Don t touch my road.

Don t touch my road. How a privatized public good may become public again: evidence from Indian political reservations. Preliminary and Incomplete VICTOIRE GIRARD CES - Universite Paris 1 Sorbonne June 2014 Abstract Public goods may be privatized, in the sense of being turned into club goods in some contexts. Caste based discrimination is a striking example where public goods are privatized. In the hindi-belt, the heart-land of India, 44.5% households members of the marginalized castes labeled Scheduled Castes (SC) declare some streets to be off-limits due to their caste in the 2006 survey used here. This exclusion rate has decreased in the previous decade as 65% SC declared to have been excluded in 1996. What have lead to this drop, and how can it be pushed further? Political representation appears to be an option. This article takes advantage of the Indian system of seats reservation for low caste members in local assemblies as it leads to exogenous variations in political outcomes. The effect of political reservation for low caste access to streets is large and significant (both statistically and economically) but does not last after teh end of the reserved term, and does not extend to the private sphere. So reservations change behaviors in the public sphere but norms and stereotypes seem harder to affect. Keywords: Roads, Public Goods, India, Inequality, Caste, Political Reservation, Discrimination JEL D63, D74, J15, O12, O53 Contact information: Victoire Girard, Maison des Sciences Economiques, 106-112 Boulevard de l Hopital, 75013 Paris. e-mail: victoire.girard@univ-paris1.fr. I thank Andrew Foster for providing the data. 1

1 Introduction 44.5% households members of the marginalized scheduled castes, henceforth SCs, declare some streets to be off-limits due to their caste in States of the "Hindi belt". 1 This observation is in line with other surveys showing that SCs are still victims of widespread discrimination, in particular in public good access. In a 2001-2002 survey by (Shah, Mander, Thorat, Deshpande, and Baviskar, 2006), households considered untouchables could not access public roads in 18% of the 565 surveyed villages (spread in 11 major states). Exclusions magnitude varies with the type of public good denied: from a high 50% of exclusion for water facilities to a low 18% for primary health center and roads, while police stations are at an intermediary 28%. More recently, (Davenport and Armstrong, 2010) report access to the village water facility to be forbidden to SCs in 29% of the 1589 surveyed Gujarati villages. These rules are not leftovers from old habits that would mechanicaly fade away. News reports abound on how SC households may be punished for not following these rules. And Bros and Couttenier (2011) show atrocities against SCs to be correlated to the type of water source used, where sources like open wells trigger more atrocities than sources less prone to ritual pollution. Yet the Indian Constitution of 1949 abolised untouchability, and public goods are a cornerstone of development. So how can public good access be improved for marginalized communities? This is the question this article is interested in. In particular, I am interested in the political representation channel. Seats are reserved for SCs and other low castes in the Indian local political assemblies. Do these reservations help increasing marginalized constituents access to public goods? The answer is an unambiguous yes. But only in the short term. Reservations allow to properly identify the impact of political representation on SC households. This article is an extension of the existing literature on the impact of political reservation and quotas. Quotas are observed worldwide as an attempt to compensate for, and eventually eradicate, a discriminatory legacy. Quotas for women political representation are the most widespread. In India, these quotas helped to increase political representation for women even in unreserved seats, through a change in womens likelihood to run for the election and voters attitudes toward women in general (Beaman, Chattopadhyay, Duflo, Pande, and Topalova, 2009). Related to the question of public good access, where the public good at stake is security, women quotas have also lead to a tremendous increase in crimes reported against women due to an increased reporting of these crimes rather than to an increase in the 1 States where Hindi is widely spoken, namely in this study sample: Bihar, Haryana, Madhya Pradesh, Rajasthan and Uttar Pradesh. 2

