Who gets the subsidy in subsidized credit programs? Some evidence from India Nishant Chadha∗ Department of Economics, University of British Columbia (UBC)

October 15, 2010

Abstract Does the caste of the agent have any bearing on his interaction with a corrupt official? Using data from India, which contains information on beneficiaries of subsidized loans, I find that lower castes pay higher bribes compared to high castes to obtain similar loans. Using a 1993 constitutional amendment, which reserved seats in local bodies for low castes, I find that low caste borrowers living in reserved villages pay signifcantly lower bribes than those living in unreserved villages. To delineate the channel through which this effect works I look at the interaction of the vote share of a low caste political party (the BSP) and the reservation status of the village. I find that in reserved villages located in higher vote share districts low caste borrowers pay lower average bribes than reserved villages located in low vote share districts. In fact all the change in bribe amounts because of reservation can be attibuted to this interaction. In a reserved village which falls in a district with zero voteshare there is no change in the bribe amount paid by SC borrowers Keywords: Caste; corruption; subsidies. JEL Classification Numbers: O17, D73, H20.

∗

Address: # 997-1873 East Mall, Vancouver, BC V6T1Z1, Canada, e-mail: [email protected]. The author is grateful to Dr. Mukesh Eswaran for advice and suggestions, and to Dr. Tomas Lemieux, Dr. Patrick Francois, and seminar participants at the University of British Columbia for helpful comments.I would also like to thank Mr. Amar Saxena and Mr. Vishwamitra Sabarwal of Lucknow, India for their help in making the 1995 panchayat election records available to me.

1

1

Introduction This paper looks at a subsidized credit scheme in India and asks a simple but pertinent

question: Does access to the subsidized credit differ by caste? To understand why this is a valid question we need to note two things. Firstly, credit is not always dispensed as laid down by the law. Corruption among bank officials is well recorded, and this gives officials the instruments to manipulate and divert resources. Secondly, public service in India is dominated by upper and middle castes. In so far as social standing of the borrower, his influence within the community and caste ties within the public services on the one hand and bribe paid to the official on the other are factors that affect the outcome of this interaction between corrupt official and borrower, the high castes have advantages over the lower castes in the first set. So one could reasonably expect the low castes to pay higher bribes to obtain loans. The idea is that the ability to rope in politically and socially influential people on your behalf and getting them to intervene with the official and paying the official a bribe are close substitutes. However getting a political leader or a senior civil servant to act on your behalf is not costless. The presence of caste networks however lessens these costs, it is cheaper to approach someone through your network than without it. Traditionally the high castes have had a decided advantage in this aspect, their caste network gives them access to people in various positions of power. The low castes on the other hand, until recently had little representation in politics and civil service and hence find it costly to engage influential people. Hence they might have to pay higher bribes to compensate for this lack of influence. In this paper I focus on a very particular interaction. One between a borrower interested in obtaining a subsidized loan and a bank official considering this application. The terms of these loans are fixed by law, leaving very little leeway for the official. This makes it ideal to study the processes of corruption. In the data I know how much bribe if any was paid to obtain the loan, and I find that the lower castes do pay higher bribes. Since the bribes increase the effective interest rate of these loans another way to put it would be that 2

lower caste borrowers get to keep less of the subsidy than a similar high caste borrower. By exploiting a policy change which potentially increased the presence of the lower castes in local politics I show that it was indeed the case that increased influence resulted in lower average bribes. Among the host of poverty alleviation schemes initiated by various Indian governments past and present subsidized credit schemes, under different names and guises, occupy a place of pride.1 . In this paper I look at the Integrated Rural Development Program(henceforth the IRDP), the latest and largest in a string of such schemes(Pulley (1989); Swaminathan (1990)). Real expenditure on IRDP has increased 14% per year on average since 1980-81 and between 1980/81-1987/88, real IRDP investment amounted to INR 77 billion.In 1987-88 alone the government spent INR 4.7 billion on subsidies associated with the program. In 1989 IRDP covered nearly 27 million rural families, approximately 135 million rural people. Credit is both important to and difficult to obtain for the poor in developing countries like India. Lacking any assets that can be put up as collateral they are excluded from formal credit markets and have to rely on informal credit where they are often exploited by moneylenders(Ghosh et al. (2001); Banerjee (2003)). The IRDP makes credit available for productive investment at lower than market interest rates and without any collateral, the idea being that the investment will help them repay the loan and increase their permanant income(Pulley (1989); Dreze (1990)). Numerous studies have identifed problems and expressed reservations with the IRDP2 . Dreze (1990) argues in favour of employment guarantee schemes (and against IRDP) citing the inherent uncertain nature of the village setting. Given the complementary nature of the different assets it is not unreasonable to believe that saddling an agent with an unproductive asset can actually make him worse off 3 . Secondly, studies (Pulley (1989); Dreze et al. 1

Subsidized credit schemes have in fact been a staple in a lot of developing countries(see Buttari (1995); Besley (1994)) 2 In this paper I am not questioning the overall viability of IRDP vis-a-vis other poverty alleviation programs, rather I am interested in seeing how effectively the IRDP managed to achieve one of it’s stated aims 3 As an example consider the following from Dreze:“Take the example of the much-discussed IRDP buffalo.

3

(1997); Dreze (1990) etc.) have documented pervasive corruption in the implementation of the scheme. Bribes are commonly paid by the beneficiaries to the officials to obtain loans. Apart from the obvious normative concerns one might have with this sort of corruption there is another, graver problem. The scheme incorporates a lower than market interest rate and a capital subsidy4 so that the farmer(borrower) can keep more of the surplus generated from the investment than he would have been able to if he had borrowed from the informal credit market. Now, if the officials (who on paper do not have the discretion of setting the interest rate which is set by the government) have an additional instrument, in the form of a bribe, available to them they can expropriate some of this surplus. In fact it is not difficult to think of models in which, under fairly reasonable conditions, the effective interest rate for the “subsidized” loan (actual interest rate corrected for the bribe) is equal to the interest rate in the informal market (Chaudhuri and Gupta (1996, 1997); Saha and Thampy (2006)). How should we think about the bribe setting process? Given the personal nature of these interactions I argue that the correct way is to look at this as a bargaining process between the official and the farmer, with the dice loaded heavily in the favour of the official. If this is indeed a reasonable depiction of reality then any characteristic of the borrower that decreases his bargaining weight is also going to lower the surplus that he gets to keep, that is increases the bribe received by the official. I posit that in this context the caste of the agents is very highly correlated to their own social and political influence and their access to other politically influential people (through their network) thus affecting their bargaining power in the afore mentioned transaction. The lower castes have the least influence in society and hence should end up paying higher bribes. This is exactly what I observe in the data. I find that when I look at the beneficiaries who have an IRDP loan and who admitted to paying a bribe to obtain the loan, the Scheduled For a farmer who already owns land and animals, it is extremely easy to make good use of an additional buffalo. For an asset less household, the same animal can be a Damocles’ sword. If the buffalo falls ill and cash is not available to take special care of it, or if it grows weak during periods on non-lactation when fodder also has to be bought, the household can easily be ruined(see also Seabright (1989)).” 4 The IRDP loan is a package: it consists of a cheap loan and a capital subsidy. I discuss this in more detail in a later section

