along linguistic lines). The sample of districts and villages selected is given in Map 1. Using this method, we examine the implications of cross-state and within-state differences in demographics, social structure and administrative and political organization for Panchayat performance. Kerala leads the four south Indian states in levels of civic engagement and literacy. In terms of social organization Karnataka villages have the highest levels of upper caste domination with Karnataka voters far more likely than those in other states to vote on caste or religious lines in Panchayat elections. In terms of administrative set up Tamil Nadu has relatively low levels of autonomy and funding available to Gram Panchayats (GPs) – power is more concentrated in higher levels of government. 3. Kerala leads in the provision of public goods at the village level. But its GPs are perceived by their constituents to have current levels of investment in public goods that are lower than the other South Indian states. Tamil Nadu, on the other hand, has the lowest provision of public goods in our sample, though its GPs are perceived to have higher levels of current activity than those in Kerala. The variation in GP performance across states that we observe seems to mirror findings from the World Bank study on panchayat finances. 4. The results suggest therefore that Kerala’s successes in promoting civic consciousness, along with fiscal and political decentralization, might have had real implications for better public service delivery. The current fiscal problems faced by the state may be contributing to the perceived slippage in the effectiveness of its GPs. These state level comparisons, however, cannot establish causal connections on the reasons behind the observed differences, including the important question of whether the Kerala model can be replicated in the other states. Here our more detailed analysis of the political economy of panchayats, which focuses more on examining variations within blocks/taluks, may be more instructive. 5. We examine the impact of caste reservations finding that when an SC/ST household resides in a village which has been reserved for an SC/ST pradhan they are 7 per cent more likely to obtain targeted benefits. This demonstrates that caste reservations help by improving the access of disadvantaged groups to government programs. It mirrors other research that has shown that women’s reservations improve the match between policy choices and the preferences of women. Thus, reservations seem to be a valuable tool to reduce traditional forms of discrimination in local government. 6. Our results on gram sabhas and political participation also have implications for policy. We find that gram sabhas are often not held regularly (25 per cent of GPs did not have

vi

even one gram sabha in the previous year), and even when they are held beneficiary selection is discussed only in 22 per cent. They are more likely to be held in larger villages with higher literacy rates. Interestingly, after conditioning on these variables, we find no state differences in the propensity to hold gram sabhas. However, we note that only 20 per cent of our household respondents have ever attended a gram sabha, with village literacy again associated with both hearing of and attending the meeting. The meetings are less likely to be attended by women, highlighting potentially important gender differences in participation. In contrast, SC/STs and landless are more likely to attend them. Furthermore, illiterates, landless and SC/STs are more likely to attend gram sabhas in villages which have higher levels of literacy. This again suggests the positive externalities from living in more literate communities. 7. We find that, when gram sabhas are held, there may be some policy benefits; Gram sabhas are associated with a better chance that landless, illiterate, and SC/ST households will obtain a BPL cards. However, while these results are suggestive we cannot conclude that they are causal. Similar results, but with weaker effects, are obtained when the village is more literate. 8. The gram sabha results are suggestive of the key role that they could play in improving the quality of panchayat government. But we find that they are often not held, and even they are held are not well attended with key issues not discussed. The findings suggest that more research into the nature and impact of gram sabhas is warranted, but the greater transparency that they engender could have positive implications. 9. The overall structure of the GP is important. The South Indian states differ in the administrative makeup of GPs, especially the number of villages per GP. We find evidence of cross village inequality in public good provision in a GP with the Pradhan’s village receiving more resources. 10. The final section examines the political economy of political selection and the determinants of politician quality. This section has three key findings. First, the political class is selected on the basis of political connections and economic advantage. Second, politicians exhibit a preference for people from their own social group in service delivery and are, on the whole, opportunistic and benefit disproportionately from public transfer programs. Third, the education level of politicians has a consistently positive effect on selection and a negative effect on opportunism. This suggests that more educated politicians are better and recognized as such by voters. However, whether education matters directly or because it is correlated with other characteristics that make an individual fit for public office cannot be discerned from our results. Nonetheless, the results add to a growing appreciation among economists that education may be important because of its role in inculcating civic values. The unique observation about its role in politics given here also offers a fresh perspective on the value of human capital investments in low income countries. 11. The results demonstrate important interplays between village level variables, the process of political selection, and the targeting of public resources. For example, increased literacy at the village level reduces political opportunism while measures of political dominance are correlated with worse targeting of resources. We also find evidence suggestive of barriers to

vii

entry – land ownership and political connections predict selection but not behavior when in office. 12. Our finding that educated politicians are better in terms of actual performance suggests that it is important to focus on factors that select better politicians as a step toward improving the quality of government. More generally, the results and analyses in the paper reinforce the observation that formal institutions of democracy are no guarantee of effective government. It is essential that preconditions exist for sorting in the right kinds of people – the talented, the virtuous and those who give political voice to the disadvantaged. There is clearly much more we can learn about this process, but these results are a first effort to study the issue empirically. 13. To summarize, we can draw the following lessons for policy from these findings:

a. Caste Reservations work by improving targeting of private transfers to schedule castes and tribes. We find that programs that provide private benefits such as toilets, housing and transfers to the poor and disadvantaged (including provision of BPL card) are more likely to reach SC/STs when the GP has a Pradhan that is reserved for an SC/ST. This suggests that caste reservations are effective in including disadvantaged groups into the purview of local government. It supplements previous research that finds that woman Pradhans in seats reserved for women tend to make decisions more in line with the needs of women. b. Pradhans prefer their home village: The home village of the pradhan tends to receive more high-spillover public goods than other villages in the GP controlling for factors such as village size and head quarter status. This result, a consequence of the incentives that underlie democracy, points to inequalities that may exist within GPs that could be persistent and may be important to address. c. Gram Sabhas may be central to effective local government but are not regularly held: When gram sabhas are held we find that benefits are better targeted to the poor and disadvantaged, and reduce political opportunism. Therefore they seem to improve the transparency of government. Further research will have to determine how this works and their implications for public goods allocation, but clearly they are potentially central to the effective and equitable functioning of GPs. The fact that they are often not held is worrying and needs attention. Also, while SC/STs are more likely to participate in gram sabhas, presumably because of their role in beneficiary selection, we find that women are far less likely to attend them. This is a potential source of gender exclusion that needs attention. d. Literacy Matters: Several results point to the importance of village literacy in improving the functioning of GPs – in reducing political opportunism, improving targeting, etc. We also find that more educated politicians are less opportunistic. Therefore, investments in human capital can be central to improving the quality of democratic governance in addition to their enhancing individual well-being.

viii

e. Finance Matters Corroborating findings from the recent World Bank report on fiscal decentralization in India (World Bank, 2004), we find that differences in the quality of local government between the four South Indian states are correlated with what we know of their levels of fiscal decentralization. In particular, Kerala has led the other states in providing public services at the local level but seems to be slipping more recently in a manner that concurs with its worsening fiscal situation. More generally we find that it is very difficult to understand the state of GP finances because of vast inconsistencies in accounting practices at the GP level. GP budgetary data is therefore very difficult to obtain and even when it is available is difficult to compare and evaluate. f. Socio-Cultural Institutions Matter We show that villages demonstrate high levels of inequality within them, and that this is inequality is both within and between castes. We find evidence showing that caste dominance tends to increase political opportunism. g. Higher salaries may reduce opportunism A result with direct policy implications is that relatively higher real wages for politicians tend to attract wealthier politicians, and improve beneficiary selection suggesting reduced political opportunism.

14. These findings provide some important insights into the political economy and the institutional setting for panchayats. In future research we hope to examine the role of land reform in reducing economic and social inequality, and the quality of government. We will also examine the determinants and implications of social, economic and political participation.

1. INTRODUCTION

1.1 The 73rd amendment to the Indian constitution, passed in 1993, has been one of the most important pieces of legislation in recent Indian history. Its goals are:

a) To systematize the functioning of Panchayati Raj Institutions (PRIs) by mandating regular elections to the three tiers of local government, and requiring states to both increase PRIs taxation and spending power, and PRIs allocation of state and central discretionary funds. At the same time there is an effort to improve the transparency of local government by requiring that gram sabhas or village councils be held at regular intervals, between four to six times a year, to discuss budgetary allocations, select beneficiaries and conduct other important panchayat business.

b) To ensure that disadvantaged groups within village communities are granted a voice in

local deliberations, the 73rd amendment also mandated that 1/3rd of all elected positions in Panchayats, including Panchayat president, be reserved for women. Similarly elected positions in Panchayats are to be reserved for Scheduled Castes and Tribes in proportion to their population share

1.2 All national governments since 1993 have been committed to the implementation of the amendment, and state governments have complied with varying degrees of commitment. The current United Progressive Alliance (UPA) government in Delhi has gone even further by substantially increasing panchayat budgets and possibly giving them the authority to administer important schemes like the Employment Guarantee Scheme. 1.3 This experiment in decentralization is, arguably, one of the most ambitious innovations in local government undertaken by a low income country. The stated aim is to improve citizens' ability to access and influence the public service delivery system and to directly tackle social exclusion by a system of political reservations. Despite the breadth of this democratic experiment, there is remarkably little quantitative evidence on how well the experiment has worked. There is, however, a large and growing qualitative and "action research" literature on Panchayats that come to a diverse set of conclusions - reflecting the difficulties of studying such a broad topic in a complex country. A comprehensive review of this literature is beyond the scope of this report but overviews can be found in World Bank (2000), Matthew and Buch (2000), and Crook and Manor (1998). 1.4 Quantitative analyses of Panchayats using large samples are rarer, however. An exception is the important work by Chattopadhyay and Duflo (2004a) on the causal impact of women's reservations on Panchayat action in Rajasthan and West Bengal. They find that reservations improve the ability of women to govern, in a way that is congruent with the desires of women in the population. Work by Alsop, Krishna and Sjoblom (2000), also on Rajasthan and Madhya Pradesh, highlights the role of reservation in reducing the systematic exclusion of women and disadvantaged groups from decision making processes at the local level. Bardhan and Mookherjee (2003) examine the role of elected village councils in affecting land reform in the Indian state of West Bengal, and Foster and Rosenzweig (2001) examine how decentralization interacts with land ownership patterns to affect public good outcomes. Finally,

1

Chaudhuri and Heller (2004) have, more recently, completed a survey studying the impact of the "People's Campaign for Decentralized Planning" in Kerala showing that it increased the level of participatory planning in panchayats, had a positive impact on development performance and on social inclusion, but that levels of participation have declined in recent years - findings that are consistent with our study. 1.5 But, given the scope of the experiment and regional focus of the existing quantitative work, a large number of open questions remain. How does the political economy of village democracy really work? What determines the quality of village politicians? How well has decentralization worked in early adopter states such as Kerala and Karnataka? What is the impact of caste reservations? Do village meetings open to all citizens (Gram Sabhas) succeed in increasing the voice of the poor and disadvantaged? Answering these questions is crucial in formulating Panchayat policy. 1.6 The above questions also point to a need for a sound, quantitative evidentiary base to provide some answers to these questions. This motivates the research that underlies this report. 1.7 The report is based on four research papers (“Panchayats and Resource Allocation: A Comparison of the South Indian States,” “The Politics of Public Good Provision: Evidence from Indian Local Governments,” “Participatory Democracy in Action: Survey Evidence from India," and “Political Selection and the Quality of Government: Evidence from South India”). We will summarize each of them, and then draw on the findings to discuss their implications for policy. We aim in these summaries to provide a sense of our findings using basic econometric tools, but for details on the theory and empirical methodology underlying our results we refer the readers to the actual papers. The actual papers can be found in the Annexes A-D.

2

2. PANCHAYATS AND RESOURCE ALLOCATION: A COMPARISON OF THE INDIAN STATES1

2.1 The four states in our sample provide an interesting contrast in their approach to panchayat decentralization. Kerala has taken decentralization the furthest among them, with forty percent of state expenditures mandated to be allocated to panchayats, with regular gram sabhas and high levels of citizen participation. Karnataka has also been a pioneer in panchayat decentralization, and was the first Indian state to mandate regular panchayat elections. Andhra Pradesh, under the former Chief Minister Chandrababu Naidu tried to find alternatives to the panchayat system via the Janmabhoomi program. Tamil Nadu, of all the states in our sample, has probably the weakest gram panchayats with much of the power held by higher levels of government. 2.2 An important question that remains in understanding the relative impact of the decentralization process in these four states is the extent to which their political history and social structure have affected the functioning of local governments. There is considerable evidence demonstrating that the Travancore region that is currently part of the state of Kerala has a long history of progressive policies (Jeffrey, 1993). Similarly Mysore state which is currently part of the state of Karnataka was also ruled by relatively autonomous rulers who placed a special emphasis on education and economic development (Bhagavan, 2003). Recent work by Banerjee and Iyer (2003) has shown that there are strong path dependencies in land tenure policies - specifically whether the region of India had a zamindari or ryotwari system in place during British Rule. These systems which were established early in the 19th century are shown to have significant contemporary impacts on a variety of indicators of development. Furthermore, scholars have argued that differences in cultural systems can have an important effect of human development (e.g. Dyson and Moore, 1983). Given these path-dependencies and the cultural differences, it is possible that Kerala is different because "Kerala is Kerala". There is something special about the state that makes it particularly hospitable to good, equitable governance. If such path -dependencies prove to be definitive, then policy options are likely to be relatively small. 2.3 The sampling strategy employed by this research project allows us to compare the states, controlling for differences that may come from historical or cultural path-dependencies. Details of the sampling strategy are available in the paper in the Appendix, but, in brief, we compare villages on either side of the current borders of the four states which belonged to the same political entity prior to the state’s reorganization in 1956. These villages have additionally been matched by majority language. Map 1 shows the districts that were selected, with each dot representing a village. Since, the villages across each pair of borders share a common history till 1956, and speak the same majority language, any differences we observe between the matched villages cannot be because of different political histories prior to 1956, or because of different language - which proxy for local kinship structure and social organization2. The differences have to be attributed to changes that have occurred after 1956. The comparison is particularly 1 This section summarizes results from the paper Tim Besley, Rohini Pande and Vijayendra Rao. “Panchayats and Resource Allocation: A Comparison of the South Indian States,” mimeo, 2005 2 Kolar district in Karnataka is an exception since it was part of old Mysore state, it was selected, however, because it shares large cultural affinities with Chithoor district in AP and Dharmapuri in Tamil Nadu.

3

interesting because the states provide an excellent comparison of differences in the implementation of the 73rd amendment. 2.4 What do we learn from our results? First, they provide more information on different aspects of Kerala's sophisticated political culture. Table 2.1 presents these results. Kerala has the highest voter turnout in all types of election among the four states. Households in Kerala are most likely to participate in political activities. Furthermore, Kerala’s electorate is among the least likely to vote for candidates based on caste or religious lines and most likely to vote based on party lines. Kerala has a more active civic culture with active participation in gram sabhas (See Chaudhuri and Heller (2004) for more on gram sabhas in Kerala) as is seen in Table 2.2. Table 2.2 also reveals an interesting composition effect: while having the highest gram sabha attendance, Kerala has at the same time the lowest attendance for beneficiary selection. This may imply that in Kerala, gram sabhas are devoted to more substantial issues. In addition, those attending the Gram Sabha in Kerala are much more likely to speak during the meeting than those in other states. Levels of land inequality are high in Kerala, as measured by the average fraction of landless households in a village, in table 2.8. However, the fraction of villages in which the upper caste holds the majority of the land is lowest in Kerala. This implies that land inequality is less likely to be driven by caste based inequality than in the other states. Kerala, perhaps influenced by this active political culture, also dominates the other states in the availability of public goods, as reflected in Table 2.3. However, all our indicators of current investments on public goods by the panchayats are lower in Kerala than in the other states (Table 2.4). Similarly we find that Kerala lags behind Andhra Pradesh in the provision of BPL cards and behind other states in public works programs (Table 2.5). To some extent this is because of Kerala's higher levels of development and lower levels of poverty. But, other evidence from the World Bank's fiscal decentralization study (World Bank, 2004) suggests that fiscal constraints have reduced the availability of funds to panchayats resulting in lower levels of GP activity. 2.5 Tamil Nadu GPs in our sample are at the other end. They lag all the other states in the provision of most public goods (other than water tanks and bus stops, Table 2.3). More importantly, current levels of activity by GPs are also below other states as seen in Table 2.4. This is also true in the provision of private benefits such as BPL cards, housing and electricity (Table 2.5). On the other hand, villagers in Tamil Nadu, are second only to those in Kerala in their political and civic participation - they are more likely to vote than villagers in AP and Karnataka, and more likely to pay taxes (Tables 2.1, and respectively 2.2). 2.6 It is interesting to note that the remaining two states, Karnataka and AP are rather similar, despite purported efforts in AP to circumvent the panchayat system. Since KA has been far ahead of AP in promoting democratic decentralization, it is interesting that this has not led to large differences in the provision of public goods (except for paved roads, Table 2.3), or indeed even in current GP activity in public goods provision (Table 2.4). On private benefits Karnataka leads all the states in overall activism - particularly in the provision of toilets and electricity. But AP leads the states in providing BPL cards and public works projects. Karnataka is the most likely state to have an NGO active in the village, but it is also the least likely to have held a gram sabha in the last six months (Table 2.6) - which can largely be attributed to drought conditions in

4

the state at the time of the survey. However, even though AP faced the same climatic conditions, it was far more likely than Karnataka to have held gram sabhas. 2.7 There are also some interesting results on the willingness to pay for public services shown in Table 2.7. Here we see that households in Kerala are much more likely to say that they are willing to pay more for public services across the board. We also observe a greater willingness to pay for public services in TN compared to the Andhra Pradesh and Karnataka. Note also that in the means, we see that in all the states except KA close to 50% of our respondents say that they are willing to pay more for one or more public services. While willingness to pay questions have important flaws, these results suggest a large gap between the demand and supply of service provision. They also point to a potential for increased participation by villagers in public good provision. 2.8 Finally, it is also interesting to note the strong caste influences in Karnataka. Karnataka villages have the highest proportion of land owned by upper castes (36 per cent, as given in Table 2.8). Perhaps as a consequence, Karnataka voters are far more likely than those in other states to vote along caste or religious lines. 2.9 Having explored broad patterns of differences across the states that reflect differences in state policies since 1956, in the next three sections we turn to a detailed examination of the political economy of panchayats.

5

Map 1: Sampling Strategy

6

Table 2.1Political participation, mean comparison

StateParticipate

political Voted GP Voted MLA Voted MP Vote group Vote partyVote

candidateAndhra 0.253 0.761 0.865 0.761 0.063 0.131 0.377

(0.435) (0.427) (0.342) (0.427) (0.242) (0.338) (0.485)Karnataka 0.053 0.713 0.782 0.713 0.142 0.053 0.370

(0.224) (0.452) (0.413) (0.452) (0.349) (0.225) (0.483)Kerala 0.311 0.844 0.902 0.844 0.079 0.392 0.133

(0.463) (0.363) (0.297) (0.363) (0.270) (0.488) (0.339)TamilNadu 0.093 0.801 0.811 0.801 0.091 0.029 0.598

(0.290) (0.399) (0.392) (0.399) (0.287) (0.168) (0.490)All 0.154 0.777 0.831 0.777 0.102 0.142 0.373

(0.361) (0.417) (0.375) (0.417) (0.303) (0.350) (0.484)Notes: standard deviations in parenthesis

Political participation, regression

StateParticipate

political Voted GP Voted MLA Voted MP Vote group Vote partyVote

candidateAndhra -0.073 -0.154 -0.082 -0.154 -0.007 -0.249 0.186

(2.311) (3.588) (2.754) (3.588) (0.286) (7.780) (3.152)Karnataka -0.265 -0.184 -0.174 -0.184 0.073 -0.339 0.211

(11.224) (6.103) (9.859) (6.103) (4.810) (11.812) (6.060)TamilNadu -0.208 -0.120 -0.155 -0.120 -0.001 -0.332 0.382

(7.264) (3.823) (9.109) (3.823) (0.089) (12.261) (9.741)Pradhan's Village 0.027 0.012 -0.005 0.012 0.021 0.007 -0.016

(1.601) (1.086) (0.372) (1.086) (1.688) (0.724) (0.898)Reserved GP -0.006 -0.004 0.003 -0.004 0.001 0.016 -0.033

(0.218) (0.183) (0.243) (0.183) (0.061) (0.871) (1.889)female -0.117 -0.002 -0.098 -0.002 -0.026 -0.046 -0.103

(7.488) (0.163) (11.553) (0.163) (2.994) (3.498) (4.952)SCST 0.043 0.013 -0.003 0.013 0.004 0.043 -0.006

(2.528) (0.731) (0.222) (0.731) (0.226) (2.234) (0.322)wealthy 0.013 -0.090 0.036 -0.090 0.003 -0.006 0.041

(1.282) (5.051) (2.913) (5.051) (0.263) (0.585) (2.858)landless -0.024 0.069 -0.019 0.069 -0.016 -0.018 0.011

(1.807) (4.344) (1.504) (4.344) (1.802) (2.035) (0.564)politician -0.188

(7.928)N 5460 5460 5460 5460 4940 4940 4940Adj R-sq 0.154 0.038 0.053 0.038 0.024 0.186 0.141Notes:1) "Participate political" is an indicator variable, equal to 1 if the household took part in any political activities, such as going to rallieshand out leaflets, give speeches, writing pamphlets, giving money or support in kind for political campaigns2)"Vote group" is an indicator variable, equal to 1 if the main reason for voting for the Pradhan candidate is his religion, caste, gender, neighborhood, or friend group 3) "Vote party" is an indicator variable, equal to 1 if the main reason for voting for the Pradhan candidate is his party4) "Vote candidate" is an indicator variable, equal to 1 if the main reason for voting for the Pradhan candidate is an individual characteristic or accomplishment: income, education, land ownership, promises, previous record, active in village, or gave the most money5)absolute values of t-statistics clustered by block in parenthesis6)block pair fixed effects included in regression

7

Table 2.2Gram Sabha participation and house tax payment, mean comparison

State Attend GS

Attend GS for

beneficiary GS speaking taxpayAndhra 0.107 0.935 0.286 0.375

(0.309) (0.248) (0.455) (0.484)Karnataka 0.141 0.900 0.036 0.873

(0.348) (0.301) (0.186) (0.333)Kerala 0.397 0.686 0.523 0.912

(0.489) (0.464) (0.500) (0.283)TamilNadu 0.131 0.806 0.252 0.890

(0.338) (0.397) (0.435) (0.313)All 0.199 0.777 0.338 0.825

(0.399) (0.416) (0.473) (0.380)Notes: standard deviations in parenthesis

Gram Sabha participation and house tax payment, regression

State Attend GS

Attend GS for

beneficiary GS speaking taxpayAndhra -0.200 0.335 -0.357 -0.646

(5.024) (9.917) (5.128) (8.702)Karnataka -0.179 0.239 -0.544 -0.147

(5.656) (11.023) (15.822) (2.772)TamilNadu -0.194 0.127 -0.247 -0.104

(6.161) (6.351) (5.927) (1.632)Pradhan's Village 0.019 0.017 0.018 0.029

(1.377) (0.591) (0.693) (1.928)Reserved GP 0.002 -0.090 -0.051 -0.014

(0.108) (2.534) (1.182) (0.625)female -0.187 -0.097 -0.074 -0.034

(11.768) (2.860) (2.737) (4.111)SCST 0.023 0.014 -0.025 -0.016

(1.344) (0.331) (0.553) (1.049)wealthy -0.011 0.023 -0.028 0.061

(0.527) (0.772) (0.949) (3.414)landless 0.014 -0.035 -0.079 -0.074

(1.214) (1.338) (1.831) (4.110)politician -0.231 0.057

(9.636) (2.519)N 5460 1054 1054 5460Adj R-sq 0.180 0.076 0.197 0.268Notes:1)absolute values of t-statistics clustered by block in parenthesis2)block pair fixed effects included in regression

8

Table 2.3Current level of public goods, mean comparison

State

Schools per 1000

inhabitants

Health facilities per

1000 inhabitants

Number drinking water

sources

Number overhead

tanks

Bus stop in village

(dummy)Proportion paved road

Proportion road with light

Andhra 1.980 0.235 3.171 0.943 0.500 0.206 0.436(1.534) (0.512) (2.713) (0.931) (0.504) (0.213) (0.258)

Karnataka 1.403 0.078 3.753 0.610 0.577 0.787 0.418(1.098) (0.210) (2.454) (0.748) (0.495) (0.182) (0.263)

Kerala 2.120 2.891 12.397 0.143 0.024 0.459 0.396(1.137) (1.621) (9.906) (0.451) (0.153) (0.200) (0.281)

TamilNadu 1.068 0.151 1.924 1.132 0.653 0.465 0.460(1.061) (0.529) (1.778) (0.821) (0.478) (0.301) (0.280)

All 1.535 0.701 5.257 0.686 0.454 0.542 0.427(1.234) (1.386) (6.652) (0.825) (0.498) (0.302) (0.272)

Notes: standard deviations in parenthesis

Current level of public goods, regression

State

Schools per 1000

inhabitants

Health facilities per

1000 inhabitants

Number drinking water

sources

Number overhead

tanks

Bus stop in village

(dummy)Proportion paved road

Proportion road with light

Andhra -0.625 -2.675 -9.111 0.678 0.449 -0.311 0.154(1.806) (10.460) (5.325) (2.497) (6.597) (5.891) (2.422)

Karnataka -1.208 -2.794 -7.714 0.506 0.576 0.248 0.148(5.083) (13.776) (4.785) (3.675) (13.091) (7.070) (3.098)

TamilNadu -1.332 -2.847 -11.178 0.998 0.727 -0.033 0.145(5.419) (9.677) (7.137) (6.715) (18.896) (0.721) (3.271)

Prad. Village -0.234 -0.011 1.190 0.345 0.173 -0.024 0.044(1.750) (0.185) (3.101) (3.830) (3.814) (1.350) (1.873)

Reserved GP 0.014 -0.055 0.718 -0.017 0.049 -0.031 -0.007(0.158) (0.572) (1.120) (0.241) (0.992) (1.138) (0.297)

N 495 495 504 504 504 501 488Adj R-sq 0.232 0.659 0.450 0.246 0.275 0.475 0.184Notes:1)absolute values of t-statistics clustered by block in parenthesis2)block pair fixed effects included in regression

9

Tabl

e 2.

4G

P a

ctiv

ity, m

eans

com

paris

on

Sta

teO

vera

ll G

P

activ

ity

GP

act

ivis

m in

sc

hool

s (c

ount

)G

P a

ctiv

ism

in

heal

th (c

ount

)G

P a

ctiv

ism

in

wat

er (c

ount

)

GP

act

ivis

m in

sa

nita

tion

(cou

nt)

GP

act

ivis

m in

tra

nspo

rt (c

ount

)G

P a

ctiv

ism

in

road

(cou

nt)

GP

act

ivis

m in

el

ectri

city

(c

ount

)

GP

act

ivis

m in

irr

igat

ion

(cou

nt)

And

hra

0.40

70.

529

0.34

30.

529

0.62

90.

214

0.94

30.

714

0.25

7(0

.227

)(0

.653

)(0

.587

)(0

.675

)(0

.802

)(0

.447

)(0

.832

)(0

.783

)(0

.530

)K

arna

taka

0.40

90.

418

0.20

30.

484

0.50

50.

132

0.87

41.

011

0.09

3(0

.291

)(0

.596

)(0

.583

)(0

.646

)(0

.663

)(0

.370

)(0

.780

)(1

.217

)(0

.327

)K

eral

a0.

438

0.33

30.

500

0.31

00.

270

0.08

70.

802

0.76

20.

143

(0.2

38)

(0.5

37)

(0.6

54)

(0.5

13)

(0.4

97)

(0.2

83)

(0.6

07)

(0.7

74)

(0.3

94)

Tam

ilNad

u0.

238

0.31

30.

278

0.39

60.

125

0.04

90.

264

0.54

90.

076

(0.2

01)

(0.5

73)

(0.5

08)

(0.5

82)

(0.3

32)

(0.2

16)

(0.5

42)

(0.6

98)

(0.2

92)

All

0.36

90.

383

0.31

40.

423

0.36

00.

109

0.69

70.

784

0.12

3(0

.260

)(0

.587

)(0

.592

)(0

.606

)(0

.601

)(0

.330

)(0

.739

)(0

.952

)(0

.372

)N

otes

:1)s

tand

ard

devi

atio

ns in

par

enth

esis

2)O

vera

ll G

P a

ctiv

ity is

the

ratio

of s

ecto

rs in

whi

ch G

P w

as a

ctiv

e, to

tota

l sec

tors

3)A

ctiv

ities

are

afte

r las

t ele

ctio

n

GP

act

ivity

, reg

ress

ions

Sta

teO

vera

ll G

P

activ

ity

GP

act

ivis

m in

sc

hool

s (c

ount

)G

P a

ctiv

ism

in

heal

th (c

ount

)G

P a

ctiv

ism

in

wat

er (c

ount

)

GP

act

ivis

m in

sa

nita

tion

(cou

nt)

GP

act

ivis

m in

tra

nspo

rt (c

ount

)G

P a

ctiv

ism

in

road

(cou

nt)

GP

act

ivis

m in

el

ectri

city

(c

ount

)

GP

act

ivis

m in

irr

igat

ion

(cou

nt)

And

hra

0.10

30.

203

-0.2

170.

336

0.33

30.

050

0.46

10.

059

0.12

4(1

.375

)(0

.824

)(1

.459

)(1

.898

)(2

.428

)(0

.551

)(2

.330

)(0

.229

)(1

.390

)K

arna

taka

0.10

70.

117

-0.2

410.

291

0.26

50.

059

0.46

10.

361

-0.0

49(1

.776

)(0

.685

)(3

.048

)(2

.040

)(2

.537

)(1

.691

)(2

.978

)(2

.009

)(0

.997

)Ta

milN

adu

-0.1

120.

018

-0.1

600.

209

-0.1

40-0

.048

-0.2

86-0

.019

0.00

7(1

.781

)(0

.094

)(1

.551

)(1

.273

)(1

.450

)(1

.349

)(1

.794

)(0

.118

)(0

.122

)P

rad.

Vill

age

0.09

20.

103

0.12

50.

153

0.11

00.

082

0.27

90.

167

-0.0

07(3

.899

)(1

.425

)(2

.299

)(2

.509

)(1

.938

)(1

.983

)(4

.252

)(2

.381

)(0

.212

)R

eser

ved

GP

-0.0

10-0

.010

-0.0

240.

028

0.04

30.

016

0.00

90.

049

0.00

1(0

.273

)(0

.178

)(0

.383

)(0

.398

)(0

.631

)(0

.476

)(0

.165

)(0

.381

)(0

.024

)N

504

504

504

504

504

504

504

504

504

Adj

R-s

q0.

215

0.04

20.

203

0.04

20.

108

0.05

30.

246

0.16

70.