actual incidence of violences (Iyer, Mani, Mishra, and Topalova, 2011). Lastly, women leaders fund public goods more in line with female constituent preferences than male leaders do (Chattopadhyay and Duflo, 2004). Be it through quotas or political representation in contested elections, constituent actions may change as a result of political representation (Beaman, Chattopadhyay, Duflo, Pande, and Topalova, 2009; Bhalotra, Clots-Figueras, and Iyer, 2013) Regarding caste-based quotas, the research has mainly focused on the impact of politicians identity on their political decisions: are the type and location of financed public goods different according to the leader s caste? The answer is yes. In particular, elected representative provide more high spillover goods in places close from their residence, while more low spillover goods are targeted toward households of their caste group (Besley, Pande, Rahman, and Rao, 2004). Fund allocation differences according to leaders identities are enacted. But the impact of political representation can also go through exposure to this new leader, because the leader has a role-model for her group members (Beaman et al., 2009), or because increased interactions between two groups are expected to reduce stereotypes between groups (following Allport (1954) seminal work). Focusing on caste-based reservations, Chauchard (2010) argues for a positive impact of political reservation on low caste constituents lives, due to the emergence of new norms of between caste interactions (after reservation) even if stereotypes may not be affected. This article adds to this literature on the impact of political reservation and how it affects constituents. Rather than studying funds allocation to different public goods, the question is whether reservation can improve access to already-existing public goods. To my knowlege, the question has not yet been addressed in the literature. Yet reservations are controversial, assessing their consequences for constituents is thus essential. Access to public goods in particular does not require to divert money from some projects to put it in others, supposed to be favored by SC households. It is a matter of overcoming traditional restrictions to the free use of pre-existing public goods. In Pande (2003) words, "at the Independence, the Indian State committed to use public policy to end caste-based discrimination, and to improve the economic status of disadvantaged groups". Quotas are a centerpiece of this commitment and their impact on street access encompass both of the two objectives cited by Pande (2003). First, having certain streets in a village declared off-limits to certain castes is a blatant case of caste-based discrimination. Second, streets are supposed to be a canonical public goods and freedom to move is a pre-requisite of economic empowerment. The combination of both objectives makes road access a great candidate to assess reservation consequences. 3

Next section presents the caste and reservation systems while section presents the data. The estimation strategy and results are in sections and. Section concludes. 2 The Indian Context 2.1 Castes Castes have carved the Indian social setting for more than 3,000 years. Three key features of castes are important to remind for this study. First, castes are hereditary, exclusive and virtually unchangeable at the household level. 2 Second, castes are ordinated on a social status ladder, that matches a purity ladder, where so-called ritual pollution may happen between people of different purity status if some rules to prevent this pollution are not followed. 3 Third, and closely linked to the second aspect, caste groups are segregated: this has lead to forms of spatial segregation, and to historically very strict matrimonial segregation. These theoretical features translate into preferences or rules affecting everyday life interactions. Up until now, castes structure business networks (Munshi, 2011), and severely curve mates selection (Banerjee, Duflo, Ghatak, and Lafortune, 2013), while revealing subjects caste in an experiment significantly affects their performances (Hoff and Pandey, 2006). Castes are recorded in the Census through four broad groups, Scheduled Castes (SC), Scheduled Tribes (ST), Other Backwrd Castes (OBC) and Other Castes (OC). Scheduled Castes households, who belong to a list of particularly backward castes, that used to be considered untouchables, still suffer from caste based discrimination. This translate in a variety of ways, including exclusion from market, or systematic deviation from market price at their disadvantage, being sold/rented out products at higher prices, and bought/rented in products at lower prices than non SC households (Thorat, Mahamanlik and Sadana s survey results in Thorat and Newman, 2010). Spatial segregation is also observed, as in a traditional Indian village SCs live in a separate neighborhood or hamlet (Deliège, 2004). Lastly, SCs suffer from widespread exclusion. Exclusion can be formal, through clear cut access denial as outlined in the introduction, or more pernicious, through a differential treatment, which has been widely documented 2 Even if some cases of identity manipulation have been recorded, these are marginal, and correspond to group level strategies, to increase a caste status in the society, or on the contrary benefit from some positive action policies (Deliège, 2004; Cassan, 2013). 3 Ritual pollution is believed to happen for example if a member of a high caste performs ritually dirty actions like scavenging or shares some intimacy (direct touching or sharing food or drinks) with a member of a low caste. In case of failure to comply by the rule, being polluted has spiritual consequences, but time or de-polluting activities usually allow to be get one s purity back 4