4

Caste(SC)5 pay a significantly higher bribe; about INR 600-7006 more depending on the specification when the average bribe in the data is about INR 1700. Then I exploit an exogenous source of change in the political influence of the SC to figure the effect of this change on bribe paid. Comparing villages belonging to Gram Panchayats (GP)7 reserved for SC pradhans 8 to villages in unreserved GPs, I find that SC households residing in the former pay lower bribes. With the position of the pradhan comes more patronage and influence, thus increasing the bargaining power of (at least some) SC households, which results in more surplus for them and lower bribes for the officials. The efficacy of this reservation policy and its effects on the welfare of SCs have been widely questioned, both in academics and the popular press(need references). So how do I find this big a change?9 I provide a novel explanation for this. Village pradhans and GPs do not operate in isolation, they interact with elected officials and bureaucrats at the block and district levels. If the only changes that occurred in the whole political setup were at the village level then we may despair of finding any discernible changes. A broader change however, with people more responsive to the needs of the SC at higher levels of government might however give enough patronage to the SC pradhans for us to find measurable differences. Following this line of thought, I exploit the fact that at the time these reservations were introduced the Bahujan Samaj Party(BSP), a political party representing the interests of SC, was on the ascent in Uttar Pradesh(UP) the state from which the data comes from. Using the vote share of BSP in a district as a measure of the political mobilization of BSP in that district: I find, that in reserved villages which lie in a district with higher vote share the average bribe amount paid by SC households is significantly lower than in a reserved village which lies in a district with lower BSP vote share. 5

In this data set I identify the lower castes as those belonging to the Scheduled Caste group In 1997 exchange rates 1 USD=37 INR 7 Literally translates into Village council 8 Pradhan in the vernacular means head or chief; in this context the position refers to the village head person. 9 I should make it clear here that I am not claiming wide distributional benefits from this change. I can only claim that on an average SC households in reserved villages are paying lower bribes. This effect could be driven by a handful of households, those related to the pradhan for example. 6

5

1.1

Related literature

Subsidized credit programs have been widely studied both in policy circles(see Buttari (1995)) and in academic circles(Besley (1994); Morduch (2000).In the Indian context the IRDP has been the subject of a huge literature dealing with the efficacy of poverty alleviation schemes(Pulley (1989); Dreze (1990); Swaminathan (1990); Seabright (1989, 1991); Dantwala (1990)). The lessons from a general evaluation of subsidized credit programs are similar to those already mentioned about the IRDP. A problem identified in most of these papers and which is also the focus of this paper is that of mis allocation of loans and poor targeting of subsidies. There seems to be a general consensus that the biggest beneficiaries of subsidized credit are the wealthier farmers, not the poorer ones. This paper takes a similar stand but by looking at the caste of the agents, which is an inherited component of social status(along with wealth, caste still determines a persons status specially in rural India). In this respect this paper ties in closely with the vast literature dealing with caste inequalities in India (Deshpande (2000, 2001)). This paper also relates to a burgeoning literature talking about how the age old traditions of caste shape the current economic realities in India(see Munshi and Rosenzweig (2006, 2009); Luke and Munshi (2007)). In terms of methodology this paper is close to others in the literature which use the nationwide reservation policy introduced in 1993 as an exogenous change, to identify effects on variables of interest10 . Two strands of literature come closest to this paper. The first is one that looks at caste affiliation and provision of public goods. Using data from South India Besley et al. (2004) find that for low spillover public goods SC households benefit when they live in a GP with SC pradhan. Bardhan and Mookherjee (2006) using data from West Bengal find that in intra village allocation of public goods lower castes and poor are not discriminated against, but

10

The seminal paper is of course Chattopadhyay and Duflo (2004) which looks at the effects of reservation for women.See also Duflo (2005); Besley et al. (2004)

6

in inter village allocation villages with a higher proportion of poor and lower castes receive less public goods. Secondly there are papers which look at outcomes of the rise of ethnic politics in India(Banerjee and Pande (2007); Munshi and Rosenzweig (2008)). Using survey data from Uttar Pradesh (UP), Banerjee and Pande (2007) find that over the last two decades as ethnic voting has increased so has the corruption level of elected candidates. In this paper I provide indirect evidence of the strategy adopted by corrupt BSP politicians of diverting resources towards their constituency, the SCs. The rest of the paper is organized as follows: Section 2 and 3 discuss the data and the setting of the study, Section 3 develops a simple model to formalize ideas, Section 5 provides empirical results and discussion and Section 6 concludes.

2

Data and Setting The data used in the paper come from the two primary sources. The household data is

taken from the Living Standards Measurement Survey(henceforth LSMS) carried out with the assistance of the World Bank in India in 1997-98. The villages of this study are located in South and South-Eastern Uttar Pradesh (henceforth UP) and North and Central Bihar11 . UP and Bihar, together with Madhya Pradesh,and Rajasthan have often been referred to as India’s “poverty belt”12 . UP and Bihar are characterized by unusually large populations with per-capita expenditure levels far below the poverty line. Eastern and Southern UP, from where the study villages were drawn, is generally poorer than the Western part of the state, and poverty levels have been rising in recent years13 . Bihar, which lies just east of UP, has the lowest per capita rural income in India, and is the most rural state in the country. Both these states have suffered from unrest, inter-caste conflict, and political violence in 11

In Up the data covers districts in Eastern UP and what is popularly called Bundelkhand These states were(are?) referred to as the BIMARU states (BIhar,MAdhya Pradesh,Rajasthan,Uttar pradesh) to designate the ‘hungry belly’ of ‘sick India’(Bose (1988); Dubey (1992)). Bimar is a Hindi word meaning ‘sick’. 13 see Jeffery and Lerche, eds (2003) for a comprehensive coverage of current issues related to UP 12