050

Not

es:

1)ab

solu

te v

alue

s of

t-st

atis

tics

clus

tere

d by

blo

ck in

par

enth

esis

2)bl

ock

pair

fixed

effe

cts

incl

uded

in re

gres

sion

10

Table 2.5Private benefits (public works and BPL cards), mean comparison

StateAny GP provision

House GP provision

Toilet GP Provision

Water GP Provision

Electricity GP provision BPL received

Received money for

public works

Andhra 0.046 0.025 0.006 0.003 0.013 0.322 0.127(0.209) (0.156) (0.074) (0.053) (0.111) (0.468) (0.334)

Karnataka 0.122 0.024 0.032 0.002 0.073 0.101 0.051(0.327) (0.154) (0.175) (0.039) (0.260) (0.302) (0.220)

Kerala 0.041 0.019 0.019 0.000 0.014 0.297 0.019(0.199) (0.138) (0.135) (0.000) (0.117) (0.457) (0.136)

TamilNadu 0.023 0.006 0.006 0.006 0.007 0.251 0.020(0.150) (0.075) (0.075) (0.075) (0.083) (0.434) (0.139)

All 0.065 0.018 0.018 0.002 0.032 0.220 0.044(0.246) (0.133) (0.133) (0.049) (0.177) (0.414) (0.205)

Notes: standard deviations in parenthesis

Private benefits (public works and BPL cards), regression

StateAny GP provision

House GP provision

Toilet GP Provision

Water GP Provision

Electricity GP provision BPL received

Received money for

public works

Andhra -0.004 0.012 0.008 0.002 -0.039 0.207 0.075(0.220) (1.344) (0.797) (0.569) (2.871) (1.981) (3.484)

Karnataka 0.077 0.009 0.032 0.000 0.033 -0.032 0.010(5.969) (1.610) (4.147) (0.139) (3.033) (0.407) (1.088)

TamilNadu -0.032 -0.015 0.001 0.006 -0.035 0.088 -0.028(3.063) (3.263) (0.122) (1.726) (3.732) (0.872) (3.536)

Pradhan's Village 0.008 0.000 0.007 0.002 0.001 -0.018 0.004(0.770) (0.081) (1.394) (0.711) (0.174) (1.195) (0.810)

Reserved GP -0.007 -0.006 0.000 0.001 -0.004 0.019 -0.007(0.942) (1.354) (0.038) (0.918) (0.778) (0.602) (0.882)

female 0.005 0.006 -0.006 0.000 0.006 -0.004 -0.005(0.943) (1.769) (1.881) (0.045) (1.440) (0.401) (0.751)

SCST 0.035 0.016 -0.001 0.000 0.025 0.128 0.043(3.020) (2.377) (0.140) (0.219) (3.103) (3.930) (3.886)

wealthy -0.043 -0.014 -0.006 0.001 -0.030 -0.096 -0.001(5.311) (3.757) (1.312) (0.468) (4.160) (4.079) (0.171)

landless 0.019 0.005 0.007 -0.001 0.010 0.074 0.014(1.914) (1.007) (1.365) (0.438) (1.554) (4.850) (2.764)

politician 0.033 -0.002 0.028 -0.003 0.018 0.092 0.059(1.889) (0.429) (2.394) (2.467) (1.483) (1.365) (2.363)

N 5460 5460 5460 5460 5460 5460 5422Adj R-sq 0.044 0.009 0.025 0.002 0.041 0.167 0.047Notes:1)absolute values of t-statistics clustered by block in parenthesis2)block pair fixed effects included in regression

11

Table 2.6Village level activities, mean comparison

State NGO activeGS held last

6mo

Andhra Pradesh 0.686 0.710(0.468) (0.457)

Karnataka 0.379 0.692(0.487) (0.463)

Kerala 0.111 0.984(0.316) (0.125)

Tamil Nadu 0.292 0.672(0.456) (0.471)

All states 0.331 0.761(0.471) (0.427)

Notes: standard deviations in parenthesis

Village level activities, regression

State NGO activeGS held last

6moAndhra 0.103 -0.217

(1.375) (1.459)Karnataka 0.107 -0.241

(1.776) (3.048)Tamil Nadu -0.112 -0.160

(1.781) (1.551)Prad. Village 0.092 0.125

(3.899) (2.299)Reserved GP -0.010 -0.024

(0.273) (0.383)N 504 504Adj R-sq 0.215 0.203Notes:1)absolute values of t-statistics clustered by block in parenthesis2)block pair fixed effects included in regression

12

Table 2.7Household willingness to pay, mean comparison

State

Willing provide roads

Willing provide

anganwadi

Willing provide

Health subc

Willing provide P.

school

Willing provide dr

waterWilling

provide anyAndhra 0.329 0.210 0.263 0.228 0.276 0.485

(0.470) (0.407) (0.440) (0.420) (0.448) (0.500)Karnataka 0.103 0.084 0.021 0.089 0.090 0.189

(0.304) (0.277) (0.144) (0.285) (0.287) (0.392)Kerala 0.333 0.362 0.369 0.337 0.401 0.550

(0.471) (0.481) (0.483) (0.473) (0.490) (0.498)TamilNadu 0.352 0.296 0.291 0.314 0.338 0.439

(0.478) (0.457) (0.455) (0.464) (0.473) (0.496)All 0.258 0.228 0.214 0.231 0.260 0.386

(0.438) (0.420) (0.410) (0.422) (0.439) (0.487)Notes: standard deviations in parenthesis

Household willingness to pay, regression

State

Willing provide roads

Willing provide

anganwadi

Willing provide

Health subc

Willing provide P.

school

Willing provide dr

waterWilling

provide anyAndhra 0.009 -0.200 -0.133 -0.140 -0.168 0.027

(0.232) (4.108) (2.683) (2.819) (3.741) (0.821)Karnataka -0.211 -0.301 -0.348 -0.251 -0.334 -0.268

(6.429) (8.444) (8.842) (7.458) (10.691) (12.258)TamilNadu 0.030 -0.066 -0.068 -0.029 -0.067 -0.044

(0.893) (1.908) (1.719) (0.851) (2.241) (1.982)Pradhan's Village 0.010 0.033 0.025 0.031 0.017 0.032

(0.613) (2.204) (2.098) (2.170) (1.040) (1.747)Reserved GP 0.002 0.008 0.008 -0.001 0.025 0.015

(0.171) (0.460) (0.502) (0.079) (1.523) (0.760)female -0.045 -0.046 -0.057 -0.052 -0.065 -0.085

(4.104) (3.871) (4.318) (4.414) (6.235) (7.324)SCST 0.001 0.003 0.006 0.001 -0.013 0.019

(0.075) (0.150) (0.400) (0.061) (0.899) (1.071)wealthy 0.025 0.038 0.032 0.038 0.015 0.057

(1.547) (2.482) (2.401) (2.912) (1.116) (3.363)landless -0.014 -0.030 -0.027 -0.028 -0.026 -0.063

(0.901) (1.952) (1.715) (2.026) (1.712) (4.208)politician -0.053 -0.039 -0.089 -0.033 -0.023 -0.003

(1.524) (1.033) (2.476) (0.865) (0.589) (0.056)N 5460 5460 5460 5460 5460 5460Adj R-sq 0.077 0.097 0.150 0.084 0.099 0.116Notes:1)absolute values of t-statistics clustered by block in parenthesis2)block pair fixed effects included in regression

13

Table 2.8Inequality and caste domination, mean comparisons

Upper caste land

dominance (dummy)

Upper caste land

proportionFraction

landless hhsAndhra Pradesh 0.171 0.255 0.286

(0.380) (0.260) (0.235)Karnataka 0.335 0.364 0.232

(0.473) (0.277) (0.188)Kerala 0.087 0.171 0.430

(0.283) (0.201) (0.247)Tamil Nadu 0.236 0.244 0.409

(0.426) (0.331) (0.283)All states 0.226 0.270 0.336

(0.419) (0.284) (0.253)Notes: 1) Upper caste land dominance is an indicator variable, equals 1 if upper castes own more than half the land in the village2)Standard deviations in parenthesis

14

3. CASTE RESERVATIONS AND THE POLITICS OF PUBLIC GOOD PROVISION3

3.1 The 73rd constitutional amendment mandated political reservation in favor of SC/ST for the Pradhan position, and required that the extent of such reservation in a state reflect the SC/ST population share in that state. The amendment also required that no GP be reserved for the same group for two consecutive elections. The choice of which GPs to reserve was left to individual states. Typically, the same fraction of GPs are reserved in every district in a state. 3.2 A GP has responsibilities of civic administration with limited independent taxation powers. On average, roughly 10 percent of a GP's total revenue come from own revenues with the remainder consisting of transfers from higher levels of government. While the ambit of GP policy influence varies across Indian states GPs typically perform (at least) two distinct policy tasks. The first is beneficiary selection for central and state welfare schemes. We consider this policy task as provision of low spill-over public goods because the benefits are likely to accrue to individual households. These are schemes which provide beneficiary households with funds to acquire household public goods such as housing and private electricity and water supply. Eligibility for these schemes is usually restricted to households below the official poverty line. In addition, most schemes require that a minimum fraction of beneficiaries be SC/ST. The second area of GP policy activism is the construction and maintenance of village public goods such as street-lights, roads and drains. Using the same logic, we consider this policy task as provision of high spill-over public goods. The GP decides the distribution of these public goods within the village, and the quality of such public good provision. 3.3 Schedule XI of the Indian Constitution defines the functional items for which states may devolve responsibility to Panchayats. Panchayat legislation requires that the Pradhan consult with villagers (via gram sabha meetings) and ward members in deciding the choice of beneficiaries and allocation of public goods. However, final decision-making powers in a GP are vested with the Pradhan. 3.4 In this section we use information from an independent audit of village facilities to construct an index of GP activity on high spill-over (i.e. village-level) public goods. This index measures whether the GP undertook any construction or improvement activity on within-village roads, drains, street-lights and water sources since the last GP election. The index is normalized to lie between 0 and 1. Roughly 79% of our sample villages experienced GP activism on at least one of these public goods. 3.5 We use data from household surveys in a random sub-sample of 193 villages to measure the provision of low spill-over (household) public goods. In every sampled village twenty one household surveys were conducted, of which four were with SC/ST households and one was with an elected Panchayat representative.

3 This section summarizes results Timothy Besley, Rohini Pande, Lupin Rahman and Vijayendra Rao. (2004a), “The Politics of Public Good Provision: Evidence from Indian Local Governments,” Journal of the European Economics Association, 2(2-3), 416-426.

15

3.6 An additional household survey was conducted with the Pradhan if s/he resided in that village, and with a ward member otherwise (in six villages both a ward member and Pradhan interview were conducted). 3.7 This gives us a total of 4059 households of which 981 were SC/ST. We measure a household's exposure to low spill-over public goods by a dummy which equals one if it had a house or toilet built under a government scheme or if it received a private water or electricity connection via a government scheme since the last GP election. Approximately 7% of the sample households fall in this category. 3.8 We are interested in the implications of political reservation and Pradhan proximity for the allocation of high and low spill-over public goods across and within villages. We capture a village's reservation status by an indicator variable which equals one if the village belongs to a GP reserved for SC/ST. We use two variables to measure the political influence of a village - the first equals one if the Pradhan resides in that village, and the second equals one if the GP headquarters are in that village. Household Level Evidence 3.9 The results are reported in Table 3.2, columns (1) through (4). In column (1) we see that, in line with scheme guidelines, household (i.e. low spill-over) public goods are targeted towards SC/ST households - on average, a SC/ST household is 6 percent more likely to receive such a public good. In column (2) we find that the extent of such targeting is enhanced by living in a reserved GP. Relative to living in a non-reserved GP, living in a reserved GP increases a SC/ST household's likelihood of getting such a public good by 7 percentage points. In columns (3) and (4) we examine whether the targeting of a SC/ST household is affected by location in the Pradhan’s village or in the GP headquarter. The results show that these two locations do not affect targeting. This suggests that enhanced targeting of SC/ST households only comes from reservation. We have seen so far that SC/ST Pradhans allocate low-spillover public goods to SC/ST households within villages. Now we move to investigate the allocation across villages.

Village Level Evidence 3.10 In our household-level regressions (columns (1)-(4)) we controlled for all village characteristics by using village fixed effects. The magnitude of the village fixed effects is in fact a village-level measure of household public goods provision. In columns (5) and (6) we examine whether village level political power influences this measure. None of our measures of political power - whether the Pradhan position is reserved for SC/ST, whether it is the Pradhan's village and/or GP headquarters - affects village-level allocation of household public goods. Household public goods have low spill-overs and are targeted towards SC/STs. Hence we expect non-SC/ST and SC/ST Pradhans' to differ in their propensity to allocate resources towards such public goods. Given this, it is unsurprising that the overall incidence of targeted public goods is unrelated to Pradhan's residence. However, it is surprising that this is also the case when the Pradhan position is reserved for SC/ST. It appears that political reservation is relevant for within-village allocation of low spill-over goods but not for overall village allocation.

16

3.11 Columns (7) and (8) consider the village incidence of high spill-over public goods, as measured by the GP activism index. We find that this index is, on average 0.04 points, higher in the Pradhan's village and not significantly different in reserved GPs. The fact that these public goods are high spill-over is consistent with the finding that the reservation status of the GP does not affect the extent of village-level provision.

17

Household Level Data Mean S.d.

Targeted Schemes 0.072 [0.258]

SC/ST Household 0.242 [0.428]

SC/ST Household*Pradhan reserved for SC/ST 0.066 [0.248]

SC/ST Household*Pradhan Village 0.098 [0.297]

SC/ST Household*GP headquarters 0.074 [0.261]

Muslim 0.044 [0.205]

Christian 0.009 [0.096]

Elected Officials' Household 0.049 [0.216]

SC/ST*Elected Officials' Household 0.010 [0.100]

Proportion Landless 0.312 [0.463]

Age of Household Head 48.001 [14.623]

Whether Household Head Literate 0.636 [0.481]

Household Size 5.336 [2.386]

Proportion Household Farmers 0.673 [0.469]

Village Level Data

Non-Targeted Schemes 0.443 [0.315]

Proportion SC/ST Households 0.298 [0.255]

Pradhan Village 0.421 [0.494]

Pradhan reserved for SC/ST 0.210 [0.408]

Pradhan Village*Pradhan reserved for SC/ST 0.094 [0.292]

GP headquarters 0.367 [0.482]

Log Total Population 7.266 [0.971]

Log Village Area 6.375 [0.978]

Proportion Area Irrigated 0.137 [0.150]

Proportion Landless 0.304 [0.248]

Literacy Rate 0.342 [0.133]

Distance From Nearest Town 19.435 [15.612]

Male Agricultural Wage Rate 48.023 [11.950]

TABLE 3.1: Summary Statistics

18

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

SC

/ST

Hou

seho

ld0.

066*

**0.

048*

**0.

041

0.03

4(0

.014

)(0

.016

)(0

.025

)(0

.025

)S

C/S

T H

ouse

hold

*Pra

dhan

rese

rved

for S

C/S

T0.

071*

*0.

071*

*0.

064*

*(0

.031

)(0

.031

)(0

.032

)S

C/S

T H

ouse

hold

*Pra

dhan

vill

age

0.03

0.03

2(0

.025

)(0

.025

)S

C/S

T H

ouse

hold

*GP

hea

dqua

rters

-0.0

19-0

.019

(0.0

25)

(0.0

25)

Pro

porti

on S

C/S

T H

ouse

hold

s-0

.007

-0.0

170.

041

0.07

7*(0

.027

)(0

.027

)(0

.042

)(0

.045

)P

radh

an V

illag

e-0

.02

-0.0

260.

048*

*0.

044*

(0.0

20)

(0.0

21)

(0.0

23)

(0.0

24)

Pra

dhan

rese

rved

for S

C/S

T-0

.003

-0.0

02-0

.003

-0.0

24(0

.012

)(0

.013

)(0

.039

)(0

.039

)P

radh

an V

illag

e*P

radh

an re

serv

ed fo

r SC

/ST

-0.0

03-0

.008

0.00

3-0

.002

(0.0

28)

(0.0

30)

(0.0

51)

(0.0

52)

GP

hea

dqua

rter

-0.0

03-0

.007

0.04

1*0.

02(0

.012

)(0

.014

)(0

.023

)(0

.025

)C

ontro

lsno

nono

yes

noye

sno

yes

Fixe

d ef

fect

svi

llage

villa

gevi

llage

villa

gebl

ock

bloc

kbl

ock

bloc

kO

bser

vatio

ns40

5940

5940

5940

5919

317

439

536

6R

-squ

ared

0.1

0.11

0.11

0.11

0.43

0.46

0.67

0.68

TAB

LE 3

.2: E

ffect

of S

C/S

T R

eser

vatio

n on

reso

urce

allo

catio

n

Not

es: T

he d

epen

dent

var

iabl

e in

col

umns

(1)-

(4) i

s a

dum

my

varia

ble

whi

ch e

qual

s on

e if

the

hous

ehol

d's

hous

e or

toile

t was

bui

lt un

der a

gov

ernm

ent s

chem

e or

if it

rece

ived

a p

rivat

e w

ater

or

elec

trici

ty c

onne

ctio

n vi

a a

gove

rnm

ent s

chem

e si

nce

the

last

GP

ele

ctio

n. T

he d

epen

dent

var

iabl

e in

col

umns

(5)-

(6) i

s th

e vi

llage

fixe

d ef

fect

from

col

umn

(4) r

egre

ssio

n (e

xclu

ding

the

cons

tant

). Th

e de

pend

ent v

aria

ble

in c

olum

ns (7

) - (8

) is

an in

dex

of w

heth

er G

P u

nder

took

any

con

stru

ctio

n or

impr

ovem

ent a

ctiv

ity o

n ro

ads,

dra

ins,

stre

etlig

hts

and

wat

er s

ourc

es a

fter t

he la

st G

P e

lect

ion.

The

S

C/S

T H

ouse

hold

dum

my

equa

ls 1

for S

C/S

T ho

useh

olds

. The

Pra

dhan

vill

age

dum

my

equa

ls o

ne if

the

Pra

dhan

resi

des

in th

e gi

ven

villa

ge. T

he G

P h

eadq

uarte

r dum

my

equa

ls 1

if th

e G

P

head

quar

ter i

s lo

cate

d in

the

villa

ge. I

ndiv

idua

l con

trols

incl

uded

are

dum

mie

s fo

r if h

ouse

hold

is M

uslim

and

Chr

istia

n, h

ouse

hold

siz

e, a

ge, l

itera

cy a

nd o

ccup

atio

n of

hou

seho

ld h

ead

and

whe

ther

it is

th

e ho

useh

old

of a

n el

ecte

d pa

ncha

yat o

ffici

al (a

lone

and

inte

ract

ed w

ith d

umm

y fo

r bei

ng a

SC

/ST

hous

ehol

d. V

illag

e co

ntro

ls in

clud

ed a

re p

ropo

rtion

of l

andl

ess

hous

ehol

ds, l

og to

tal v

illag

e po

pula

tion,

log

villa

ge a

rea,

pro

porti

on o

f irr

igat

ed la

nd,

villa

ge li

tera

cy ra

te, d

ista

nce

from

nea

rest

tow

n, a

nd d

aily

mal

e ag

ricul

tura

l wag

e ra

te.

All

villa

ge c

ontro

ls e

xcep

t for

the

agric

ultu

ral w

ages

are

from

199

1 C

ensu

s of

Indi

a. A

gric

ultu

ral w

ages

are

from

sur

vey

data

. Var

iatio

n in

sam

ple

Rob

ust s

tand

ard

erro

rs in

bra

cket

s. *

sig

nific

ant a

t 10%

; **

sig

nific

ant a

t 5%

; ***

sig

nific

ant a

t 1%

Vill

age

fixed

effe

ctV

illag

e pu

blic

goo

dsH

ouse

hold

pub

lic g

oods

Hou

seho

ld re

gres

sion

Vill

age

leve

l

19

4. GRAM SABHAS AND POLITICAL PARTICIPATION4

4.1 The gram sabha is the lynchpin of the panchayat system. It has the potential to structure democratic institutions to ensure a fair and efficient allocation of public funds. The idea that encouraging citizen participation can improve the workings of a democracy is also echoed in the political science literature. One role for participation emphasized in that literature is to improve the flow of information into the political process beyond that available by electing representatives. Thus, Verba et al.(1995) characterize political participation as "information rich" acts and observe that:

"From the electoral outcome alone, the winning candidate cannot discriminate which of dozens of factors, from the position taken on a particular issue to the inept campaign run by the opposition ..., was responsible for the electoral victory." (page 10).

4.2 This paper studies an institution to encourage political participation among the poor and to improve the quality of governance in an Indian context - Gram Sabha meetings. These are village meetings called by the elected local government (Gram Panchayat) to discuss resource allocation decisions in the village. 4.3 The 73rd Constitutional Amendment Act of India in 1993 made it mandatory for Indian states to hold elections for Gram Panchayats and to give them policy-making powers. 4.4 There are two main ways in which such meetings may improve the workings of government. First, relative to elected representatives, these meetings may better reflect citizens' preferences on issues such as how to target resources to the neediest groups. Second, by providing a forum for monitoring the actions of elected representatives they may reduce agency problems in politics, and the extent of corruption. 4.5 While holding Gram Sabhas is compulsory, their frequency and content owes a lot to the discretion of elected officials. Officials from the State or District administration can also have a role in this by choosing not to attend, and therefore making the gram sabha less attractive to hold. It is also the case that a well-attended meeting may have no bite on policy decisions. We exploit our household and village surveys to examine the determinants of participation in Gram Sabhas, and whether having a Gram Sabha affects beneficiary selection for welfare programs.

4.6 While there is much interest in how participation improves the quality of governance in the developing world (see, for example, Manor (2004)), evidence on the determinants of participation at the household level is thin, especially compared to the extensive studies available for the advanced democracies. Moreover, the literature is replete with concerns about elite dominance of democratic institutions. (Bardhan and Mookherjee (2000) and Platteau and Abraham (2004))

4 This paper summarizes results from: Besley, Timothy, Rohini Pande and Vijayendra Rao, [2005], “Participatory Democracy in Action: Survey Evidence from India," forthcoming in the Journal of the European Economics Association.

20

4.7 This raises the specter of participatory institutions being a veil which have little impact on the well-being of the poor. Here, however, we find that it is the most disadvantaged groups who attend village meetings and that holding such meetings improves the targeting of resources towards the neediest groups. 4.8 Our findings contribute to a broader debate about the role of decentralized governance in improving the quality of government in the developing world. The merits of decentralization have been widely debated -- see, for example, Bardhan (2002) and Triesman (2002). However, it is clear that many institutional details, even within decentralized governance, can be important. The use of village meetings of the kind studied here is one. It is important to understand how these institutional differences affect the way in which government operates. 4.9 In our survey, in every village, we conducted group meetings in which we obtained information on the last Gram Sabha meeting, and also village-level demographic and economic variables. In a random sub-sample of 259 villages we conducted twenty household surveys, and obtained information on Gram Sabha attendance and household beneficiary status. 4.10 Table 4.1 reports descriptive statistics. The average village has 328 households, of which 34 percent are landless. Twenty percent belong to the traditionally well-off upper castes and 28 percent to the historically disadvantaged scheduled castes and tribes (now on, SC/ST). According to the 1991 census literacy rate in our sample villages averaged 41 percent, but as is well known was much higher in Kerala villages. Seventy five percent of the villages had at least one Gram Sabha meeting in the last year, and in 22 percent of these meetings beneficiary selection was discussed. 4.11 In our household data-set we observe that while over 50 percent of the respondents had heard of a Gram Sabha only 20 percent had ever attended a Gram Sabha meeting. We also collected information on a household's beneficiary status, as defined by whether it has a `Below Poverty Line' (BPL) card. The GP, in collaboration with state government officials, is supposed to identify (via a census) households with income below the poverty line, and to give these households a BPL card. Possession of this card makes the household eligible for an array of government schemes, ranging from subsidized food through the public distribution system to free hospitalization. The list of BPL households, and subsequent selection of beneficiary households under various schemes is supposed to be ratified in Gram Sabha meetings.

4.12 The analysis is in two parts. We first study the determinants of holding a Gram Sabha meeting and who attends. We then look for evidence that holding a Gram Sabha meeting affects public resources allocation.

Determinants of holding a Gram Sabha and who attends: 4.13 The results of the analysis are reported in Table 4.2, column (1). More populous villages are more likely to have had a Gram Sabha meeting, and there is weak evidence that villages with higher literacy rate are more likely to hold Gram Sabha meetings. Interestingly,

21

after conditioning on matched block pair effects we don't observe any significant state differences in the decision to have a Gram Sabha. 4.14 In Columns (2)-(5) we use our household data to examine who has heard of, and who attends Gram Sabha meetings. 4.15 Village literacy rate is positively correlated with both hearing of the Gram Sabha and attending it. We also find evidence of significant state effects, with respondents from Kerala much more likely to have both heard of Gram Sabha meetings and participated in them. However, in the case of individual characteristics we observe significant differences in who has heard of and who attends Gram Sabha meetings. Moreover, various measures of economic and social disadvantage have a differential impact on the propensity to attend Gram Sabhas. Women and illiterates are less likely to both hear of and attend these meetings. In contrast, SC/STs and the landless are more likely to attend Gram Sabha meetings but no more likely to have heard of Gram Sabhas. Wealthy and upper castes, on the other hand, are more likely to have heard of Gram Sabhas but not to attend. 4.16 In column (4) we show that the individual characteristics continue to have a significant effect even when we control for all village characteristics, through fixed effects. Finally, in column (5) we examine whether village literacy particularly affects the likelihood of the disadvantaged to attend. The results of the estimation with interaction terms imply that higher village literacy increases the likelihood of illiterates, landless, and to a lesser extent SC/STs to participate. Women however, are not more likely to participate in Gram Sabhas in higher literacy villages. 4.17 These findings are notable for two reasons. First, there is some suggestion of a political externality from living in a more literate community. Second, Gram Sabha meetings seem to be a forum used by some of the most disadvantaged groups in the village - landless and scheduled castes/tribes. This suggests that these groups find the Gram Sabha useful and that Gram Sabha meetings may play some role in moving policy in a direction favored by these groups. We now look for evidence of the latter.

Does participation matter? 4.18 There are many who argue that participation in the political process has an intrinsic benefit. It builds trust in government and legitimizes state action. Unfortunately, our data do not permit us to look at these issues. However, we will look at the possibility that participation in Gram Sabhas yields instrumental (i.e. policy) benefits. These could be community-wide or by targeting resources to more specific groups. Here, we will focus on the latter, examining whether targeting of public programs are related to whether a Gram Sabha meeting has been held in the past twelve months. 4.19 We focus on an important specific policy administered at the village level -- access to a below poverty line (BPL) card. Beneficiary selection for such cards is influenced by the GP. As discussed earlier, possession of this card gives a villager access to an array of public benefits. We estimate a household regression which exploits within village variation in individual

22

characteristics to examine whether the targeting of BPL cards differs depending on whether the village had a Gram Sabha in the last year. 4.20 The results are reported in Table 4.3. In column (1) we report the baseline regression which does not include any interaction terms. This shows, not surprisingly, that BPL cards are targeted towards landless, illiterate and SC/ST households. In column (2) we include interactions between measures of disadvantage and whether the village had a Gram Sabha meeting. We find targeting of landless and illiterate individuals is more intensive in villages that had held a Gram Sabha meeting. Moreover, these effects are economically significant with an 8-10% increase in the probability of receiving a BPL card in a village that held a Gram Sabha. We find similar, but statistically insignificant, evidence for SC/STs. 4.21 These results do show persuasively that there is heterogeneity in targeting BPL cards across villages. Moreover, it would be tempting to attribute this to whether a Gram Sabha meeting is held. However, some caution is warranted. In column (3), we interact the characteristics that represent disadvantage - illiteracy, landlessness and schedule caste/tribe -- with the village literacy rate instead of whether the village had a Gram Sabha meeting. All three of these interactions are significant. This does raise the possibility that holding a Gram Sabha meeting is correlated with other village characteristics that are important in shaping the way in which public resources are targeted. Therefore we cannot say that holding a Gram Sabha has a causal effect on targeting. This is not an issue we can resolve with the existing data. However, these encouraging results on Gram Sabhas clearly deserve further careful investigation. 4.22 In conclusion, while this paper focuses on a specific institution -- the Gram Sabha, the results contribute to a wider debate on how institution design can shape public resource allocation and how the poor can increase their voice in public institutions. It is frequently remarked that poverty is much more than material deprivation and that the poor may receive much less voice in the political process. Moreover, a good deal of cynicism attends initiatives to strengthen that voice. 4.23 While the context is very specific, our results sound a more optimistic note. The illiterate, landless and SC/STs are significantly more likely to attend Gram Sabha meetings than other groups. Moreover, there appears to be more targeting towards these groups where Gram Sabha meetings are held. The results are also suggestive of some externalities from literacy in the political process at the village level. 4.24 Less optimistically, it is clear that Gram Sabhas are not a forum for women in their current form. Women respondents are around 20% less likely to attend a Gram Sabha than men. Whether this has significant consequences for public resource allocation needs further investigation. But it is clear the representativeness of Gram Sabhas is likely to be affected by this. Other tools such as gender reservation in Panchayat representation may go some way towards remedying this (see Chattopadhyay and Duflo (2004a) and Besley et. al. (2004b)).

23

Table 4.1:Descriptive Statistics

Overall Andhra Pradesh Karnataka Kerala Tamil NaduVillage level dataTotal households 328.10 305.50 365.80 401.10 227.40

Fraction of households which are 0.34 0.25 0.23 0.48 0.41landlessFraction of households which are 0.28 0.23 0.41 0.21 0.22SC/STFraction of households which are 0.20 0.13 0.32 0.12 0.19Upper casteLiteracy Rate in 1991 0.41 0.24 0.37 0.63 0.35

Fraction of villages which had a 0.76 0.71 0.68 0.98 0.67Gram Sabha in last yearFraction of Gram Sabhas at which 0.22 0.21 0.33 0.30 0.02beneficiary selection was discussed

Household level dataHeard of Gram Sabha 0.53 0.29 0.42 0.93 0.37

Ever attended Gram Sabha 0.20 0.11 0.14 0.40 0.13

Possess a BPL Card 0.22 0.32 0.10 0.30 0.25All variables based on survey data, except the village literacy rate which is from the 1991 Census of India

24

Table 4.2: Gram Sabha: Occurrence and AttendanceVillage had Household data: Gram Sabha

Gram sabha Heard of Attended (1) (2) (3) (4) (5)

Literacy Rate in 1991 0.328 0.323*** 0.235***(0.246) (0.118) (0.073)

Total number of households 0.093*** -0.001 0.006(0.030) (0.014) (0.010)

Fraction landless households 0.044 -0.017 -0.067**(0.086) (0.047) (0.032)

Fraction upper caste households -0.079 0.056 -0.011(0.116) (0.047) (0.032)

Fraction SC/ST households 0.03 0.021 -0.019(0.104) (0.041) (0.029)

Pradhan position reserved 0.01 0.043** -0.003(0.042) (0.020) (0.015)

Village Had Gram Sabha 0.026 0.030**(0.023) (0.014)

Illiterate -0.129*** -0.027** -0.030** -0.103***(0.015) (0.012) (0.013) (0.028)

Illiterate*literacy rate in 1991 0.183**(0.078)

SCST 0.001 0.021 0.034** -0.029(0.019) (0.016) (0.017) (0.040)

SCST*literacy rate in 1991 0.139(0.097)

Landless -0.012 0.041*** 0.030** -0.073**(0.014) (0.012) (0.012) (0.029)

Landless*literacy rate in 1991 0.232***(0.066)

Female -0.214*** -0.182*** -0.187*** -0.086***(0.014) (0.012) (0.014) (0.030)

Female*literacy rate in 1991 -0.242***(0.076)

Upper caste 0.035** 0.013 -0.004 -0.007(0.018) (0.016) (0.017) (0.018)

Wealthy 0.057*** -0.049*** -0.035** -0.027*(0.016) (0.014) (0.015) (0.016)

Andhra Pradesh -0.018 -0.171*** -0.168***(0.091) (0.048) (0.035)

Karnataka -0.089 -0.153*** -0.156***(0.063) (0.033) (0.032)

Tamil Nadu 0.019 -0.161*** -0.188***(0.061) (0.037) (0.029)

Fixed effects Block pair Block pair Block pair Village VillageObservations 476 4445 4935 5455 5240R-squared 0.22 0.39 0.17 0.25 0.25

Standard errors in brackets clustered at GP level in column (1) and at village level in all other regressions. Wealthy is a dummy for consumer durable ownership. Columns (2)-(4) also include respondent age and age squared as controls.* denotes significant at 10%; ** significant at 5%; *** significant at 1%

25

Table 4.3: Gram Sabha Occurrence and Beneficiary SelectionReceived BPL card

(1) (2) (3)Illiterate 0.028* -0.042* -0.057*

(0.015) (0.026) (0.030)Illiterate*Gram Sabha held 0.091***in last year (0.030)Illiterate* literacy rate in 1991 0.206***

(0.072)SCST 0.150*** 0.094** -0.03

(0.020) (0.042) (0.044)SCST*Gram Sabha held 0.062in last year (0.047)SCST* literacy rate in 1991 0.430***

(0.097)Landless 0.075*** 0.018 -0.098***

(0.016) (0.030) (0.035)Landless* Gram Sabha held 0.067*in last year (0.035)Landless*literacy rate in 1991 0.386***

(0.081)Female -0.011 -0.009 -0.005

(0.010) (0.010) (0.010)Upper caste -0.028* -0.028* -0.036**

(0.017) (0.016) (0.017)Wealthy -0.082*** -0.079*** -0.066***

(0.014) (0.014) (0.014)Fixed effects Village Village Village

Number of observations 5455 5364 5039R-squared 0.4 0.4 0.42Robust standard errors, clustered by village, in brackets. All regressions include respondent age and age squared as controls. * significant at 10%; ** significant at 5%; *** significant at 1%

26

5. POLITICAL SELECTION AND THE QUALITY OF GOVERNMENT5: 5.1 Common sense discussions of political life often place the quality of politicians at center stage. Yet the modern political economy literature remains dominated by a paradigm in which good policy is achieved solely by getting incentives right rather than by improving the quality of the political class. While incentives are important, personal qualities of politicians such as honesty, integrity and competence are also potentially important, especially in environments where politicians face limited formal sanctions. 5.3 We test these ideas using the data that we have collected from both politician and non-politician households. We also have information on a host of village institutions. We examine institutions which affect the identity of the politically dominant group, those determining returns to politics and finally, those affecting information flows. 5.4 There are two main components to our empirical analysis. We begin by studying politician characteristics -- the "selection equation" for politicians, and how these are affected by village institutions. Second, we look at which characteristics make politicians better policy makers -- specifically, showing less opportunism in relation to public programs. Here again, we examine the role of village institutions. 5.5 Our paper also contributes to a growing empirical literature on decentralized government in the developing world. There is emerging evidence that decentralization affects resource allocation. Faguet (2004) finds that decentralization improved targeting in Bolivia. Bardhan and Mookherjee (2003) examine the role of elected village councils in affecting land reform in the Indian state of West Bengal. Chattopadhyay and Duflo (2004a) show political reservation for women affected public good allocation in two Indian states. Finally, Foster and Rosenzweig (2001) examine how decentralization interacts with land ownership patterns to affect public good outcomes. None of these papers focuses on how politician's characteristics affect this process. 5.6 The results are presented in two parts. We first examine determinants of politician selection, and then at how policy is determined. In Table 5.1 we report some descriptive statistics. Politicians have 3.1 more years of education and 3.7 more acres of land than the average respondent. Furthermore, politicians are almost 4 times more likely to have a member of family in politics than the average respondent. Only 36 percent of the respondents believe that their pradhan kept their electoral promises. 8.7 percent say that their voting choices were most importantly determined by group identity (religion, caste, gender or region), while 36 percent state that they vote on their perception of the candidate’s quality. Note also that 21% of households possess a BPL card. As for village characteristics , it can be noted that a dominant caste – one that comprises at least 40 percent of the villagers – exists in more than half the villages (51.9 percent). In 78 percent of the villages a Gram Sabha was held in the last year and the average literacy rate across these villages is 42 percent. 5 This section summarizes results from Timothy Besley, Rohini Pande and Vijayendra Rao, “Political Selection and the Quality of Government: Evidence from South India,” mimeo, May 2005

27

Political Selection: 5.8 We start by examining the household data to see if particular types of individuals and households are more likely to become politicians – these regressions are reported in Table 5.2. All regressions in this table either control for village or GP fixed effects, thus the results examine variation within a village or GP. In columns (1) and (2) the dependent variable is whether the respondent is an elected GP politician( i.e. a Pradhan or ward member). Being eligible for reservation is not significantly correlated with being a politician. However, years of education and land ownership are positively correlated with being a politician. In addition, a respondent from a family with a history of political participation is 12% more likely to be a politician, and years of education and more land are both associated with a higher chance of being a politician. 5.9 In columns (2) and (3) we restrict the sample to the groups eligible for reservations – women and SC/STs. For both groups we observe positive selection of education, but not on land. We find that family political history is correlated with selection only for women. The absence of any impact from political history for SC/STs possibly reflects their relative lack of political experience. Columns (4) – (6) conduct the same analysis restricting the dependent variable to becoming a Pradhan and the results are similar. 5.10 In Table 5.3 we look at how village institutions affect the process of political selection. We are specifically interested in how different measures of political dominance influence the characteristics of elected politicians. We do this by interacting the institutional variable with individual characteristics and examining the effect of the interaction on the likelihood of being elected politician. In column (1) we consider the existence of a dominant caste and its effect on selection. The positive and significant coefficient on the interaction with land owned is evidence that in villages which have a dominant caste, individuals owning more land are more likely to be elected politicians. Columns (2) and (3) examine the effect of reservations – examining women’s reservation and SC/ST reservation in turn. In both cases we observe that, relative to other politicians, reserved politicians are less educated, own less land, and are less likely to come from families with political experience. This, we think, reflects the historic economic, social and political disadvantages faced by low castes and women. 5.11 In column (4) we examine the impact of the pradhan’s salary on selection to see whether higher formal returns from electoral politics cause more affluent politicians to enter politics. We observe that politicians in villages with relatively higher Pradhan salary own more land. In column (5) we see if there is a macro information effect that comes from belonging to a more literate village on political selection. We see that relatively more educated politicians are likely to be elected from more literate villages. Further, respondents belonging to groups eligible for reservation are more likely to enter politics in such villages. However, this effect is not significant. 5.12 Overall, the results suggest that village institutions that reduce the dominance of major castes increase the presence of economically disadvantaged groups in politics. Further higher returns to political office encourages the selection of wealthier politicians. In addition, more literate villages elect better educated leaders.