in the case of schooling. 4 Beside the moral aspect of the question, universal access to public goods is central for SCs economic empowerment, and thereby to India s achievements in poverty reduction. 2.2 Reservations In 1993, the 73rd amendment to the Constitution of India has institutionalized local level political councils called Gram Panchayat. These councils have decision power on the fixing and building of local public goods such as roads or water devices. They also decide which households are entitled to social programs. The 1993 reform is also important because it puts quotas in place as a tool of positive discrimination. Heads seats in some village councils are reserved to low caste members and/or to women. Gender quotas are randomly imposed in a third of all constituencies. Regarding caste-based quotas, low caste who enjoy reservations are SCs, STs, and in some States OBCs. The proportion of seats reserved for each caste in each State is proportional to the number of that caste in that State population. Which villages are concerned by the council head (called Pradhan) seat reservation is a decision made by the administration at the State level. Rules change from one Sate to the other. Some State allocate reserved seats in a purely random way, others rank villages according to the proportion of their population being SC ST and OBC, and rotate from the most populated to the least populated villages. In the sample, out of the 178 terms considered (two election terms in 89 villages), 43 SC Pradhans are elected, and 63% of these SC Pradhans have been elected in a reserved seat. Villages where SC are elected in the absence of a SC quota are likely to have very specific features, as political reservations have precisely been implemented because of SC under-representation. What makes these villages so special? Descriptive statistics show that out of 19 terms where a SC Pradhan is elected on a seat unreserved for a SC, 15 are elected on seats reserved for a women, and 3 are elected after a reservation for SC took place, leaving one observation where a SC has been elected out of any reservation treatment. The main driver of SC election out of SC quotas is thus gender quotas. Gender quotas seem to change villages political landscapes: table 1 shows that Pradhans caste distribution in women-reserved seats converges toward the population caste-distribution. table 1 statistics are computed for the 139 terms where neither a SC nor a ST nor a OBC seat reservation was imposed. Outside gender-quotas, OC (the high castes) win twice more Pradhan seats than their share in the population, while SC and ST are poorly represented. The contrast with caste distribution on women reserved seats is striking. Caste distribution on women reserved seats closely mirrors the population caste composition. Some women may be elected 4 Low caste students can be ignored or physically punished by teachers (PROBE report, 1999), or receive less good grade for similar answers (Hanna and Linden, 2012). 5

only because their husband could not run for office, 5 but another parallel dynamic is at play, where seat reservation for women imposes to reconsider who can be legitimate to run for office, and low cast women, in particular SC, appear to have taken advantage of this situation. 6 Table 1: Caste composition of elected Pradhan and population for terms without any caste-based quota. unreserved seat reserved woman (not sc) population share SC 8.3 22.4 19.3 ST 0 4.5 6.6 OBC 45.8 46.3 50.8 OC 45.8 26.8 22.7 total 100 100 100 Villages where a low caste Pradhan has purposefully been elected, be it after a gender-quota, are likely to differ from others on several respects, and in particular in the way low castes are treated in these villages. I thus leave villages with SC Pradhan elected on unreserved seats in the control group, and focus the analysis on reserved seats.indeed, the reservation timing can be endogenous as a village level outcome, but the interaction term of the reservation status with households SC identity is exogenous (this identification strategy is inspired by Besley et al., 2004). As for ST reservation, the sample is too small to allow a proper study: only three villages display both variations in ST street access declarations and reservation status. 7 The study thus focuses on SC quotas, implemented in 24 of the 89 villages surveyed. SC reservation is observed only once in each village. 5 This idea is widely spread, but the evidence show that women election on reserved seats does change things for women, be it through priorities re-orientation, or attitudes change (Chattopadhyay and Duflo, 2004; Beaman et al., 2009; Iyer et al., 2011). 6 This mirrors(buch, 2013, p. 52) affirmation after a field research in 1,200 Panchayats of Madhya Pradesh, Rajasthan and Uttar Pradesh that Women who entered new panchayat showed a wider participation across social and economic class/caste. Their occupation, income levels and household assets indicated presence of hiterto unrepresented sections. 7 At least one ST is present in 47 of the 89 villages of the study, but only 5 villages have experienced reservation of the Pradhan seat to a ST over one of the two terms, and in two of these villages no ST ever declare any street exclusion. Also, four other villages have had an ST reservation in both terms but the impact is not time variant in that case so it will be captured by village effects. 6