7

recent times. Overall poverty levels are even higher in Bihar than in Uttar Pradesh, and highest in the Northern region. The focus of this study is the state of UP. The state has done very poorly in terms of social development and this is closely related to the entrenched class and caste structure in the state. The upper class and upper caste elite have been uncompromising in not easing their dominant political position and accommodating emerging peasant and lower caste movements (Hasan (1998)). Violent defense of caste, class and gender privilege is commonplace in the state, as are clear political fault lines along class and caste. State spending on public health and education is limited, and spending on the poorest social groups even more so. Instead, rent seeking by people in power, their followers and the bureaucrats they promote is the norm as conflicts between different elites have spilled over into the bureaucracy (Jeffery and Lerche, eds (2003)). The result is an inefficient bureaucracy, where personal enrichment is more important than actual completion of tasks. This contributes significantly to the low performance of the state in sectors such as health and education (Dreze and Gazdar (1996)). The field survey was administered in villages drawn at random from 12 districts in UP and 13 districts in Bihar. A total of 120 villages, with an overall sample size of 2250 households, were sampled; 57 villages in Bihar and 63 in UP. Although small, mostly household-based industries such as wood gathering, bidi making, rope making and liquor brewing exist, the economies in these areas are primarily dependent on agriculture. About 70% of the HHs listed agriculture as a major souce of income. Majority of these households derive their income from two or more activities. The data from UP (which has 63 villages in the sample) has information on the caste and gender of the pradhan and whether the pradhan belongs to a Scheduled caste. Of the 63 pradhans, 16 are from the Scheduled caste and 18 are women. In 3 villages the pradhan is a woman and belongs to the Scheduled caste. I do not have information on these variables for Bihar, so most of the results that I present later are for the state of UP. Table 2 on page 29 gives the caste composition of the sample (only for UP). In most of the estimation results

8

I present later I do not present the results for the middle castes and the Muslim HHs. I only present results for the four major castes. The comparison group are always the Higher castes. The Backward Agricultural Castes (BAC), comprising mostly of Yadavs, a backward peasant caste, are a politically mobilized group and perform as well as the Upper castes on most parameters. In the following analysis I use this information coupled with election data14 that I obtained from the Election Commission of India15 .

2.1

The Integrated Rural Development Program

The IRDP program was started as a pilot program in selected districts in 1978 and from 1980 onwards it covered all blocks in the country. The essence of the program is to provide a capital subsidy and complementary credit at below market interest rates to HHs below the official poverty line to finance productive investment in income generating assets. The loans are disbursed through a “lead” commercial bank selected for each district. Banks and block officials tend to limit the choice of investment to milch animals, bullock carts, pump sets, retails shops and other micro-enterprises. Moreover, most loans are not distributed in cash but in kind. There is a ceiling on the HH income and landholdings for a HH to be eligible for a loan. To achieve better targeting of loans at least 30% of the beneficiaries have to be SC/ST and 30% women. Selection of beneficiaries is entrusted to block level staff who are instructed to survey HHs, prepare a list of qualified beneficiaries and submit the list to the GP for approval (Pulley (1989)). The final decision on the loans is made by the bank. Interest rates are fixed at 10% and repayment periods, minimum financing and type of investment are also predetermined by the National Bank for Agriculture and Rural Development(NABARD). NABARD provides an automatic refinance at 6.5% for all IRDP loans. At loan approval bank credit is matched by the government capital subsidy which varies from 25% to 50% 14 15

Election results and vote share of different political parties for various election years www.eci.nic.in accessed on October 15, 2010

9

depending on the socio-economic status of the HH. As mentioned previously it is often the case that the highest subsidies go to the richest HHs

16

.

Overdue s and default in loans are substantial. Studies mention default rates of up to 68%. In fact “Banks perceived priority sector lending as the social cost of doing business and seldom exerted themselves for recovery”. The problem is compunded by populist politicians. Loan forgiveness is one of the most common policy promises regularly made by politicans around elections (Cole (2009)). There is rampant corruption in the implementation of the scheme. “Consideration money” (Pulley (1989)) is regularly paid by loan beneficiaries to bank officials to get them to consider their applications favourably. Below I provide some summary statistics about IRDP loans and their beneficiaries from my data. 10.14% of the HHs in the sample of 2250 households are IRDP loan beneficiaries. Out of these beneficiaries about 39% belong to the Scheduled Castes(SC). The complete breakdown of these HHs by caste is given in the Table below. The average IRDP loan amount is Rs. 9728.31, whereas the average bribe paid to obtain the loans is Rs. 1735.29. On an average the bribes as a percentage of the loan amount obtained are about 24%. 185 out of 227 HHs who had obtained loans admitted to having paid a bribe. Table 1: Caste breakdown of Loan Beneficiaries High Caste BAC Backward Other SC

Number of HHs 27 55 25 88

16

Percent of all beneficiaries 11.89 24.23 11.01 38.77

For example Pulley (1989) finds that in UP small farmers,the best off occupational group received an average subsidy of INR 1640 while casual landless labourers received only INR 1425

10

3

The 1994 amendment and the rise of the BSP

3.1

The 1994 constitutional amendment

Gram Panchayats(henceforth GP) are the lowest tier in the three-tiered organization of local self government in India17 . GPs are popularly elected village councils consisting of a Pradhan “elected by persons registered in the electoral rolls for the territorial constituencies of the Panchayat area from among themselves,” and nine to fifteen members varying according to the population of the panchayat18 (In UP every village or a group of villages with a population of 1000 or more is constituted into a panchayat. So there is a large number of GPs in the state, more than 54,000)19 . The passage of the 73rd and 74th constitutional amendments in 1994 provided urban and rural local governments with a constitutional status they previously lacked, and strengthened it by mandating regular elections to these bodies. Further seats in these newly reorganized local governments were reserved for oppressed groups. According to Section 5(a) of the UP Panchayat Raj Act 1947 (amended in 1994),seats in the Gram Panchayat at all levels are reserved for Scheduled Castes, Scheduled Tribes and Other Backward Classes in proportion to their respective population in the Panchayat area, subject to a ceiling of twenty seven percent of total seat for the Other Backward Classes. Of the seats reserved for Scheduled Castes, Scheduled Tribes and Other Backward Classes, not less than one-third of the total number of seats are to be reserved for women belonging to these respective groups. Furthermore, in terms of section 5(c), not less than one-third of the total number of seats in the Gram Panchayat, including these reserved for women belonging to Scheduled Castes, Scheduled Tribes and Other Backward Classes, are reserved for women. All the reserved seats are allotted by rotation to different territorial constituencies in a Gram Panchayat. In addition to that there is a mandatory reservation of the position of Pradhan at all levels for SC,ST 17

The three tiers being the districts, blocks and villages Quoted from http://gov.ua.nic.in/sfc/sfcenglish/ANNEXURE%204.2.htm 19 (Dhar and Gupta, 2003, Chapter 2 p.32) 18

11

and Other Backward classes in proportion to their population in the state (see Chaudhari (2003))20 . This was an important piece of legislation in the decentralization of the process of development. The impact of this regulation was two fold. Firstly, it infused an age old tradition of self government with a new importance in the political structure. The regulation devolved functional responsibilities on GPs as well as regularizing the election process. The panchayats were made responsible for the implementation of development and social justice schemes at the village level. Responsibility of selection of beneficiaries for credit, housing and employment schemes, provision and location of public goods in the village were all given to the GP. This made them an important piece in local politics. Secondly, by reserving seats for SC/ST it shook up the existing power structure within villages. The fact that a SC or ST had to occupy the position of the pradhan gave these castes a bargaining chip against the established elites who were usually from the upper castes. The elite had to accommodate these new castes, maybe not to the fullest extent but as I demonstrate it did had some effect on the outcomes of the SC.