28

Policy Effects

5.13 We now examine how political selection affects the targeting of private goods, namely, BPL cards, provided by GPs. Table 5.4, column (1) demonstrates that BPL cards are indeed targeted towards economically disadvantaged households. Specifically, a SC/ST household is 16% more likely to get a BPL card while households with a more educated head and/or more land holdings are less likely. Finally, households with a family political history are no more likely to get a BPL card. But, being a politician helps. In column (2) we observe that a politician household is 7.5% more likely to have a BPL card. This is all the more striking in view of the results in Table 5.2 which demonstrated that politician households are more likely to be landed and educated. In column (3) we examine the role of politician characteristics. Political opportunism is invariant to most politician characteristics, except education. Political opportunism is lower among more educated politicians. An extra year of education for a politician makes him or her 1.4% less likely to have a BPL card. 5.16 In Table 5.5 we examine the role of village institutions in constraining political opportunism focusing once again on the probability of obtaining a BPL card. Column (1) considers the implications of caste dominance. The presence of dominant caste in the village make it more likely that a politician will have a BPL card. Column (2) looks at women’s reservation. The likelihood of a politician having a BPL card is higher when the pradhan is a woman. In contrast, column (3) shows that SC/ST reservations make it more likely that SC/ST households will have a BPL card, and more likely that reserved politicians will also have one. This shows the salience of SC/ST reservations in improving the access of SC/STs to anti-poverty programs. 5.17 Column (4) examines the implications of variations in the pradhan’s real salary. Higher salaries are associated with no change in political opportunism, but the targeting of socially and economically disadvantaged groups is improved. In column (5) we see whether holding a gram sabha meeting, which in theory should increase transparency, has an impact on opportunism. We see a significant reduction in opportunism when a gram sabha is held. Similar effects obtain with increases in village literacy as shown in column (6). This extends some of the results we reported earlier in our analysis of gram sabhas. Columns (5) and (6) also show that improved literacy and having a Gram Sabha improves targeting of disadvantaged household. SC/ST and economically disadvantaged household are more likely to receive a BPL card in villages with higher literacy and in villages where Gram Sabha was held. 5.18 Taken together these results demonstrate the importance of incentives, transparency, and education in affecting public resource allocation. More educated leaders are less opportunistic, as are those who are paid higher salaries and belong to GPs that hold gram sabhas and have higher levels of literacy.

29

Summarizing the Results on Political Selection 5.26 This section has three key findings. First, the political class is selected on the basis of political connections and economic advantage. Second, politicians are on the whole opportunistic and benefit disproportionately from public transfer programs. Third, the education level of politicians has a consistently positive effect on selection and a negative effect on opportunism. This suggests that more educated politicians are better. However, whether education matters directly or because it is correlated with other characteristics that make an individual fit for public office cannot be discerned from our results. Nonetheless, the results add to a growing appreciation among economists that education may be important because of its role in inculcating civic values. The unique observation about its role in politics given here also offers a fresh perspective on the value of human capital investments in low income countries. 5.27 The results demonstrate important interplays between village level variables and the process of political selection, and the targeting of public resources. For example, increased literacy at the village level reduces political opportunism while measures of political dominance are correlated with targeting of resources. We also find evidence suggestive of barriers to entry, as individuals owning less land or having no political connections are less likely to be elected. Land ownership and political connections predict selection but not behavior when in office. 5.28 Our finding that educated politicians are better in terms of actual performance suggests that it is important to focus on factors that select better politicians as step toward improving the quality of government. More generally, the results and analyses in the paper reinforce the observation that formal institutions of democracy are no guarantee of effective government. It is essential that preconditions exist for sorting in the right kinds of people – the talented, the virtuous and those who give political voice to the disadvantaged. There is clearly much more we can learn about this process, but these results are a first effort to study the issue empirically.

30

Mean s.d.Respondent characteristicsYears of Education All 4.49 (4.54)

Politicians 7.58 (4.51)Land owned (in acres) All 2.26 (4.77)

Politicians 5.98 (8.87)Eligible for reservation (%) All 60.90 (48.81)

Politicians 48.70 (50.07)Family political history (%) All 6.70 (25.04)

Politicians 25.30 (43.54)Beneficiary StatusBPL card (%) All 21.70 (41.20)

Politicians 24.20 (42.80)

Perceptions and Voting Behavior (% non-politicians) Pradhan looks after village needs 38.40 (48.63)Pradhan keeps election promises 36.10 (48.03)

Vote for group identity 8.72 (28.22)Vote for candidate quality 36.08 (48.02)Institutions (% villages)Dominant caste 51.93 (50.05)

Pradhan reserved for Female 15.89 (36.63)

Pradhan reserved for SC/ST 16.66 (37.34)

Literacy rate 42.20 (18.35)

Gram Sabha 77.95 (41.53)Notes:

Table 5.1: Descriptive Statistics

1. Years of education refer to respondent's years of education. Land owned is amount of land, in acres, owned by respondent's household. A respondent is eligible for reservation if female or SC/ST. A respondent has a family political history if any member of his/her household holds or as held a political position. BPL card refers to whether the household has a BPL card.

2. Vote dummies refer to GP election. Vote for group identity=1 if respondent says she voted for the candidate with the same caste/religion/gender/place of residence. Vote for candidate quality=1 if respondent says she voted for candidate with good policy promises/candidate active in the village/good reputation.

3. A Village has a Dominant caste if over 40 percent of villagers belong to a single caste. Literacy rate is the 1991 census village literacy rate. Gram Sabha is a dummy for whether the village had a Gram Sabha meeting in the last year.

31

Table 5.2: Individual Characteristics and Politician Selection

Sample All Female SC/ST(1) (2) (3)

Eligible for 0.008reservation (0.007)

Education 0.008*** 0.007*** 0.012***(0.001) (0.001) (0.002)

Land owned 0.007*** 0.003 0.002(0.002) (0.002) (0.003)

Family political 0.119*** 0.135*** 0.062history (0.020) (0.032) (0.044)

Fixed effects Village Village GP

R-squared 0.09 0.12 0.12

N 5397 2644 1245

2. The dependent variable is an indicator variable=1 if the respondent is a politician.

3. All regressions include control for respondent age and age squared. The Pradhan regressions restrict the sample to the Pradhan and non politician households in the Pradhan's village.

4. Eligible for reservation is an indicator variable which equals one if respondent is female or SC/ST. Land ownership is the land (in acres) owned by the respondent's household. Education refers to respondent's years of education. Family political history is an indicator variable which equals one if any family member of respondent has held/holds a political position.

Politician

Notes:

1. OLS regressions with standard errors, clustered by village, in parentheses. * significant at 10%; ** at 5%; *** at 1%.

32

Inst

itutio

nD

omin

ant C

aste

Fem

ale

Pra

dhan

R

eser

vatio

nS

C/S

T P

radh

an

Res

erva

tion

Sal

ary

Lite

racy

Rat

e(1

)(2

)(3

)(4

)(5

)E

ligib

le fo

r res

erva

tion

0.01

3-0

.013

**-0

.009

-0.0

37-0

.012

(0.0

09)

(0.0

06)

(0.0

06)

(0.0

64)

(0.0

16)

Elig

ible

for r

eser

vatio

n*-0

.007

1.03

2***

1.03

2***

0.02

70.

05V

illag

e C

hara

cter

istic

(0.0

12)

(0.0

06)

(0.0

07)

(0.0

38)

(0.0

34)

Edu

catio

n0.

008*

**0.

006*

**0.

006*

**-0

.001

0.00

5**

(0.0

01)

(0.0

01)

(0.0

01)

(0.0

07)

(0.0

02)

Edu

catio

n*-0

.001

-0.0

06**

*-0

.003

***

0.00

50.

007*

Vill

age

Cha

ract

eris

tic(0

.002

)(0

.001

)(0

.001

)(0

.004

)(0

.004

)

Land

ow

ned

0.00

5**

0.00

6***

0.00

8***

-0.0

27*

0.00

2(0

.002

)(0

.002

)(0

.002

)(0

.015

)(0

.004

)La

nd o

wne

d*0.

005*

-0.0

06**

*-0

.007

***

0.02

1**

0.01

6V

illag

e C

hara

cter

istic

(0.0

03)

(0.0

01)

(0.0

02)

(0.0

10)

(0.0

11)

Fam

ily p

oliti

cal h

isto

ry0.

112*

**0.

083*

**0.

111*

**0.

037

0.06

7(0

.030

)(0

.019

)(0

.020

)(0

.216

)(0

.051

)Fa

mily

pol

itica

l his

tory

*0.

013

-0.0

76**

*-0

.131

***

0.05

00.

104

Vill

age

Cha

ract

eris

tic(0

.040

)(0

.020

)(0

.022

)(0

.132

)(0

.108

)

Fixe

d ef

fect

sV

illag

eV

illag

eV

illag

eV

illag

eV

illag

eR

-squ

ared

0.09

0.25

0.26

0.09

0.09

N53

9753

9753

9753

7651

87

3. R

egre

ssio

ns in

clud

e re

spon

dent

age

and

age

-squ

ared

as

a co

ntro

l var

iabl

e. E

xpla

nato

ry v

aria

bles

are

def

ined

in n

otes

to T

able

s 1

and

2.

4. D

omin

ant c

aste

is a

n in

dica

tor e

qual

to 1

if th

e vi

llage

has

a d

omin

ant c

aste

, as

defin

ed in

tabl

e 5.

1; S

alar

y is

log

Pra

dhan

sal

ary/

log

mal

e ag

ricul

tura

l wag

e; L

itera

cy is

the

villa

ge li

tera

cy ra

te in

the

1991

cen

sus

Tabl

e 5.

3: V

illag

e C

hara

cter

istic

s an

d P

oliti

cian

Sel

ectio

n

Not

es:

1. O

LS re

gres

sion

s w

ith s

tand

ard

erro

rs, c

lust

ered

by

villa

ge, i

n pa

rent

hese

s. *

sig

nific

ant a

t 10%

; **

at 5

%; *

** a

t 1%

.

2. T

he d

epen

dent

var

iabl

e is

an

indi

cato

r var

iabl

e=1

if th

e re

spon

dent

is a

pol

itici

an.

33

(1) (2) (3)SC/ST household 0.164*** 0.162*** 0.166***

(0.019) (0.019) (0.019)Household head's -0.008*** -0.008*** -0.008***education (0.002) (0.002) (0.002)Respondent's education -0.003* -0.003** -0.003*

(0.001) (0.001) (0.002)Land owned -0.004*** -0.004*** -0.003*

(0.001) (0.001) (0.001)Family political history -0.012 -0.021 -0.029

(0.020) (0.020) (0.019)Politician 0.075** 0.199**

(0.033) (0.080)Politician*Reserved -0.105

(0.071)Politician*Education -0.014**

(0.007)Politician*Land owned 0.001

(0.003)Politician*Family political 0.069history (0.083)Fixed effects Village Village VillageR-squared 0.36 0.36 0.36N 5366 5366 5366

3. All regressions include as household controls: household size, head's age and age squared, fraction eldeand fraction children. Other variables are as defined in Table 2 notes.

Table 5.4: Politician Characteristics and BPL Beneficiary Selection

Notes:

1. OLS regressions with standard errors, clustered by village, in parentheses. * significant at 10%; ** at 5%; *** at 1%.2. The dependent variable is an indicator variables which equals one if the respondent s household has a Bcard.

34

Inst

itutio

nD

omin

ant c

aste

Fem

ale

Pra

dhan

re

serv

atio

nS

C/S

T P

radh

an

rese

rvat

ion

Sal

ary

Lite

racy

rate

Gra

m S

abha

(1)

(2)

(3)

(4)

(5)

Pol

itici

an-0

.01

0.06

9*0.

101*

*0.

483*

0.39

9***

0.28

2***

(0.0

53)

(0.0

39)

(0.0

40)

(0.2

83)

(0.0

98)

(0.0

95)

Pol

itici

an*

0.18

5**

0.49

8**

-0.3

77*

-0.2

39-0

.746

***

-0.2

42**

Vill

age

Cha

ract

eris

tic(0

.079

)(0

.219

)(0

.209

)(0

.171

)(0

.188

)(0

.105

)R

eser

ved

polit

icia

n0.

035

-0.0

28-0

.098

1.06

9*-0

.144

-0.3

43**

(0.0

93)

(0.0

77)

(0.0

76)

(0.5

95)

(0.1

76)

(0.1

42)

Res

erve

d po

litic

ian*

-0.1

94-0

.547

**0.

409*

-0.4

01*

0.21

0.35

9**

Vill

age

Cha

ract

eris

tic(0

.135

)(0

.243

)(0

.232

)(0

.222

)(0

.338

)(0

.161

)S

C/S

T ho

useh

old

0.18

0***

0.14

5***

0.11

9***

-0.4

01*

-0.0

440.

108*

**(0

.025

)(0

.019

)(0

.026

)(0

.222

)(0

.040

)(0

.039

)S

C/S

T ho

useh

old*

-0.0

210

0.11

2**

0.34

7**

0.51

2***

0.07

2V

illag

e C

hara

cter

istic

(0.0

40)

(0.0

00)

(0.0

55)

(0.1

37)

(0.0

93)

(0.0

45)

Eco

nom

ic D

isad

vant

age

0.01

10.

092*

**0.

096*

**-0

.201

-0.0

180.

060*

**(0

.027

)(0

.015

)(0

.014

)(0

.158

)(0

.031

)(0

.019

)E

cono

mic

Dis

adva

ntag

e*-0

.001

-0.0

05-0

.065

0.18

3*0.

271*

**0.

045*

Vill

age

Cha

ract

eris

tic(0

.051

)(0

.020

)(0

.050

)(0

.098

)(0

.076

)(0

.025

)Fa

mily

pol

itica

l his

tory

-0.0

51*

-0.0

37*

-0.0

22-0

.149

0.02

20.

016

(0.0

28)

(0.0

22)

(0.0

21)

(0.2

02)

(0.0

42)

(0.0

35)

Fam

ily p

oliti

cal h

isto

ry*

0.04

80

-0.0

920.

076

-0.1

03-0

.058

Vill

age

Cha

ract

eris

tic(0

.040

)(0

.046

)(0

.065

)(0

.125

)(0

.096

)(0

.042

)Fi

xed

effe

cts

Vill

age

GP

GP

Vill

age

Vill

age

Vill

age

R-s

quar

ed0.

360.

30.

30.

370.

380.

36N

5369

5369

5369

5348

5159

5287

3. R

egre

ssio

ns in

clud

e th

e ho

useh

old

cont

rols

def

ined

in n

otes

to T

able

4. E

cono

mic

dis

adva

ntag

e is

a d

umm

y w

hich

equ

als

one

if th

e ho

useh

old

head

is il

liter

ate

or la

ndle

ss. O

ther

var

iabl

e de

finiti

ons

are

in n

otes

to T

able

s 1

and

2.

Form

al re

turn

s, li

tera

cy, a

nd in

form

atio

nP

oliti

cal d

omin

ance

Tabl

e 5.

5: V

illag

e C

hara

cter

istic

s an

d B

enef

icia

ry S

elec

tion

for B

PL

card

s

Not

es

1. O

LS re

gres

sion

s w

ith s

tand

ard

erro

rs, c

lust

ered

by

villa

ge, i

n pa

rent

hese

s. *

sig

nific

ant a

t 10%

; **

at 5

%; *

** a

t 1%

.2.

The

dep

ende

nt v

aria

ble

is a

n in

dica

tor v

aria

bles

whi

ch e

qual

s on

e if

the

resp

onde

nt's

hou

seho

ld h

as a

BP

L ca

rd.

35

6. POLICY IMPLICATIONS 6.1 The results from the four papers reported above have some important lessons for policy. We list some of them below:

a. Caste Reservations work by improving targeting of private transfers to schedule castes and tribes.

We find that programs that provide private benefits such as toilets, housing and transfers to the poor and disadvantaged (including provision of BPL card) are more likely to reach SC/STs when the GP has a Pradhan that is reserved for an SC/ST. This suggests that caste reservations are effective in including disadvantaged groups into the purview of local government. It supplements previous research that finds that woman Pradhans in seats reserved for women tend to make decisions more in line with the needs of women (Chattopadhyay and Duflo, 2004a). b. Pradhans prefer their home village: The home village of the pradhan tends to receive more high-spillover public goods than other villages in the GP controlling for factors such as village size and head quarter status. This result, a consequence of the incentives that underlie democracy, points to inequalities that may exist within GPs that could be persistent and may be important to address. c. Gram Sabhas may be central to effective local government but are not regularly held: When gram sabhas are held we find that benefits are better targeted to the poor and disadvantaged, and reduce political opportunism. Therefore they seem to improve the transparency of government. Further research will have to determine how this works and their implications for public goods allocation, but clearly they are potentially central to the effective and equitable functioning of GPs. The fact that they are often not held is worrying and needs attention. Also, while SC/STs are more likely to participate in gram sabhas, presumably because of their role in beneficiary selection, we find that women are far less likely to attend them. This is a potential source of gender exclusion that needs attention. d. Literacy Matters: Several results point to the importance of village literacy in improving the functioning of GPs – in reducing political opportunism, improving targeting, etc. We also find that more educated politicians are less opportunistic and perceived as better performing. Therefore, investments in human capital can be central to improving the quality of democratic governance in addition to their enhancing individual well-being. e. Finance Matters Corroborating findings from the recent World Bank report on fiscal decentralization in India (World Bank, 2004), we find that differences in the quality of local government between the four South Indian states are correlated with what we know of their levels of fiscal decentralization. In particular, Kerala has led the other states in providing public services at the local level but seems to be slipping more recently in a manner that concurs with its worsening fiscal situation. More generally we find that it is very difficult to understand the state of GP finances because of vast inconsistencies in accounting practices at the GP level.

36

GP budgetary data is therefore very difficult to obtain and even when it is available is difficult to compare and evaluate. f. Socio-Cultural Institutions Matter We show that villages demonstrate high levels of inequality within them, and that this is inequality is both within and between castes. We find evidence showing that caste dominance tends to increase political opportunism. g. Higher salaries may reduce opportunism A result with direct policy implications is that relatively higher real wages for politicians tend to attract wealthier politicians and to improve targeting of disadvantaged groups which suggests a reduction in political opportunism.

6.2 These findings provide some important insights into the political economy and the institutional setting for panchayats. In future research we hope to examine the role of land reform in reducing economic and social inequality, and the quality of government. We will also examine the determinants and implications of social, economic and political participation.

37

REFERENCES Alsop, Ruth, Anirudh Krishna, and Disa Sjoblom. (2000) “Inclusion and Local Elected Governments: The

Panchayat Raj System in India.” Social Development Paper 37. World Bank, South Asia Region, Social Development Unit, Washington, D.C.

Banerjee, Abhijit and Laksmi Iyer (2003). “History, Institutions and Economic Performance: The Legacy

of Colonial land Tenure Systems in India.” mimeo, MIT. Bardhan, Pranab, (2002) “Decentralization of Government and Development,” Journal of Economic

Perspectives 16(4), 185-205. Bardhan, Pranab and Dilip Mookherjee. (2000) “Capture and Governance at Local and National Levels.”

American Economic Review. 90(2), 135-139. Bardhan, Pranab and Dilip Mookherjee, (2003) “Political Economy of Land Reforms in West Bengal

1978-98 ." mimeo Boston University. Besley, Timothy, Rohini Pande, Lupin Rahman and Vijayendra Rao. (2004a), “The Politics of Public

Good Provision: Evidence from Indian Local Governments,” Journal of the European Economics Association, 2(2-3), 416-426.

Besley, Timothy, Rohini Pande, Vijayendra Rao and Radu Ban, (2004b), “Tokenism or Agency? The

Impact of Women’s Reservation on Panchayats in South India.” mimeo, Development Research Group, The World Bank

Besley, Timothy, Rohini Pande and Vijayendra Rao, [2005a], “Participatory Democracy in Action:

Survey Evidence from India," forthcoming in the Journal of the European Economics Association. Besley, Timothy, Rohini Pande and Vijayendra Rao.[2005b] “Panchayats and Resource Allocation: A

Comparison of the South Indian States,” mimeo Besley, Timothy, Rohini Pande and Vijayendra Rao [2005c]. “Political Selection and the Quality of

Government: Evidence from South India,” mimeo. Bhagavan, Manu. (2003) Sovereign Spheres: Princes, Education and Empire in Colonial India, Oxford

University Press, New Delhi. Chattopadhyay, Raghabendra and Esther Duflo, (2004a), “Women as Policy Makers: Evidence from a

India-Wide Randomized Policy Experiment,” Econometrica, 72(5), 1409-1444. Chaudhuri, Subham and Patrick Heller, (2004) "Building Local Democracy: Evaluating the Impact of

Decentralization in Kerala," Paper presented to Netsappe III, Paris, June 2004 Crook, Richard C. and James Manor. (1998) Democracy and Decentralisation in South Asia and West

Africa : participation, accountability, and performance. Cambridge ; New York : Cambridge University Press.

38

Dyson, Tim and Mick Moore (1983) ‘On Kinship Structure, Female Autonomy, and Demographic Behavior in India’ Population and Development Review Vol.9, No. 1, pp 35-60.

Faguet, Jean Paul, (2004), “Does Decentralization Increase Responsiveness to Local Needs? Evidence

from Bolivia,” Journal of Public Economics, 88: 867-894. Foster, Andrew and Mark Rosenzweig, (2001), “Democratization, Decentralization and the Distribution

of Local Public Goods in a Poor Rural Economy,” mimeo, Brown. Jeffrey, Robin (1993), Politics, Women and Well-Being: How Kerala Became "A Model,” Oxford

University Press, Delhi Manor, James. (2004) “Democratization with Inclusion: political reforms and people’s empowerment at

the grassroots,” Journal of Human Development, 5(1), 5-29. Matthew, George and Nirmala Buch. (2000) Status of Panchayati Raj in the States and Union Territories

of India 2000, Institute for Social Studies, Delhi. Platteau, Jean-Philippe and Anita Abraham, [2002], “Participatory Development in the Presence of

Endogenous Community Imperfections,” Journal of Development Studies, 39(2), 104-136. Triesman, Daniel. (2002) “Decentralization and the Quality of Government,” mimeo, UCLA. Verba, Sidney, Kay Lehman Scholzman and Henry E. Brady, (1995) Voice and Equality: Civic

Voluntarism in American Politics Cambridge Mass: Harvard University Press. World Bank (2004) "India: Fiscal Decentralization to Local Governments," Report No. 26654-IN, Rural

Development Unit, South Asia Region. World Bank (2000) Overview of Rural Decentralization in India, Vol 1, 2, and 3, Rural Development

Unit, South Asia Region.

39

ANNEX A:

Panchayats and Resource Allocation: A Comparison of the South Indian States

Panchayats and Resource Allocation: A

Comparison of the South Indian States�

Timothy Besley

LSE

Rohini Pande

Yale University

Vijayendra Rao

World Bank

Draft: April 2005

Contents

1 Introduction 3

2 Methodology 7

2.1 Sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2.2 Questionnaires . . . . . . . . . . . . . . . . . . . . . . . . . . 11

3 State Comparisons 12�Acknowledgements: We are grateful to Radu Ban, Lupin Rahman, Siddharth Sharma

and Jillian Waid for research assistance, and the IMRB sta¤ for conducting the survey.

We thank the World Bank�s Research Committee and the South Asia Rural Development

Unit for �nancial support. The opinions in the report are those of the authors and do not

necessarily re�ect the points of view of the World Bank or its member countries.

1

3.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

3.2 Evidence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

3.2.1 Cross Village Resource Allocation . . . . . . . . . . . . 21

3.2.2 Household targeting . . . . . . . . . . . . . . . . . . . 24

3.2.3 Participation, Information and Socio-Political Structure 26

4 Conclusions 32

2

1 Introduction

The 73rd amendment to the Indian constitution, passed in 1993, has been

one of the most important pieces of legislation in recent Indian history. Its

goals are:

a) To systematize the functioning of Panchayati Raj Institutions (PRIs)

by mandating regular elections to the three tiers of local self government,

and requiring states to both increase PRI taxation and spending power, and

PRIs allocation of state and central discretionary funds.

b) To ensure that disadvantaged groups within village communities are

granted a voice in local deliberations, the 73rd amendment also mandated

that 1/3rd of all elected positions in Panchayats, including Panchayat pres-

ident, be reserved for women. Similarly elected positions in Panchayats are

to be reserved for Scheduled Castes and Tribes in proportion to their popu-

lation share. No elected post should be reserved for the same group for two

consecutive elections.

All National governments since 1993 have been committed to the im-

plementation of the amendment, and State governments have complied with

varying degrees of commitment. The current United Progressive Alliance

(UPA) government in Delhi has gone even further and asserted in its Com-

mon Minimum Program that:

1) After consultations with States, the UPA Government will

ensure that all funds given to States for poverty alleviation and

rural development schemes by Panchayats are neither delayed nor

diverted. Monitoring will be strict. In addition, after consultation

3

with States, the UPA Government will consider crediting elected

Panchayats such funds directly.

2) Devolution of funds will be accompanied by a similar devo-

lution of functions and functionaries as well. Regular elections to

Panchayat bodies will be ensured and the amended Act in respect

of the Fifth and Sixth Schedule Areas will be implemented.

3) The UPA Government will ensure that the Gram Sabha is

empowered to emerge as the foundation of Panchayati Raj. "

Thus, there is likely to be a renewed emphasis on PRIs as a means of

providing public services to the poor, and thereby ensuring that rural com-

munities can bene�t from the gains to economic growth.

This experiment in decentralization is, arguably, one of the most ambi-

tious experiments in redesigning governance structures undertaken by a low

income country. The stated aim was to improve citizens�ability to access

and in�uence the public service delivery system and to directly tackle social

exclusion by a system of political reservations. Despite the breadth of this

democratic experiment, there is remarkably little quantitative evidence on

how well the experiment has worked. There is, however, a large and growing

qualitative and "action research" literature on Panchayats that come to a

diverse set of conclusions - re�ecting the di¢ culties of studying such a broad

topic in a such a complex country. A comprehensive review of this literature

is beyond the scope of this report but overviews can be found in World Bank

(2000), Matthew and Buch (2000), and Manor (1998).

Qualitative work has important strengths, but it also has important weak-

nesses (Rao and Woolcock 2003), central among which is its relative inability

4

to generate generalizable �ndings which are essential to a policy dialogue.

It is also more suited to demonstrating correlations or "a¢ nities" rather

than clear causal connections - for instance on the important question of the

impact of the reservations policy. Therefore, an informed policy dialogue

requires both qualitative and quantitative information.

Quantitative analysis of Panchayats using large samples are rare. An ex-

ception to this is the important work by Chattopadhyay and Du�o (2004a)

on the causal impact of women�s reservations on Panchayat action in Ra-

jasthan and West Bengal. They �nd that reservations improve the ability of

women to govern, in a way that is congruent with the desires of women in

the population. Work by Alsop, Krishna and Sjoblom (2000), also on Ra-

jasthan and Madhya Pradesh, highlights the role of reservation in reducing

the systematic exclusion of women and disadvantaged groups from decision

making processes at the local level. Chaudhuri and Heller (2004) have, more

recently, completed a survey studying the impact of the "People�s Campaign

for Decentralized Planning" in Kerala showing that it increased the level of

particpatory planniong in panchayats, had a positive impact on development

performance and on social inclusion, but that levels of participation have

declined in recent years - �ndings that are consistent with our study.

But, given the scope of the experiment and regional focus of the existing

quantitative work, a large number of open questions remain. How well has de-

centralization worked in early adopter states such as Kerala and Karnataka?

How do village Panchayats raises resources and implement policies? What

is the impact of caste reservations? Do village meetings open to all citizens

(Gram Sabha meetings) succeed in increasing the voice of the poor and dis-

5

advantaged? Answering these question are crucial in formulating Panchayat

policy.

The above questions also point to a crucial need for a sound, quantitative

evidentiary base to provide some answers to these questions. Quantitative

data collection can also allow us to establish a baseline regarding functioning

of PRIS that will permit researchers and policymakers to identify how public

service delivery via PRIs changes as PRIs get more resources and more powers

over time. These observations motivate the research that underlies this

report.

This paper is based on survey evidence collected by the authors in con-

junction with the World Bank in four Indian states (Andhra Pradesh, Tamil

Nadu, Karnataka and Kerala) in 2002. The survey focussed on the local tier

of elected self government �Gram Panchayats (GP).

Section 3 describes the sampling methodology and survey design in detail.

Section 4 describes the institutional di¤erences in PRIs across our four sample

states, and studies the di¤erences in the e¤ectiveness of GP Institutions. This

analysis is informative of the extent to which states di¤er in the provision

of public services at the village level, and how GP activism di¤ers in the

four states. Section 5 summarizes �ndings from a research program which

uses these data to conduct in-depth analysis of the political economy of GP,

reservations for women, reservations for Scheduled Castes and Tribes, and the

e¤ectiveness of Gram Sabha meetings. Section 6 draws out the implications

of the �ndings from this analysis for policy.

6

2 Methodology

Our data come from a village- and household- level survey conduced in

Andhra Pradesh (AP), Karnataka (KA), Kerala (KE) and Tamil Nadu (TN).

The survey was conducted between September-November 2002.

The administrative unit below the state in India is the district. Each

Indian district is divided into blocks. Every block consists of multiple GPs.

A GP typically consists of 1-5 revenue villages, and its demarcation is done

on a population basis. The Panchayat Act of every Indian states mandates

the population criteria to be followed in that state.1

Sampling was done in multiple stages, and consisted of purposive sam-

pling up to the level of blocks and random sampling within these blocks.

Our �nal sample consists of 527 villages belonging to 201 elected GPs. In

a random sub-sample of 259 villages, 20 household surveys per village were

conducted, giving a sample of 5,180 households. In addition, a household

survey was also �elded to an elected member of the GP in every village (with

precedence given to the GP head if he/she lived in that village) - this gives

us an additional household sample of 544 elected o¢ cials. We describe the

stages of our sampling below.

2.1 Sampling

� District sample: for each pair of states two districts (one per state)

that shared a common boundary were selected. One district in KA

1In Andhra Pradesh and Kerala, it is a (revenue) village irrespective of its size. In

Tamil Nadu it is a revenue village with population of 500 or more. In Karnataka it is a

group of villages with population between 5 and 7 thousand.

7

(Kolar) that shared boundaries with both AP and TN entered the

sample twice. The same holds for one district in AP (Chithoor) This

gives us nine unique districts - 2 districts each in AP, KE and TN

and 3 in KA. The district pairs were selected, with one exception, to

focus on districts that had belonged to same administrative unit during

colonial rule, but had been transferred to di¤erent units when the states

were reorganized in 1956. These are the districts of Bidar and Medak

from the erstwhile state of Hyderabad, now in KA and AP respectively,

Pallakad, Coimbatore, Kasargod, Dakshin Kanada, Dharmapuri, and

Chithoor, all from erstwhile Madras state and now in KE, TN, KE,

KA, TN and AP respectively.

In KA, we also sampled Kolar district. This was a part of erstwhile

Mysore state, the precursor to modern KA, and thus does not follow the

colonial- rule matching process described above. However, its inclusion

increases variation when we compare the other three states with KA.

Furthermore, Kolar has common borders with both Chithoor in AP and

Dharmapuri in TN - which allows for a three part comparison within

the same geographic area. Map 1 provides a graphical description of

this matching.

� Block sample: For each district pair (which shared a common bound-

ary) 3 pairs of blocks were selected (that is, 3 blocks in each of the two

districts). If one district was matched with 2 di¤erent districts then

6 blocks were chosen from it (three per match). In one block in KE

an additional block was sampled as a check on our language matching.

This gave us a total of 37 blocks (12 in KA, 9 in AP and TN and 7 in

8

KE).2

For each pair of districts the three pairs of blocks which were the most

�linguistically similar�, in terms of the mother tongue of individuals

living in the block, were chosen. Language is a good proxy in these

regions for cultural di¤erences given the prevalence of caste and lin-

guistic endogamy. Hence, language matching allows us to partially

control for "unobservable" socio-cultural di¤erences. Linguistic simi-

larity was computed using 1991 census block level language data. The

historical and administrative similarity of linguistically matched blocks

was checked using princely state maps and the Report of the States

Reorganization Committee. Details on how the linguistic and historic

matching was implemented are in Appendix II.

� GP sample: In AP, KA and TN we randomly sampled 6 GPs per

block. In KE the population per GP in KE is roughly double that in

the other three states. For this reason, in KE we instead sampled 3

GPs in every block. This procedure gave a total of 201 GPs.

� Village sample: In every sampled GP in AP, KA and TN we sampled

all villages if the GP had 3 or fewer villages. If it had more than

three villages, then we selected the Pradhan�s village and randomly

selected two other villages. We excluded all villages with less than 200

persons from our sampling frame. All hamlets with population over

200 were considered as independent villages in drawing the sample. In

KE, we directly sampled wards instead of villages (as villages in KE

2The additional block was sampled in Kerala as a check on our sampling strategy.

9

tend to be very large) - we sampled 6 wards per GP. This gave us

a �nal village sample size of 527 villages.3 For sampled villages, any

associated hamlets were also included as part of the sample.

� Household village sample: In every block in AP, KA and TN we

randomly selected 3 of our 6 sampled GPs and conducted household in-

terviews in all sampled villages falling in these GPs. In KE we randomly

selected 2 GPs in one block and one GP in the other block. Within

sampled GPs we conducted household interviews in all sampled wards.

Overall this gave us a �nal sample size of 5180 households.4

� Choice of households within a village: Twenty households were

sampled, of which four were always SC/ST. The survey team leader

in every village walked the entire village to map it and identify total

number of households. This was used to determine what fraction of

households in the village were to be surveyed. The start point of the

survey was randomly chosen, and after that every Xth household was

surveyed such that the entire village was covered (going around the

village in a clockwise fashion).

� Elected o¢ cial sample: In every village in our sample an interview

was conducted with an elected Panchayat o¢ cial - if the Pradhan lived

in the village he/she was interviewed, otherwise a ward member was

randomly selected. In some cases, the Pradhan was not available at

3The state-wise break up is AP: 69 villages, KA: 182 villages, KE: 126 wards; TN 129

villages.4Number of villages for household sample were: AP: 32 villages, KA: 90 villages, KE

66 villages, TN 71 villages.

10

�rst visit and a ward member was selected. However, in these cases the

investigator usually went back and interviewed the Pradhan. Hence our

sample of elected o¢ cials is larger than the number of sampled villages

- and stands at 544.

2.2 Questionnaires

Four di¤erent questionnaires were used to collect data at the Village, Politi-

cian and Household level (see Appendix B for the questionnaires).

At the village level two questionnaires were used. First, we administered a

questionnaire using Participatory Rapid Appraisal (PRA) techniques (Cham-

bers 2003) to a group of men selected to represent di¤erent caste groups in

the village. The PRA questionnaire assessed villagers views on problems in

the village, and the work done by the GP. The PRA was also used to collect a

detailed listing of castes within the village, and land distribution both within

and between castes. The PRA respondents were also asked to construct an

oligarchy matrix for the village - listing the extent to which prominent ac-

tivities in the village were controlled by the Pradhan, former Pradhan and

the Vice-Pradhan. A short PRA-based questionnaire was separately �elded

to a (i) a group of women and (ii) a group of SC/ST individuals. These

PRA obtained separate measures of women�s and SC/ST problem ranking

vis-a-vis public service delivery.

The second village-level questionnaire was an audit of all public goods

in the village. This was an independent audit conducted by an investigator

who visually assessed the quality of schools, clinics, roads, drinking water,

and sanitation and also identi�ed the extent of GP involvement improving

11

these facilities.

In 259 villages we �elded household surveys. Twenty households were

surveyed per village, with 10 male and 10 female respondents. Four SC/ST

households were purposively selected in every village. The household ques-

tionnaire obtained information on household�s socio-economic status, house-

hold structure, views and use of public services in the village, private govern-

ment bene�ts. Respondents were also asked to rank-order problems in the

village. Since the sample is divided between male and female and SC/ST and

non-SC/ST respondents this provides yet another source of information on

gender and caste di¤erences on preferences about village problems. In each

of the 522 sample villages a household survey was also conducted with one

elected GP o¢ cial. In addition to all the questions on the household ques-

tionnaire politicians were also asked a series of questions about their conduct

of GP activities.

3 State Comparisons

The empirical analysis in this report focusses on comparing GPs in the four

South Indian states of Andhra Pradesh, Karnataka, Kerala and Tamil Nadu.