3 Data 3.1 Dataset Data are extracted from the last round of the Rural Economic and Development Survey (REDS), undertaken in 2006 by the National Council for Applied Economic Research (NCAER). It contains both a survey at the village level and a complete census of the households of the surveyed villages. The dataset contains information about 242 villages of 17 major Indian States, while the households census covers 115,000 households. This article is focused on the Hindi belt States. The Hindi belt is the heartland of India, which encompasses on average poorer States and display more tensed inter-caste relationships than southern India. In the surveyed villages, the 1996 exclusion figure for SCs in Hindi Belt States is 65% against 12% in non Hindi Belt States. Politics is also more caste-oriented in these States (?).The area is thus particularly prone to studying how caste based political reservation affect caste based exclusion. The final sample encompasses 48,219 households, spread in 89 villages. 3.2 Street Exclusions The household census contains information on whether certain streets of the village are off-limit for respondents due to their caste or religious identities (with one question for caste-based exclusion and another for religion-based exclusion). Information are collected both for the survey moment and 10 years before (recall question). Descriptive statistics on the extent of street exclusions and how they vary with group identities in table 2 show that SC are, as expected, the ones who declare most street exclusions. ST appear to declare less street exclusions although their social status is comparable to the SCs status. While ST are only 7% to suffer from street exclusion in 2006 the corresponding figure for SC is 44,5%. But SC and ST display very different settlement patterns. ST more often than SC live in places where they belong to the majority or are a very small minority: 60% of the STs live in villages where STs represent less than 10% or more than 50% of the population, while 10% of the SCs live in villages where SCs are less than 10 or more than 50% of the population. 7

Table 2: Street Access and group shares in the whole sample caste group population share excluded 1996 excluded 2006 nb hh gained access between 96 and 06 SC 20,0% 65,0% 44,5% 2076 ST 6,5% 31,4% 7,2% 779 OBC 50,8% 5,5% 4,0% 520 OC 22,4% 1,6% 1,3% 62 4 Estimation Strategy The aim is to assess whether reservation of the Pradhan seat for a SC household can change SC households access to streets. A first specification, in cross section is: Street_access iv = α 1 + α 2 Reserv_SC v SC i + α 3 + X i + δ v + ε iv The coefficient of interest is α 2, telling whether low caste households access to streets is affected by the fact that their Pradhan is low caste too. X i accounts for household level controls with the household caste category, its exclusion from streets in 1996, the head s sex, education level (above or below primary), age, income and occupation with a dummy for whether the primary income source comes from agriculture. δ v are village level fixed effects to account for the endogeneity of the reservation timing. The error term, ε iv is clustered at the level of caste categories within villages. A second specification, taking advantage of the data panel structure is: Street_access ivt = β 1 + β 2 Reserv_SC vt SC i + δ vt + δ gt + δ i + η ivt Here δ i are household fixed effects to account for time invariant unobservables at the household level while δ vt accounts for possible changes over time at the village level, which may be correlated to the reservation status of the village. Following Besley et al. (2004), this strategy allows to identify the impact of reservation through within village variations. Changes over time at the caste group level, which are independent of village level reservations, are additionally accounted for through δ gt, which stands for general amelioration trends at the group level. The error term η ivt is two way clustered, to account for shocks at the household level and at the caste group level within each village and year (as this is the scale of the interaction term of interest). The first specification gives an idea of the broad picture. However, if village-year fixed effects are 8