3.2

The rise of the Bahujan Samaj Party and the Samajwadi Party

Bahujan Samaj Party21 :Established in 1984 by Mr. Kanshi Ram the BSP is a political party exclusively representing the interests of the SC (especially the Chamars a subcaste)22 . The genesis of the party lay in trade union style low caste based movements. This meant that the initial leadership of the party came from educated, middle class, salaried SC many of whom had taken advantage of the affirmative action policies adopted since independence in India. Initially a bit player on the electoral scene the party made rapid gains in the 1990s using an “us against them” strategy. The campaign slogans used by the party reflect 20

This paper does not specifically deal with the UP Panchayat Raj Act. Instead it analyzes the Act of the Central government that forced the states to amend their respective panchayat raj acts (Panchayats are a subject on the state list). 21 Literally translated Party of the majority 22 This has been changing in recent times, see Chandra (2004).

12

this:Brahmins, Thakurs and Banias are thieves, the rest belong to the oppressed group and 85% ruled by 15%, this won’t last. Figure 1 shows the rapid rise of BSP in the 1990’s. Under the leadership of it’s curent leader Ms. Mayawati BSP was part of coalition governments in 1995, 1997 and 2003, and in 2007 it formed the government on it’s own majority. The following quote taken from the BSP website amply demonstrates it’s ideology The ideology of the Bahujan Samaj Party (BSP) is “Social Transformation and Economic Emancipation” of the “Bahujan Samaj”, which comprises of the Scheduled Castes (SCs), the Scheduled Tribes (STs), the Other Backward Classes (OBCs) and Religious Minorities such as Sikhs, Muslims, Christians, Parsis and Buddhists and account for over 85 per cent of the country’s total population. The people belonging to all these classes have been the victims of the “Manuwadi” system in the country for thousands of years, under which they have been vanquished, trampled upon and forced to languish in all spheres of life. In other words, these people were deprived even of all those human rights, which had been secured for the upper caste Hindus under the age-old “Manuwadi Social System”. (Source:www.bspindia.org accessed on March 6, 2010) Whenever the BSP was in power it undertook development projects targeted at the SC23 , and money was diverted from other projects towards those targeted at the SC. It massively expanded the already existing Ambedkar Village Project(AVP), a program for the upliftment of villages with a substantial SC population. The AVP, in short, during the period when the BSP was in power, provided extra funds to the Dalits(Scheduled Caste) in our sample villages which they could utilize due to special reservations provided in the new panchayats. Samajwadi Party: Established in 1991 by Mulayam Singh Yadav, is a party representing the interests of the yadavs a Backward Agricultural but nevertheless landowning peasant caste. The yadavs were one of the biggest beneficiaries of the land reforms which transferred land from absentee landlords to the tenants actually farming the land. Thus even though both these parties represent low caste political movements their interests are antagonistic. The BSP represents marginal farmers or landless casual labour whereas the SP represents a 23

Note that this is not a statement about the efficacy of such investments but more a populist measure

13

relatively richer landowning group which usually employs the former. Thus the interests of these two parties are opposed. This can be used to provide a good robustness check for my results. Again figure 1 shows the increase in SP vote share over time. In the next section I provide the remaining empirical results and a discussion of the results.

4

Theoratical framework In this section I outline a very simple theoratical model to fix ideas the presented. There

are three economic agents in the model: the borrower, bank official and the politician. The timing and the structure of the game they play is as follows. The borrower has a project which requires funding. He approaches the official with a request for a loan of amount L. The official asks for a part of the loan amount as bribe b as a condition for approving the loan. At this stage the borrower has three options. He can end this transaction here, in which case the game ends and there is no credit transaction. Alternatively he can agree to the bribe amount proposed by the official, pay it and get the rest of the money. Finally the borrower can ask a politician to intervene on his behalf. This costs the borrower an amount c. The politician can successfully intervene with the official (that is use his influence over the official) with a probability p. If the politican is successful in using his influence over the official then the borrower does not have to pay a bribe, he gets the full loan amount and the corrupt official has to pay a fine F 24 . The game ends at this stage. This game can be thought of as a bargaining game with the dice heavily loaded in favour of the official. The borrower’s only bargaining chip here is the threat(and his ability to act on that threat) to get the politican involved. I assume that all agents are risk neutral, so their utilities are represented by their respective payoffs. The above game can be summarized by the payoffs to the agents after each of 24

This fine can be thought of as reduced form way to model reduction in the official’s lifetime earnings because of incurring the politician’s displeasure, because of a transfer to a less lucrative position for example.

14

the choices made by the by the borrower. Let the utilities of the agents be given by ub , uo , up repectively for the borrower, the official and the politician. If the borrower ends the transaction, then none of the involved parties get anything out of the transaction. So ub = uo = up = 025 . If the borrower agrees to pay the asked bribe amount then ub = L − b, uo = b, up = 0. The official’s payoff is the bribe he gets, the borrower’s is the remaining amount from the loan and since the politician isn’t involved at this stage his payoff I assume to be zero. The borrower refuses to pay the bribe and pays a cost c to get the politician involved on his behalf. This however does not mean that the politician will always be able to impose a penalty and get the official to sanction the loan. This happens with probability p. The borrower gets his loan sanctioned, does not pay a bribe and the offical has to pay a fine F . With a probability 1 − p however the politician cannot influence the official, and the borrower has the choice of paying the bribe and getting the loan or ending the transaction. Using this we can write the utilities of the agents as follows

ub = L − c − (1 − p)b uo = (1 − p)b − pF up = c

25

This is a rather stylized view of things. In reality if the subsidized loan was refused then the borrower might try his luck in the informal credit market. However, what matters is that the utility to the borrower be greater when borrowing from the government than when he borrows from the informal market. This is indeed a reasonable assumption for most transactions in the informal market with their really high interest rates.