An important question that remains in understanding the relative impact of

the decentralization in these four states is the extent to which their political

history and social structure have a¤ected the functioning of local govern-

ments. There is considerable evidence demonstrating that the Travancore

region that is currently part of the state of Kerala has a long history of

progressive policies since (Je¤rey, 1992). Similarly Mysore state which is

12

currently part of the state of Karnataka was also ruled by relatively au-

tonomous rulers who placed a special emphasis on education and economic

development (Bhagavan, 2003). Recent work by Banerji and Iyer (2003)

has shown that there are strong path dependencies in land tenure policies -

speci�cally whether the region of India had a zamindari or ryotwari system in

place during British Rule. These systems which were established early in the

19th century are shown to have signi�cant contemporary impacts on the a

variety of indicators of development. Furthermore, scholars have argued that

di¤erences in cultural systems can have an important e¤ect of human devel-

opment (e.g. Dyson and Moore, 1983). Given these path-dependencies and

the cultural di¤erences, it is possible that Kerala is di¤erent because "Kerala

is Kerala". There is something special about the state that makes it partic-

ularly hospitable to good, equitable governance. If such path -dependencies

prove to be de�nitive, then policy options are likely to be relatively small.

The sampling strategy outlined above allows us to compare the states,

controlling for di¤erences that may come from historical or cultural path-

dependencies. We will compare villages on either side of the current borders

that originally belonged to the same political entity, and which have also

been matched by majority language. Thus, any di¤erences we observe be-

tween these matched villages cannot be because of di¤erent political histories

prior to 1956, or because of di¤erences langauge - which is a proxy for local

kinship structure and social organization. The di¤erences have to attributed

to di¤erences that have emerged after 1956. The comparison is particularly

interesting because the states provide an excellent contrast of di¤erences in

the implementation of the 73rd amendment. In this section, we will brie�y

13

highlight these di¤erences.5

In the last two decades there have been important di¤erences in how

states have structured panchayats. Some of these di¤erences are summarized

in Table 1. Consider the data on village funding. The data are from 1997

and likely to be considerably di¤erent today, but some di¤erences that are

consistent with the above discussion can be discerned. It is clear that the four

states di¤er considerably in the availability of funding to GPs. Much of the

funding is tied to particular programs and the level of discretionary funding

di¤ers even more across the states. Kerala clearly dominates, followed by

Karnataka. These two states are the subject of an excellent recent report on

�scal decentralization (World Bank 2004) which makes clear that Kerala and

Karnataka are rather di¤erent from in each other in many respects. While

Kerala followed a learning-by-doing strategy of progressively increasing the

responsibility of GPs with a signi�cant decentralization program, Karnataka

has been more cautious in its approach with more authority in the hands

of the state government. Both these states, despite being better than other

Indian states, do not have good accounting systems which does not permit for

much transparency in local funding decisions. An important conclusion of the

report is the fact Kerala has faced signi�cant �scal problems in recent years.

This has caused considerable strain in GP �nances, with promised allocations

from state governments not being sanctioned to GPs. Thus, while Kerala has

over the years been leading in giving GPs considerable �scal authority and

power, in recent years this authority has su¤ered considerable strains. Thus,

the report concludes "a necessary condition for a well-functioning system

5The state di¤erences are derived from a note prepared by Geeta Sethi (SASRD).

14

of �scal decentralization is a healthy �nancial position." This suggests that

shifts in the e¤ectiveness of GPs may not entirely be because of historical

and cultural factors, but because of current trends.

3.1 Background

Kerala �Strong Fiscal Decentralization The Kerala state government

has, to a large extent, embraced the principal of decentralization and has

taken an active role in ensuring the e¤ective implementation of legislation

and the state�s vision in this respect. The Kerala Panchayat Raj Act (1994)

introduced a three-tier PRI system with a signi�cant element of political

and �scal decentralization �distinct from early experiments in decentralized

planning in the 1970s.

In 1996 this legislation was amended according to the recommendations

of the Committee on the Decentralization of Powers which took some bold

steps towards creating local self-government. Unlike other states, where de-

velopment decisions are taken by the state government and local government

implements the works, in Kerala locally elected leaders have been given full

power to prepare and implement development projects based on their func-

tional jurisdiction, the needs of the people and the resources available to

them. To ensure integration of funds allocated to sectors and schemes with

the plans of local bodies at all levels, �nancial and taxation powers have been

devolved to local government. In addition, approximately 35-40% of plan ex-

penditure is earmarked to development projects prepared by local bodies,

which makes Kerala the most �scally decentralized state in India.

Administrative decentralization is also underway, albeit to a lesser extent.

15

At present, there is a dual system of control over line agency sta¤ which, to-

gether with technical complexities inherent in the structure of development

planning, have meant that elected o¢ cials have yet to gain e¤ective control

over line o¢ cials. However, the state is committed to tackling these prob-

lems with initiatives for administrative reorganization and statutory changes

which extend the power of elected leaders and institutionalize the process of

local level planning and plan implementation.

In addition to legislation, various informal mechanisms have been pro-

moted to encourage participation at the grass-roots level and foster devel-

opment planning from below. These include informal governance structures

such as neighborhood groups and bene�ciary selection committees and the

Campaign for Decentralized Planning. This was a drive to empower lo-

cal bodies to prepare, plan and implement development projects and har-

ness Kerala�s vast human resources by forming expert advisory committees

manned by quali�ed volunteers.

The state government has also sought to improve accountability and

transparency and to stem capture of elected institutions by bureaucrats and

the local elite. These structures together with state legislation have made the

Kerala model of decentralization an e¤ective tool to foster local development

planning and bring about the wider goal of democratic decentralization.

Karnataka� Strong political decentralization Karnataka has a long

history of democratic decentralization with three distinct periods of pan-

chayat legislation and a well-organized and politically conscious rural soci-

ety. The system in place prior to 1983 was largely ine¤ectual with panchayats

16

having little real power.

In 1983 the Government of Karnataka, led by the Janata Party, passed a

radical decentralization act which legislated a two-tier panchayati raj system

with reservations for women and SCs and STs and a local participatory insti-

tution called the Gram Sabha. This was taken as a basis for the subsequent

Karnataka Panchayat Raj Act (1993) passed in order to bring state legisla-

tion in line with the 73rd and 74th Amendments to the Constitution. The

Act introduced a three-tier system with the aim of empowering representative

local government and fostering local participation in rural development to-

gether with the formation of the District Planning Committee (DPC) whose

main function involved overseeing the development plan for the district as

a whole. Particular focus was placed on distributing political power within

PRIs to improve accountability and reduce elite-group capture by introducing

rotation of leadership between elected members.

Andhra Pradesh� Weak Political Decentralization Since 1958, Andhra

Pradesh has incorporated the PRI system in its state legislation, the most

recent being the Andhra Pradesh Panchayat Raj Act (1994). In addition to

constitutional requirements this act introduced reservation of seats for the

Backward Classes and party-based elections for the top two tiers of local

government.

In practice, the state vision of PRIs, and their role with respect to develop-

ment planning and local governance, is mixed. While several sub-committees

have been formed to examine decentralization with regards to panchayats,

the state legislator has done little to empower them. Identi�cation of PRI

17

functions at the local level has not fully taken place which, together with the

lack of a District Planning Committee, implies that panchayats at all levels

have no major role in development planning or implementation, except in

bene�ciary selection. This problem is ampli�ed by a lack of �scal decentral-

ization. Legislated taxation powers have not been e¤ectively devolved and

the majority of PRI funds are earmarked grants for central or state sponsored

schemes. Together this has lead to a serious mismatch between the limited

functions entrusted to panchayats and the �nances available to them that

has acted to compromise political decentralization and accountability.

In addition, administrative decentralization has not taken place with par-

allel structures at the Rural Development and Panchayat Raj departments

remaining largely separate. PRIs thus form a small and marginalized com-

ponent in the state�s vision of rural development which has fostered local

participation and community development through other means. The most

prominent of these is the Janmabhoomi program which is a participatory

development initiative focussing on the creation of stake-holder groups, man-

aged and controlled by state civil servants.

The degree of government commitment and amount of local development

funds channeled through such programs indicate that the Andhra Pradesh

government has in e¤ect by-passed PRIs and the concept of democratic de-

centralization and is undertaking rural development without signi�cant loss

of central control.

Tamil Nadu�Weak on political, administrative and �scal decen-

tralization The State of Tamil Nadu has a volatile tradition of local rep-

18

resentative institutions dating from the 1860s, and was one of the few states

to voice concern over the 73rd and 74th Constitutional Amendments. Wide-

spread state-level reluctance to comply with this legislation is re�ected in the

Tamil Nadu Panchayats Act (1994) which did little to devolve state powers

and empower PRIs, even to the extent that past legislation had done. Elec-

tions were delayed to such an extent that central government threatened to

withdraw all funds for rural development and were �nally held in 1996 when

the DMK party came in power and embraced democratic decentralization as

one of its political mandates.

The experience of decentralization in Tamil Nadu is therefore in �ux

with greater devolution of powers to local government expected in the future.

Under the current legislation, political and functional decentralization is very

limited. PRIs fall under the jurisdiction of state o¢ cials (who have the power

to dissolve them) and there are virtually no state schemes and functionaries

transferred to local government.

The Gram Sabha, till recently, was a defunct institution for community

decision-making with its bene�ciary selection function being carried out by

line or elected o¢ cials at higher levels. However, e¤orts at the grass-roots

level to mobilize democracy in decision-making and rural participation in

development are going some way to improve its e¤ectiveness.

The main locus of state development planning and �nance is still through

the District Rural Development Agency which is a registered body controlled

by state bureaucrats with little connection with PRIs. Panchayats are also

bypassed in rural development planning by the growth of independent state

and central schemes such as the MPs and MLAs Area Development scheme.

19

Lack of �scal autonomy means that local bodies are largely dependent on

the meager state and central government for their resources which are pre-

assigned, state grants for local bodies being 8% of the share in tax collection.

There is also insu¢ cient administrative decentralization, which compromises

accountability. Line o¢ cials working in panchayat bodies are declared gov-

ernment o¢ cials and do not come under the management of local bodies. As

a result local elected o¢ cials cannot supervise their activities or contribute

to their projects except in service delivery.

As mentioned above, it is expected that democratic decentralization will

come to the forefront of rural development planning in the near future with

the change in government. Already recent community training drives and

capacity building for participatory planning at the village level indicate that

major initiatives are underway to strengthen PRIs at all levels.

3.2 Evidence

We now examine cross-state di¤erences in public good provision as a means of

examining whether these di¤erences mirror the institutional di¤erences that

we discussed above. We also discuss whether public good outcomes vary with

reservation and whether it is the Pradhan�s village. Our mode of analysis is

two fold. First, we present cross-tabulations. Second, we report the results

from a basic regression which includes state dummies, dummy for whether

the Pradhan�s post is reserved, a dummy for Pradhan�s village and dummies

for each matched block pair. As discussed above, there is ample reason to

believe that matched blocks share common historical and cultural traits. In

the following discussion we abbreviate Andhra Pradesh to AP, Karnataka to

20

KA, Kerala to KE, Tamil Nadu to TN.

3.2.1 Cross Village Resource Allocation

Levels of public goods As is well known, KE has long been the leading

Indian state with respect to human development indicators. Table 2a reports

state-wise means for our sample villages from the 1991 census to see whether

this is true for our villages. Table 2b provides the regression analogue, where

we include block-pair �xed e¤ects. Here, the state dummy variables focus

on di¤erences between the states within each block pair. It is clear that on

almost all indicators KE was well ahead of the other states in 1991 in our

sampled villages. One important exception is schooling, but this may be en-

tirely due to the sampling method which sampled wards in KE and villages

everywhere else. Thus, it was di¢ cult to assign census village level informa-

tion to our sampled wards in KE. Since schools service large populations -

and are generally available for entire villages, any missed village in the cen-

sus would result in an underestimate of the number of schools. KE is also

behind AP and KA on the provision of domestic electricity. This is unlikely

to be due to a sampling anomaly. We �nd no relationship with reservations,

which is consistent with the fact that choice of reserved GPs is intended to

be random. There is a generally positive e¤ect of school outcomes with

Pradhan�s village, but this e¤ect disappears when you control of population

size and variation within the block.

Moving to the public goods data from our survey, which was conducted

11 years after the census, we see that the patterns are both similar and

di¤erent. Table 3a and 3b report information from the facilities survey with

21

state averages and block-pair �xed e¤ects regressions respectively. KE clearly

dominates the other states on schools, health facilities and drinking water

sources. But it is behind on the number of overhead tanks, bus stops in

the village, and the proportion of households with electric lights. Overhead

tanks are easy to explain since KE probably has di¤erent mechanisms of

water delivery than the other states, but the lack of bus stops in the village

and electric lights may suggest that KE has put a much greater emphasis

on basic investments in education, health and water than on other services.

Looking at di¤erences between the other three states we see that generally

TN lags behind AP and KA, which was not the case in 1991. This suggests

that there has been a reduction in investments in public services in TN in

the last decade, in comparison with AP and KA. Since these results compare

the variation within block pairs, geography should not play a big role in

explaining the di¤erences between states. These di¤erences should re�ect

public investments made since 1956 when the states were reorganized along

linguistic lines. Note again that reservations have no e¤ect, while the current

Pradhan�s home village has better public services.

The fact that KE is ahead in levels of public investments should not

be surprising given the size of allocation to GPs and its e¤orts on �scal

decentralization. But the fact that KA is no di¤erent than AP may lead one

to speculate that KA�s e¤orts on political decentralization have not translated

into results on the ground. TN�s distinctly worsening situation from 1991 to

2002 is also consistent with the fact that it has poorly funded PRIs that lack

authority.

22

GP activism The analysis of di¤erences in the levels of public good avail-

ability re�ect the history of investments in public services since 1956 by each

state. We are unable to distinguish between investments made directly by

the state government, and those made via PRIs. In order to get more insights

into this, we now move to a direct examination of levels of GP activism since

the last election in each of the states6. This Panchayat "activism" in our data

is measured from two di¤erent sources:

a)The facilities survey: Where after making an assessment of a facility

the interviewer asked households living close to the facility about changes

made since the last election.

b) The PRA: Where a detailed set of questions were asked about the

activities of the Panchayat since the last election.

We consider the PRA data to be more accurate than the facilities survey

data on this topic, because the PRA re�ects the results of a consensus view

from a moderated group discussion from a representative sample of knowl-

edgeable people, while the facilities information is more ad hoc. Nevertheless,

we report results from both sources of information. Table 4 begins with the

facilities survey results. The clearest result here is that TN signi�cantly

lags behind the other three states in overall activism, and in investments in

schools, anganwadis, health, drinking water, roads, and street lights. AP and

KA do not show any signi�cant di¤erences with KE in overall activism or in

schools. But they lag behind KE in anganwadis, health and drinking water.

AP does better than all the other states on roads and street lights, while KA

6Since GP elections in AP were held a few months before the survey, the AP results

re�ect activism from the previous election.

23

does no di¤erently than KE on these investments.

In the PRA data, presented in Table 5 the contrast with KE is even more

striking. AP and KA do better, or no di¤erent, than KE on all investments

other than health. In particular KE lags behind these two states in drinking

water investments, sanitation, roads, and electricity. AP and KA are not

very di¤erent from one another, and TN lags behind all three states in overall

activism and road investments. Additionally,the Pradhan�s village bene�ts

from increased activism across the board - an e¤ect that remains after several

more village level controls are added. Also note that again that we observe

no impact of reservations.

What can we learn from these results? First, and perhaps most impor-

tantly, KE is slipping. This is consistent with the �ndings from the World

Bank report on �scal decentralization showing problems with KE�s �nancing

of PRIs - which is a result of its �scal problems at the state level. It is also

consistent with the recent work by Chaudhuri and Heller (2004) on Kerala

panchayats.

Our results reinforce the point TN has generally very inactive GPs. Also

note that KA and AP are rather similar to one another. Since KA has been

far ahead of AP in promoting democratic decentralization, with AP under

the Naidu government even making attempts to entirely bypass PRIs, it is

interesting that this has not led to large di¤erences in GP activism.

3.2.2 Household targeting

From public goods we move the provision of private goods since an important

function of Panchayats is to target poor families with schemes to provide

24

private bene�ts such as housing, private water supply and toilets. The data

in these tables come from the household surveys explained above. We will use

the same block-pair matching structure as in the public goods analysis, but in

addition to controlling for state dummies, Pradhan�s village and reservations

we will include household level variables - whether the respondent is female,

SC/ST, whether the household is wealthy, landless, or if the respondent is a

local politician. We examine six di¤erent types of investments by Panchayats

in private goods, and an indicator of overall activism in Table 6.

Once again TN�s relatively poor performance on this is obvious - it lags

behind the other states on every indicator except water. However, we again

see that KE generally lags behind both AP and KA. KA leads all the states

in overall activism - particularly in the provision of toilets and electricity,

while AP leads the states in providing BPL cards and public works projects.

These results are similar to the public goods activism results from the PRA

and again provide some teeth to the argument that KE�s PRI initiatives have

been slipping. It could also indicate that since KE has a lower incidence of

poverty, it will have fewer potential bene�ciaries of targeted schemes than

the other states.

Note, however, that targeting is not entirely bad. SCSTs bene�t greatly

from all the schemes, as one would hope since many schemes are designed

with them in mind. Wealthy households are much less likely to bene�t, while

landless households are more likely to bene�t - particularly by receiving BPL

cards and public works programs. A worrying result is that politicians bene�t

with a higher incidence of overall targeting and from the provision of toilets,

and public works programs - though they also receive fewer investments in

25

drinking water. Since, controlling for SCST and indicators of wealth and

land, politicians should be treated no di¤erently than anyone else - this result

suggest that there there me be some private appropriation of public schemes.

This issue is examined in greater detail in Besley, Pande and Rao (2005b).

3.2.3 Participation, Information and Socio-Political Structure

We now turn to an examination of some institutional dimensions of gover-

nance at the village level. The data we collected are rich in information

about participation both at the village and individual level, and on measures

of political and social inequality.

Village level participation: Table 7 reports results on village level par-

ticipation beginning with whether an NGO is active in a village. NGOs have

over the years become increasingly active in South India and we see that 33%

of the villages in our sample have NGOs present. Controlling for block-pair

�xed e¤ects, we see that Karnataka has the highest level of NGO activity

while Tamil Nadu lags behind the other states. Interestingly, we also see a

high degree of CBO activity in all the states, but after controlling for block-

pairs no state dominates. Gram Sabha activity also shows some interesting

patterns. Looking at Gram Sabha meetings held in the last twelve months

KE is behind all the other states. However, the picture changes in looking at

Gram Sabhas held in the last six months where KE is ahead of all the other

states. This is partly because our survey was conducted during a drought in

parts of KA, AP and TN and Gram Sabhas were not held as regularly in these

states - perhaps a way of preventing villagers from voicing complaints about

26

drought-alleviation work. The Pradhan village always does better than other

villages, as in the other results and no reservations e¤ects are observed.

Gram Sabha Participation by households Table 8 reports �ndings

from some indicators of Gram Sabha participation at the individuals level.

We see that individuals are far more likely to attend to Gram Sabhas in

KE, and to speak in them. But, attending the Gram Sabha to seek private

bene�ts is much more likely in the other states. This suggests two things -

one that households in KE are better o¤ and therefore do not need to seek

private bene�ts from the government as much, and - two - that that KE has

a more politically sophisticated population. KE�s citizens are likely to use

Gram Sabhas to have a say in decisions over public goods and services. Wor-

ryingly, the data also demonstrate social exclusion in Gram Sabhas. Women

are much less likely to attend Gram Sabhas or to speak in them. And land-

less individuals are also less likely to speak in Gram Sabhas. Interestingly,

politicians also say that they are less likely to attend Gram Sabhas possibly

because they have little to gain by attending them - unless they are in o¢ ce.

Gram Sabhas are examined in more detail in one of the papers in the appen-

dix. The impact of participation in Gram Sabhas is examined in more detail

in Besley, Pande and Rao (2005a and 2005b).

Household Information and participation KE�s much higher level of

civic sophistication is also apparent in Table 9. Here we see that KE house-

holds are much more likely to be regular readers of newspapers and to pay

taxes. But, individuals in AP and TN are more likely to know the name of

the Chief Minister, possibly because of the personality cults around Chan-

27

drababu Naidu and Jayalalitha the then chief ministers of these states. All

these indicators of civic sophistication are higher for the wealthy and for

politicians. However, these indicators are lower for the landless and much

lower for women - again demonstrating exclusion.

We further examine these themes with indicators of political participa-

tion in Table 10. Looking at various indicators - whether a member of the

household is politically active, voted in GP elections, MLA elections and

Parliamentary elections - we again see that KE dominates the other states.

Interestingly voters in KE are more likely to vote along political lines, while

in the other states they are more likely to base their decisions on the char-

acteristics of individual politicians. This is not surprising given the level to

which elections in AP and TN are based on the personalities of politicians,

but the fact that KA also shows less party-based voting than KE demon-

strates that KE elections are signi�cantly more determined by party politics.

Note again that women are much less likely to vote. However, other excluded

groups like SC/STs are more likely to participate in political activities and

the landless are also more likely to vote. This suggests that the political

process could provide a means for less advantaged groups to exercise their

preferences. Note that we also see a perverse e¤ect of politicians claiming

that they are likely to be a¢ liated with a political activity, and to participate

in politics - this suggests that the responses of politicians to the questions

we asked may also be driven by political motives.

Moving to more material forms of participation we examine the extent

to which households contribute in cash or kind to the provision of public

goods in Table 11. Note that material participation is higher in KE for

28

roads and health, but lower for schools and drinking water. This is consis-

tent with the �ndings on panchayat activism we observed above suggesting,

unsurprisingly, that household contributions may be driven by the extent of

panchayat involvement in these activities. Note again that the wealthy are

more likely to contribute, and politicians also say that they are more likely

to contribute. Women show a lower incidence of contributions but this may

be because of their lower levels of earning and lack of individual agency in

making decisions.

These results can be contrasted with the results on willingness to pay

for public goods, reported in Table 12. Here we see that households in KE

are much more likely to say that they are willing to pay more for public

services across the board. We also observe a greater willingness to pay in

TN compared to the other states. Similarly wealthy households indicate a

greater willingness to pay. Note also that in the means, we see that in all the

states except KA close to 50% of our respondents say that they are willing to

pay more for one or more public services. While willingness to pay questions

have important �aws, these results do suggest a gap between the demand

and supply of service provision.

Inequality Finally we examine various indicators of economic, social and

political inequality in these villages. We examine these indicators, reported

in Tables 13 and 14, merely by looking at mean di¤erences across the states

(rather than the block-pair �xed e¤ect regressions which are less relevant

here). We should note that the data for these indictors was collected entirely

by using PRA methods. For indicators of caste and land inequality the PRA

29

group was asked to list all the caste and religious groups living in the village

showing how many households belonged to each group. Then, for each

group, they were asked to place the households in di¤erent broad categories

of land ownership. This, method, allows us to obtain not only a detailed

caste listing for every village, and measures of village land inequality, but

to decompose land inequality in each village to its between and within-caste

components. The results of this decomposition can be observed in Table 13.

Interestingly KE shows the highest level of land inequality overall with a gini

of 0.66. This is probably because its higher level of economic development

makes non-farm incomes more salient and land less important as a measure

of overall inequality. The other three states have land ginis that are not

signi�cantly di¤erent from each other - ranging from 0.52 in KA to 0.58 in

TN.

We also use the Theil entropy measure of inequality because it can be

decomposed into between and within-caste components. Measuring this

presents a challenge because the number of castes per village varies consider-

ably. Villages with more castes would have arti�cially higher between-caste

inequality. To correct for this we group all castes into three categories - high

(which include "forward" castes and castes considered "dominant" in the

state, low (castes classi�ed as SC, ST, and "backward" but not "other back-

ward"), and middle which is the residual category. Decomposing inequality

into these three groups we see that 13% of inequality can be explained by

between-caste inequality in KE, which increases to 18% in AP. We should

note that between-within decompositions of inequality consistently tend to

hover around 15% regardless of the nature of the data and the type of group

30

(Kanbur and Venables, 2003), so it would not be valid to contrast these

results with racial or spatial inequality observed in other data. But, the

comparison across states in our own data are valid and suggest that caste

is a much less salient indicator of inequality in KE than in the other states.

This is further demonstrated in Table 14 where we see that only 17% of

land is controlled by upper castes in KE, compared with 36% in KA. Caste

dominance is therefore much more prevalent in KA.

Some villages also tend to be under the control of a few families. In order

to construct a measure of this type of oligarchy, we asked the PRA group to

construct another matrix showing whether the Pradhan, the ex-Pradhan and

the vice-Pradhan controlled some important categories of political and eco-

nomic power - such as whether they were the biggest landowners, or whether

they owned the largest factory in the village. The proportion of "yes" an-

swers in this matrix, then provides us a measure of oligarchy. State level

di¤erences in this measure are reported in Table 14 where we see that TN

has higher oligarchy than the other states, followed by KE.

To summarize, we see that villages are characterized by a great deal of

inequality and social heterogeneity within them. It has not been possible

to measure this with previous data from India, or indeed, most parts of the

world. The extent to which these variables a¤ect public services remains an

open question - which, to some extent, we examine in other papers.

31

4 Conclusions

What can we learn from these results? Fist, they have some relevance for

our understanding of the "Kerala model." Our �ndings provide more �esh

on the well-known fact of Kerala�s sophisticated political culture. Kerala has

the highest voter turnout among the four states in all types of election. Fur-

thermore, Kerala�s electorate is among the least likely to vote for candidates

on caste or religious lines. It also has a more active civic culture with active

participation in gram sabhas. While levels of land inequality are high, per-

haps because of the reduced salience of land as an indicator of wealth, these

inequalities are less likely to be because of caste based di¤erences than in

the other states. Kerala, perhaps in�uenced by this active political culture,

also dominates the other states in the availability of public goods. However,

consistent with other recent work, we �nd that Kerala is slipping. All our in-

dicators of current investments on public goods by the panchayats are lower

in Kerala than in the other states. Similarly we �nd that Kerala lags behind

Andhra Pradesh in the provision of BPL cards and public works programs.

To some extent this is because of Kerala�s higher levels of development and

lower levels of poverty. But, other evidence from the World Bank�s �scal

decentralization study suggests that a �scal constraints have reduced the

availability of funds to panchayats resulting in lower levels of GP activity.

Tamil Nadu GPs in our sample are at the end of the distribution. They

lag all the other states in the provision of most public goods (other than

water tanks and bus stops). More importantly, current levels of activity by

GPs are also behind the other states. This is also true in the provision of

private bene�ts such as BPL cards, housing and electricity. On the other

32

hand, villagers in Tamil Nadu, are second only to those in Kerala in their

political and civic participation - they are more likely to pay taxes than

villagers in AP and Karnataka, and more likely to vote.

It is interesting to note that the remaining two states, Karnataka and

AP are rather similar to one another. Since KA has been far ahead of AP

in promoting democratic decentralization, with AP under the Naidu govern-

ment even making attempts to entirely bypass PRIs, it is interesting that

this has not led to large di¤erences in the provision of public goods (except

for paved road), or indeed even in current GP activity in public goods provi-

sion. On private bene�ts Kartnataka leads all the states in overall activism

- particularly in the provision of toilets and electricity. But AP leads the

states in providing BPL cards and public works projects. Karnataka is the

most likely state to have an NGO active in the village, but it is also the

least likely to have held a gram sabha in the last six months - which can

largely be attributed to drough conditions in the state at the time of the

survey. However, even though AP faced the same climatic conditions, it was

far more likley than Karnatka to have held gram sabhas.

Finally, it is also interesting to note the strong caste in�uences in Kar-

nataka. Karnataka villages have the highest proportion of land owned by

upper castes (36 per cent), with 34 per cent of villages having over half their

land owned by upper castes. Perhaps as a consequence, Karnataka voters

are far more likely than those in other states to vote along caste or religious

lines.

33

References

[1] Bardhan, Pranab, [2002], �Decentralization of Government and Devel-

opment,� Journal of Economic Perspectives, 16(4), 185-205.

[2] Bardhan, Pranab and Dilip Mookherjee, [2000], �Capture and Gov-

ernance at Local and National Levels,� American Economic Review,

90(2), 135-139.

[3] Besley, Timothy and Stephen Coate, [1997], �An Economic Model of

Representative Democracy,�Quarterly Journal of Economics, 112(1),

85-114.

[4] Besley, Timothy, Rohini Pande, Lupin Rahman and Vijayendra Rao,

[2004a], �The Politics of Public Good Provision: Evidence from Indian

Local Governments,�Journal of the European Economics Association,

2(2-3), 416-426.

[5] Besley, Timothy, Rohini Pande, Lupin Rahman and Vijayendra Rao,

[2004b], �Decentralization in India: A Survey of South Indian Panchay-

ats,�mimeo, LSE.

[6] Besley, Timothy, Rohini Pande, Vijayendra Rao and Radu Ban, [2005a],

"Participatory Democracy in Action: Survey Evidence from South In-

dia," Journal of the European Economics Association (forthcoming).

[7] Besley, Timothy, Rohini Pande, and Vijayendra Rao, [2005b], "Political

Selectiona and the Quality of Government: Evidence from South India,"

mimeo, LSE

34

[8] Besley, Timothy, Rohini Pande, Vijayendra Rao and Radu Ban, [2004c],

�Tokenism or Agency? The Impact of Women�s Reservation on Pan-

chayats in South India,� mimeo, Development Research Group, The

World Bank

[9] Bhagavan, Manu, Sovereign Spheres: Princes, Education and Empire in

Colonial India, Oxford University Press, New Delhi, 2003

[10] Chattopadhyay, Raghabendra and Esther Du�o, [2004b], �E¢ ciency

and Rent-seeking in Local Governments,�mimeo MIT.

[11] Chattopadhyay, Raghabendra and Esther Du�o, [2004a], �Women as

Policy Makers: Evidence from a India-Wide Randomized Policy Exper-

iment,�Econometrica, 72(5), 1409-1444.

[12] Chaudhuri, Shubham, [2003], �What di¤erence does a constitutional

amendment make? The 1994 Panchayati Raj Act and the attempt to

revitalize rural local government in India,� typescript, Columbia Uni-

versity

[13] Chaudhuri, Subham and Patrick Heller, "Building Local Democracy:

Evaluating the Impact of Decentralization in Kerala," Paper presented

to Netsappe III, Paris, June 2004

[14] Dyson, Tim and Mick Moore (1983) �On Kinship Structure, Female

Autonomy, and Demographic Behavior in India�Population and Devel-

opment Review Vol.9, No. 1, pp 35-60.

35

[15] Enikolopov, Ruben and Ekaterina Zhuravskaya, [2004], �Decentraliza-

tion and Political Institutions,�mimeo, Institute for Advanced Study,

Princeton.

[16] Faguet, Jean Paul, [2004], �Does Decentralization Increase Responsive-

ness to Local Needs? Evidence from Bolivia,� Journal of Public Eco-

nomics, 88: 867-894.

[17] Foster, Andrew and Mark Rosenzweig, [2001], �Democratization, De-

centralization and the Distribution of Local Public Goods in a Poor

Rural Economy,�mimeo, Brown.

[18] Je¤rey, Robin (1993), Politics, Women and Well-Being: How Kerala

Became�A Model,�Oxford University Press, Delhi

[19] Kanbur, Ravi and Anthony Venables, "Spatial Inequality and Develop-

ment," mimeo, Cornell University, October 2003

[20] Manor, James, [2004], �Democratization with Inclusion: political re-

forms and people�s empowerment at the grassroots,�Journal of Human

Development, 5(1), 5-29.

[21] Matthew, George and Nirmala Buch, Status of Panchayati Raj in the

States and Union Territories of India 2000, Institute for Social Studies,

Delhi, 2000

[22] Platteau, Jean-Philippe and Anita Abraham, [2002], �Participatory De-

velopment in the Presence of Endogenous Community Imperfections,�

Journal of Development Studies, 39(2), 104-136.

36

[23] Rao, Vijayendra and Michael Woolcock, �Integrating Qualitative and

Quantitative Approaches in Program Evaluation,�Chapter 8 in Fran-

cois Bourguignon and Luiz A. Pereira da Silva (editors), The Impact

of Economic Policies on Poverty and Income Distribution: Evaluation

Techniques and Tools, World Bank and Oxford University Press, 2004.

[24] Triesman, Daniel, [2002], �Decentralization and the Quality of Govern-

ment,�mimeo, UCLA.

[25] Verba, Sidney, Kay Lehman Scholzman and Henry E. Brady, [1995],

Voice and Equality: Civic Voluntarism in American Politics, Cambridge

Mass: Harvard University Press.

[26] World Bank, [2000], World Development Report 2000/2001: Attacking

Poverty, Washington, DC, The World Bank

[27] World Bank (2004), "India: Fiscal Decentralization to Local Govern-

ments," Report No. 26654-IN, Rural Development Unit, South Asia Re-

gion.

[28] World Bank (2000), Overview of Rural Decentralization in India, Vol 1,

2, and 3, Rural Development Unit, South Asia Region.

37

A Comparison of Gram Panchayats across the Sampled States Andhra Pradesh Karnataka Kerala Tamil Nadu Year of passing State Panchayati Raj Act April 21, 1994 April 30, 1994 April 23, 1994 April 24, 1994

Year of 1st regular election 2001 1993 1995 1996

Minimum Size for a GP area

A revenue village, irrespective of size

Village(s) with population between 5000-7000 A village, irrespective of its size A revenue village with population

upwards of 500 Reservation for Backward Castes One-third of total seats About one-thirds of total seats No reservation No reservation

Election of chairman Direct Indirect Indirect Direct Committee System Agricultural Committee, Public

Health and Sanitation Committee, Communications Committee

Production Committee , Social Justice Committee , Amenities

Committee

Functional Committee, for different subjects like agriculture, sanitation, communication, pubic

health and education

No provision for Committees

Finances : Obligatory Taxes

House Tax, Tax on produce sold in village, Property Transfer Duty,

Advertisement Tax

Tax on buildings/houses, Tax on non-agricultural lands

Entertainment Tax, Taxes for services, Duty on property transfer, House/Building tax, Tax on non-

agricultural land, Water Tax, Lighting Tax, Conservancy fee,1 Drainage Tax, Sanitation Tax for

public latrines,

House/building Tax , Surcharge on Stamp Duty, Tax on Professionals

Finances : Obligatory Non-Tax Sources

Tax devolution from higher levels of government, Income from

endowments, trust or panchayat investments, Income from village fisheries and woods, Unclaimed

deposits, Grants from higher levels of Government, Share of fines

imposed on Village, share of stamp duty

Share of land revenue, Grant of 1 lakh rupees per annum, Rent/sales

proceeds

Grant-in-aid, Basic tax grants, Income from remunerative

enterprises, Income from trusts and endowments, Unclaimed deposits,

Fines, Income from ferries

House Tax matching grant from Government, Grants from higher Panchayat levels, income from endowments and trusts, Income

from fisheries, Share of entertainment tax, Vehicle fee

Finances: Discretionary Taxes

Vehicle Tax, Tax on Agricultural Land for a specific purpose, Land tax, Tax on Education level, tax on construction of public works,

Entertainment Tax, Tax on non-motor vehicles, Advertisement

Tax, Lump sum levy on factories in lieu of taxes

Special tax on construction of public works, Professional Tax,

Advertisement Tax

Special tax on construction of

public works, Pilgrim tax, trade and tourist bus tax

Finances: Discretionary Non-Tax Sources

Fee for the use of community land and resources, fees for use of buildings and property under

Panchayat or government control, street cleaning fee,2 Market/bazaar fee (committee), lump sum levy on

factories in lieu of taxes

Water rate, fee on buses, taxis and auto-stands, fees for use of

buildings under Panchayat control, Fee for the use of community land and resources, Market/bazaar fee (vendor), Fees on animals sold,

pilgrim fee

Panchayat may raise loans, Government grants and loans, Fee for the use of community land and

resources, collection from beneficiaries of institutions

governed or financed by Panchayat, fees for use of

buildings under Panchayat control

Income from ferries, Income from unclaimed deposits, Drainage fee, Sanitation fee for public latrines, fees for use of buildings under

Panchayat or government control, Market/bazaar fee

Ability to approve schemes without External Sanction

Yes, up to Rs. 10,000 Yes, up to Rs. 10,000 Yes, no monetary limit Yes, if scheme is financed by the panchayat’s own funds

Estimated Village Panchayat Expenditure per Capita (1997)i

Rs. 55.71 Rs. 72.48 Rs. 198.55 Rs. 61.53

Estimated Village Panchayat Revenue per Capita (1997)i

Rs. 58.22 Rs. 69.50 Rs. 335.41 Rs. 72.35

Table compiled using data from PRIA. “The State of Panchayats,” Government of India. “India Panchayati Raj Report 2001,” and Government of Karnataka. “The Karnataka Panchayat Raj (Grama Panchayat Taxes and Fees) Rules, 1994.”

i Calculated using data from Government of India documents “Population Projections for India and States 1996-2016” and the “Report of the Eleventh Finance Commission.”