not accounted for, the effect of reservation for villages reserved in 2006 can be mixed with other features of the village, as the reservation timing is not exogenous in every State, and street access has risen over time. Hence the interest of the second specification. 5 Results 5.1 Reservations ans Street Access Reservation raises SC access to streets. In 2006 the effect offsets SCs initial disadvantage (column 1, table 3). If one moves to the panel specification the effect magnitude is the same. For SC households, having a SC Pradhan makes street exclusions 9% less likely (column 2, table 3). This is huge compared to the observed 20% point decrease in SC street exclusion declarations. 9

Table 3: (1) (2) (3) (4) (5) caste exclu caste exclu caste exclu caste exclu religion exclu SC 0.0884*** (0.0168) res_sc -0.00845 (0.0849) SC*res_SC -0.0868*** -0.0901*** -0.140*** -0.0891*** (0.0325) (0.0230) (0.0259) (0.0187) SC*pradh_SC 0.0544** (0.0237) SC*res_OBC 0.0313 (0.0200) Mus*res_SC -0.0332 (0.0360) Observations 40,047 79,420 79,420 79,154 79,420 village fix effects yes no no no no household controls yes no no no no time*village fix effects no yes yes yes yes time*caste fix effects no yes yes yes yes time*religion fix effects no no no no yes household fix effects no yes yes yes yes R-squared 0.58 0.183 0.184 0.183 0.045 Robust standard errors in parentheses. Errors are clustered by village caste groups in column 1, with controls included: household caste category, it s exclusion from streets in 1996, the head s sex, education level (above or below primary), age, income and occupation with a dummy for whether the primary income source comes from agriculture. Errors are two way clustered for time*village*caste and household levels in the rest of the table. *** p<0.01, ** p<0.05, * p<0.1. The previous estimate leaves in the control group 19 villages where a SC Pradhan has been elected out of a reserved seat. SC Pradhan election out of reservation are controlled for in column 3. The variable 10

has an unexpected significantly positive sign. SC Pradhans elected outside SC quotas seem to worsen SCs situation. However, please recall that 15 of these 19 Pradhans have been elected on seats reserved for women. And gender and caste effects may interact in unexpected ways (?). 8 In the sample, gender quotas in general do not worsen SCs situation as an interaction term between SC households and woman reservation is not significant (results for women reservations along side Pradhan caste and caste-based reservations are in appendix table 6). Insufficient variation in the sample does unfortunately not allow to use interaction terms to disentangle the effect of gender from that of caste when female Pradhan seats are won by a low caste outside a caste-quota. A blunt alternative is to omit from the sample the 74 terms where the Pradhan seat is reserved for a woman. Women reservations are supposed to take place in a randomly chosen third of each village sample. In this study sample, 43% terms are reserved for a women on average, but only 8% of the SC reserved terms. Results displayed in appendix table 6 should thus be interpreted with caution. All the more as SC elections, unlike SC reservations, are endogenous. One can still notice that once gender questions are taken away, elected SC Pradhan cease to impact SC households outcomes, although the coefficient magnitude is comparable. As for the impact of reservation, it is is virtually unchanged compared to the whole sample results in table 3. The gender of SC Pradhans elected outside SC quotas thus seems to play a part for SC Pradhans to have a significantly positive coefficient in table 3. Coming back to the general sample and table 3 results interpretation, the election backlash effect (having a low caste representative worsen low caste situations) can pertain to the fact that SC women elected on reserved seats are elected after a contest in opposition to other women, that may have been high caste women. This competition between caste is not observed for seats reserved for SC, where high castes have to accept that a SC is to be Pradhan. Please recall table 1: outside reserved seats, high castes win nearly half elections, although they represent a fifth of the electorate, while on women reserved seats, high castes win only a quarter of all elections. Women reservation may favor the emergence of new political personalities (Buch, 2013) or alliances among lower castes to increase their access to power (an hypothesis backed by the observation that OBC street access, just as SC street access, is negatively affected by the election of a SC on a women reserved seat in appendix table 6). Woman Pradhan, even if low caste, may also prioritize gender-based requests over caste-based requests. 9 The rest of the article 8 The situation studied in (?) mirros the one studied in this article, as she shows that the gender effect is different for low caste women (elected on SC SCT reserved seats) than for women eleced on unreserved seats. Her interpretation relies on the idea that gender and class effects can be contradictory. 9 All the more when elected on a women reserved seat. Indeed, in the lab, insisting on Asiatic women ethnicity vresus gender identities affect their math and verbal performances in opposite directions, each time consistent with stereotypes (Shih et al., 2006). So in the field, being elected on a seat made accessible due to one s gender may strengthen gender identity. Additionally, 11