15

4.1

Solution

We are looking for a subgame perfect solution to the above game. I assume that L − c − b > 0 so that the borrower always wants to finish the credit transaction rather than end it at an earlier stage. This implies that the borrower has to decide between paying the asked bribe or enagaging the politican. The borrower will engage the politician if his utility from approaching the politican is higher than his utility from paying the bribe outright, i.e if

L − c − (1 − p)b > L − b

(1)

or if b>

c p

The result is fairly intuitive, if the official asks for a very high bribe then it is optimal for the borrower to pay the cost to engage the politician and take the chance that he can successfully influence the official. For the official it is always the case that his utility is lower if the borrower does go to the politician, since

(1 − p)b − pF = b − p(b + F )
So in equilibrium he will choose the highest bribe he can without sending the borrower

16

to the politician. Hence

b∗ =

c p

(2)

Thus on the equilibrium path, the politican never gets involved. The borrower asks for a loan amount L, the official asks for a bribe b = pc . The borrower pays the bribe and obtains the loan. From the expression for equilibrium bribe, the following are obvious ∂b∗ ∂c

> 0, so the lower the cost of approaching the politician the lower is the bribe

paid in equilibrium. ∂b∗ ∂p

< 0, which implies that the lower the probability of intervention the higher is

the bribe paid. In particular as p → 0, b∗ → L − c, the highest amount the official can extract. The chief assumption of this paper is that caste networks are important in politics. It is easier(cheaper) to approach a politician in your caste network than it is to approach a politician who does not belong to your caste. That is cowncaste < cothercaste . The implication for the bribe amount is that b∗owncaste < b∗othercaste . This is my preferred explanation for my empirical results presented later in the paper. Traditionally the politicians and the bureaucracy were all upper or middle castes, but the recent rise of low caste parties means that there are more politicans from the lower castes. This means that the cost of approaching politicians has fallen for the low castes, thus reducing the equilibrium bribe for them.

5

Empirical Estimation In the traditional setting where the high castes still occupy most positions of power I

expect that the cost of approaching a politician c is different for different groups. It is lower for high castes than it is for SCs. Hence the bribe paid should also be different. In the next 17

paragraph I provide results for the following estimating equation 3

Bribeamountij = α0 + α1 casteij + α2 Xij + ωj + ij

(3)

Here the dependent variable is the bribe paid to obtain the loan. casteij are the dummies for the different caste groups, Xij are various HH level controls and ωj are village level controls. The results of the estimation are provided in table 5. Columns(1)-(3) show the results without HH level controls. In column (1) and (2) I use the full sample while column(3) only uses the data from UP. Also columns (2) and (3) drop those observations for which the bribe amount was bigger than the loan amount. The key result of these regressions is that the SC households pay significantly higher bribes when compared to higher castes and the BAC26 . The next two columns provide estimation results with a full set of controls. The results for the caste dummies are very similar to the specification without controls. SC households pay significantly higher bribes even after controlling for demographics, income and assets. Column (5) also contains loan amount as a control to account for the possible correlation between the amount of the bribe and the size of the loan. It is worth noting that like previous studies I too find that wealthier people get more of the subsidy. The coefficient on the Income and dummy of whether the HH owns a house are both negative and significant. HH with a higher income and a house pay lower bribes, hence keep more of the subsisy. The point estimates are also significant economically; depending on the specification SC pay INR 700-900 more in bribes to obtain similar loan amounts. Given that the average bribe amount is about INR 1750 this amounts to about 40-50% of the average amount. For the estimation strategy to be valid it must be that the selection of villages be random or based on observables. Failing this it could be that whatever unobservables are driving selection are also driving the different outcomes for SC borrowers in reserved villages. Table 4 presents the results of a comparision between reserved and unreserved villages. The villages 26

Statistically there is no difference between the Backward other castes and the SC

18

appear to be the same on all characteristics except the population of SC households. The proportion of SC households is significantly higher in villages with a SC pradhan. The selection of villages for reservation seems to have been motivated politically. To see why think of a low caste party in power at state level; now it is in the interest of this party to reserve seats for SC in villages where the SC are not in a numerical majority. This way it maximizes the number of seats on which it can get SC representatives, as all the reserved seats have SC pradhans plus on some unreserved seats SC candidates may win as well. The opposite is true for a party representing the interests of caste groups other than SC. This is not a problem for the estimation as the estimates conditional on SC proportion can be obtained.

27

. In the estimations that follow I control for this proportion. I think that this

is an important caveat when using this particular regulation regarding GP reservations to tease out potentially causal effects, and that it has not been focused on in earlier studies. Also given the particular social setting, it is unlikely that a low caste person would win the elections for Pradhan in the absence of reservations. The average term for a GP is 5 years; in the data all SC pradhans but one are in their first term (they have served less than 5 years in office). The average years in office for an SC pradhan is 3.02 whereas the average years in office for the other pradhans is 4.79. This seems to support the idea that these SC pradhans are in office only because of a state regulation, hence this political change can be considered exogenous to the village setting. Now arguably in the panchayats reserved for SC pradhans the SC borrowers have cheaper access to a politician or other person of influence than in the other villages. Since this reservation policy was exogenously introduced it can be used to estimate a causal effect of increased influence (which translates into more bargaining power) on the bribe amount paid. To do so I estimate equation 4

27

The nice babus at the Election commission accepted that not all was Kosher with the reservation of GPs and there was political intervention. Given that there are 54,000 odd GPs in the state the proportion of SC is the easiest proxy to condition on.

19

Bribeamountij = α0 + α1 dj + α2 Vj + ij

(4)

Here i, j index the HH and village respectively, dj is a dummy which is 1 if the village is reserved and Vj contains village level characteristics (most importantly the proportion of SC in the village). I expect α1 < 0 as in reserved villages, SC households should be paying a lower bribe on average. Next to look at the effect of interaction of reservations with the wider BSP movement I estimate equation 5

Bribeamountijk = α0 + α1 dj + α2 V Sk + α3 (dj XV Sk ) + α4 Vj + ijk

(5)

Here i, j, k respectively index the HH, village and district. As before, dj is a dummy which is 1 if the village is reserved and Vj contains village level characteristics. V Sk is the BSP vote share in the district in which the village j is located. I expect that α3 the coefficient on the interaction is negative. If a reserved village falls in district with a higher BSP vote share I expect that SC households in that village pay lower bribes on average than a village which falls in a district with a lower BSP vote share. Table 6 shows the results for equation 4. Columns(1),(2),(4) use only the SC sample and estimate an equation exactly like 4, whereas columns (5)-(6) use the pooled sample and include dummies for SC and an interaction for SC living in reserved villages. In columns (1),(2) and (4) the coefficient on the dummy for reserved village is negative and significant, even after controlling for the proportion of SC households in the village. In the pooled regression the coefficients are similar to those obtained with only the SC sample. This implies that on an average SC households living in reserved villages pay INR 640-1000 less in bribes than those living in unreserved villages. Given that from the first set of estimation results we know that on an average SC pay about 700-800 in bribes reservation has a substantial effect on bribes paid. In column (3) I present a result of a similar estimation but for the other