1 Sanitary levy for the cleaning of privately owned latrines/cesspools 2 For those who own a pet dog

Table 2a: 1991 Levels of Public Goods, simple mean comparison

State

Schools per 1000

inhabitants

Health facilities per

1000 inhabitants

Taps available (dummy)

Tube well available (dummy)

Bus stop in village

(dummy)

Pucca approach road

(dummy)

Kacha approach road

(dummy)

Domestic electricity (dummy)

Fraction land irrigated

Andhra 1.457 0.124 0.091 0.328 0.508 0.410 0.635 0.723 0.023(1.137) (0.422) (0.290) (0.473) (0.504) (0.496) (0.485) (0.451) (0.060)

Karnataka 2.009 0.077 0.214 0.126 0.747 0.725 0.269 0.000 0.136(1.149) (0.215) (0.411) (0.333) (0.436) (0.448) (0.445) (0.000) (0.113)

Kerala 0.750 0.304 0.726 0.887 0.976 0.976 1.000 0.073 0.314(0.405) (0.254) (0.448) (0.318) (0.154) (0.154) (0.000) (0.260) (0.231)

TamilNadu 1.093 0.108 0.178 0.000 0.876 0.690 0.349 0.713 0.196(0.567) (0.189) (0.384) (0.000) (0.331) (0.464) (0.478) (0.454) (0.190)

All 1.390 0.148 0.320 0.309 0.808 0.740 0.518 0.296 0.181(1.019) (0.271) (0.467) (0.462) (0.394) (0.439) (0.500) (0.457) (0.189)

Notes: standard deviations in parenthesis

Table 2b: 1991 Levels of Public Goods, regression

State

Schools per 1000

inhabitants

Health facilities per

1000 inhabitants

Taps available (dummy)

Tube well available (dummy)

Bus stop in village

(dummy)

Pucca approach road

(dummy)

Kacha approach road

(dummy)

Domestic electricity (dummy)

Fraction land irrigated

Andhra 0.340 -0.211 -0.498 -0.606 -0.390 -0.578 -0.395 0.352 -0.113(1.383) (2.522) (4.484) (6.851) (4.233) (5.683) (4.002) (4.108) (2.416)

Karnataka 0.885 -0.223 -0.404 -0.786 -0.146 -0.111 -0.897 -0.286 -0.026(5.845) (3.926) (4.169) (12.041) (2.648) (1.534) (12.594) (3.935) (0.622)

TamilNadu 0.216 -0.212 -0.505 -0.907 0.029 -0.203 -0.732 0.469 -0.031(1.498) (3.115) (4.732) (14.676) (0.495) (2.572) (9.755) (6.103) (0.595)

Prad. Village -0.084 0.031 0.084 0.048 0.099 0.118 -0.106 -0.018 0.003(0.879) (1.203) (2.372) (1.700) (3.094) (3.602) (3.273) (0.936) (0.290)

Reserved GP 0.043 0.004 0.027 -0.014 0.054 0.064 -0.037 -0.008 -0.050(0.503) (0.117) (0.582) (0.399) (1.348) (1.430) (0.817) (0.225) (2.288)

N 477 476 472 481 478 478 480 482 481Adj R-sq 0.295 0.140 0.372 0.569 0.177 0.281 0.442 0.676 0.353Notes:1)absolute values of t-statistics clustered by census code in parenthesis2)block pair fixed effects included in regression

Table 3a: Current level of public goods, simple mean comparison

State

Schools per 1000

inhabitants

Health facilities per

1000 inhabitants

Number drinking water

sources

Number overhead

tanks

Bus stop in village

(dummy)Proportion paved road

Proportion road with light

Andhra 1.980 0.235 3.171 0.943 0.500 0.206 0.436(1.534) (0.512) (2.713) (0.931) (0.504) (0.213) (0.258)

Karnataka 1.403 0.078 3.753 0.610 0.577 0.787 0.418(1.098) (0.210) (2.454) (0.748) (0.495) (0.182) (0.263)

Kerala 2.120 2.891 12.397 0.143 0.024 0.459 0.396(1.137) (1.621) (9.906) (0.451) (0.153) (0.200) (0.281)

TamilNadu 1.068 0.151 1.924 1.132 0.653 0.465 0.460(1.061) (0.529) (1.778) (0.821) (0.478) (0.301) (0.280)

All 1.535 0.701 5.257 0.686 0.454 0.542 0.427(1.234) (1.386) (6.652) (0.825) (0.498) (0.302) (0.272)

Notes: standard deviations in parenthesis

Table 3b: Current level of public goods, regression

State

Schools per 1000

inhabitants

Health facilities per

1000 inhabitants

Number drinking water

sources

Number overhead

tanks

Bus stop in village

(dummy)Proportion paved road

Proportion road with light

Andhra -0.625 -2.675 -9.111 0.678 0.449 -0.311 0.154(1.806) (10.460) (5.325) (2.497) (6.597) (5.891) (2.422)

Karnataka -1.208 -2.794 -7.714 0.506 0.576 0.248 0.148(5.083) (13.776) (4.785) (3.675) (13.091) (7.070) (3.098)

TamilNadu -1.332 -2.847 -11.178 0.998 0.727 -0.033 0.145(5.419) (9.677) (7.137) (6.715) (18.896) (0.721) (3.271)

Prad. Village -0.234 -0.011 1.190 0.345 0.173 -0.024 0.044(1.750) (0.185) (3.101) (3.830) (3.814) (1.350) (1.873)

Reserved GP 0.014 -0.055 0.718 -0.017 0.049 -0.031 -0.007(0.158) (0.572) (1.120) (0.241) (0.992) (1.138) (0.297)

N 495 495 504 504 504 501 488Adj R-sq 0.232 0.659 0.450 0.246 0.275 0.475 0.184Notes:1)absolute values of t-statistics clustered by block in parenthesis2)block pair fixed effects included in regression

Table 4a: GP activism, from facilities questionnaire, simple mean comparison

StateOverall GP

activity (dummy)

GP activism in schools (dummy)

GP activism in anganwadi (dummy)

GP activism in health (dummy)

Nr Drinking Water Sources built/Improved

Proportion road built/improved

Proportion road with light

built/improvedAndhra 0.871 0.343 0.014 0.014 1.157 0.436 0.820

(0.337) (0.478) (0.120) (0.120) (1.708) (0.332) (0.343)Karnataka 0.901 0.407 0.253 0.027 0.407 0.162 0.254

(0.299) (0.493) (0.436) (0.164) (0.814) (0.198) (0.388)Kerala 0.984 0.651 0.698 0.087 2.159 0.187 0.315

(0.125) (0.479) (0.461) (0.283) (4.787) (0.186) (0.324)TamilNadu 0.243 0.104 0.042 0.007 0.083 0.019 0.053

(0.430) (0.307) (0.201) (0.083) (0.383) (0.081) (0.195)All 0.736 0.374 0.270 0.034 0.841 0.165 0.287

(0.441) (0.484) (0.444) (0.183) (2.610) (0.232) (0.397)Notes: standard deviations in parenthesis2)activities are after last election

Table 4b: GP activism, from facilities questionnaire, regression

StateOverall GP

activity (dummy)

GP activism in schools (dummy)

GP activism in anganwadi (dummy)

GP activism in health (dummy)

Nr Drinking Water Sources built/Improved

Proportion road built/improved

Proportion road with light

built/improvedAndhra 0.038 -0.011 -0.560 -0.123 -1.429 0.289 0.582

(0.476) (0.116) (12.771) (4.248) (1.982) (5.651) (6.016)Karnataka 0.057 0.036 -0.336 -0.082 -2.147 0.009 0.049

(1.110) (0.512) (8.155) (3.399) (3.454) (0.321) (0.845)TamilNadu -0.725 -0.421 -0.574 -0.121 -2.722 -0.209 -0.251

(13.263) (5.449) (13.478) (4.951) (4.481) (9.643) (4.368)Prad. Village 0.057 0.089 0.045 0.043 0.283 0.024 0.033

(1.924) (2.062) (1.356) (2.530) (1.153) (1.277) (1.128)Reserved GP -0.099 -0.076 0.017 0.019 0.005 -0.041 -0.090

(3.287) (1.282) (0.581) (1.499) (0.018) (1.560) (2.224)N 504 504 504 504 504 501 484Adj R-sq 0.497 0.256 0.327 0.046 0.144 0.342 0.557Notes:1)absolute values of t-statistics clustered by block in parenthesis2)block pair fixed effects included in regression

Table 5a: GP activity, from PRA, simple means comparison

StateOverall GP

activity

GP activism in schools (count)

GP activism in health (count)

GP activism in water (count)

GP activism in sanitation (count)

GP activism in transport (count)

GP activism in road (count)

GP activism in electricity (count)

GP activism in irrigation (count)

Andhra 0.407 0.529 0.343 0.529 0.629 0.214 0.943 0.714 0.257(0.227) (0.653) (0.587) (0.675) (0.802) (0.447) (0.832) (0.783) (0.530)

Karnataka 0.409 0.418 0.203 0.484 0.505 0.132 0.874 1.011 0.093(0.291) (0.596) (0.583) (0.646) (0.663) (0.370) (0.780) (1.217) (0.327)

Kerala 0.438 0.333 0.500 0.310 0.270 0.087 0.802 0.762 0.143(0.238) (0.537) (0.654) (0.513) (0.497) (0.283) (0.607) (0.774) (0.394)

TamilNadu 0.238 0.313 0.278 0.396 0.125 0.049 0.264 0.549 0.076(0.201) (0.573) (0.508) (0.582) (0.332) (0.216) (0.542) (0.698) (0.292)

All 0.369 0.383 0.314 0.423 0.360 0.109 0.697 0.784 0.123(0.260) (0.587) (0.592) (0.606) (0.601) (0.330) (0.739) (0.952) (0.372)

Notes:1)standard deviations in parenthesis2)Overall GP activity is the ratio of sectors in which GP was active, to total sectors3)Activities are after last election

Table 5b: GP activity, from PRA, regressions

StateOverall GP

activity

GP activism in schools (count)

GP activism in health (count)

GP activism in water (count)

GP activism in sanitation (count)

GP activism in transport (count)

GP activism in road (count)

GP activism in electricity (count)

GP activism in irrigation (count)

Andhra 0.103 0.203 -0.217 0.336 0.333 0.050 0.461 0.059 0.124(1.375) (0.824) (1.459) (1.898) (2.428) (0.551) (2.330) (0.229) (1.390)

Karnataka 0.107 0.117 -0.241 0.291 0.265 0.059 0.461 0.361 -0.049(1.776) (0.685) (3.048) (2.040) (2.537) (1.691) (2.978) (2.009) (0.997)

TamilNadu -0.112 0.018 -0.160 0.209 -0.140 -0.048 -0.286 -0.019 0.007(1.781) (0.094) (1.551) (1.273) (1.450) (1.349) (1.794) (0.118) (0.122)

Prad. Village 0.092 0.103 0.125 0.153 0.110 0.082 0.279 0.167 -0.007(3.899) (1.425) (2.299) (2.509) (1.938) (1.983) (4.252) (2.381) (0.212)

Reserved GP -0.010 -0.010 -0.024 0.028 0.043 0.016 0.009 0.049 0.001(0.273) (0.178) (0.383) (0.398) (0.631) (0.476) (0.165) (0.381) (0.024)

N 504 504 504 504 504 504 504 504 504Adj R-sq 0.215 0.042 0.203 0.042 0.108 0.053 0.246 0.167 0.050Notes:1)absolute values of t-statistics clustered by block in parenthesis2)block pair fixed effects included in regression

Table 6: Levels of activism, means

StateAny GP provision

House GP provision

Toilet GP Provision

Water GP Provision

Electricity GP provision BPL received

Received money for

public works

Andhra 0.046 0.025 0.006 0.003 0.013 0.322 0.127(0.209) (0.156) (0.074) (0.053) (0.111) (0.468) (0.334)

Karnataka 0.122 0.024 0.032 0.002 0.073 0.101 0.051(0.327) (0.154) (0.175) (0.039) (0.260) (0.302) (0.220)

Kerala 0.041 0.019 0.019 0.000 0.014 0.297 0.019(0.199) (0.138) (0.135) (0.000) (0.117) (0.457) (0.136)

TamilNadu 0.023 0.006 0.006 0.006 0.007 0.251 0.020(0.150) (0.075) (0.075) (0.075) (0.083) (0.434) (0.139)

All 0.065 0.018 0.018 0.002 0.032 0.220 0.044(0.246) (0.133) (0.133) (0.049) (0.177) (0.414) (0.205)

Levels of activism, regression

StateAny GP provision

House GP provision

Toilet GP Provision

Water GP Provision

Electricity GP provision BPL received

Received money for

public works

Andhra -0.004 0.012 0.008 0.002 -0.039 0.207 0.075(0.220) (1.344) (0.797) (0.569) (2.871) (1.981) (3.484)

Karnataka 0.077 0.009 0.032 0.000 0.033 -0.032 0.010(5.969) (1.610) (4.147) (0.139) (3.033) (0.407) (1.088)

TamilNadu -0.032 -0.015 0.001 0.006 -0.035 0.088 -0.028(3.063) (3.263) (0.122) (1.726) (3.732) (0.872) (3.536)

Pradhan's Village 0.008 0.000 0.007 0.002 0.001 -0.018 0.004(0.770) (0.081) (1.394) (0.711) (0.174) (1.195) (0.810)

Reserved GP -0.007 -0.006 0.000 0.001 -0.004 0.019 -0.007(0.942) (1.354) (0.038) (0.918) (0.778) (0.602) (0.882)

female 0.005 0.006 -0.006 0.000 0.006 -0.004 -0.005(0.943) (1.769) (1.881) (0.045) (1.440) (0.401) (0.751)

SCST 0.035 0.016 -0.001 0.000 0.025 0.128 0.043(3.020) (2.377) (0.140) (0.219) (3.103) (3.930) (3.886)

wealthy -0.043 -0.014 -0.006 0.001 -0.030 -0.096 -0.001(5.311) (3.757) (1.312) (0.468) (4.160) (4.079) (0.171)

landless 0.019 0.005 0.007 -0.001 0.010 0.074 0.014(1.914) (1.007) (1.365) (0.438) (1.554) (4.850) (2.764)

politician 0.033 -0.002 0.028 -0.003 0.018 0.092 0.059(1.889) (0.429) (2.394) (2.467) (1.483) (1.365) (2.363)

N 5460 5460 5460 5460 5460 5460 5422Adj R-sq 0.044 0.009 0.025 0.002 0.041 0.167 0.047Notes:1)absolute values of t-statistics clustered by block in parenthesis2)block pair fixed effects included in regression

Table 7a: Village participation, simple means comparison

State NGO active CBO activeGS held last

6moGS held last12mo

Nr education comitees

Nr total comitees

Andhra Pradesh 0.686 0.243 0.710 0.696 0.557 0.557

(0.468) (0.432) (0.457) (0.464) (1.016) (1.016)

Karnataka 0.379 0.819 0.692 0.538 0.264 0.330

(0.487) (0.386) (0.463) (0.500) (0.466) (0.657)

Kerala 0.111 0.389 0.984 0.984 0.675 2.341

(0.316) (0.489) (0.125) (0.125) (0.470) (1.550)

Tamil Nadu 0.292 0.590 0.672 0.664 0.021 0.056

(0.456) (0.493) (0.471) (0.474) (0.143) (0.308)

All states 0.331 0.575 0.761 0.702 0.335 0.770

(0.471) (0.495) (0.427) (0.458) (0.578) (1.304)

Table 7B: Village participation, regressions

State NGO active CBO activeGS held last

6moGS held last12mo

Nr education comitees

Nr total comitees

Andhra 0.103 0.203 -0.217 0.336 0.333 0.050(1.375) (0.824) (1.459) (1.898) (2.428) (0.551)

Karnataka 0.107 0.117 -0.241 0.291 0.265 0.059(1.776) (0.685) (3.048) (2.040) (2.537) (1.691)

Tamil Nadu -0.112 0.018 -0.160 0.209 -0.140 -0.048(1.781) (0.094) (1.551) (1.273) (1.450) (1.349)

Prad. Village 0.092 0.103 0.125 0.153 0.110 0.082(3.899) (1.425) (2.299) (2.509) (1.938) (1.983)

Reserved GP -0.010 -0.010 -0.024 0.028 0.043 0.016(0.273) (0.178) (0.383) (0.398) (0.631) (0.476)

N 504 504 504 504 504 504Adj R-sq 0.215 0.042 0.203 0.042 0.108 0.053Notes:1)absolute values of t-statistics clustered by block in parenthesis2)block pair fixed effects included in regression

Table 8a Gram Sabha participation, means

State Attend GS

Attend GS for

beneficiary GS speaking

Andhra 0.107 0.935 0.286

(0.309) (0.248) (0.455)

Karnataka 0.141 0.900 0.036

(0.348) (0.301) (0.186)

Kerala 0.397 0.686 0.523

(0.489) (0.464) (0.500)

TamilNadu 0.131 0.806 0.252

(0.338) (0.397) (0.435)

All 0.199 0.777 0.338

(0.399) (0.416) (0.473)

Table 8b Gram Sabha participation, regression

State Attend GS

Attend GS for

beneficiary GS speaking

Andhra -0.200 0.335 -0.357

(5.024) (9.917) (5.128)

Karnataka -0.179 0.239 -0.544

(5.656) (11.023) (15.822)

TamilNadu -0.194 0.127 -0.247

(6.161) (6.351) (5.927)

Pradhan's Vill 0.019 0.017 0.018

(1.377) (0.591) (0.693)

Reserved GP 0.002 -0.090 -0.051

(0.108) (2.534) (1.182)

female -0.187 -0.097 -0.074

(11.768) (2.860) (2.737)

SCST 0.023 0.014 -0.025

(1.344) (0.331) (0.553)

wealthy -0.011 0.023 -0.028

(0.527) (0.772) (0.949)

landless 0.014 -0.035 -0.079

(1.214) (1.338) (1.831)

politician -0.231

(9.636)

N 5460 1054 1054

Adj R-sq 0.180 0.076 0.197

Notes:

1)absolute values of t-statistics clustered by block in parenthesis

2)block pair fixed effects included in regression

Table 9 Household information and tax payment, means

State Read news Knows CM taxpay

Andhra 0.233 0.689 0.375

(0.423) (0.463) (0.484)

Karnataka 0.295 0.403 0.873

(0.456) (0.491) (0.333)

Kerala 0.550 0.626 0.912

(0.498) (0.484) (0.283)

TamilNadu 0.300 0.683 0.890

(0.459) (0.466) (0.313)

All 0.353 0.572 0.825

(0.478) (0.495) (0.380)

Household information and tax payment, regression

State Read news Knows CM taxpay

Andhra -0.138 0.330 -0.646

(3.136) (7.124) (8.702)

Karnataka -0.097 0.031 -0.147

(2.250) (0.846) (2.772)

TamilNadu -0.125 0.238 -0.104

(3.496) (8.966) (1.632)

Pradhan's Vill 0.032 0.026 0.029

(2.078) (1.667) (1.928)

Reserved GP 0.004 0.046 -0.014

(0.238) (2.385) (0.625)

female -0.303 -0.307 -0.034

(19.951) (17.735) (4.111)

SCST -0.068 -0.037 -0.016

(3.179) (1.629) (1.049)

wealthy 0.180 0.179 0.061

(12.951) (10.628) (3.414)

landless -0.030 -0.033 -0.074

(2.200) (1.738) (4.110)

politician 0.266 0.258 0.057

(12.337) (8.656) (2.519)

N 5460 5460 5460

Adj R-sq 0.283 0.308 0.268

Notes:

1)absolute values of t-statistics clustered by block in parenthesis

2)block pair fixed effects included in regression

Table 10 Political participation, means

StateHH member

politicalHH party affilieated

Participate political Voted GP Voted MLA Voted MP Vote group Vote party

Vote candidate

Andhra 0.054 0.642 0.253 0.761 0.865 0.761 0.063 0.131 0.377

(0.227) (0.480) (0.435) (0.427) (0.342) (0.427) (0.242) (0.338) (0.485)

Karnataka 0.050 0.064 0.053 0.713 0.782 0.713 0.142 0.053 0.370

(0.218) (0.244) (0.224) (0.452) (0.413) (0.452) (0.349) (0.225) (0.483)

Kerala 0.053 0.429 0.311 0.844 0.902 0.844 0.079 0.392 0.133

(0.223) (0.495) (0.463) (0.363) (0.297) (0.363) (0.270) (0.488) (0.339)

TamilNadu 0.052 0.247 0.093 0.801 0.811 0.801 0.091 0.029 0.598

(0.222) (0.431) (0.290) (0.399) (0.392) (0.399) (0.287) (0.168) (0.490)

All 0.052 0.279 0.154 0.777 0.831 0.777 0.102 0.142 0.373

(0.222) (0.449) (0.361) (0.417) (0.375) (0.417) (0.303) (0.350) (0.484)

Political participation, regression

StateHH member

politicalHH party affilieated

Participate political Voted GP Voted MLA Voted MP Vote group Vote party

Vote candidate

Andhra 0.011 0.468 -0.073 -0.154 -0.082 -0.154 -0.007 -0.249 0.186

(0.809) (3.805) (2.311) (3.588) (2.754) (3.588) (0.286) (7.780) (3.152)

Karnataka 0.008 -0.147 -0.265 -0.184 -0.174 -0.184 0.073 -0.339 0.211

(0.944) (1.351) (11.224) (6.103) (9.859) (6.103) (4.810) (11.812) (6.060)

TamilNadu 0.006 0.020 -0.208 -0.120 -0.155 -0.120 -0.001 -0.332 0.382

(0.929) (0.162) (7.264) (3.823) (9.109) (3.823) (0.089) (12.261) (9.741)

Pradhan's Vill 0.011 -0.002 0.027 0.012 -0.005 0.012 0.021 0.007 -0.016

(1.568) (0.151) (1.601) (1.086) (0.372) (1.086) (1.688) (0.724) (0.898)

Reserved GP -0.001 0.037 -0.006 -0.004 0.003 -0.004 0.001 0.016 -0.033

(0.149) (1.078) (0.218) (0.183) (0.243) (0.183) (0.061) (0.871) (1.889)

female -0.014 -0.097 -0.117 -0.002 -0.098 -0.002 -0.026 -0.046 -0.103

(2.249) (7.945) (7.488) (0.163) (11.553) (0.163) (2.994) (3.498) (4.952)

SCST 0.009 0.060 0.043 0.013 -0.003 0.013 0.004 0.043 -0.006

(1.368) (2.723) (2.528) (0.731) (0.222) (0.731) (0.226) (2.234) (0.322)

wealthy 0.046 0.010 0.013 -0.090 0.036 -0.090 0.003 -0.006 0.041

(5.623) (0.722) (1.282) (5.051) (2.913) (5.051) (0.263) (0.585) (2.858)

landless -0.026 -0.018 -0.024 0.069 -0.019 0.069 -0.016 -0.018 0.011

(2.954) (1.170) (1.807) (4.344) (1.504) (4.344) (1.802) (2.035) (0.564)

politician -0.079 -0.326 -0.188

(12.695) (6.822) (7.928)

N 5460 5460 5460 5460 5460 5460 4940 4940 4940

Adj R-sq 0.019 0.316 0.154 0.038 0.053 0.038 0.024 0.186 0.141

Notes:

1)absolute values of t-statistics clustered by block in parenthesis

2)block pair fixed effects included in regression

Table 11 Household participation in cash or kind, means

StateProvision for

roadsProvision for anganwadi

Provision for Health subc

Provision for P. School

Provision for dr. water

Any provision

Andhra 0.128 0.033 0.017 0.060 0.089 0.208

(0.334) (0.180) (0.128) (0.237) (0.285) (0.406)

Karnataka 0.072 0.037 0.002 0.100 0.055 0.179

(0.258) (0.188) (0.039) (0.300) (0.228) (0.384)

Kerala 0.346 0.139 0.044 0.073 0.085 0.415

(0.476) (0.346) (0.206) (0.260) (0.279) (0.493)

TamilNadu 0.059 0.018 0.010 0.068 0.104 0.183

(0.236) (0.132) (0.097) (0.252) (0.305) (0.387)

All 0.145 0.057 0.016 0.079 0.080 0.244

(0.352) (0.232) (0.127) (0.270) (0.272) (0.429)

Household participation in cash or kind, regression

StateProvision for

roadsProvision for anganwadi

Provision for Health subc

Provision for P. School

Provision for dr. water

Any provision

Andhra -0.130 -0.029 -0.037 0.121 0.124 0.079

(2.823) (1.127) (3.289) (2.575) (4.313) (1.265)

Karnataka -0.193 -0.041 -0.047 0.133 0.082 0.013

(3.965) (1.560) (4.892) (2.724) (4.999) (0.193)

TamilNadu -0.230 -0.067 -0.043 0.076 0.092 -0.061

(5.263) (2.779) (5.433) (1.992) (6.154) (1.099)

Pradhan's Vill 0.003 0.008 0.006 0.012 -0.001 -0.013

(0.250) (1.246) (1.536) (1.339) (0.121) (0.761)

Reserved GP 0.014 0.007 -0.001 0.028 -0.005 0.021

(0.672) (0.638) (0.141) (1.539) (0.471) (0.781)

female -0.033 -0.016 -0.005 -0.035 -0.022 -0.058

(3.305) (2.618) (1.362) (4.489) (2.796) (5.771)

SCST 0.000 -0.008 -0.004 -0.030 0.010 -0.029

(0.007) (1.572) (1.465) (3.055) (0.965) (1.538)

wealthy 0.052 0.031 0.010 0.044 0.034 0.091

(4.587) (3.909) (4.005) (3.587) (2.851) (6.305)

landless -0.060 -0.026 -0.005 -0.037 -0.007 -0.072

(3.278) (2.853) (1.156) (5.115) (0.970) (4.053)

politician 0.138 0.089 0.050 0.119 0.156 0.228

(5.760) (4.721) (2.936) (4.230) (4.973) (6.259)

N 5460 5460 5460 5460 5460 5460

Adj R-sq 0.173 0.094 0.044 0.105 0.065 0.164

Notes:

1)absolute values of t-statistics clustered by block in parenthesis

2)block pair fixed effects included in regression

Table 12: Household willingness to pay, means

State

Willing provide roads

Willing provide

anganwadi

Willing provide

Health subc

Willing provide P.

school

Willing provide dr

waterWilling

provide any

Andhra 0.329 0.210 0.263 0.228 0.276 0.485

(0.470) (0.407) (0.440) (0.420) (0.448) (0.500)

Karnataka 0.103 0.084 0.021 0.089 0.090 0.189

(0.304) (0.277) (0.144) (0.285) (0.287) (0.392)

Kerala 0.333 0.362 0.369 0.337 0.401 0.550

(0.471) (0.481) (0.483) (0.473) (0.490) (0.498)

TamilNadu 0.352 0.296 0.291 0.314 0.338 0.439

(0.478) (0.457) (0.455) (0.464) (0.473) (0.496)

All 0.258 0.228 0.214 0.231 0.260 0.386

(0.438) (0.420) (0.410) (0.422) (0.439) (0.487)

Household willingness to pay, regression

State

Willing provide roads

Willing provide

anganwadi

Willing provide

Health subc

Willing provide P.

school

Willing provide dr

waterWilling

provide any

Andhra 0.009 -0.200 -0.133 -0.140 -0.168 0.027

(0.232) (4.108) (2.683) (2.819) (3.741) (0.821)

Karnataka -0.211 -0.301 -0.348 -0.251 -0.334 -0.268

(6.429) (8.444) (8.842) (7.458) (10.691) (12.258)

TamilNadu 0.030 -0.066 -0.068 -0.029 -0.067 -0.044

(0.893) (1.908) (1.719) (0.851) (2.241) (1.982)

Pradhan's Vill 0.010 0.033 0.025 0.031 0.017 0.032

(0.613) (2.204) (2.098) (2.170) (1.040) (1.747)

Reserved GP 0.002 0.008 0.008 -0.001 0.025 0.015

(0.171) (0.460) (0.502) (0.079) (1.523) (0.760)

female -0.045 -0.046 -0.057 -0.052 -0.065 -0.085

(4.104) (3.871) (4.318) (4.414) (6.235) (7.324)

SCST 0.001 0.003 0.006 0.001 -0.013 0.019

(0.075) (0.150) (0.400) (0.061) (0.899) (1.071)

wealthy 0.025 0.038 0.032 0.038 0.015 0.057

(1.547) (2.482) (2.401) (2.912) (1.116) (3.363)

landless -0.014 -0.030 -0.027 -0.028 -0.026 -0.063

(0.901) (1.952) (1.715) (2.026) (1.712) (4.208)

politician -0.053 -0.039 -0.089 -0.033 -0.023 -0.003

(1.524) (1.033) (2.476) (0.865) (0.589) (0.056)

N 5460 5460 5460 5460 5460 5460

Adj R-sq 0.077 0.097 0.150 0.084 0.099 0.116

Notes:

1)absolute values of t-statistics clustered by block in parenthesis

2)block pair fixed effects included in regression

Table 13: Simple mean comparisons, inequality variables

Gini GE (a=1)GE(1) within caste groups

GE(1) between

caste groups

Prop GE(1) b/w caste

groups

Andhra Pradesh 0.532 0.734 0.615 0.120 0.180

(0.189) (0.567) (0.554) (0.137) (0.156)

Karnataka 0.522 0.629 0.527 0.102 0.170

(0.155) (0.349) (0.325) (0.100) (0.156)

Kerala 0.658 1.049 0.905 0.144 0.129

(0.139) (0.569) (0.507) (0.198) (0.141)

Tamil Nadu 0.580 0.921 0.768 0.153 0.135

(0.204) (0.635) (0.540) (0.247) (0.168)

All states 0.572 0.825 0.696 0.129 0.152

(0.179) (0.550) (0.492) (0.180) (0.157)

Table 14: Simple mean comparisons, caste dominance and oligarchy variables

Nr castes

Landed percentage,

1951

Upper caste land

dominance (dummy)

Upper caste land

proportionFraction

landless hhs Oligarchy

Andhra Pradesh 11.643 0.658 0.171 0.255 0.286 0.057

(4.872) (0.122) (0.380) (0.260) (0.235) (0.059)

Karnataka 11.192 0.722 0.335 0.364 0.232 0.058

(5.399) (0.116) (0.473) (0.277) (0.188) (0.066)

Kerala 11.556 0.288 0.087 0.171 0.430 0.079

(3.850) (0.082) (0.283) (0.201) (0.247) (0.118)

Tamil Nadu 7.465 0.670 0.236 0.244 0.409 0.096

(5.077) (0.179) (0.426) (0.331) (0.283) (0.109)

All states 10.312 0.594 0.226 0.270 0.336 0.073

(5.199) (0.218) (0.419) (0.284) (0.253) (0.094)

ANNEX B:

The Politics of Public Good Provision: Evidence from Indian Local Governments

The Politics of Public Good Provision: Evidence

from Indian Local Governments∗†

Timothy Besley

London School of Economics

Rohini Pande

Yale University

Lupin Rahman

IMF

Vijayendra Rao

World Bank

Abstract

This paper uses village and household survey data from South India

to examine how political geography and politician identity impacts

on public good provision. We provide evidence that the nature of

this relationship varies by type of public goods. For high spill-over

public goods residential proximity to elected representative matters. In

contrast, for low spill-over public goods sharing the politician’s group

identity is what matters.

JEL Classifications: D78, H40,

∗Acknowledgments: We thank Ian Gascoigne for research assistance. Funding was

provided by World Bank RSB grant P077385, and from the South Asia Rural Department

of the World Bank. The views in this paper are those of the authors and should not be

attributed to the World Bank or the IMF.†Email addresses: Besley: t.besley@lse.ac.uk; Pande: Rohini.Pande@Yale.edu; Rah-

man: lrahman@imf.org; Rao: vrao@worldbank.org.

1

1 Introduction

Making the state more relevant to the interests of the poor is an increasingly

important theme in discussions of anti-poverty policies. Yet there is little

consensus on the appropriate way to develop governance structures that are

responsive to the interests of the poor. Whether greater decentralization of

political power can achieve this remains unclear. On the one hand, it may

enhance the accountability of elected representatives and amplify the politi-

cal voice of poor people while, on the other, it may enhance the influence of

local elites (Bardhan and Mookherjee (2000)) Moreover, whether decentral-

ized public good provision better represents the needs of the local population

remains sensitive to assumptions about heterogeneity of preferences in the

local population and extent of spill-overs associated with different public

goods (Besley and Coate (2003)).

This paper uses survey data on village governments in South India to pro-

vide some evidence on these issues. In India, a 1993 constitutional amend-

ment made a three-tier elected local government obligatory throughout the

country.1 Our focus is on the lowest tier of this local self-government. This

is a popularly elected village council — the Gram Panchayat (from now on,

GP). The constitutional amendment also required state governments to del-

egate certain policy-making powers to these local governments. The specific

choice of these policies was left up to states. States have typically dele-

gated responsibility for the construction and maintenance of village public

goods and beneficiary selection for various central and state-funded wel-

1The three tiers are defined at different administrative levels with the village being the

lowest, then the block and finally the district. Matthew and Buch (2003) provide more

details about how this was implemented.

2

fare schemes to these bodies (see Chaudhuri (2003 and Matthew and Buch

(2000)) for overviews of the diverse experience of Indian states).

The Indian decentralization experiment is unique on many fronts — of

main interest to us are the facts that it mandated political representation

via reservation for socially and economically disadvantaged groups and gave

representatives elected by villagers decision-making power over an array of

village-level public goods.2

We focus on reservation for the post of the head of the GP in favor

of scheduled castes/scheduled tribes (SC/ST). SC/STs include castes and

tribes which have historically suffered economic and social discrimination.3

In GPs where the post is reserved for SC/STs only SC/ST individuals can

stand for election. The composition of the electorate is unaffected by polit-

ical reservation.