thus focus on SC election on SC reserved seats. Table 3 also shows that not all seat reservations are beneficial to SCs: reservations for OBCs has no impact on SCs access to streets (column 4, table 3). If they had one, it would be negative. 10 Lastly, I perform a falsification test to check how Muslims street exclusions, grounded in their religious identity, reacts to SC reservation. It should a priori not be the case, as Muslim exclusions are grounded in a Hindu/Muslim divide while reservations for low castes aim at reducing the low/high caste gap. Table 3 last column confirms that the logic at play with SC reservations is indeed only a caste logic. Although the percentage of Muslim households excluded from streets dropped from 27 to 16% over the period, Muslim households access to streets is not improved by reservation of the Pradhan s seat to a SC. SC households access to streets is thus improved by reservation of the Pradhan s seat to a SC. The effect is observed only for reservations for a fellow SC member, reservation to castes with an intermediate status has no effect, nor do SC Pradhans matter for religion based exclusions. 5.2 Does access improvement last through time and why? Several stories are consistent with previous section results. Street access for SC may be imposed by the Pradhan, or non SC personal beliefs on SC may be updated (as is found for women in Beaman et al., 2009), or the social norm among non SC priors on what is acceptable evolves away from segregation (and so does it among SC but this would lead to an increase reporting of segregation, not the lowering observed here Chauchard, 2010). These stories have different implications on the long run, after the reserved term is over, and on the private versus public spheres. If increased access is only due to enforcement, the effect should not last after reservation is over. Priors on the contrary, once they have evolved, are not supposed to evolve back to segregation after reservation is over. Non-SC priors evolution may also have different impacts. If deep beliefs on SC and segregation are affected, the effect should be the same in the public and in the private sphere. If instead it is only the social norm that has evolved, telling what is acceptable to show or say, while deep beliefs have remained unaffected, the evolution should have consequences primordially in the public sphere and not so much in the private sphere. Table 4 summarizes these three roads are a gendered concern in Chattopadhyay and Duflo (2004), although which gender is more preoccupied by roads varies in the two studied States. 10 This could be interpreted directly, if OBCs are willing to insist on their traditionally higher caste status than SCs and enforce untouchability practices more once elected. This can also be interpreted indirectly as having the Pradhan position reserved for and OBC prevents any SC to be elected and elected SC improve SC access to streets on average (even if the impact is lead by Pradhan on SC reserved seats). In line with the last interpretation suggested, if the interaction term for SC seats reservation is omitted, the OBC reservation remains at 0.05 but becomes highly significant (result not included). 12