20

castes this time. There is no effect on the bribe amount. This tells us two things; firstly, caste ties matter, as only SC households gain from SC pradhans and secondly, that it is not the case that the new SC pradhans are in any way more efficient or effective at governance as the bribes are not reduced across the board. Next I present results for equation 5. Before that I need to take care of a potential omitted variable bias. To see why, take a look at the timing of the events. The panchayati raj regulation was introduced in 1994 and the elections were held in 1995. The data comes from 1997-98. Now the closest parliamentary elections (vote share data comes from the parliamentary elections, which I take as a proxy for political mobilization) were those held in 1996, which is after the GP elections. This means that the regulation could potentially have affected both the 1996 vote share and the bribe amount. I deal with this in two ways: • Use the 1991 vote share as a proxy for BSP’s political mobilization28 . • Using an instrumental variable strategy. I use 1991 and 1989 BSP vote shares as an IV for the 1991 vote share. Table 7 presents the results with the 1991 vote share as proxy. The first thing to note is that the coefficient on the reservation status dummy is insignificant. So all the effect that reservation for SC pradhans has comes from this interaction, there would be no effect on the bribe amount paid by SC households living in a district with zero BSP vote share. Also in all the specifications the coefficient on the interaction is negative and significant. The coefficient implies that moving from a reserved village located in a district with average BSP vote share to a reserved village located in a district with vote share one standard deviation away from the mean would change the bribe amount by about INR 240, a substantial number. Table 8 presents the results for the IV estimates. Column (1) provides uses only one instrument, the 1991 vote share. In column (2) I instrument with the 1991 and 1989 vote shares. The results are very similar. 28

This is reasonable since vote share has a lot of persistence. In fact the correlation between the 1991 and 1996 vote shares is 0.91

21

How should we understand this interaction? Thinking back to the model it appears to be the case that SC pradhans themselves have a very low value of p (they are not very effective in getting the officials to comply). However, when there is more SC politicians and leaders in the upper levels of district politics such as MLAs or MPs, the borrower can apply to them for patronage through the pradhan. The interaction is easier to understand as a combination of two channels. Elected offcials in India (legislators and parliametarians) and politicians of any sort in general have an unhealthy amount of interference in the bureaucracy and public service. Transfers of civil servants from one posting to another is often used by politicians as a means of getting their way (Banik (2001)). The main argument has been that rapid transfers have demoralised the IAS. Instead of posting ocers to underdeveloped areas on merit and proven ability, unwanted officers are sent on ’punishment’ assignments. When theseofficers are aware that they have been ’victimised’, it is dfficult for them to remain committed to their duties and become involved in the genuine welfare of the area. In such situations the officer has the option of aligning himself with a political patron who can get him a more stable and desirable posting. All this leads to a sharp deterioration in the standards of development administration and results in little appreciation of the citizen’s point of view. There is some evidence that BSP governments have resorted to mass transfers of civil servants to suit their political needs29 . Thus it is not unreasonable that when the BSP has more political presence in a district BSP politicians have more influence over the public officials. The BSP on it’s part is more responsive to SC pradhans because it 29

In June 2009, the Supreme Court of India issued notices to the Union and the BSP led UP government regarding mass transfers of IAS and IPS officers in the state after the poor performance of BSP in the state elections. The contention of the petioner, an IAS offcer was that the large-scale transfers of senior bureaucrats in the state, particularly the demotion of State Principal Secretary Dinesh Singh by Chief Minister Mayawati was purely politically motivated and violated the All India Services Act and the Supreme Courts judgment that senior bureaucrats must have a minimum tenure of two years and the transfer policy must conform to the laws laid down by the apex court in the matter of transfer. According to the petitioner, Ms Mayawati had centralised all powers of transfer in herself after the poll debacle she faced and resorted to large-scale politically motivated transfers of bureaucrats posted in the districts and constituencies from where the BSP candidates had lost. Source:http://www.lawyersclubindia.com/news/SC-issues-notice-to-UP-govt-on-transfer-ofbureaucrats/5033/ accessed on 20 April,2010

22

needs them and the GPs to target its rural constituency. So why did the BSP need the panchayats? As already discussed it used the reservations and installation of SC pradhans to target the SC voters in villages through schemes like the Ambdkar Village Project (AVP). It was part of it’s mobilization strategy to solidify it’s rural base of voters, and the presence of SC pradhans fit right into this strategy. Given this, it might be the case that SC pradhans can procure more resources for their villages from a BSP government. These resources, which under the BSP could be openly targeted towards the SC, would reduce the competition between SC households thus reducing the bribes. I provide some suggestive evidence for this in Table 9 where I estimate an equation similar to 6 but with the total IRDP allocation to a village as a dependent variable. I do find some evidence that reserved villages get more funds but it is weak. It does seem to be the case that reserved villages get more resources, but the interaction with the BSP vote share is insignificant. Below I provide some robustness checks for my results.

5.1

Alternative explanations and robustness checks

Using Samajwadi vote share: As already discussed the SP is another low caste (Yadavs, a peasant caste belonging to the Other Backward Caste group, form the base of the party) based political party that emerged in the 1990’s in response to the lack of representation of yadavs in established parties. This is in some ways very similar to the BSP, but the important difference is that while one party represents the landed peasantry (the SP) the other represents the landless workers (the BSP). Thus in some ways the parties’ interests are antagonistic to each other and the caste identification of the two groups to their respective parties is really strong. Now if it was the case that caste identification of parties to their constituency (my preferred explanation for the BSP-SC borrower interaction) wasn’t important but what mattered 23

was a change in the established political system (high caste led parties) and an increase in the political representation of low castes, then we should expect to find similar results if I use SP vote share instead of BSP vote share. In districts where the SP vote share is high, that is the new political movement has a stronger base, SC borrowers should on an average pay lower bribes in reserved villages. Table 10 provides the results of estimating equation 5 but with SP vote share instead of BSP vote share. There is absolutely no effect on the bribe amount paid by the SC. This shows it is the caste networks in politics that are important. Is the panchayat really needed? It might be the case that all SC households living in districts with a higher BSP vote share are better off, regardless of whether they live in reserved or unreserved villages. Then picking any village in a high vote share district and comparing it with a lower vote share district would give results similar to mine. To see this I look at the correlation between average bribe amount and BSP vote share. Figure 2 shows the results. There does not seem to be any significant correlation. So it is the interaction between the lowest level of government and the broader political movement that is important. This makes sense for without the presence of a middleman, the village pradhan, it would be prohibitively costly for a poor SC farmer living in a village to engage a district level politician on his behalf. The SC are more likely to default: On a more general level it is most likely true that the poor and lower castes are bigger risks for banks as borrowers, hence the need for subsidized and targeted credit schemes. Even so it may be the case that banks are more reluctant to lend to SC borrowers as they have a higher probability of default hence they have to pay a higher bribe to otain loans. This could explain the difference between the bribes observed. In the data the borrowers were asked if they had started repaying their IRDP loans. There is no other information on loan default. On the basis of responses to the above question I fail to find any pattern in loan default by caste. SC are no more or less likely to have started repaying than other caste groups. Still, even if a higher risk of default by the SC was the

24

reason for higher bribes for them it would not explain why the bribe amounts fall in reserved villages. The SC borrowers are very unlikely to become more productive in reserved villages in a matter of 2-3 years (the first GP elections after the reservations were held in 1995 and the data comes from 1997-98). The much more likely explanation is the one advanced in the paper, that the lower bribes are a result of incresed SC political presence.