Previous work on political reservation suggests that political reserva-

tion for a group leads to a higher incidence of policies preferred by and/or

targeted towards that group (see Pande (2003) for state-level evidence in

the case of SC/ST, and Chattopadhyay and Duflo (2002) for village-level

evidence in the case of women). Our contribution is to point to the im-

portance of public good technology and political geography in shaping the

policy impact of political reservation.

The head has the ability to shape resource allocation, and hence may do

so in a direction that favors his own village. How village members benefit

from this depends on the technology of the public good. With high spill-over

2As expenditure levels of village governments are largely set by state governments our

main focus is on distribution.3See Pande (2003) for a description of which castes/tribes belong to these categories,

and Gupta (2000) for an overview of caste-based discrimination.

3

public goods such as the access road to a village or an overhead tank for

water, the whole village benefits. However, for low spill-over goods such as

programs targeted towards specific groups within the villages, it is less clear.

We may expect this to depend on the underlying preferences and sympathies

of the head.

Our analysis incorporates insights from the local public finance literature

— this concerns the allocation of public spending across geographical units

within a polity. In the well-knownWeingast, Shepsle, Johnsen (1980) model,

the problem is to allocate pure local public goods to a variety of districts,

each of whose interest is represented by a legislator. They propose that

resource allocation will obey a “norm of universalism” in which each district

gets what they want as long as all other districts are allowed to do the

same. In their model, there is excessive spending, but the allocation is

equal. This contrasts with agenda setting models of resource allocation

where the propose is able to get an advantage in getting his/her preferred

outcome (see Romer and Rosenthal (1978)) or a minimum winning coalition

model in which the winning group is able to get an outcome that it favors

(see Baron (1993)). Our findings suggest that agenda setting models can

better explain public good allocation in South Indian villages.

This paper fits into a wider literature studying the social and politi-

cal context of public spending. A variety of studies place weight on the

relationship between heterogeneity and public goods provision — see, for ex-

ample, Alesina, Baqir and Easterly (1999) and Miguel and Gugerty (2002).

It is also related to the large literature on political determinants of resource

allocation, see for example, Knight (2003).

The remainder of the paper is organized as follows. In the next section,

we describe the institutional setting. In section three, we discuss a simple

4

model which motivates our results. Section four we describe our survey and

present results. Section five concludes.

2 Institutional background

The GP is the lowest tier of local self-government in India and is a popularly

elected village council. Depending on village population, a GP may cover

between 1 and 5 revenue villages. Every GP consists of up to twenty wards.4

Elections are at the ward-level, and the elected ward members constitute the

GP council. The head of this council is the Pradhan.5

The 73rd constitutional amendment mandated political reservation in

favor of SC/ST for the Pradhan position, and required that the extent of

such reservation in a state reflect the SC/ST population share in that state.

The amendment also required that no GP be reserved for the same group

for two consecutive elections. The choice of which GPs to reserve was left to

individual states. Typically, the same fraction of GPs are reserved in every

district in a state.

A GP has responsibilities of civic administration with limited indepen-

dent taxation powers.6 While the ambit of GP policy influence varies across

Indian states GPs typically perform (at least) two distinct policy tasks. The

first is beneficiary selection for central and state welfare schemes. These are

schemes which provide beneficiary households with funds to acquire house-

hold public goods such as housing and private electricity and water supply.

4For our sample states the population per ward varies between 300 and 800.5In Andhra Pradesh and Tamil Nadu the Pradhan is directly elected, while in Kar-

nataka he/she is nominated from the pool of elected ward members.6On average, roughly 10 percent of a GP’s total revenue come from own revenues with

the remainder consisting of transfers from higher levels of government.

5

Eligibility for these schemes is usually restricted to households below the

official poverty line. In addition, most schemes require that a minimum

fraction of beneficiaries be SC/ST. The second area of GP policy activism

is the construction and maintenance of village public goods such as street-

lights, roads and drains. The GP decides the distribution of these public

goods within the village, and the quality of such public good provision.7

Panchayat legislation requires that the Pradhan consult with villagers

(via village meetings) and ward members in deciding the choice of beneficia-

ries and allocation of public goods. However, final decision-making powers

in a GP are vested with the Pradhan.

3 Theory

We start with a theoretical model which is intended to think through the

issues.8 Consider a GP comprising of two villages indexed j ∈ {1, 2}. Eachvillage has two caste groups indexed k ∈ {s, n}, where s denotes the SC/STgroup and n the non SC/ST group. The share of group s in village j is πj.

For simplicity, assume a single public good is provided to each group

within a village. Let gjk ∈ [0,G] denote the level of public good provisionfor caste k in village j. This public good may have positive spill-overs

for villagers belonging to the other caste group, −k. Hence individuals

(potentially) care about the level of public goods provided to both caste

groups in a village. Specifically:

7Schedule XI of the Constitution defines the functional items for which states may

devolve responsibility to Panchayats.8The model is very similar in many respects to Besley and Coate (2003).

6

V jk

³gjk, g

j−k´= log(gjk) + λ log(gj−k) + y

jk

λ ≥ 0 measures the extent of spill-overs in public good provision. Privategoods are captured in the term yjk. If λ = 1, then it is a pure village-level

public good, while if λ = 0, then the good only benefits the group to whom

it is provided.

Public goods are funded from a fixed pot of tax revenue, T . We normalize

the price of public good provision to one. Thus, the budget constraint is:

g1s + g2s + g

1n + g

2n = T .

Group-wise allocation of public goods is determined by elected GP rep-

resentatives. Each village elects one villager as representative, one of whom

is the Pradhan. We adopt the convention that village one is the Pradhan’s

village, that is it has the Pradhan as the representative. The GP is reserved

if only SC/ST individuals can run for election in village 1. For expositional

ease we assume that, absent reservation, SC/ST individuals never run for

election.9 We do not explicitly model the decision making procedure but

assume that it maximizes a weighted sum of the utility of the two repre-

sentatives where a weight µ > 1/2 is applied to the utility of the Pradhan.

Let ` (j) ∈ {s, n} be the type of the Panchayat representative in village j.9This assumption is in line with reality — Chattopadhyay and Duflo (2003) show that

this can be explained by the minority group having higher costs of running for election,

while Pande (2003) shows that this can also be explained by inadequate minority repre-

sentation in political parties.

7

Then, the public good allocation will solve:

µV 1`(1)

³g1`(1), g

1−`(1)

´+ (1− µ)V 2`(2)

³g2`(2), g

2−`(2)

´subject to

g1s + g2s + g

1n + g

2n = T

It is easy to check that the solution to this is:

g1`(1) =µ1+λT g1−`(1) =

µλ1+λT

g2`(2) =(1−µ)1+λ T g2−`(2) =

(1−µ)λ1+λ T.

Thus the village/caste group allocation depends on the decision-making pro-

cess as represented by µ and the extent of spill-overs in public good provision

as represented by λ. Comparison of the public good level across groups yields

the following empirically testable predictions.

Claim 1 Pradhan effects — Relative to non-Pradhan village, public good al-

location is higher in Pradhan’s village.

Claim 2 Caste effects — Relative to non SC/ST group, the public good al-

location for the SC/ST group is higher when the GP is reserved.

Claim 3 Spill-overs — The impact of reservation on public good allocation

diminishes as spill-overs increase.

4 Evidence

In this section we use survey data from India to provide evidence on the

impact of Pradhan residence and political reservation on the provision of

low and high spill-over public goods.

8

4.1 Data and Survey Design

Our data comes from a survey we conducted in three South Indian states —

Andhra Pradesh, Karnataka and Tamil Nadu — between September-November

2002. At this point at least one year had lapsed since the last GP election

in each of our sample states.10 The survey covered 396 villages across 181

GPs in thirty blocks (a block is the administrative unit below a district in

a state).11 Summary statistics are provided in Table 1 (for details of the

survey, see Besley, Pande, Rahman and Rao (2003)).

We use information from an independent audit of village facilities to

construct an index of GP activity on high spill-over (i.e. village-level) public

goods. This index measures whether the GP undertook any construction or

improvement activity on within-village roads, drains, street-lights and water

sources since the last GP election. The index is normalized to lie between 0

and 1. Roughly seventy-nine percent of our sample villages experienced GP

activism on at least one of these public goods.

We use data from household surveys in a random sub-sample of 193 vil-

lages to measure the provision of low spill-over (household) public goods.

In every sampled village twenty one household surveys were conducted, of

which four were with SC/ST households and one was with an elected Pan-

chayat representative.12 This gives us a total of 4059 households of which

10The second round of GP elections in these states occurred in August 2001 in Andhra

Pradesh, February 2000 in Karnataka, and October 2001 in Tamil Nadu.11The survey was also conducted in Kerala. Kerala, however, has a different adminis-

trative structure — for instance, a Kerala Gram Panchayat covers a population of 30,000

as against 5-10,000 in the other states.12An additional household survey was conducted with the Pradhan if s/he resided in

that village, and with a ward member otherwise (in six villages both a ward member and

Pradhan interview were conducted).

9

981 were SC/ST. We measure a household’s exposure to low spill-over public

goods by a dummy which equals one if it had a house or toilet built under a

government scheme or if it received a private water or electricity connection

via a government scheme since the last GP election. Approximately seven

percent of the sample households fall in this category.

We are interested in the implications of political reservation and Pradhan

proximity for the allocation of high and low spill-over public goods across

and within villages. We capture a village’s reservation status by a dummy

variable which equals one if the village belongs to a GP reserved for SC/ST.

We use two dummy variables to measure the political influence of a village

— the first equals one if the Pradhan resides in that village, and the second

equals one if the GP headquarters are in that village.

4.2 Household Level Evidence

Let yivg be an indicator variable which equals one if household i in village v

in GP g has received a low spill-over public good since the last GP election.

We estimate a regression of the form:

yivg = αv + γ1Civg + γ2Civg ×Rg + γ3Civg × Pvg + γ4Civg ×Gvg + φXivg + εivg

(1)

where Civg is a SC/ST dummy, Rg the SC/ST reservation dummy and Pvg

and Gvg the Pradhan’s village and GP headquarter dummies respectively.

αv are village fixed effects and Xivg is a set of household level controls (see

notes to Table 2 for details). Inclusion of a village fixed effect implies that

we identify the effect of reservation on public good provision solely from

within village variation in allocation.

10

The results are in Table 2, columns (1) through (4). In column (1) we

see that, in line with scheme guidelines, household public goods are tar-

geted towards SC/ST households — on average, a SC/ST household is six

percent more likely to receive such a public good. In column (2) we find

that the extent of such targeting is enhanced by living in a reserved GP.

Relative to living in a non-reserved GP, living in a reserved GP increases a

SC/ST household’s likelihood of getting such a public good by seven per-

centage points. Columns (3) and (4) demonstrate that this effect is robust

to including interactions with Pradhan village and GP headquarter since

Pradhans may belong to the GP headquarter, and that neither interactions

are significant. This suggests that enhanced targeting of SC/ST households

only comes from reservation.

4.3 Village Level Evidence

We now turn to the determinants of village-level allocation of public goods.

Our model suggests that overall, relative to non-Pradhan villages, the Prad-

han’s village will be allocated more public goods. The difference in alloca-

tion will, however, vary by type of public good. In the case of low spill-over

public goods we will expect higher provision of public goods in Pradhan

village if the GP is reserved for SC/ST, and lower otherwise. As the spill-

overs associated with the public good increase the difference between levels

of provision in reserved and non-reserved Pradhan villages should diminish.

For high spill-over public goods, irrespective of GP reservation status, we

should observe higher allocation in Pradhan’s village.

To examine these predictions we turn to a village-level analysis. First, to

examine the village-level determinants of household public good incidence we

11

recover the village fixed effects from (1) and regress these on village char-

acteristics. Second, to examine the determinants of high spill-over public

goods we use our index of GP activism on village public goods.13.

Our empirical model for village level regressions is:

yvg = αb + γ1Rg + γ2Pvg + γ3Gvg + γ4Rg × Pvg + φXvg + εvg

where αb are block dummy variables andXvg are village level controls. These

regressions rely on within-block variation in the explanatory variables for

identification purposes.

In our household-level regressions (columns (1)-(4)) the village fixed ef-

fects were jointly significant. In columns (5) and (6), Table 2 we examine

whether village level measures of political power underlie the statistical sig-

nificance of the village fixed effects. However, none of our measures of polit-

ical power — whether the Pradhan position is reserved for SC/ST, whether it

is the Pradhan’s village and/or GP headquarters — affects village-level alloca-

tion of household public goods. Household public goods have low spill-overs

and are targeted towards SC/ST. Hence we expect non-SC/ST and SC/ST

Pradhans’ to differ in their propensity to allocate resources towards such

public goods. Given this, it is unsurprising that the overall incidence of

targeted public goods is unrelated to Pradhan’s residence. However, it is

surprising that this is also the case when the Pradhan position is reserved

for SC/ST. It appears that political reservation is relevant for within-village

allocation of low spill-over goods but not for overall village allocation.

Columns (7) and (8) consider the village incidence of high spill-over

public goods, as measured by the GP activism index. We find that this

13As the public good audit was conducted in every village while household surveys were

conducted in only half the villages we have twice as many observations in the latter case

12

index is, on average 0.04 points, higher in the Pradhan’s village. In term’s of

our theory, this underlines our assumption that µ exceeds one half — so that

the Pradhan enjoys agenda setting power in resource allocation. Moreover,

the fact that these public goods are high spill-over is consistent with the

finding that the reservation status of the GP does not affect the extent of

village-level provision.14

5 Concluding Remarks

This paper takes a preliminary look at resource allocation by elected village

governments using data from three Indian states. We motivated the em-

pirical analysis with a simple model of resource allocation based on three

aspects — the effect of Pradhan’s group identity on policy, the agenda setting

powers of the Pradhan and the extent of spill-overs associated with different

types of public goods. The evidence speaks to the relevance of these ideas.

The results add to a growing body of evidence which looks at decision mak-

ing at the local level and its impact on the well-being of the poor. However,

much remains to be done to gain a complete picture of democracy works in

low income contexts.

References

Alesina, Alberto, Reza Baqir and William Easterly, “Public Goods and Eth-

nic Divisions,” Quarterly Journal of Economics, 114(4), pp. 1243-1284.

Bardhan, Pranab and Dilip Mookherjee, [2000], “Capture and Governance

at Local and National Levels,” American Economic Review, 90(2), pp. 135-

14While the effect of being in the Pradhan’s village is interesting, we cannot rule out the

possibility that the Pradhan’s village is politically powerful for reasons other than having

the Pradhan.

13

39

Baron, David P., [1993], “Government Formation and Endogenous Parties,”

American Political Science Review ; 87(1), pp. 34-47.

Besley, Timothy and Stephen Coate, [2003], “Centralized versus Decentral-

ized Provision of Local Public Goods: A Political Economy Analysis,” Jour-

nal of Public Economics; 87(12), pp. 2611-2637.

Besley, Timothy, Rohini Pande, Lupin Rahman and Vijayendra Rao, [2003]

“Decentralization in India: A Survey of South Indian Panchayats”, Mimeo,

World Bank.

Chattopadhyay, Raghabendra and Esther Duflo, [2001], “Women as Pol-

icy Makers: Evidence from a India-Wide Randomized Policy Experiment,”

Mimeo, Department of Economics, MIT.

Chaudhuri, Shubham, [2003], “What difference does a constitutional amend-

ment make? The 1994 Panchayati Raj Act and the attempt to revitalize

rural local government in India,” typescript, Columbia University.

Gupta, Dipankar (2000), Interrogating Caste: Understanding hierarchy &

difference in Indian Society, Penguin Books, New Delhi

Mathew, George and Nirmala Buch [2003], Status of Panchayati Raj in the

States and Union Territories of India, Concept Publishing Company, New

Delhi

Miguel, Ted and Mary Kay Gugerty [2002], “Ethnic Diversity, Social Sanc-

tions and School Funding in Kenya”, Mimeo, Department of Economics,

University of California, Berkeley.

Oates, Wallace, [1972], Fiscal Federalism, Harcourt Brace: New York.

Pande, Rohini, [2003], “Minority Representation and Policy Choices: The

Significance of Legislator Identity,” American Economic Review ; 93(4), pp.

1132-1151.

14

Romer, Thomas and Howard Rosenthal, [1978], “Political Resource Allo-

cation, Resource Allocation and the Status Quo,” Public Choice, 33, pp.

27-43.

Weingast, Barry, Kenneth Shepsle, and C. Johnsen, [1981], “The Political

Economy of Benefits and Costs: A Neo-classical Approach to Distributive

Politics,” Journal of Political Economy, 89, pp. 642-64.

15

Household Level Data Mean S.d.

Targeted Schemes 0.072 [0.258]

SC/ST Household 0.242 [0.428]

SC/ST Household*Pradhan reserved for SC/ST 0.066 [0.248]

SC/ST Household*Pradhan Village 0.098 [0.297]

SC/ST Household*GP headquarters 0.074 [0.261]

Muslim 0.044 [0.205]

Christian 0.009 [0.096]

Elected Officials' Household 0.049 [0.216]

SC/ST*Elected Officials' Household 0.010 [0.100]

Proportion Landless 0.312 [0.463]

Age of Household Head 48.001 [14.623]

Whether Household Head Literate 0.636 [0.481]

Household Size 5.336 [2.386]

Proportion Household Farmers 0.673 [0.469]

Village Level Data

Non-Targeted Schemes 0.443 [0.315]

Proportion SC/ST Households 0.298 [0.255]

Pradhan Village 0.421 [0.494]

Pradhan reserved for SC/ST 0.210 [0.408]

Pradhan Village*Pradhan reserved for SC/ST 0.094 [0.292]

GP headquarters 0.367 [0.482]

Log Total Population 7.266 [0.971]

Log Village Area 6.375 [0.978]

Proportion Area Irrigated 0.137 [0.150]

Proportion Landless 0.304 [0.248]

Literacy Rate 0.342 [0.133]

Distance From Nearest Town 19.435 [15.612]

Male Agricultural Wage Rate 48.023 [11.950]

TABLE 1: Summary Statistics

(1) (2) (3) (4) (5) (6) (7) (8)SC/ST Household 0.066*** 0.048*** 0.041 0.034

(0.014) (0.016) (0.025) (0.025)SC/ST Household*Pradhan reserved for SC/ST 0.071** 0.071** 0.064**

(0.031) (0.031) (0.032)SC/ST Household*Pradhan village 0.03 0.032

(0.025) (0.025)SC/ST Household*GP headquarters -0.019 -0.019

(0.025) (0.025)Proportion SC/ST Households -0.007 -0.017 0.041 0.077*

(0.027) (0.027) (0.042) (0.045)Pradhan Village -0.02 -0.026 0.048** 0.044*

(0.020) (0.021) (0.023) (0.024)Pradhan reserved for SC/ST -0.003 -0.002 -0.003 -0.024

(0.012) (0.013) (0.039) (0.039)Pradhan Village*Pradhan reserved for SC/ST -0.003 -0.008 0.003 -0.002

(0.028) (0.030) (0.051) (0.052)GP headquarter -0.003 -0.007 0.041* 0.02

(0.012) (0.014) (0.023) (0.025)Controls no no no yes no yes no yesFixed effects village village village village block block block blockObservations 4059 4059 4059 4059 193 174 395 366R-squared 0.1 0.11 0.11 0.11 0.43 0.46 0.67 0.68

TABLE 2: Effect of SC/ST Reservation on resource allocation

Notes: The dependent variable in columns (1)-(4) is a dummy variable which equals one if the household's house or toilet was built under a government scheme or if it received a private water or electricity connection via a government scheme since the last GP election. The dependent variable in columns (5)- (6) is the village fixed effect from column (4) regression (excluding the constant). The dependent variable in columns (7) - (8) is an index of whether GP undertook any construction or improvement activity on roads, drains, streetlights and water sources after the last GP election. The SC/ST Household dummy equals 1 for SC/ST households. The Pradhan village dummy equals one if the Pradhan resides in the given village. The GP headquarter dummy equals 1 if the GP headquarter is located in the village. Individual controls included are dummies for if household is Muslim and Christian, household size, age, literacy and occupation of household head and whether it is the household of an elected panchayat official (alone and interacted with dummy for being a SC/ST household. Village controls included are proportion of landless households, log total village population, log village area, proportion of irrigated land, village literacy rate, distance from nearest town, and daily male agricultural wage rate.All village controls except for the agricultural wages are from 1991 Census of India. Agricultural wages are from survey data. Variation in sample Robust standard errors in brackets. * significant at 10%; ** significant at 5%; *** significant at 1%

Village fixed effectVillage public goodsHousehold public goods

Household regression Village level

ANNEX C:

Participatory Democracy in Action: Survey Evidence from India

Participatory Democracy in Action:

Survey Evidence from South India∗

Tim Besley (LSE) Rohini Pande(Yale)

and Vijayendra Rao (World Bank)†

Abstract

We use household and village survey data from South India to exam-

ine who participates in village meetings called by elected local govern-

ments, and what effect these meetings have on beneficiary selection

for welfare programs. Our main finding is that members of socially

and economically disadvantaged groups, specifically landless and low

caste individuals, are both more likely to attend these meetings and

be chosen as beneficiaries in villages which have village meetings.

JEL Classification: H40, H42, O20∗Acknowledgements We thank Lupin Rahman, Radu Ban, Siddharth Sharma and Jil-

lian Waid for research assistance, and the IMRB staff for conducting the survey. We are

grateful to the World Bank’s Research Committee and the South Asia Rural Development

Unit for financial support. The opinions in the paper are those of the authors and do not

necessarily reflect the points of view of the World Bank or its member countries.†Email addresses: Besley <t.besley@lse.ac.uk>; Pande<rohini.pande@yale.edu>;

Rao<vrao@worldbank.org>

1

1 Introduction

How to structure democratic institutions to ensure a fair and efficient allo-

cation of public funds is a central issue in the political economy of devel-

opment. The new governance agenda has emphasized citizen empowerment

as a tool for improving the workings of democratic institutions.1 But such

terms can easily be dismissed as empty rhetoric unless embodied in workable

institutional solutions.

The idea that encouraging citizen participation can improve the work-

ings of a democracy is also echoed in the political science literature. One

role for participation emphasized in that literature is to improve the flow

of information into the political process beyond that available by electing

representatives. Thus, Verba et (1995) characterize political participation

as “information rich” acts and observe that:

”From the electoral outcome alone, the winning candidate cannot

discriminate which of dozens of factors, from the position taken

on a particular issue to the inept campaign run by the opposition

..., was responsible for the electoral victory.” (page 10).

This paper studies an institution aimed at encouraging political partici-

pation among the poor and improving the quality of governance in an Indian

context – Gram Sabha meetings. These are village meetings called by the

elected local government (Gram Panchayat) to discuss resource allocation

decisions in the village.2 There are two main ways in which such meetings1ee, for example, World Bank (2000).2he 73rd Constitutional Amendment Act of India in 1993 made it mandatory for Indian

states to hold elections for Gram Panchayats and to give them policy-making powers.

2

may improve the workings of government. First, relative to elected repre-

sentatives, these meetings may better reflect citizens’ preferences on issues

such as how to target resources to the neediest groups. Second, by provid-

ing a forum for monitoring the actions of elected representatives they may

reduce agency problems in politics, and the extent of corruption.

While holding Gram Sabhas is compulsory, their frequency and content

owes a lot to the discretion of elected officials.3 Moreover, even a well-

attended meeting may have no bite on policy decisions. Here, we exploit a

large household and village survey of local governments in the four South

Indian states to examine of participation in Gram Sabhas, and whether

having a Gram Sabha affects beneficiary selection for welfare programs.

While there is much interest in how participation improves the quality

of governance in the developing world (see, for example, Manor (2004)), evi-

dence on the determinants of participation at the household level is thin, es-

pecially compared to the extensive studies available for the advanced democ-

racies. Moreover, the literature is replete with concerns about elite dom-

inance of democratic institutions.4 This raises the specter of participatory

institutions being a veil which have little impact on the well-being of the

poor. Here, however, we find that it is the most disadvantaged groups who

attend village meetings and that holding such meetings improves the tar-

geting of resources towards the neediest groups.

Our findings contribute to a broader debate about the role of decentral-

ized governance in improving the quality of government in the developing

world. The merits of decentralization have been widely debated – see, for3State or District admininstration officials can also affect this by choosing not to attend,

and therefore making the Gram Sabha less attractive to hold.4see, for example, Bardhan and Mookherjee (2000) and Platteau and Abraham (2002).

3

example, Bardhan (2002) and Triesman (2002). However, it is clear that

many institutional details, even within decentralized governance, can be im-

portant. The use of village meetings of the kind studied here is one. It is

important to understand how these institutional differences affect the way

in which government operates.

The paper is organized as follows. In the next section, we describe the

context for our study and our data. Section three contains the analysis, and

Section four concludes.

2 Context

Our focus is on the lowest level of self government in India, the Gram Pan-

chayat (GP). Each GP covers between 1-5 villages. The Gram Sabha is a

village-level body consisting of persons registered in the electoral rolls of a

GP. It was intended to be a supervisory body that audits and regulates the

functioning of the GP. Specifically, it is supposed to ratify the GP budget,

and identify and approve of beneficiaries for welfare schemes implemented

by the GP. To achieve these tasks, most Indian states require that the Gram

Sabha meet (roughly) four times a year.

Between September -November 2002 we conducted a village and house-

hold survey of 522 villages and over 5000 households in the four South Indian

States of Andhra Pradesh, Karnataka, Kerala and Tamil Nadu . For admin-

istrative purposes Indian states are divided into districts, and then blocks.

For each state pair we selected two districts which shared a common state

boundary. The district pair belonged to the same political entity during the

200 years of British colonial rule, prior to 1956 when all Indian states were

reorganized along linguistic lines. This allows us to estimate state differ-

4

ences while controlling for common colonial history. For each district pair

we selected the 3 most ’linguistically similar’ block pairs (that is, 3 blocks

in each of the two districts). We defined linguistic similarity in terms of

the mother tongue of individuals living in the block, and computed it using

1991 census block level language data. 5 In total, we had 18 block pairs.

In each block we randomly sampled 3 GPs, and per GP up to 3 villages. In

Kerala, we sampled wards rather than villages as ward size approximates

village size in other states.

In every village, we conducted group meetings in which we obtained

information on the last Gram Sabha meeting, and also village-level demo-

graphic and economic variables. In a random sub-sample of 259 villages we

conducted twenty household surveys, and obtained information on Gram

Sabha attendance and household beneficiary status.

Table 1 reports descriptive statistics. The average village has 328 house-

holds, of which 34 percent are landless. Twenty percent belong to the tradi-

tionally well of upper castes and 28 percent to the historically disadvantaged

scheduled castes and tribes (hereafter SC/ST). According to the 1991 census

literacy rate in our sample villages averaged 41 percent, but as is well known

was much higher in Kerala villages. Seventy five percent of the villages had

at least one Gram Sabha meeting in the last year, and in 22 percent of these

meetings beneficiary selection was discussed.

In our household data-set we observe that while over 50 percent of the

respondents had heard of a Gram Sabha only 20 percent had ever attended

a Gram Sabha meeting. We also collected information on a household’s ben-5The historical and administrative similarity of linguistically matched blocks was

checked using princely state maps and the Report of the States Reorganization Com-

mittee (for details on sampling procedure, see Besley, Pande, Rahman and Rao, 2004b).

5

eficiary status, as defined by whether it has a ‘Below Poverty Line’ (BPL)

card. The GP, in collaboration with state government officials, is supposed

to identify (via a census) households with income below the poverty line,

and to give these households a BPL card. Possession of this card makes

the household eligible for an array of government schemes, ranging from

subsidized food through the public distribution system to free hospitaliza-

tion. The list of BPL households, and subsequent selection of beneficiary

households under various schemes is supposed to be ratified in Gram Sabha

meetings.

3 Analysis

The analysis is in two parts. We first study the determinants of holding

a Gram Sabha meeting and who attends. We then look for evidence that

holding a Gram Sabha meeting affects public resources allocation.

3.1 Determinants of holding a Gram Sabha and who attends

To study which villages have Gram Sabha meetings we estimate a linear

probability regression of the following form:

Svbs = αb + γs + δxvbs + εvbs

here Svbs is an indicator variable denoting whether village v in block

pair b and state s had a Gram Sabha in the past twelve months, alphab are

dummies for matched block pairs (18 in total) and γs are state fixed effects.

The variables xvbs are village level characteristics (number of households,

literacy rate in 1991, fraction landless, fraction SC/ST, fraction upper caste

6

and whether the position of Pradhan is reserved for a women or SC/ST).

We cluster the standard error at the GP level.

The results are in Table 2, column (1). More populous villages are more

likely to have had a Gram Sabha meeting, and villages with a higher literacy

rate are weakly more likely to hold Gram Sabha meetings. Interestingly,

after conditioning on matched block pair effects, we don’t observe significant

state differences in the decision to have a Gram Sabha.

In Columns (2)-(5) we use our household data to examine who has heard

of, and who attends Gram Sabha meetings. Columns (2) and (3) estimate

regressions of the form:

givbs = αb + γs + δxvbs + λcivbs + εivbs

here givbs indicates whether individual i (in village v in block pair b in state

s) has heard of the Gram Sabha in column (2), and whether he/she has ever

attended a Gram Sabha meeting in column (3). The variables civbs denote

a vector of respondent characteristics (whether respondent is an SC/ST,

female, illterate, landless, upper caste, to comes from a wealthy household

as measured by durables ownership).6

Village literacy rate is positively correlated with both hearing of the

Gram Sabha and attending it. We find evidence of significant state effects,

with respondents from Kerala more likely to have both heard of Gram Sabha

meetings and participated in them. However, in the case of individual char-

acteristics we observe significant differences in who has heard of and who

attends Gram Sabha meetings. Moreover, various measures of economic and

social disadvantage have a differential impact on the propensity to attend6he equation is estimated allowing for clustering of the error terms varepsiloniv at the

village level.

7

Gram Sabhas. Women and illiterates are less likely to both hear of and

attend these meetings. In contrast, SC/STs and the landless are more likely

to attend Gram Sabha meetings but no more likely to have heard of Gram

Sabhas. In contrast, the wealthy and upper castes are more likely to have

heard of Gram Sabhas but not to attend.

In column (4) we show that the effect of individual characteristics on

participation is robust to the inclusion of village fixed effects. Again, land-

less and SC/ST respondents report themselves more likely to attend a Gram

Sabha. Finally in column (5) we examine whether village literacy, in addi-

tion to affecting overall participation in a Gram Sabha meeting, also affects

the propensity of the disadvantaged to attend. We estimate the participa-

tion regression with village fixed effects and include the interactions between

village literacy rates and measures of individual economic and social disad-

vantage. Illiterate, landless and SC/ST individuals, but not women, are

more likely to participate in higher literacy villages.

These findings are notable for two reasons. First, there is some sug-

gestion of a political externality from living in a more literate community.

Second, Gram Sabha meetings seem to a be a forum used by some of the

most disadvantaged groups in the village – landless, illiterates and scheduled

castes/tribes. This suggests that these groups find the Gram Sabha useful

and that Gram Sabha meetings may play some role in moving policy in a

direction favored by these groups. We now look for evidence of the latter.

3.2 Does participation matter?

There are many who argue that participation in the political process has an

intrinsic benefit. It builds trust in government and legitimizes state action.

8

Unfortunately, our data do not permit us to look at these issues. However,

we are able to look at the possibility that participation in Gram Sabhas

yields instrumental (i.e. policy) benefits. These could be community wide

or by targeting resources to more specific groups. Here, we will focus on

the latter, examining whether targeting of public programs are related to

whether a Gram Sabha meeting has been held in the past twelve months.

We focus on an important specific policy administered at the village level

– access to a below poverty line (BPL) card. Beneficiary selection for such

cards is influenced by the GP. As discussed earlier, possession of this card

gives a villager access to an array of public benefits. We estimate a household

regression which exploits within village variation in individual characteristics

to examine whether the targeting of BPL cards differs depending on whether

the village had a Gram Sabha in the last year. Our key equation is:

biv = βv + ξciv + θ (civ ∗ Sv) + εiv

here βv is a village level fixed effect and εiv is adjusted for clustering at

the village level. The coefficients on household characteristics civ represent

the way in which access to BPL cards is targeted at the household level.

Our main interest is in the coefficients on θ which interacts household char-

acteristics with whether a Gram Sabha meeting was held in the past twelve

months – the indicator variable Sv. If θ is significantly different from zero,

then this suggests that some household types are favored in villages that

hold Gram Sabha meetings.

The results are reported in Table 3. In column (1) we report the base-

line regression which does not include any interaction terms, θ. This shows,

not surprisingly, that BPL cards are targeted towards landless, illiterate and

SC/ST households. In column (2) we include interactions between measures

9

of disadvantage and whether the village had a Gram Sabha meeting. We

find targeting of landless and illiterate individuals is more intensive in vil-

lages that have held a Gram Sabha meeting. Moreover, these effects are

economically significant with an 8-10% increase in the probability of receiv-

ing a BPL card in a village that held a Gram Sabha. We find similar, but

statistically insignificant, evidence for SC/STs.

These results do show persuasively that there is heterogeneity in target-

ing BPL cards across villages. Moreover, it would be tempting to attribute

this to whether a Gram Sabha meeting is held. However, some caution

is warranted. In column (3), we interact the characteristics that repre-

sent disadvantage – illiteracy, landlessness and schedule caste/tribe – with

the village literacy rate instead of whether the village had a Gram Sabha

meeting. All three of these interactions are significant. However, the

point estimate of the effect evaluated at the mean literacy rate is substan-

tially smaller than the effects in columns (2)-(4). But this does raise the

possibility that holding a Gram Sabha meeting is correlated with other vil-

lage characteristics that are important in shaping the way in which public

resources are targeted. Unfortunately, this is not an issue that we can re-

solve. However, these encouraging results on Gram Sabhas clearly deserve

further careful investigation.

4 Concluding Comments

While this paper focusses on a specific institution – the Gram Sabha –

the results contribute to a wider debate on how institution design can shape

public resource allocation and how the poor can increase their voice in public

institutions. It is frequently remarked that poverty is much more than

10

material deprivation and that the poor may receive much less voice in the

political process. Moreover, a good deal of cynicism attends initiatives to

strengthen that voice.

In this regard, our results sound a more optimistic note. The illiterate,

landless and SC/STs are significantly more likely to attend Gram Sabha

meetings than other groups. Moreover, there appears to be more targeting

towards these groups where Gram Sabha meetings are held. The results are

also suggestive of some externalities from literacy in the political process at

the village level.

Less optimistically, it is clear that Gram Sabhas are not a forum for

women in their current form. Women respondents are around 20% less

likely to attend a Gram Sabha than men. Whether this has significant

consequences for public resource allocation needs further investigation. But

it is clear the representativeness of Gram Sabhas is likely to be affected by

this. Other tools such as gender reservation in Panchayat representation

may go some way towards remedying this.7

Going forward, it is important to refocus debates on decentralization

more clearly on the institutional form that this takes. To this end, the kind

of study undertaken here should be useful in assessing the way in political

institutions are used. There are grounds for viewing participation may be

important in its own right. However, it may also have instrumental benefits

to groups who participate. Either way, it is clear that household surveys

have much potential in studying these issues.