channels and their expected consequences. Table 4: Why do reservation affect SC street access declarations? during ongoing term in public or or persistent? private spheres? Reduced prejudices both both Updated social norm both public>private Street access enforcement ongoing public>private Table 5 results allow to assess table 4 cases relevance. First, SC reservation appears to play a part only during the reserved term: having had the Pradhan seat reserved for a SC during the previous term does not have a significant impact on SC access to streets after the reserved term is over (column 1, table 5). Please note that year 1996 ongoing panchayats were the first elected panchayats after the 1993 Constitution amendment, so the variable disclosing previous reservation is always equal to zero in year 1996. The interpretation of the previous term reservation variable thus rests on Panchayat reservations in intermediary terms, between 1996 and 2006. That being said, if reservations had a long lasting impact, the variable should be significant. The absent impact of previous term reservation fits in the enforcement channel rather than in the prejudice or social norm channel. But how do these results extend to questions beyond roads? Information on caste based exclusion from temples is available for years 1996 and 2006. Column 2 in table 5 shows that SC access to temple raises with reservation for a SC Pradhan. Temple exclusion, just as street exclusion, drop by almost 20 percentage points over the period. Reservation appears to be a push factor in this amelioration as reservation of the Pradhan seat for an SC makes SC 6% less likely to be excluded form a temple. The effect is again observed only for ongoing reservations. What about private outcomes? The survey gives information on perceived job market discrimination, asking to household heads whether themselves or a family member have been discriminated against because of their caste or religion while seeking a job. The question is asked separately for the household head and it s family members, and for the survey moment en 10 years before. Job market discrimination drops by one to three percent over the period if one considers respondent discrimination or family member discrimination. But if anything, job market discrimination raises with reservations or is unaffected (columns 3 and 4 of table 5). This may be interpreted either as a backlash effect, if non SC report on the private sphere the resentment unexpressed in the public sphere, or as a social norm effect if SC become more sensitive to discrimination in general once a SC Pradhan is elected. However, the positive effect is observed only for ongoing reservations, at the same moment as the effect on streets or temple better 13

accessibility. This observation fits better in a backlash effect story than in a story on SC perceptions. Reservation of the Pradhan s seat to a SC makes SC less likely to be excluded from streets or temples, but does not affect, or with the opposite sign, labor market discrimination. So reservations may change behaviors, but this is true only in the short run, and the improvement is restricted to the public good sphere. These results fit better in an enforcement story than in stories of social norms or prior evolutions. Table 5: (1) (2) (3) (4) street temple head labor family labor exclusion exclusion discrimination discrimination SC*res_SC -0.0944*** -0.0566*** 0.0142* 0.00918 (0.0190) (0.0171) (0.00833) (0.00962) SC*previously_res_SC 0.00487-0.0185-0.0102 0.0142 (0.0378) (0.0404) (0.0164) (0.0145) Observations 79,384 79,420 79,420 79,418 village fix effects no no no no time*village fix effects yes yes yes yes time*caste fix effects yes yes yes yes household fix effects yes yes yes yes R-squared 0.183 0.173 0.012 0.031 Robust standard errors in parentheses two way clustered by time*village*caste and household levels. *** p<0.01, ** p<0.05, * p<0.1. 6 Conclusion Political reservations aim at improving some groups political representation, to improve their social status and their economic achievements. Access to roads combines both concerns. I find political reservations at the village level to be a powerful policy tool to improve low caste households access to public goods. But the impact appears to be restricted to ongoing reservations and can on the contrary lead to a backlash effect in the private sphere, where SC household heads appear to suffer from more job market discrimination during reserved terms. SC Pradhans can manage to improve 14

SC constituents situation in the public sphere for the time of their term. But it is due to harsh or gentle (a bargained) enforcement rather that deep priors (with low caste stigmatization) or social norms (what is acceptable to do in public) evolutions. In the end, reservation may help re-allocate public benefits while it lasts, with long term consequences. It also changes caste relations in the short run. But one shot reservation is not enough to update neither caste-based stereotypes nor social norms. These findings call for replication with data on more than three electoral terms as reservation repetition may change things. But if the impact of reservation is confirmed to be temporary after replications, it questions the legislator purpose to use positive action as a transitory tool, to be abandoned once practices and beliefs are updated. 15