References Banerjee,

A. V. and R. Pande, “Parochial Politics: Ethnic Preferences and Politi-

cian Corruption,” Working Paper Series Harvard University, John F. Kennedy School of Government, 2007. Banerjee, A. V., “Credit Markets and Economic Development,” September 2003. MIT Dept. of Economics Working Paper No. 02-17. Bardhan, P. and D. Mookherjee, “Pro-poor targeting and accountability of local governments in West Bengal,” Journal of Development Economics, 2006, 79 (2), 303–327. Besley, T., “How do market failures justify interventions in rural credit markets?,” The World Bank Research Observer, 1994, 9 (1), 27. , R. Pande, L. Rahman, and V. Rao, “The politics of public good provision: Evidence from indian local governments,” Journal of the European Economic Association, 2004, 2 (2-3), 416–426. Bose, A., From Population to People, Vol. 1, Delhi: BR Publishers, 1988. Buttari, J., “Subsidized Credit Programs: The Theory, the Record, the Alternatives,” USAID Evaluation Special Study, 1995, 75. Chandra, K., Why ethnic parties succeed, Cambridge University Press, 2004.

25

Chattopadhyay, R. and E. Duflo, “Women as policy makers: Evidence from a randomized policy experiment in India,” Econometrica, 2004, pp. 1409–1443. Chaudhari, S., “What difference does a Constitutional Amendment make? The 1994 Panchayati Raj Act and the Attempt to Revitalize Rural Local Government in India,” 2003. Working paper Columbia University. Chaudhuri, S and M.R Gupta, “Delayed formal credit, bribing and the informal credit market in agriculture: A theoretical analysis,” Journal of Development Economics, December 1996, 51 (2), 433–449. and

, “Formal Credit, Corruption and the Informal Credit Market in Agriculture: A

Theoretical Analysis,” Economica, May 1997, 64 (254), 331–343. Cole, S., “Fixing market failures or fixing elections? Agricultural credit in India,” American Economic Journal: Applied Economics, 2009, 1 (1), 219–250. Dantwala, M.L., “Problems in Critique of IRDP,” Economic and Political Weekly, 1990, 25 (51), 2806–2806. Deshpande, A., “Does Caste Still Define Disparity? A Look at Inequality in Kerala, India,” in “The American Economic Review,” Vol. 90 Papers and Proceedings of the One Hundred Twelfth Annual Meeting of the American Economic Association May 2000, pp. 322–325. , “Caste at Birth? Redefining Disparity in India,” Review of Development Economics, 2001, 5 (1), 130–144. Dhar, T.N and S.P Gupta, Promises to keep: Panchayati Raj Institutions in Uttar Pradesh, Lucknow: Society for Himalayan Environmental Rehabilitation and Peoples Action in collaboration with Indian Institute of Public Administration, 2003. Dreze, J., “Poverty in India and the IRDP delusion,” Economic and Political Weekly, 1990, 25 (39), 95–104. 26

and H. Gazdar, “Uttar Pradesh: The burden of inertia,” Indian development: Selected regional perspectives, 1996, pp. 33–128. , P. Lanjouw, and N. Sharma, “Credit in Rural India: A case study,” LSE STICERD Reserch Paper No. DERPS06, September 1997. Dubey, K.N., Processes of Socio-Economic Development, Jaipur: Rawat Publishers, 1992. Duflo, E., “Why Political Reservations?,” Journal of the European Economic Association, 2005, 3 (2-3), 668–678. Ghosh, P., D. Mookherjee, and D. Ray, “Credit Rationing in Developing Countries: An Overview of the Theory,” in “Readings in the Theory of Economic Development” 2001, chapter 11. Hasan, Z., Quest for power: Oppositional movements and post-Congress politics in Uttar Pradesh, Oxford University Press, 1998. Jeffery, R. and J. Lerche, eds, Social and political change in Uttar Pradesh: European perspectives, Manohar Pubns, 2003. Luke, N. and K. Munshi, “Social affiliation and the demand for health services: Caste and child health in South India,” Journal of development economics, 2007, 83 (2), 256–279. Morduch, J., “The Microfinance Schism,” World Development, 2000, 28 (4), 617 – 629. Munshi, K. and M. Rosenzweig, “Traditional institutions meet the modern world: Caste, gender, and schooling choice in a globalizing economy,” American Economic Review, 2006, 96 (4), 1225–1252. and

, “The efficacy of parochial politics: caste, commitment, and competence in Indian

local governments,” NBER Working Paper, 2008.

27

and

, “Why is mobility in India so low? Social insurance, inequality, and growth,”

NBER Working Paper, 2009. Pulley, R. V., “Making the Poor Cerditworthy A Case Study of the Integrated Rural Development Program in India,” World Bank Discussion Papers, 1989, (58). Saha, B. and T. Thampy, “Extractive bribe and default in subsidized credit programs,” Journal of Economic Behavior & Organization, 2006, 60 (2), 182–204. Seabright, P., “Failure of livestock investments under IRDP: evidence from two villages in Tamil Nadu,” Economic and Political Weekly, 1989, 24 (39), 2203–2208. , “Quality of livestock assets under selective credit schemes:: Evidence from South Indian data,” Journal of Development Economics, 1991, 37 (1-2), 327–350. Swaminathan, M., “Village level implementation of IRDP: comparison of West Bengal and Tamil Nadu,” Economic and Political Weekly, 1990, 25 (13), 17–27.

28

Figure 1: Vote share in sample districts weighted by valid votes cast

Table 2: Caste composition of sample

High Caste Middle caste BAC Backward Other SC Muslim Upper Muslim Lower

Number of HHs 157 19 365 237 352 33 53

29

Percent of Sample 12.91 1.56 30.02 19.49 28.95 2.71 4.36

Figure 2: Bribe amount vs. district vote share

30

Table 3: Summary statistics Variable Panel A: HH characterstics

Mean

Std. Dev.

N

HH size ln(Income per HH member Total Land owned (in acres) Irrigated Land (in acres) Education Own a pump (Yes=1,No=0) Num of daughters Medical emergency during RP (Yes=1,No=0) Is the HH a lender? (Yes=1,No=0) Outstanding Debt Panel B: Village characterstics

6.381 5.845

3.477 0.49

1217 1214

3.193

5.705

1036

2.289

3.976

988

4.61 0.221

2.581 0.415

1217 1022

0.832 0.636

1.1 0.481

1217 1210

0.11

0.313

1217

3520.429

10060.627

1190

0.242 0.017 0.058 0.792 0.125 39.586

0.43 0.129 0.235 0.408 0.332 25.88

120 120 120 120 120 116

0.529 257.283 0.274

0.501 163.680 0.208

119 120 120

Telephone Police station Bank Public primary school Public health center Percentage of HHs with off farm employment Electricity No. of HH in village Proportion of SC