7ee Chattopadhyay and Duflo (2004) and Besley et. al. (2004c)

11

References

Bardhan, Pranab, and Dilip Mookherjee (2000). ”Capture and Gover-

nance at Local and National Levels.” American Economic Review,

90(2), 135-139.

Bardhan, Pranab (2002). “Decentralization of Government and Devel-

opment.” Journal of Economic Perspectives, 16(4), 185-205.

Besley, Timothy, Rohini Pande, Lupin Rahman, and Vijayendra Rao

(2004a).“The Politics of Public Good Provision: Evidence from In-

dian Local Governments.” Journal of the European Economics As-

sociation, 2(2-3), 416-426.

Besley, Timothy, Rohini Pande, Lupin Rahman, and Vijayendra Rao,

[2004b]. “Decentralization in India: A Survey of South Indian Pan-

chayats.” mimeo, LSE.

Besley, Timothy, Rohini Pande, Vijayendra Rao, and Radu Ban,

(2004c). “Tokenism or Agency? The Impact of Women’s Reserva-

tion on Panchayats in South India.” mimeo Development Research

Group, The World Bank.

Chattopadhyay, Raghabendra, and Esther Duflo, (2004). “Women as

Policy Makers: Evidence from a India-Wide Randomized Policy

Experiment.” Econometrica, 72(5), 1409-1444.

Manor, James, (2004). “Democratization with Inclusion: political re-

forms and people’s empowerment at the grassroots.” Journal of Hu-

man Development, 5(1), 5-29.

Platteau, Jean-Philippe, and Anita Abraham, (2002). “Participatory

Development in the Presence of Endogenous Community Imperfec-

12

tions.” Journal of Development Studies, 39(2), 104-136.

Triesman, Daniel, (2002). “Decentralization and the Quality of Govern-

ment.” mimeo, UCLA.

Verba, Sidney, Kay Lehman Scholzman, and Henry E. Brady,

(1995).Voice and Equality: Civic Voluntarism in American Politics.

Cambridge Mass: Harvard University Press.

World Bank, (2000). World Development Report 2000/2001: Attacking

Poverty. Washington, DC, The World Bank.

13

Table 1:Descriptive Statistics

Overall Andhra Pradesh Karnataka Kerala Tamil NaduVillage level dataTotal households 328.10 305.50 365.80 401.10 227.40

Fraction of households which are 0.34 0.25 0.23 0.48 0.41landlessFraction of households which are 0.28 0.23 0.41 0.21 0.22SC/STFraction of households which are 0.20 0.13 0.32 0.12 0.19Upper casteLiteracy Rate in 1991 0.41 0.24 0.37 0.63 0.35

Fraction of villages which had a 0.76 0.71 0.68 0.98 0.67Gram Sabha in last yearFraction of Gram Sabhas at which 0.22 0.21 0.33 0.30 0.02beneficiary selection was discussed

Household level dataHeard of Gram Sabha 0.53 0.29 0.42 0.93 0.37

Ever attended Gram Sabha 0.20 0.11 0.14 0.40 0.13

Possess a BPL Card 0.22 0.32 0.10 0.30 0.25All variables based on survey data, except the village literacy rate which is from the 1991 Census of India

Table 2: Gram Sabha: Occurrence and AttendanceVillage had Household data: Gram Sabha

Gram sabha Heard of Attended (1) (2) (3) (4) (5)

Literacy Rate in 1991 0.328 0.323*** 0.235***(0.246) (0.118) (0.073)

Total number of households 0.093*** -0.001 0.006(0.030) (0.014) (0.010)

Fraction landless households 0.044 -0.017 -0.067**(0.086) (0.047) (0.032)

Fraction upper caste households -0.079 0.056 -0.011(0.116) (0.047) (0.032)

Fraction SC/ST households 0.03 0.021 -0.019(0.104) (0.041) (0.029)

Pradhan position reserved 0.01 0.043** -0.003(0.042) (0.020) (0.015)

Village Had Gram Sabha 0.026 0.030**(0.023) (0.014)

Illiterate -0.129*** -0.027** -0.030** -0.103***(0.015) (0.012) (0.013) (0.028)

Illiterate*literacy rate in 1991 0.183**(0.078)

SCST 0.001 0.021 0.034** -0.029(0.019) (0.016) (0.017) (0.040)

SCST*literacy rate in 1991 0.139(0.097)

Landless -0.012 0.041*** 0.030** -0.073**(0.014) (0.012) (0.012) (0.029)

Landless*literacy rate in 1991 0.232***(0.066)

Female -0.214*** -0.182*** -0.187*** -0.086***(0.014) (0.012) (0.014) (0.030)

Female*literacy rate in 1991 -0.242***(0.076)

Upper caste 0.035** 0.013 -0.004 -0.007(0.018) (0.016) (0.017) (0.018)

Wealthy 0.057*** -0.049*** -0.035** -0.027*(0.016) (0.014) (0.015) (0.016)

Andhra Pradesh -0.018 -0.171*** -0.168***(0.091) (0.048) (0.035)

Karnataka -0.089 -0.153*** -0.156***(0.063) (0.033) (0.032)

Tamil Nadu 0.019 -0.161*** -0.188***(0.061) (0.037) (0.029)

Fixed effects Block pair Block pair Block pair Village VillageObservations 476 4445 4935 5455 5240R-squared 0.22 0.39 0.17 0.25 0.25

Standard errors in brackets clustered at GP level in column (1) and at village level in all other regressions. Wealthy is a dummy for consumer durable ownership. Columns (2)-(4) also include respondent age and age squared as controls.* denotes significant at 10%; ** significant at 5%; *** significant at 1%

Table 3: Gram Sabha Occurrence and Beneficiary SelectionReceived BPL card

(1) (2) (3)Illiterate 0.028* -0.042* -0.057*

(0.015) (0.026) (0.030)Illiterate*Gram Sabha held 0.091***in last year (0.030)Illiterate* literacy rate in 1991 0.206***

(0.072)SCST 0.150*** 0.094** -0.03

(0.020) (0.042) (0.044)SCST*Gram Sabha held 0.062in last year (0.047)SCST* literacy rate in 1991 0.430***

(0.097)Landless 0.075*** 0.018 -0.098***

(0.016) (0.030) (0.035)Landless* Gram Sabha held 0.067*in last year (0.035)Landless*literacy rate in 1991 0.386***

(0.081)Female -0.011 -0.009 -0.005

(0.010) (0.010) (0.010)Upper caste -0.028* -0.028* -0.036**

(0.017) (0.016) (0.017)Wealthy -0.082*** -0.079*** -0.066***

(0.014) (0.014) (0.014)Fixed effects Village Village Village

Number of observations 5455 5364 5039R-squared 0.4 0.4 0.42Robust standard errors, clustered by village, in brackets. All regressions include respondent age and age squared as controls. * significant at 10%; ** significant at 5%; *** significant at 1%

ANNEX D:

Political Selection and the Quality of Government: Evidence from South India

Political Selection and the Quality of

Government: Evidence from South India∗

Timothy Besley (LSE) Rohini Pande (Yale)

and Vijayendra Rao (World Bank)†

Abstract

This paper uses household data from India to examine the economic and social

status of village politicians, and how individual and village characteristics affect

politician behavior while in office. Education increases the chances of selection

to public office and reduces the odds that a politician uses political power

opportunistically. In contrast, land ownership and political connections enable

selection but do not affect politician opportunism. At the village level, changes

in the identity of the politically dominant group alters the group allocation of

resources but not politician opportunism. Improved information flows in the

village, however, reduce opportunism and improve resource allocation.

∗We thank numerous seminar participants, and Joseph Altonji, Penny Goldberg, Asim Khwaja,

Dominic Leggett, Barry Weingast and, especially, Chris Udry for comments. We also thank Lupin

Rahman, Radu Ban, Sarah Goff, Siddharth Sharma and Jillian Waid for research assistance, and

IMRB staff for conducting the survey. We thank World Bank’s Research Committee and the

South Asia Rural Development Unit for financial support. The opinions in the paper are those

of the authors and do not necessarily reflect the points of view of the World Bank or its member

countries.†Email addresses: Besley <t.besley@lse.ac.uk>; Pande<rohini.pande@yale.edu>;

Rao<vrao@worldbank.org>

1

“The nature of the workings of government depends ultimately on the

men who run it. The men we elect to office and the circumstances we

create that affect their work determine the nature of popular government.

Let there be emphasis on those we elect to office.” V.O. Key (1956).

“A Hindu’s public is his caste.” B.R. Ambedkar (1937).

1 Introduction

Common sense discussions of political life often place the quality of politicians at

center stage. For example, Thomas Jefferson believed that a key role of elections

was to create a “natural aristocracy” of the talented and virtuous (Jefferson (1813)).

Yet the modern political economy literature remains dominated by a paradigm in

which good policy is achieved solely by getting incentives right rather than by im-

proving the quality of the political class. While incentives are important, personal

qualities of politicians such as honesty, integrity and competence are potentially im-

portant, especially in environments where politicians face limited formal sanctions.

Equally, in environments where ethnicity is central to the economic organization of

the society, a politician’s group identity is likely to matter.

This paper uses household data from Indian villages to examine how individuals’

economic and group characteristics affect political selection, and politician behavior

in office. Further, we study how village characteristics which alter the political

dominance of different population groups, and the extent of information flows in a

village, affects these relationships.

Our analysis makes use of a remarkable political experiment in India. The 73rd

amendment of the Indian constitution in 1993 created a new tier of local govern-

ment which, by the year 2000, had led to the constitution of 227,698 new village

governments, Gram Panchayats (GP), staffed by over two million elected represen-

tatives. In an effort to infuse fresh blood into the political class, the amendment

2

mandated that close to half of these elected positions be reserved for traditionally

disadvantaged population groups (lower caste groups and women). These village

governments enjoy wide-ranging responsibility for beneficiary selection for govern-

ment welfare programs (Matthew and Buch 2000).

One of the most important GP responsibilities, and one we use to identify politi-

cian quality, is the targeting of ‘Below Poverty Line’ cards (BPL). Ownership of

a BPL card provides a household with access to subsidized food via the Indian

public distribution system. It is also typically an eligibility requirement for other

government welfare schemes, e.g. housing schemes. The Indian Planning Com-

mission estimates that there were 45 million BPL households in 2000-01, and that

the effective annual income gain of owning a BPL card was Rs. 415 per household.

Further, it estimates that the public distribution system only reaches fifty seven per-

cent of BPL households and over twenty percent of BPL card holders are not poor,

suggesting substantial mis-targeting by, among others, village politicians (Planning

Commission, 2005).1

We develop a simple model of political selection to understand how the political

selection process in a village can affect the allocation of BPL cards. Politicians differ

along two dimensions – the group interest they represent and their quality as policy

makers. Higher quality politicians better target BPL cards. Voters favor higher

quality politicians, but also have group preferences. Bad politicians are relatively

more likely to enter when formal returns to politics are low and/or returns to polit-

ical opportunism are high. They are more likely to be selected if information about

politician quality is limited, and voters vote along group lines. At the village level,

political reservation of the village chief’s position changes the identity of the polit-1The estimated income gain is based on an All India household survey, and worked out as follows:

the differential between the average market and PDS price of the grains was multiplied with the

average quantity given to a cardholder (done separately for rice and wheat and then added up).

Their findings on targeting were based on a comparison of the number of households with BPL

cards with independent estimates of the number of poor.

3

ically dominant group, and thereby the group targeting of BPL cards. If prior to

political reservation no group of villagers were politically dominant, then reservation

will also reduce coordination costs and thereby the likelihood of bad politicians. We

also examine the role of aggregate information flows in the village, and find that they

reduce the likelihood of bad politicians and improve the targeting of BPL cards.

We test the empirical relevance of these ideas using survey data from the four

South Indian states. The survey, which was designed by the authors and conducted

in 2002, surveys both politician and non-politician households.

The empirical analysis has two components. First, we estimate a “selection

equation” for politicians and investigate how selection is affected by individual and

village characteristics. Political selection in our sample is based on economic advan-

tage and political connections – politicians are more likely to be educated, own land

and have family political connections. Village characteristics that prevent the polit-

ical dominance of the traditional village elite, in particular via political reservation

for women and low castes, reduce the extent of such selection. In addition, villages

with higher literacy rates select more educated politicians.

Second, we examine politician quality as measured by BPL card status. On

average, politicians are opportunistic – relative to a non-politician household, a

politician household is more likely to have a BPL card. Individual and village

characteristics affect the extent to which this is true. Better educated politicians

exhibit less political opportunism. This is not true for land ownership or political

connections. Turning to village characteristics, political reservation of the village

chief changes the identity of the politically dominant group and the group allocation

of BPL cards. However, it does not reduce political opportunism. Finally, politicians

in villages with a relatively higher literacy rate, or which hold village meetings,

exhibit lower political opportunism.

The remainder of the paper is organized as follows. In the next section, we

discuss related work. Section three develops a simple model to identify why political

4

selection may fail to produce good politicians. Section four introduces the data and

develops the empirical tests. Results are in section five, and section six concludes.

2 Related Literature

The Downsian model of politics, which has dominated political economy for over a

generation, has no role for political selection. The role of politics is to seek out

the policy position of the median voter, and not to examine who implements that

policy. Until recently, political selection was also absent from political agency models

– the classic analyses being due to Barro (1973) and Ferejohn (1986). They focus

exclusively on the problem of moral hazard in politics and the role of elections in

restraining politicians.2 The problem of incentives embodied in constitution design

is also the main theme in the Public Choice literature pioneered by Buchanan.3

More recent work has emphasized the importance of politician characteristics in

explaining political behavior. This puts greater weight on the political selection

mechanism. The citizen-candidate approach of Besley and Coate (1997) and Os-

borne and Slivinski (1996) characterizes political competition as a three-stage game

of entry, voting and policy making. The model explains endogenously who enters,

and who succeeds, in politics. This approach can be used either to study selection2Recent political agency models study the implications of good and bad politicians for policy

outcomes where these types are unobserved. For example, Coate and Morris (1995) draw out

implications for the quality of public decisions and Maskin and Tirole (2004) contrast appointing

versus electing judges in this framework. Besley (2004) uses this framework to study equilibrium

quality of the pool of politicians as a function of the rewards to politicians.3The following quote from Buchanan captures this idea clearly:

“To improve politics, it is necessary to improve or reform rules, the framework

within which the game of politics is played. There is no suggestion that improvement

lies in the selection of morally superior agents who will use their powers in some ‘public

interest’ ” (Buchanan (1989, page 18)).

5

on policy preferences (or “identity ”) or selection on valence characteristics such as

talent or virtue.

The citizen-candidate approach has been applied to study the effect of political

reservation by Pande (2003) and Chattopadhyay and Duflo (2004). Both argue

that reservation matters by changing the identities of those elected to office. Lee,

Moretti and Butler (2004) argue that this framework explains the U.S. data. The

focus in all these cases is on how politics changes spatial policy preferences.

The quality dimension in political selection has been studied in this framework

by Caselli and Morelli (2002), Poutvarra and Takalo (2003) and Besley and Coate

(1997). Caselli and Morelli (2002) argue that the key issue is to understand factors

which affect the supply of bad politicians, such as the rents that they can earn while

in office. Imperfect information may also affect the incidence of bad politicians by

making it difficult to spot candidate quality. Poutvarra and Takalo (2003) develop

a model in which the value of holding office impinges on candidate quality via its

effect on election campaigns. Besley and Coate (1997) consider the implications of

coordination problems among voters. Gehlbach and Sonin (2004) apply a citizen

candidate framework to ask when economic elites (such as businessmen) will run

for political office. Running for office is in this world an alternative to lobbying

for influence. They argue that business candidates lead to greater misuse of public

office, and suggest that such use of office is more likely in developing countries.

Empirical work on the quality of government using cross-country data, such

as La Porta, Lopez-de-Silanes, Shleifer and Vishny (1999), is typically unable to

decompose the quality of government into problems of selection or incentives. How-

ever, recent work by Jones and Olken (2005) uses death of national leaders in office

as a source of exogenous variation to show that unexpected changes in national

leadership affect economic growth. This effect is strongest in autocratic polities,

suggesting that personal qualities of leaders matter. Moreover, the weaker effect

in democracies suggests that political selection may have some virtuous properties

6

when conducted in the more open entry processes of a democracy.

Our paper also contributes to a growing empirical literature on decentralized

government which finds that decentralization affects resource allocation in low in-

come countries. Faguet (2004) finds that decentralization improved targeting in

Bolivia. Bardhan and Mookherjee (2003) examine the role of elected village coun-

cils in affecting land reform in the Indian state of West Bengal. Chattopadhyay and

Duflo (2004) show political reservation for women affected public good allocation in

two Indian states. Finally, Foster and Rosenzweig (2001) show that decentralization

interacted with land ownership patterns across Indian villages to affect public good

outcomes. None of these papers, however, focus on how politicians’ characteristics

affect the workings of decentralized governments. But an important difference be-

tween politics at the local and national level could well be in terms of the kind of

people who hold public office.

3 The Model

We use a simple citizen-candidate model of politics to identify possible reasons why

low quality politicians can be elected to office. This will be useful in motivating the

empirical analysis below.

3.1 The Environment

Consider a village populated by N individuals, each eligible to be elected as a politi-

cian. Politicians enjoy policy authority over the allocation of public resources, here

BPL cards. For simplicity, we focus on election of a single politician.

Each citizen belongs to a group j. There are M such groups with a fraction πj of

citizens in group j. These groups can be thought of as representing policy interests

of different groups, such as gender, caste or wealth. If elected, an individual’s

group identity will be important if she cannot commit to policy outcomes before

7

the election. Conflict of interest in policy priorities between groups creates spatial

political competition to holding office. Each group member prefers a politician from

her own group.

In addition to her group identity, a politician (once elected) can be good or bad.

Relative to a bad politician, a good politician better targets BPL cards towards

the deserving. We do not need to be specific about the exact interpretation of what

makes for a good politician – honesty or competence. We assume politician quality

is a valence issue, i.e. one on which all citizens (regardless of their group identity)

have the same ranking. We denote this characteristic by τ ∈ {g, b} where g stands

for ‘good’ and b for ‘bad’.

We do not model the policy process explicitly. Hence, preferences are in reduced

form – preferences over politicians rather than policy. Let k denote a politician’s

group identity. A type {k, τ} politician gives citizen i from group j a payoff of:

λj (k)− C (τ, I, k)

Thus, preferences are separable with λj (k) a group identity component and C (τ, I, k)

a quality component. Bad politicians are costly as C (g, I, k) = 0 < C (b, I, k)∀k.

The variable I indexes the extent to which village characteristics prevent dishon-

est politicians from imposing a cost on the other citizens. “Good” characteristics

reduce C (b, I, k). We will return to this below.

Politicians are citizens, with similar preferences. The difference is that politi-

cians may enjoy a private “benefit” from holding office. Thus a type (j, τ) politician

receives utility

λj (j) + B (τ, I)

from holding office. The term B (τ, I), which is also affected by characteristics I,

is a group-independent benefit from holding public office. It would, for example,

depend on politician wages and the returns to opportunism when in office. We

concentrate on the case where B (b, I) ≥ B (g, I) , which implies that bad politicians

8

have a higher demand for public office than good ones.4

3.2 The Political Process

We model the electoral process as a two-stage citizen-candidate game. At stage one

candidates decide whether to enter, and at stage two voters cast their votes. We

consider non-cooperative entry and voting decisions, and analyze the two stages of

the political process in reverse order.

Voting The group characteristic k is observed by voters before they cast their vote.

However, we allow for imperfect information with respect to candidate quality – τ .

For simplicity, assume that τ is revealed to all voters during the election campaign

with probability q (∈ (0, 1)) (Hence, voters are always symmetrically informed).

Voting decisions form a Nash equilibrium from among the candidates who enter.

Following Besley and Coate (1997), we refine the voting equilibrium by eliminating

weakly dominated strategies. This implies that voting is sincere in two-candidate

elections, but puts relatively little structure on multi-candidate voting. We assume

that indifferent voters abstain and that in the event of a tie, the winning candidate

is picked at random from among those who have the most votes.

Entry Each citizen faces a group-specific cost of running for office δj . Let vj (0)

be the utility of a citizen of type j when nobody runs for public office. We assume

everyone prefers to avoid a situation in which nobody runs for office, i.e. vj (0) <

λj (k)∀ (j, k) = 1, ...,M.. Each citizen’s pure strategy, denoted by σi ∈ {0, 1},is whether to enter as a candidate. A collection of such decisions (one for each

citizen) must form a Nash equilibrium in pure or mixed strategies.4This inequality may be reversed in societies that have a strong ethic of public service so that

good politicians earn relatively higher rents such that B (g, I) is large.

9

3.3 Political Equilibrium

A political equilibrium is an equilibrium in the entry and voting stages of the game.

Rather than providing an exhaustive description of equilibria, we use the model to

examine various reasons why equilibria can result in bad politicians being elected.

We begin by studying an important case – when there is a politically dominant

group. This occurs if a citizen from some group can defeat a citizen from any other

group in a pairwise comparison. This includes the case where one group comprises

more than half the population, but it can happen more generally if preferences

are appropriately ordered.5 In our data, political reservation, by reserving some

seats for citizens from particular groups, creates a politically dominant group. Let

the dominant group be denoted by d, and assume at least one candidate from the

dominant group is willing to run rather than having nobody in office, i.e.:

λd (d)− vd (0) + B (τ, I) > δd for τ ∈ {g, b} .

The existence of a dominant group relaxes competition in the spatial dimension.6

This allows the selection process to focus on within-group competition between good

and bad candidates. From a social point of view, a single good candidate from the

dominant group standing for office is preferable.7 Thus, the main focus is on whether

bad candidates enter, and have any chance of being elected.

We start with the entry process. As a first pass, consider the incentive for a bad

candidate to run given that there are only good candidates in the race. Since q < 1,

voters will not detect that he is bad some of the time. Thus, he faces a positive

probability of being elected and capturing B (b, I). Whether he does so depends on

the probability of capturing B (b, I) relative to the entry cost. Specifically:5This is possible if there is a group k such that a “good” candidate drawn from group k is a

Condorcet winner among the set of all types.6However, for this to be true, it has to be the case that even a bad candidate from the dominant

group will win against a candidate from any other group.7The only reason for multiple good candidates to run is if B (g, I) is high relative to δd.

10

Proposition 1 With a politically dominant group d, if B (b, I) is high enough, there

is no pure strategy equilibrium in which only good candidates of type d enter.

The intuition is straightforward – if bad candidates earn sufficiently high rents,

then at the point that no more good candidates wish to enter, it is worthwhile for

a bad candidate to enter if there is some chance that she will be elected. Thus, to

sustain equilibria with only good candidates the rents must be sufficiently low for

bad candidates. This is true if institutions restrain consumption or rents by bad

candidates sufficiently. Further, the threshold ratio of rents for bad and good candi-

dates is increasing in the information about candidates. Thus, better information

makes it more likely that only good candidates enter.

We next ask whether an equilibrium with only bad candidates is possible. Sup-

pose that a single bad candidate is running for office. Then, if a good candidate

enters, he will win as long as he is identified as good, i.e. with probability q. Thus

for only bad candidates to run, it must be that no good candidate wishes to enter.

Here, the source of political dominance matters. For reserved jurisdictions we need

only check that a good candidate from the reserved group would not enter. How-

ever, without reservation, we also need to consider entry by candidates who are not

from the politically dominant group. We consider each case in turn.

Proposition 2 Suppose the political position is reserved for group d. Then a pure

strategy Nash equilibrium with only bad candidates of type d exists if entry costs are

sufficiently large so that:

δd >

(1 + q

2

)[B (g, I) + C (b, I, d)] .

The required condition reflects the two motives for a good candidate to hold

office – the personal benefit to running [B (g, I)] and the gain from not having a

bad candidate in office [C (b, I, d)]. If, relative to entry costs, these are sufficiently

11

weak (reflecting the fact that winning is only probabilistic), then good candidates

will not enter.8

This kind of equilibrium is most likely when information is poor (q close to zero)

and when B (g, I) + C (b, I, d) is low relative to entry costs. Thus, high wages and

good information (q close to one) improve the quality of politicians by destroying

the equilibrium in which only bad candidates stand.

Extending this to politically dominant groups in general requires an additional

condition:

Proposition 3 Suppose that there is a politically dominant group d and

λk (k)− λk (d) > C (b, I, k)∀k 6= d.

Then a pure strategy Nash equilibrium exists with only bad candidates of type d if:

δd >

(1 + q

2

)[B (g, I) + C (b, I, d)]

The extra condition says that citizens prefer to vote on the basis of their group

identity rather than candidate quality.9 If group attachment is weak, then it is

not possible to construct an equilibrium where all candidates are bad, as voters will

switch to good candidates even if they are not from their group.

Propositions 2 and 3 both rest on entry costs in politics being non-negligible

relative to private benefits. More generally, they suggest two important issues in

affecting candidate quality: (i) the relative returns to holding office among good

and bad candidates and (ii) the probability of detecting bad candidates in electoral

competition. These are the main forces at work with a politically dominant group.8The proposition illustrates a somewhat extreme case – more generally there can be pure or

mixed strategy equilibrium comprising good and bad candidates.9It is feasible to work with weaker, but less straightforward to state, conditions. We require

that when contrasting a type k (6= d) candidate with a type d candidate the set of types for which

candidate quality is salient is a population minority.

12

If political reservation simply changes the type of political dominance, then the

reservation status of a village need not affect the probability of selecting a bad

politician. However, if politicians have group preferences that affect the policies

they implement, then the group allocation of resources should change.

In the absence of political dominance, it is hard to say much concretely about

the likelihood of bad politicians. However, one further important effect may arise

in such cases. This is the possibility of a coordination failure among voters as

illustrated by Besley and Coate (1997). They construct an equilibrium in which a

two candidate equilibrium between sufficiently polarized candidates can be sustained

by voters’ beliefs that insufficiently many other voters will support a high quality

candidate if he or she enters.10 This kind of example gives a further reason to

believe that polarization can result in low quality candidates holding office, as voter

coordination is not a issue when polarization is low.

We have assumed that bad politicians have no extra electoral power to influ-

ence elections. The likelihood of observing bad politicians would be strengthened if

bad candidates can directly influence voting outcomes and prevent citizens voting

for good candidates through bribery, intimidation or manipulation of information

flows. This can be incorporated in our model as implying lower (net) benefits for

good candidates from holding office. Although we do not have evidence of electoral

intimidation, we find that candidates’ economic and political power affect their like-

lihood of selection but not their performance. This is suggestive of extra electoral

power or barriers to entry for the politically and economically disadvantaged.

Our analysis ignores the role of parties. In reality, parties may also influence

outcomes. The coordination failure result of Besley and Coate (1997) cannot arise

if parties coordinate political entry among groups 1 and 2. However, in situations

where bad candidates can also corrupt parties, then we would not expect parties to

resolve the issues raised above.10This can be formalized in the framework described here in the case of two groups.

13

3.4 Empirical Implications

Our model of the political process identifies channels through which village char-

acteristics that alter political dominance, politician rents ex post, and information

flows in the village, should affect politician quality. Here, we briefly outline how we

will test the empirical relevance of these channels.

The main vehicle for testing the model is through the allocation of BPL cards,

one of the main ways of targeting transfers in our villages. While this is only one of

the many policies that are dealt with by village politicians, BPL card allocation is a

possible source of political rents. Moreover, having well-targeted transfer programs

is likely to be of interest to a wide group of citizens within a village.

If we suppose that good politicians make a bona fide effort to reach the poorest

groups, then the cost of a bad politician C (b, I, k) is (partly) that an eligible indi-

vidual from group k does not receive a BPL card. The private benefits of holding

public office B (b, I) could also be partly due to politicians targeting BPL cards to

themselves when they are not eligible for one.

Our model predicts that political institutions and village characteristics which

improve targeting and diminish the power of the politician (or make him more

accountable) affect the extent of BPL card mis-targeting. If institutions of restraint

through monitoring were perfect, then we would not expect the politician’s type to

affect the targeting rule.

In our empirical analysis we examine how individual, and village, characteristics

that alter political dominance and information flows affect who is selected as a

politician, and the selection of BPL card holders. If, as predicted by the model,

differences in politician performance are systematically linked to politician quality, as

measured by characteristics such as education, and group identity, then institutions

which alter the extent of selection on these characteristics should have a predictable

impact on policy outcomes. We look for such evidence.

14

4 Data and Empirical Analysis

We begin by describing the institutional context for our analysis. We then describe

the survey data and our empirical specification.

4.1 Institutional Context

The 73rd constitutional amendment of India, passed in 1993, created a three-tier

elected local government in every state. We focus on the lowest tier – a popularly

elected village council called the Gram Panchayat (GP). GPs are demarcated on a

state-specific population basis, and may consist of multiple villages. A GP is divided

into wards, with elections held at the ward-level. The GP council consists of elected

ward members, and is headed by an elected Pradhan.11

The 73rd constitutional amendment mandated political reservation of a certain

fraction of elected GP positions in favor of two groups – scheduled castes and tribes

(hereafter, SC/ST) and women. Only individuals belonging to the group benefitting

from reservation can stand for election in a reserved position. The constitutional

amendment required that SC/ST reservation in a state be proportional to the group’s

population share, while women’s reservation equal one-third of all positions. No

position can be reserved for the same group for two consecutive elections.

A GP has responsibilities of civic administration with limited independent tax-

ation powers. Here, we focus on the allocation of BPL cards by GP politicians.

Since 1997 the Indian government has used a targeted public food distribution sys-

tem which provides BPL card holders subsidized food while charging a near market11A state’s Panchayat Act mandates the population or geographic criteria for GP demarcation.

This is the (revenue) village in Andhra Pradesh and Kerala, and a revenue village with 500 or more

persons in Tamil Nadu. In Karnataka it is a group of villages with between 5,000 and 7,000 persons.

The population per ward varies between 300 and 800 for these states.There is also variation in mode

of Pradhan election. In Andhra Pradesh and Tamil Nadu the Pradhan is directly elected, while

Karnataka and Kerala she is nominated from the pool of elected ward members.

15

price for the others. In 2000-01 the annual income gain per household from having a

BPL card for our sample states was roughly 5% of an agricultural labor household’s

annual expenditure (using 1999 NSS figures).12 In addition to subsidized food, most

GP administered welfare schemes, e.g. employment and housing schemes, restrict

eligibility to BPL households.

The central government uses the Planning Commission’s poverty estimates (which

are based on the National Sample Survey) to determine the number of BPL house-

holds per state, and accordingly releases foodgrain. The state government allocates

district-wise “quota” of BPL cards. Similarly, within a district, a “quota” of BPL

households is determined at the GP level with the GP bearing much of the respon-

sibility for allocating these BPL cards.

States are required to conduct a household survey to identify eligible house-

holds. GP politicians bear substantial responsibility for conducting this survey.

They choose the village surveyors, and tabulate the results bearing in mind the

quota allocated to the GP. The result is a preliminary ‘BPL’ list of recipients. The

list is supposed to be finalized at a village meeting comprising all citizens registered

on the GP’s electoral roles (called a Gram Sabha). This Gram Sabha meeting also

arranges household names in the order of priority. The same procedure is supposed

to be used when choosing households from among BPL households for other welfare

schemes.

In reality GP officials enjoy substantial discretion in determining the final BPL

list. In our data, for example, only 76% of villages had held a Gram Sabha in the

past year and only 20% of households report ever having attended a Gram Sabha.

Moreover, beneficiary selection was reported as discussed in only 22% of Gram Sabha12Under the public food distribution system 20 kg of food grains per month is provided at 50%

economic cost to BPL households. The effective annual income gain was Rs. 1025 in Andhra

Pradesh, Rs. 520 in Karnataka, Rs. 1414 in Kerala and Rs. 809 in Tamil Nadu We describe how

this income gain was calculated in footnote 1. (Planning Commission, 2005)

16

meetings (See Besley, Pande and Rao (2005)). Further, of the 540 politicians we

surveyed, only 9% stated that the Gram Sabha decided final inclusions or exclusions

from the BPL list; in contrast, 87% believed that this power lay with a Panchayat

official.

4.2 Data

Our analysis uses household survey and village meeting data which we collected

between September and November 2002. Our sample covered 259 villages in the four

southern states of India – Andhra Pradesh, Karnataka, Kerala and Tamil Nadu.13

Our sample includes nine boundary districts in these states. Indian districts are

divided into blocks. In each district we sampled 3 blocks, and six randomly sampled

GPs within each block. In GPs with three or fewer villages, we sampled all villages;

otherwise, we sampled the Pradhan’s village and two randomly selected villages.14 In

each sample village we conducted twenty-one household surveys. Household selection

was random, and we alternated between male and female respondents. In every

village, we required that four of the sampled households be SC/ST households and

one be an elected Panchayat official, preferably the Pradhan.15 Our final household

sample size is 5180 non-politician and 265 politician households (100 politicians are

from reserved jurisdictions).

Table 1 provides descriptive statistics. The average respondent has slightly over

4 years of education. Politicians are significantly more educated. Average land

holdings are 2.4 acres; however, when we restrict the sample to politicians this figure

rises to 5.7 acres. Roughly sixty percent of our respondents are either SC/ST or

female, and therefore eligible for reservation. In terms of political experience, seven13At the time of our survey at least one year had lapsed since the last GP election in every state.14In Kerala to account for the higher GP population we sampled 3 GPs per block and 6 wards

per GP – the Pradhan’s ward and five randomly selected wards.15We always interviewed the Pradhan, and in non-Pradhan villages we interviewed a randomly

selected ward member.

17

percent of our respondents have/had a family member with a political position.

Finally, twenty-one percent of our households possess a BPL card.

Voter turnout in GP elections is high, with 85 percent of our respondents report-

ing having voted in the last GP election. Eight percent of our respondents stated

that candidate group identity (defined along religion, caste, gender or regional lines)

was the most important reason for their candidate choice in the GP election, while

over thirty percent stated that the candidate’s quality (in terms of reputation or

policy promises) determined their vote. However, less than forty percent of the re-

spondents believed that their Pradhan has either kept his/her election promises or

looked after their needs.

Our model suggests that increases in formal returns to politics, improvements

in information flows, and reductions in cost of entry should lower the incidence of

bad politicians. Political reservation would reduce the incidence of bad politicians

if it causes a previously undominated village to become politically dominated. Oth-

erwise, its main effect should be to change the group allocation of resources.

Our choice of village characteristics is aimed at testing these mechanisms. We

are unable to examine the formal returns to politics due to a lack of within-state

variation. We proxy for information flows in the village by the 1991 village literacy

rate, and whether the village had at least one Gram Sabha meeting in the last

year. Both variables were positively correlated with household survey measures of

individual information. By focussing on literacy rates from before the Panchayat

system was introduced, we can avoid the concern of Panchayat activism causing

educational change. However, we recognize that our information variables may be

correlated with other unobserved village characteristics, and later we discuss the

implications of this for our results.

For political reservation, we use data on the reservation status of our surveyed

politicians, and on whether the position of the Pradhan is reserved. The Pradhan

position is reserved for women and SC/STs in roughly 16% of our GPs each. Within

18

a block, reservation of the Pradhan position and of wards within a village, is deter-

mined by a rotational system and is exogenous to village characteristics.16 Finally,

we define a village as having a dominant caste if the fraction of households belonging

to the single largest non SC/ST caste exceeds the median caste dominance in our

village sample (this stands at 40%). Unlike political reservation, having a dominant

caste need not imply political dominance. However, a large anthropological liter-

ature suggests that barriers to entry for minority groups are often higher in such

villages, and it is also more likely that the largest caste group is politically dominant

(see, for instance, Wade 1988). Low migration rates across Indian villages imply that

village caste structure is relatively stable.