7 Appendix Table 6: SC elected Prdhan and women reservations (1) (2) (3) (4) VARIABLES full sample no women reservations SCxres_SC -0.153*** -0.154*** -0.110*** -0.163*** (0.0259) (0.0255) (0.0281) (0.0234) SCxpradh_SC 0.0727*** 0.0743*** 0.0691 (0.0211) (0.0206) (0.0527) SCxres_woman 0.00962 (0.0293) STxres_SC 0.0762 0.0905 (0.0520) (0.0626) STxpradh_SC 0.0371 0.0250 (0.0472) (0.0520) OBCxres_SC -0.0192-0.0217 (0.0182) (0.0191) OBCxpradh_SC 0.0260* 0.0265* (0.0147) (0.0147) OBCxres_woman -0.00723 (0.00998) Observations 79,420 79,420 27,058 27,058 time*village fix effects yes yes yes yes time*caste fix effects yes yes yes yes household fix effects yes yes yes yes R-squared 0.185 0.185 0.235 0.235 16

References Allport, G. (1954). The nature of prejudice. Addison-Wesley Pub. Co. Banerjee, A., E. Duflo, M. Ghatak, and J. Lafortune (2013). Marry for what? caste and mate selection in modern india. American Economic Journal: Microeconomics 5(2), 33 72. Beaman, L., R. Chattopadhyay, E. Duflo, R. Pande, and P. Topalova (2009, November). Powerful women: Does exposure reduce bias? Quarterly Journal of Economics 124(4), 1497 1540. Besley, T., R. Pande, L. Rahman, and V. Rao (2004). The politics of public good provision: Evidence from indian local governments. Journal of the European Economic Association 2(2-3), 416 426. Bhalotra, S., I. Clots-Figueras, and L. Iyer (2013, November). Path-Breakers: How Does Women s Political Participation Respond to Electoral Success. Economics Discussion Papers 740, University of Essex, Department of Economics. Bros, C. and M. Couttenier (2011). Is blood thicker than water? untouchability and public infra-structure. Manuscript. Université de Paris, Centre de Sciences Humaines: Delhi. Buch, N. (2013). From Oppression to Assertion: Women and Panchayats in India. Taylor & Francis. Cassan, G. (2013). Identity based policies and identity manipulation: Evidence from colonial punjab. CEPREMAP Working Papers (1306). Chattopadhyay, R. and E. Duflo (2004). Women as policy makers: Evidence from a randomized policy experiment in india. Econometrica 72(5), 1409 1443. Chauchard, S. (2010). Can the experience of political power by a member of a stigmatized group change the nature of day-to-day interpersonal relations? evidence from rural india. Unpublished Manuscript. Davenport, C. and D. Armstrong (2010, Feb). Understanding Untouchability: A Comrehensive Study of Practices and conditions in 1589 villages (Navsarjan Trust (Ahmedabad, India), Robert F. Kennedy Memorial Center for Human Rights ed.). Deliège, R. (2004). Les castes en Inde aujourd hui (1st ed.). Paris: Presses universitaires de France. Hanna, R. N. and L. L. Linden (2012). Discrimination in grading. American Economic Journal: Economic Policy 4(4), 146 68. 17

Hoff, K. and P. Pandey (2006, May). Discrimination, social identity, and durable inequalities. American Economic Review 96(2), 206 211. Iyer, L., A. Mani, P. Mishra, and P. Topalova (2011). The power of political voice. Harvard Business School Working Paper. Munshi, K. (2011, March). Strength in numbers: Networks as a solution to occupational traps. The Review of Economic Studies 78(3), 1069 1101. Pande, R. (2003). Can mandated political representation increase policy influence for disadvantaged minorities? theory and evidence from india. American Economic Review 93(4), 1132 1151. PROBE report (1999). Public report on basic education in india. PROBE Team (India) and Delhi School of Economics. Centre for Development Economics. Shah, G., H. Mander, S. Thorat, S. Deshpande, and A. Baviskar (2006, August). Untouchability in Rural India. New Delhi ; Thousand Oaks, Calif: SAGE Publications Pvt. Ltd. Shih, M., T. L. Pittinsky, and A. Trahan (2006). Domain-specific effects of stereotypes on performance. Self and Identity 5(1), 1 14. Thorat, S. and K. S. Newman (2010). Blocked by caste : economic discrimination in modern India. New Delhi: Oxford University Press. 18