31

Table 4: Summary statistics Variable Reserved Unreserved No. of HH in village 242.875 228.17 Proportion of SC 0.426 0.246 Avg. income in village(Rs.) 2519.78 2308.443 Avg. land holding(acres) 3.094 3.083 Percentage of HHs 51.4 37.261 with off farm employment Telephone 0.313 0.128 Police station 0 0.021 Bank 0 0.043 Public primary school 0.75 0.68 Public health center 0.125 0.043 Electricity 0.563 0.617 p-value in parentheses

32

Difference 14.407(0.734) 0.180(0.002) 211.337(0.315) 0.012(0.986) 14.14(0.095) 0.185(0.095) -0.021(0.564) -0.042(0.41) 0.069(0.61) 0.082(0.25) -0.054(0.706)

33

N R squared

constant

Loan amount

Pucca dwelling

Own a house

Income per HH member Livestock owned

Irrigated Land

Total Land

HH size

Education

Sex

Age

SC

Back Other

BAC

188 .039

182 .033

1241.177 (278.551)∗∗∗

1283.333 (266.203)∗∗∗

716.490 (331.009)∗∗

770.148

(412.151)

(324.788)∗∗

270.728

(403.763)

(356.065)

(346.400)

228.571

Bribe amount (Bribe/Loan) < 1 223.171

105 .154

(358.241)∗∗∗

1444.444

(419.771)∗∗

849.423

(571.735)

-39.899

(446.025)

Bribe amount UP and (Bribe/Loan) < 1 11.956

6725.796 139 .164

(2671.355)∗∗

139 .187

(2658.947)∗∗∗

7072.635

(.014)

.023

6.977 (164.852)

44.431 (165.076)

-683.609 (296.282)∗∗

(389.518)

(397.718)

-623.730

372.412

387.434

(290.796)∗∗

-871.863 (379.977)∗∗

-814.405 (386.183)∗∗

.429 (.892)

(.830)

(45.840)

1.108

13.048 (41.689)

(36.697)

-6.994

-8.850 (37.708)

(72.894)

-5.883

-71.864

(72.679)

(687.902)

(675.618)

-66.813

-316.908

-152.283

1.381 (11.312)

-.279

(381.839)∗∗

919.393

(494.275)

-.359

(310.231)

Bribe amount Full sample 141.314

(11.741)

(387.972)∗∗

771.779

(492.825)

-127.686

(306.188)

Bribe amount Full Sample -23.605

Table 5: Dep. Variable: Bribe amount paid to obtain IRDP loan

Bribe amount Full sample 181.014

34

N R squared

Constant

Pucca dwelling

Livestock owned

Income per HH member Total Land

Education

Proportion of SC

SC X SC pradhan

Dummy SC pradhan (Yes=1, No=0) SC

57 .036

57 .049

2192.364 (739.937)∗∗∗

2662.500

54 .008

(450.791)∗∗∗

1992.091

(1206.655)

(380.400)∗∗∗

-759.991

(1719.353)

(640.926)

(474.949)∗

(431.334)

1298.805

Bribe amount Other Castes (3) 17.655

Bribe amount SC (2) -814.730

47 .149

(3935.307)∗∗

8781.159

111 .164

(327.950)∗∗∗

94 .301

(2490.625)∗∗∗

7328.693

29.188 (186.005)

-5.216

(332.778)

(165.406)

-85.951

(21.350)

(92.085)

(309.581)

29.909

104.162 136.136

-907.091 (452.691)∗∗

(681.208)

(101.150)

-1087.168

-119.217

(916.585)

(178.040)

1500.000

(490.038)∗

737.833

-936.894

(424.478)

(569.760)

-642.935

899.477

(462.513)∗∗

Bribe amount SC (6)

1162.500

Bribe amount SC (5)

-128.157

(1755.963)

1250.232

(505.327)∗∗

Bribe amount SC (4) -1016.071

Table 6: Bribe amount paid by SC on reservation status of village Bribe amount SC (1) -642.935

Table 7: Interaction of reservation status with BSP vote share in 1991

Dummy SC pradhan (Yes=1, No=0) Dummy X Vote share of BSP proportion of SC Constant N R squared

Bribe amount

Bribe amount (2) -290.323

Bribe amount Bribe <1 Loan (3) -237.848

Bribe amount Bribe < 0.6 Loan (4) 126.180

(1) -158.893 (487.966)

(492.273)

(534.310)

(375.819)

-30.436

-33.815

-35.316

-32.831

(15.156)∗∗

(17.968)∗

(21.049)∗

(18.450)∗

1399.822

2324.302

1142.526

(1755.205)

(1501.913)

(989.881)

2662.500

2155.798

1648.766

1824.044

(383.906)∗∗∗

(755.267)∗∗∗

(521.772)∗∗∗

(474.931)∗∗∗

57 .045

57 .06

53 .084

50 .045

Table 8: Interaction of reservation status with BSP vote share in 1996 Dummy SC pradhan Dummy SC pradhan X BSP vote share 1996 Proportion of SC Constant Hansen J statistic N R squared

Bribe amount (1) 213.969

Bribe amount (2) 269.338

(635.459)

(734.209)

-3990.265

-4205.039

(2149.894)∗

(2370.957)∗

1355.531

1358.584

(1665.229)

(1660.505)

2171.830

2170.725

(716.859)∗∗∗

(715.436)∗∗∗

57 .062

p value=0.125 57 .062

Table 9: Dependant variable: Total IRDP allocation Dummy SC pradhan

Total IRDP (1) 25169.540

Total IRDP (2) 23024.900

(12430.740)∗∗

(14562.850)

Dummy SC pradhan X BSP vote share 1991 No.of HH

Total IRDP (3) 40363.740

Total IRDP (4) 39947.760

(28183.310)

(29205.770)

-1248.356

-1309.633

(1483.130)

(1422.499)

51.765 (35.740)

Proportion constant N R squared

1674.487

7524.679

(28975.860)

(24567.430)

18615.430

6025.215

18615.430

16720.470

(3719.120)∗∗∗

(7229.879)

(3720.649)∗∗∗

(6282.528)∗∗∗

1219 .115

1219 .159

1219 .152

1219 .154

35

Table 10: Interaction of reservation status with Samajwadi Party vote share in 1996 Dummy SC Pradhan (Yes=1, No=0) Dummy X Vote share of SP proportion of SC Constant N R squared

Bribe amount (1) -518.006

Bribe amount (2) -704.379

Bribe amount (3) -669.044

Bribe amount (4) -221.457

(508.767)

(547.923)

(567.315)

(402.630)

-9.166

-7.766

-9.493

-11.988

(14.037)

(14.554)

(17.588)

(17.272)

1264.818

2164.856

1054.978

(1735.056)

(1494.694)

(958.726)

2662.500

2204.666

1706.785

1854.960

(383.906)∗∗∗

(746.036)∗∗∗

(517.285)∗∗∗

(465.817)∗∗∗

57 .039

57 .051

53 .073

50 .033

36