4.3 Empirical Specification

In our household data we observe who is ultimately elected, but not who stands.

Suppose that being elected depends upon some underlying candidate quality, eij ,

for politician i in village j. Further, suppose that candidate quality depends on a

vector of candidate characteristics xij so that:

eij = βxij + ψij (1)

where ψij is a component of candidate electability that may be observable to voters,

but not to us. The parameters β can be thought of as true “production function”

parameters for candidate quality.

We suppose that there is some unobserved threshold e∗j in village j for i to be

elected to office. This subsumes the quality of challengers for public office, and the

distribution of different voting groups in village j. Then, we observe candidate i in

village j if:

eij > e∗j16No political position can be reserved for the same group for two consecutive elections. In

Besley, Pande, Rao and Rahman (2004) we show that public good provision in 1991 was statistically

indistinguishable in GPs with and without a reserved Pradhan.

19

or

βxij + ψij + ηij > e∗j

where ηij is a shock which affects how the candidate is perceived by voters in village

j. Treating e∗j as an unobserved village effect, and assuming a linear probability

model, this yields:

pij = αj + ρxij + εij . (2)

where pij is a dummy variable for whether the respondent is a politician and αj is

a village fixed effect. The parameters ρ do not only reflect the production function

if the variance of the shock ηij depends on xij . The fact that the variance of

εij depends on village characteristics, Ij , may also justify interacting ρ with such

characteristics in equation (2).

Estimating (2) allows us to examine political selection on observables, and how

this varies with village characteristics. We consider village literacy rate in 1991,

whether the Pradhan’s position is reserved and whether the village has a dominant

caste (the last may reflect barriers to entry rather than dominance per se).

To test whether politician quality and group identity matters for policy making,

we examine household access to BPL cards. Let bij be the probability that household

i in village j has a BPL card. We model this empirically as:

bij = αj + λpij + τpijeij + γxij + ηij (3)

where, as above, eij is politician “quality”. If politicians are opportunistic we expect

λ > 0, but if quality matters, then we expect τ < 0.

The above selection model tells that we expect

eij = θxij + φIj + νij (4)

where θ is the “reduced form” effect of candidate characteristics on quality working

both through the production function (1) and the probability that a candidate with

20

characteristics xij is selected. Substituting (??) into (3), we get the reduced form

model:

bij = αj + λpij + χ1 (xij ∗ pij) + χ2 (pij ∗ Ij) + γxij + µij . (5)

The coefficients χ1 = τθ and χ2 = τφ. Hence, observing that characteristic xij

enters negatively is indicative of τ < 0 and θ > 0, i.e. this is associated with being a

good politician. The latter can also be related to (2) since we would expect that a

good politician characteristic xij would have ρ > 0, if that characteristic is valued by

voters. Similarly, Ij entering negatively is associated with being a good institution.

5 Results

The results are presented in three parts. We first examine the determinants of

politician selection, and then those of beneficiary selection. Finally, we examine

how voters perceive politicians in our sample.

5.1 Selection of Politicians

We start by asking whether individual characteristics affect the likelihood that a

respondent is an elected politician. The results from estimating (2) are in Table 2.

In column (1) the dependent variable is whether the respondent is an elected GP

politician (i.e. a Pradhan or ward member). Eligibility for reservation is uncorre-

lated with being a politician. However, years of education and land ownership are

positively correlated with being a politician. An additional year of education, and

owning an additional acre of land, increase the likelihood of being a politician by

roughly 0.7% each. A respondent belonging to a family with a history of political

participation is 12% more likely to be a politician.17

17We have estimated these regressions including party affiliation variables. A respondent affiliated

with the party in power in the state is roughly 7 percent more likely to be a politician.

21

In columns (2) and (3) we restrict the sample to the groups eligible for political

reservation, women and SC/ST respectively. For both groups we observe a positive

selection on education, but not land ownership. Family political history and selection

are positively correlated only for women. For SC/ST households the absence of

selection on land and political history reflects their relative landlessness, and recent

entry into politics on the back of reservation.18 In columns (4)-(6) we restrict the

sample to Pradhan villages, and the dependent variable to whether the respondent is

the Pradhan. We observe very similar patterns of selection. Overall, the data points

to the political selection process favoring economically advantaged and politically

connected individuals.

Table 3 explores political selection in village j as a function of village character-

istics Ij . We estimate:

pij = αj + βxij + γ (xij ∗ Ij) + εij . (6)

where xij are the individual characteristics considered in Table 2. For expositional

ease we focus on the sample of all politicians.

In column (1) we observe the presence of a dominant caste increases elitism

among politicians – caste dominance is correlated with elected politicians owning

relatively more land and increased selection on family political history. Columns

(2) and (3) examine how Pradhan reservation affects selection. We distinguish be-

tween reservation open to all women, and reservation for SC/STs. Unsurprisingly,

eligibility for reservation is a near perfect predictor of selection on gender and caste.

Relative to other politicians, reserved politicians are less educated, own less land

and are less likely to have a family political history of participation. This reflects

the historical legacy of the economic, social and political disadvantage faced by these

groups. Column (4) considers the literacy rate as a proxy for information flows in18In our sample mean landholding for SC/ST households is 1.14 acres and for non SC/ST house-

holds 2.79 acres.

22

a village. Relatively more educated respondents are selected as politicians in vil-

lages with higher literacy rates. Further, respondents belonging to groups eligible

for reservation are more likely to enter politics in such villages.

Overall, the results suggest that village characteristics that reduce the domi-

nance of major castes increase the presence of economically disadvantaged groups

in politics, while those that improve information flows (as proxied for by literacy)

enhance the selection of more educated politicians.

5.2 Selection of Beneficiaries

We now examine how political selection affects the targeting of BPL cards. In Table

4, we report results from estimating regressions of the form (5) where pij = 1 if the

household has a BPL card.

In column (1) we observe that, as intended by the program, BPL cards are

targeted towards economically disadvantaged households. An SC/ST household is

16% more likely to get a BPL card while a household with a more educated head

and/or more land holdings is less likely to have a BPL card. A household’s political

history does not affect its propensity to have a BPL card. However, a politician

household is 7% more likely to have a BPL card (column (2)). This is all the

more striking in view of the results in Table 2 which demonstrated that politician

households are more likely to be landed and educated.

In column (3) we examine whether reserved politicians behave differently, and

find mixed evidence. The point estimate suggests no significant differences between

reserved and unreserved politicians. However, we cannot reject the hypothesis that

reserved politicians exhibit no political opportunism.19 Column (4) examines the19As our regressions include village fixed effects we identify the effect of reservation off villages

where reserved and unreserved politicians were interviewed. This is a relatively small sample,

hence the noisiness of our estimates. If we run separate regressions for the sample of reserved and

unreserved politicians, the BPL effect is limited to the unreserved politician sample.

23

role of politician characteristics. Politician opportunism is invariant to most politi-

cian characteristics, save education. Political opportunism is lower among more

educated politicians. An extra year of education for a politician makes him or her

1.4% less likely to have a BPL card.20

Table 5 examines the role of village characteristics in constraining political op-

portunism. These regressions include controls for household demographics. For

expositional ease we replace the controls for landownership and education, by a

disadvantage dummy which equals one if the household head is illiterate or the

household is landless. In column (1) we observe that politicians are more likely to

have a BPL card in a village with a dominant caste. Strikingly, this effect is limited

to unreserved politicians. Having a dominant caste, however, does not alter the

targeting of BPL cards among villagers.

Columns (2) and (3) in Table 5 consider Pradhan reservation (these regressions

include GP fixed effects as reservation varies by GP). The likelihood that a politi-

cian has a BPL card is higher with a female Pradhan. This could reflect personal

aggrandizement on part of the Pradhan or a more limited ability to monitor other

politicians. Once again the targeting of BPL allocation among villagers is unaf-

fected. In contrast, column (3) shows that SC/ST reservation makes it more likely

that SC/ST households and reserved politicians have a BPL card. This points to

SC/ST Pradhans having preferences that favor members of their own group.

Columns (4) and (5) of Table 5 consider the impact on targeting of village literacy

and whether the village had a Gram Sabha meeting in the last year. Gram Sabha

meetings are intended as a forum at which villagers can discuss their problems with

the GP officials, and also monitor GP activities. Higher village literacy and holding

a Gram Sabha meeting significantly reduces the likelihood that a politician has a20We have also examined party affiliation. Sharing the affiliation of the main party in the state

does not matters. In contrast, a non-politician household with the same party affiliation as the

Pradhan is 8% more likely to get a BPL card. This effect is absent among politicians.

24

BPL card and improves targeting.21

Taken together these results illustrate the importance of selection and incentives

in affecting public resource allocation. Selection is manifested in more educated

politicians being less opportunistic. Incentives are shaped by village meetings in

which villagers ratify beneficiary lists chosen by politicians.

One key idea of the theory is that bad politicians impose a cost on other citizens.

Table 6 looks at one aspect of this by seeing whether politicians with BPL cards

target other groups differently. We do this by interacting the household character-

istics which in Table 4 made it more likely that a household gets a BPL card with

whether a politician has a BPL card and the politician’s years of education.

In column (1) we find that politicians with BPL cards, who tend to come from

unreserved seats (and hence, are not SC/ST) target fewer resources to SC/ST house-

holds. The flip side of this evidence is presented in column (2) of Table 6 which shows

that more educated politicians target more BPL cards towards SC/ST households.

This suggests that the main cost of having a bad politician is borne by the histor-

ically disadvantaged population group of SC/ST citizens. Given this, it is worth

noting that the main effect of political reservation for SC/ST seems to be to shift

resource allocation in their favor.

5.3 Robustness and Validation

This section looks at whether political opportunism is apparent for other public

transfer programs – government financed house improvements and participation in

public works programs. We also examine whether opportunistic politicians are per-

ceived as “bad” politicians. Finally, we examine whether citizens’ stated basis for

voting correlates with politician opportunism.21In Besley, Pande and Rao (2005) we show that villages with higher literacy are more likely

to hold Gram Sabha meetings. Importantly, economically disadvantaged households are relatively

more likely to attend these meetings.

25

Table 7 presents results on political opportunism for other public transfer pro-

grams. Columns (1) and (2) consider whether any household member worked on a

public works project during the last year. A politician household is four percentage

points more likely to have someone who does so. Once again, this effect is stronger

among unreserved politicians. Family political history is also a positive predictor

of participation in public works. Other politician characteristics do not, however,

explain such participation.

Columns (3) and (4) in Table 7 consider whether since the last election, the house-

hold had any home improvements under a government scheme. These include house

construction and repair, having a toilet constructed or drinking water or electricity

provided. Roughly seven percent of our households had such an improvement. Once

again, while economically disadvantaged households are targeted by this scheme,

politicians behave opportunistically. However, in this case, political opportunism is

limited to reserved politicians; see column (4). This is explained by the fact that

many home improvement schemes restrict eligibility to SC/ST households. It also

reflects the fact that unreserved politicians come from richer households which have

such home improvements (such as toilets) already. These two observations also

underlie the fact that politicians from politically connected families are less likely to

enjoy these home improvements.

We now examine how voters perceive the performance of opportunistic Pradhans.

If voters dislike opportunism, then politicians with BPL cards should be less popular.

This issue is explored in Table 8 where we use data on villagers’ perceptions of the

quality of their Pradhan. The survey asked whether households thought that their

Pradhan “looked after village needs” and whether they “kept their promises”.

Columns (1) and (3) demonstrate that Pradhans who have a BPL card are

perceived as worse on both indicators of Pradhan quality (the regressions include

block fixed effects since variation in Pradhan data is at GP-level). This is consistent

with our interpretation of politician participation in government transfer programs

26

as being a form of rent-seeking which is disapproved of by citizens. Columns (2)

and (4) show that educated Pradhans are better regarded by villagers in their GP –

again consistent with our earlier result on education. That said, female and SC/ST

Pradhans are regarded as worse even though we did not find any evidence of greater

opportunism among these groups of politicians. This may, therefore, be due to the

fact that these groups have specific policy agendas. It could also be a reflection of

respondents at large being biased against traditionally disadvantaged groups.22

The second issue is motivated by an observation from the theory – that voting

along group lines diminishes the extent to which politician quality is reflected in

voting decisions. Hence, bad politicians are more likely when villagers vote along

lines of group identity. To test this idea, we examine the relationship between

citizens’ self-reported basis for voting and whether the Pradhan holds a BPL card

and is educated. We restrict attention to Pradhan elections, as our survey asked

only about voting in GP Pradhan elections.

We construct two measures of citizens’ voting preferences. First, we use respon-

dents’ report of whether they voted for a candidate based on their caste, gender,

religious or regional identity to identify the fraction of citizens who voted on the

basis of group identity. Second, we use responses to a question asking whether

respondents used the candidate’s qualifications/previous work in the village as their

basis for voting. We conjecture that more group based voting measured this way

should lead to lower quality Pradhans, and voting based on candidate quality as

leading to higher quality Pradhans.

The results are in Table 9. We run our regressions at the GP level (that is,

we construct and use GP level averages), and include district fixed effects. Greater

group based voting is correlated with Pradhans who take BPL cards and have fewer

years of education. There is, however, little evidence that reported voting on22Duflo and Topolova (2004) also find that, despite no observable differences in performance,

women Pradhans are perceived as being of worse quality.

27

candidate quality makes a difference. While the evidence is only suggestive, it is

consistent with the interpretation of the results in the previous two sections.

6 Concluding Comments

This paper has three key findings. First, the political class is selected on the

basis of political connections and economic advantage. Second, in targeting public

resources politicians exhibit group preferences and are opportunistic (in that they

benefit disproportionately from public transfer programs). Third, individual and

village characteristics mediate the extent of opportunism.

Among individual characteristics, we find that the education level of politicians

has a consistently positive effect on selection and a negative effect on opportunism.

This suggests that the more educated make better politicians and are recognized

as such by voters. However, whether education matters directly or because it is

correlated with other characteristics that make an individual fit for public office

cannot be discerned from our results. Nonetheless, the results add to a growing

appreciation among economists that education may be important because of its

role in inculcating civic values (See, for example, Dee (2004) and Milligan et al

(2004)). The unique observation about its role in politics given here also offers a

fresh perspective on the value of human capital investments in low income countries.

For the most part, our findings for village characteristics are consistent with

the theory laid out in section 3 and suggest an important interplay between village

characteristics and the process of political selection and the targeting of public re-

sources. Increased literacy at the village level reduces political opportunism while

political reservation is correlated with targeting of resources. There is some sug-

gestion of most villages being politically dominated, so that political reservation

changes the type of political dominance rather than causing political dominance.

We also find evidence suggestive of barriers to entry – while land ownership and

28

political connections predict selection they do not predict behavior when in office.

The results also cast light on the process of decentralization as it is occurring

throughout the developing world. This has attached a lot of weight in the virtues

of local decision making processes in targeting beneficiaries. Our results show that

targeting is heterogeneous and depends on those who are selected to run this process.

It further emphasizes the need to have adequate models of the political economy of

targeting to shed light on the merits of decentralization.

Our finding that educated politicians are better both in terms of both actual and

perceived performance suggests, in line with the opening quote from V.O. Key, that

it is important to focus on factors that select better politicians as a step towards

improving the quality of government. Equally, as predicted by the father of the

Indian constitution, B.R. Ambedkar, group identity remains a significant predictor

of politician behavior in India. Overall, we see the results and analysis in the paper

reinforcing the observation that formal institutions of democracy are no guarantee

of effective government. It is essential that the preconditions exist for sorting in

the right kinds of people – the talented, the virtuous and those who give political

voice to the disadvantaged. This paper is a first effort to use household level data

to study this issue empirically. But clearly there is much more to be done to gain

a deeper understanding of political selection in democratic settings.

29

References

[1] Ambedkar, B.R, [1937], “The Annihilation of Caste with a Reply to Mahatma

Gandhi,”.

[2] Bardhan, Pranab and Dilip Mookherjee, [2003], “Political Economy of Land

Reforms in West Bengal 1978–98 ,” mimeo Boston University.

[3] Barro, Robert, [1973], “The Control of Politicians: An Economic Model,” Public

Choice, 14, 19-42.

[4] Besley, Timothy, [2004], “Paying Politicians: Theory and Evidence”, Joseph

Schumpeter Lecture 2004, Journal of the European Economics Association

[5] Besley, Timothy and Stephen Coate, [1997], “An Economic Model of Represen-

tative Democracy,” Quarterly Journal of Economics, 112(1), 85-114.

[6] Besley, Timothy, Rohini Pande and Vijayendra Rao, [2005], “Participatory

Democracy in Action: Survey Evidence from India,” Journal of the European

Economics Association, 3(2-3):648-657.

[7] Buchanan, James M., [1989], “The Public-Choice Perspective,” in Essays on

the Political Economy, Honolulu: University of Hawaii Press.

[8] Caselli, Francesco and Massimo Morelli, [2004], “Bad Politicians,” Journal of

Public Economics, 71(3), 829-53.

[9] Chattopadhyay, Raghabendra and Esther Duflo, [2004a], “Women as Policy

Makers: Evidence from a India-Wide Randomized Policy Experiment,” Econo-

metrica , 72(5), 1409-1443.

[10] Duflo, Esther and Petia Topolova[2004b], “Unappreciated Service: Perfor-

mance, Perceptions and Women Leaders in India,” mimeo MIT.

30

[11] Coate, Stephen and Stephen Morris, “On the form of transfers to special inter-

ests,” Journal of Political Economy, 103, 1210-35.

[12] Dee, Thomas S. [2004], “Are there civic returns to education?”, Journal of

Public Economics, Volume 88 (9-10) 1697-1720.

[13] Faguet, Jean Paul, [2004], “Does Decentralization Increase Responsiveness to

Local Needs? Evidence from Bolivia,” Journal of Public Economics, 88, 867-

894.

[14] Ferejohn, John, [1986], “Incumbent Performance and Electoral Control,” Public

Choice, 50, 5-25.

[15] Foster, Andrew and Mark Rosenzweig, [2001], “Democratization, Decentraliza-

tion and the Distribution of Local Public Goods in a Poor Rural Economy,”

mimeo, Brown.

[16] Gehlbach, Scott and Konstantin Sonin, [2004], Businessman Candidates: Spe-

cial Interest Politics in Weakly Institutionalized Environments, unpublished

typescript.

[17] Jefferson, Thomas, [1813], “The Natural Aristocracy.” Letter to John Adams.

October 28.

[18] Jones, Benjamin and Benjamin Olken, [2005], “Do Leaders Matter? National

Leadership and Growth since World War II”, Quarterly Journal of Economics,

August.

[19] Key, V.O., [1956], American State Politics: An Introduction, New York, NY:

Alfred A. Knopf.

[20] La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer and Robert W.

Vishny, [1999], “The Quality of Government,” Journal of Law, Economics and

Organization, 15(1), 222-79.

31

[21] Lee, David S., Enrico Moretti, and Matthew Butler, [2004], “Do Voters Af-

fect or Elect Policies? Evidence from the U.S. House,” Quarterly Journal of

Economics, 119(3), 807-860.

[22] Maskin, Eric and Jean Tirole, [2004], “The Politician and the Judge”, American

Economic Review,

[23] Matthew, George and Nirmala Buch, [2000] Status of Panchayati Raj in the

States and Union Territories of India, Institute for Social Studies, Delhi.

[24] Milligan, Kevin, Enrico Moretti and Philip Oreopoulos, [2004], “Does educa-

tion improve citizenship? Evidence from the United States and the United

Kingdom,” Journal of Public Economics, 88 (9-10), 1667-1695.

[25] Osborne, Martin J. and Al Slivinski, [1996], “A Model of Political Competition

with Citizen Candidates,” Quarterly Journal of Economics, 111(1), 65-96.

[26] Padro-i-Miquel, Gerard, [2005], “The Control of Politicians in Divided Societies:

The Politics of Fear,” Mimeo, MIT.

[27] Planning Commission, [2005], “Performance Evaluation of Targeted Public Dis-

tribution System,” Programme Evaluation Organisation, Planning Commis-

sion, Government of India, New Delhi.

[28] Poutvaara, Panu and Tuomas Takalo, [2003], “Candidate Quality,” unpublished

typescript.

[29] Pande, Rohini, [2003], “Can Mandated Political Representation Provide Dis-

advantaged Minorities Policy Influence? Theory and Evidence from India,”

American Economic Review ; 93(4), 1132-1151.

[30] Wade, Robert, [1988], “Village Republics: Economic Conditions for Collective

Action in Southern India,” ICS.

32

Appendix A: Theory

Proof of Proposition 1: Suppose not. Then the number of good candidates in the race

is:

mg = int

(δd

B (g, I)

)≥ 1.

This uses the fact that all good candidates win with equal probability in any voting equi-

librium. We require that no bad would wish to enter. This requires that:

1− q

mg + 1B (b, I) < δd.

But clearly this cannot hold for large enough B (b, I) – a contradiction. QED

Proof of Proposition 2: This is a special case of Proposition 3.

Proof of Proposition 3: We first show that a least one bad candidate of type d would

wish to enter. This follows from the fact that:

λd (d)− λd (0) + B (b, I) > δd.

We now show that there is a voting equilibrium in which no good candidate would wish to

enter. Suppose that there is a single bad candidate in the race. If a good candidate of d

chooses to enter and is identified as such, then he will win in any voting equilibrium which

eliminates weakly dominated strategies. If he is not identified as good, he will win with

probability one half. We now look at the incentives of such a candidate to enter. He will

wish to enter if :

λd (d) +[q +

1− q

2

]B (g, I)− 1− q

2C (b, I, d)− δd > λd (d)− C (b, I, d) .

which reduces to the condition in the Proposition. The condition holds a fortiori if there is

more than one bad candidate in the race.

Suppose that a candidate who is not of type d enters and is identified as good. Then

since:

λk (k)− λk (d) > C (b, I, k)∀k 6= d,

we can construct a voting equilibrium in which the bad candidate from group d wins in any

voting equilibrium which eliminates weakly dominated strategies. (This follows from the

definition of political dominance.) Thus, no good candidate will choose to enter. QED.

33

Table 1: Descriptive StatisticsMean s.d.

Respondent characteristicsYears of Education All 4.49 (4.54)

Politicians 7.58 (4.51)Land owned (in acres) All 2.26 (4.77)

Politicians 5.98 (8.87)Eligible for reservation (%) All 60.90 (48.81)

Politicians 48.70 (50.07)Family political history (%) All 6.70 (25.04)

Politicians 25.30 (43.54)Beneficiary Status (% households)BPL card All 21.70 (41.20)

Politicians 24.20 (42.80)Perceptions and Voting Behavior (% non-politicians) Pradhan looks after village needs 38.40 (48.63)Pradhan keeps election promises 36.10 (48.03)

Vote for group identity 8.72 (28.22)Vote for candidate quality 36.08 (48.02)Institutions (% villages)Dominant caste 51.93 (50.05)

Pradhan reserved for Female 15.89 (36.63)

Pradhan reserved for SC/ST 16.66 (37.34)

Literacy rate 42.20 (18.35)

Gram Sabha 77.95 (41.53)Notes:1. Years of education refer to respondent's years of education. Land owned is amount of land, in acres, owned by respondent's household. A respondent is eligible for reservation if female or SC/ST. A respondent has a family political history if any member of his/her household holds or has held a political position. BPL card refers to whether the household has a BPL card.

2. Vote dummies refer to GP election. Vote for group identity=1 if respondent says she voted for the candidate with the same caste/religion/gender/place of residence. Vote for candidate quality=1 if respondent says she voted for candidate with good policy promises/candidate active in the village/good reputation.

3. A Village has a Dominant caste if over 40 percent of villagers belong to a single caste. Literacy rate is the 1991 census village literacy rate. Gram Sabha is a dummy for whether the village had a Gram Sabha meeting in the last year.

Table 2: Individual Characteristics and Politician SelectionPolitician Pradhan

Sample All Female SC/ST All Female SC/ST(1) (2) (3) (4) (5) (6)

Eligible for 0.008 0.002reservation (0.007) (0.010)

Education 0.008*** 0.007*** 0.012*** 0.006*** 0.005** 0.007*(0.001) (0.001) (0.002) (0.001) (0.002) (0.004)

Land owned 0.007*** 0.003 0.002 0.008*** 0.002 0.033**(0.002) (0.002) (0.003) (0.002) (0.002) (0.014)

Family political 0.119*** 0.135*** 0.062 0.095*** 0.086** 0.057history (0.020) (0.032) (0.044) (0.029) (0.039) (0.090)

Fixed effects Village Village GP Village Village GP

R-squared 0.09 0.12 0.12 0.09 0.11 0.23

N 5397 2644 1245 2065 1011 436Notes:1.OLS regressions with standard errors, clustered by village, in parentheses. * significant at 10%; ** at 5%; *** at 1%.2.The dependent variable is an indicator variable=1 if the respondent is a politician.3.All regressions include control for respondent age and age squared. The Pradhan regressions restrict the sample to the Pradhan and non politician households in the Pradhan's village.

4.Eligible for reservation is an indicator variable which equals one if respondent is female or SC/ST. Land ownership is the land (in acres) owned by the respondent's household. Education refers to respondent's years of education. Family political history is an indicator variable which equals one if any family member of respondent has held/holds a political position.

Table 3: Village Characteristics and Politician Selection

Institution Dominant CasteFemale Pradhan

ReservationSC/ST Pradhan

Reservation Literacy Rate(1) (2) (3) (4)

Eligible for reservation 0.013 -0.013** -0.009 -0.012(0.009) (0.006) (0.006) (0.016)

Eligible for reservation* -0.007 1.032*** 1.032*** 0.05Village Characteristic (0.012) (0.006) (0.007) (0.034)

Education 0.008*** 0.006*** 0.006*** 0.005**(0.001) (0.001) (0.001) (0.002)

Education* -0.001 -0.006*** -0.003*** 0.007*Village Characteristic (0.002) (0.001) (0.001) (0.004)

Land owned 0.005** 0.006*** 0.008*** 0.002(0.002) (0.002) (0.002) (0.004)

Land owned* 0.005* -0.006*** -0.007*** 0.016Village Characteristic (0.003) (0.001) (0.002) (0.011)

Family political history 0.112*** 0.083*** 0.111*** 0.067(0.030) (0.019) (0.020) (0.051)

Family political history* 0.013 -0.076*** -0.131*** 0.104Village Characteristic (0.040) (0.020) (0.022) (0.108)

Fixed effects Village Village Village VillageR-squared 0.09 0.25 0.26 0.09N 5397 5397 5397 5187Notes:1. OLS regressions with standard errors, clustered by village, in parentheses. * significant at 10%; ** at 5%; *** at 1%.2. The dependent variable is an indicator variable=1 if the respondent is a politician.3. Regressions include respondent age and age-squared as a control variable. Explanatory variables are defined in notes to Tables 1 and 2.

Table 4: Politician Characteristics and BPL Beneficiary Selection(1) (2) (3) (4)

SC/ST household 0.164*** 0.162*** 0.164*** 0.166***(0.019) (0.019) (0.019) (0.019)

Household head's -0.008*** -0.008*** -0.008*** -0.008***education (0.002) (0.002) (0.002) (0.002)Respondent's education -0.003* -0.003** -0.003** -0.003*

(0.001) (0.001) (0.001) (0.002)Land owned -0.004*** -0.004*** -0.004*** -0.003*

(0.001) (0.001) (0.001) (0.001)Family political history -0.012 -0.021 -0.02 -0.029

(0.020) (0.020) (0.020) (0.019)Politician 0.075** 0.109*** 0.199**

(0.033) (0.041) (0.080)Politician*Reserved -0.087 -0.105

(0.069) (0.071)F-test 0.16 1.48

[ 0.685] [0.22]Politician*Education -0.014**

(0.007)Politician*Land owned 0.001

(0.003)Politician*Family political 0.069history (0.083)Fixed effects Village Village Village VillageR-squared 0.36 0.36 0.36 0.36

N 5366 5366 5366 5366Notes:1. OLS regressions with standard errors, clustered by village, in parentheses. * significant at 10%; ** at 5%; *** at 1%.2. The dependent variable is an indicator variable which equals one if the respondent's household has a BPL card.3. All regressions include as household controls: household size, head's age and age squared, fraction elderly and fraction children. Other variables are as defined in Table 2 notes.

Table 5: Village Characteristics and BPL Beneficiary Selection

Institution Dominant casteFemale Pradhan

reservationSC/ST Pradhan

reservation Literacy rate Gram Sabha(1) (2) (3) (4) (5)

Politician -0.01 0.069* 0.101** 0.399*** 0.282***(0.053) (0.039) (0.040) (0.098) (0.095)

Politician* 0.185** 0.498** -0.377* -0.746*** -0.242**Village Characteristic (0.079) (0.219) (0.209) (0.188) (0.105)Reserved politician 0.035 -0.028 -0.098 -0.144 -0.343**

(0.093) (0.077) (0.076) (0.176) (0.142)Reserved politician* -0.194 -0.547** 0.409* 0.21 0.359**Village Characteristic (0.135) (0.243) (0.232) (0.338) (0.161)SC/ST household 0.180*** 0.145*** 0.119*** -0.044 0.108***

(0.025) (0.019) (0.026) (0.040) (0.039)SC/ST household* -0.021 0 0.112** 0.512*** 0.072Village Characteristic (0.040) (0.000) (0.055) (0.093) (0.045)Economic Disadvantage 0.011 0.092*** 0.096*** -0.018 0.060***

(0.027) (0.015) (0.014) (0.031) (0.019)Economic Disadvantage* -0.001 -0.005 -0.065 0.271*** 0.045*Village Characteristic (0.051) (0.020) (0.050) (0.076) (0.025)Family political history -0.051* -0.037* -0.022 0.022 0.016

(0.028) (0.022) (0.021) (0.042) (0.035)Family political history* 0.048 0 -0.092 -0.103 -0.058Village Characteristic (0.040) (0.046) (0.065) (0.096) (0.042)Fixed effects Village GP GP Village VillageR-squared 0.36 0.3 0.3 0.38 0.36N 5369 5369 5369 5159 5287Notes

1. OLS regressions with standard errors, clustered by village, in parentheses. * significant at 10%; ** at 5%; *** at 1%.2. The dependent variable is an indicator variable which equals one if the respondent's household has a BPL card.3. Regressions include the household controls defined in notes to Table 4. Economic disadvantage is a dummy which equals one if the household head is illiterate or landless. Other variable definitions are in notes to Tables 1 and 2.

Table 6: Politician Characteristics and BPL Beneficary SelectionPolitician Characteristic Has BPL card Years of education

(1) (2)Politician -0.147*** 0.264***

(0.023) (0.100)Politician* 1.076*** -0.020**Politician Characteristic (0.051) (0.009)Reserved politician -0.095** -0.115

(0.042) (0.141)Reserved politician* 0.118 0.002Politician Characteristic (0.087) (0.014)SC/ST household 0.169*** 0.110***

(0.019) (0.035)SC/ST household* -0.295*** 0.008**Politician Characteristic (0.083) (0.004)Economic Disadvantage 0.090*** 0.055**

(0.013) (0.027)Economic Disadvantage* -0.064 0.006*Politician Characteristic (0.062) (0.003)Family political history -0.044** 0.063

(0.017) (0.043)Family political history* 0.062 -0.010**Politician Characteristic (0.068) (0.004)Fixed effects Village VillageR-squared 0.42 0.37N 5369 5328Notes1. OLS regressions with standard errors, clustered by village, in parentheses. * significant at 10%; ** at 5%; *** at 1%.

2. The dependent variable is an indicator variable which equals one if the respondent's household has a BPL card.3. Regressions include the household controls defined in notes to Table 4. Other variable definitions are in notes to Tables 1 and 2.

Table 7: Politicians and Beneficiary Selection: Other public transfersPublic works Home improvements

(1) (2) (3) (4)Politician 0.044** 0.054 -0.004 -0.028

(0.022) (0.045) (0.014) (0.033)Politician*Reserved 0.026 0.033 0.065* 0.084**

(0.042) (0.041) (0.036) (0.035)F-test 3.65 2.13 3.22 0.58

(0.05) (0.144) (0.07) ( 0.44)SC/ST household 0.053*** 0.053*** 0.057*** 0.057***

(0.012) (0.012) (0.013) (0.013)Household head's 0 0 -0.002** -0.002**education (0.001) (0.001) (0.001) (0.001)Respondent's education -0.001 -0.001 0 0

(0.001) (0.001) (0.001) (0.001)Land owned 0 -0.001 -0.002*** -0.003***

(0.001) (0.001) (0.001) (0.001)Family political history 0.017 0.021* -0.011 0.002

(0.013) (0.013) (0.013) (0.015)Politician*Education -0.004 0.001

(0.005) (0.004)Politician*Land owned 0.004 0.005*

(0.003) (0.003)Politician*Family political -0.024 -0.084***history (0.040) (0.030)Fixed effects Village Village Village VillageR-squared 0.13 0.13 0.11 0.11N 5335 5335 5366 5366Notes:

1. OLS regressions with standard errors, clustered by village, in parentheses. * significant at 10%; ** at 5%; *** at 1%.

2. The dependent variables are dummies: Public works=1 if a member of the respondent's household worked on a public works project in the last 365 days. Home improvements=1 if respondent's house had a GP financed improvement since last election,

3. All regressions include the household controls defined in notes to table 4. Other variables are as defined in Table 2 notes.

Table 8: Pradhan Characteristics and Villager Perceptions

Looks after village needs Keeps election promises(1) (2) (3) (4)

Pradhan has BPL card -0.079** -0.098***(0.033) (0.031)

Pradhan eligible for reservation -0.075** -0.068**(0.029) (0.028)

Pradhan's education 0.005* 0.004(0.003) (0.003)

Pradhan's land ownership -0.001 -0.001(0.002) (0.002)

Pradhan's family political history 0.006 -0.01(0.040) (0.042)

Individual controls Yes Yes Yes YesOther controls Yes Yes Yes YesFixed effect Block Block Block BlockR-squared 0.18 0.18 0.18 0.18N 4854 4854 4854 4854Notes:

1. OLS regressions with standard errors, clustered by GP, in parentheses. * significant at 10%; ** at 5%; *** at 1%.2. The dependent variables are dummies: Looks after village needs=1 if respondent says Pradhan looks after village needs; Keeps election promises=1 if respondent believes Pradhan keeps his election promises. 3.Other controls includes number of villages in GP, village literacy rate, pradhan village dummy, GP headquarter dummy, total households in village and fraction SC/ST households.

Table 9: Pradhan Characteristics and Voting Patterns BPL card Years of Education

(1) (2)Group identity voting 1.265** -22.859***

(0.632) (4.505)Candidate quality voting -0.206 3.416

(0.283) (2.879)GP literacy rate -0.319 13.196***

(0.330) (2.963)Control District DistrictR-squared 0.09 0.3

N 90 90Notes:

1. GP-level OLS regressions with standard errors, clustered by block, in parentheses. Regressions are weighted by fraction SC/ST households in GP (averaged across sample villages). *significant at 10%; ** at 5%; *** at 1%.2. Dependent variables are a dummy for whether Pradhan has a BPL card and years of education of Pradhan. Group identity voting and Candidate characteristic voting are fraction of villagers in GP who report the most important reason for their vote as candidate's group identity and quality, respectively.