expected, given <strong>the</strong> random selection <strong>of</strong> GPs, <strong>the</strong>re are no important differences between reservedand unreserved GPs. 9 Note that very few villages (3% among <strong>the</strong> unreserved GPs) have tap water,<strong>the</strong> most common sources <strong>of</strong> dr<strong>in</strong>k<strong>in</strong>g water be<strong>in</strong>g hand-pumps and tube wells. Most villages areaccessible only by a dirt road. N<strong>in</strong>ety-one percent <strong>of</strong> villages have a primary school, but very fewhave any o<strong>the</strong>r type <strong>of</strong> school. Irrigation is important; 43% <strong>of</strong> <strong>the</strong> cultivated land is irrigated, withat least some land be<strong>in</strong>g irrigated <strong>in</strong> all villages. Very few villages (8%) have any public healthfacilities.We collected <strong>the</strong> data <strong>in</strong> two stages. First, we conducted an <strong>in</strong>terview with <strong>the</strong> GP Pradhan.We asked each one a set <strong>of</strong> questions about his or her family background, education, previouspolitical experience, and political ambitions, as well as a set <strong>of</strong> questions about <strong>the</strong> activities <strong>of</strong> <strong>the</strong>GP s<strong>in</strong>ce his or her election <strong>in</strong> May 1998 (with support from written records). We <strong>the</strong>n completeda survey <strong>of</strong> three villages <strong>in</strong> <strong>the</strong> GP: Two villages randomly selected <strong>in</strong> each GP, as well as <strong>the</strong>village <strong>in</strong> which <strong>the</strong> GP Pradhan resides. Dur<strong>in</strong>g <strong>the</strong> village <strong>in</strong>terview, we drew a resource map <strong>of</strong><strong>the</strong> village with a group <strong>of</strong> 10 to 20 villagers. <strong>The</strong> map featured all <strong>the</strong> available <strong>in</strong>frastructure <strong>in</strong><strong>the</strong> village, and we asked whe<strong>the</strong>r each <strong>of</strong> <strong>the</strong> available equipment items had been built or repaireds<strong>in</strong>ce May, 1998. Previous experience <strong>of</strong> one <strong>of</strong> <strong>the</strong> authors, as well as experimentation dur<strong>in</strong>g<strong>the</strong> pre-test<strong>in</strong>g period, suggested that this method yields extremely accurate <strong>in</strong><strong>for</strong>mation about<strong>the</strong> village. We <strong>the</strong>n conducted an additional <strong>in</strong>terview with <strong>the</strong> most active participants <strong>of</strong> <strong>the</strong>mapp<strong>in</strong>g exercise, <strong>in</strong> which we asked <strong>in</strong> more detail about <strong>in</strong>vestments <strong>in</strong> various public goods. Wealso collected <strong>the</strong> m<strong>in</strong>utes <strong>of</strong> village meet<strong>in</strong>gs, and asked whe<strong>the</strong>r women and men <strong>of</strong> <strong>the</strong> village hadexpressed compla<strong>in</strong>ts or requests to <strong>the</strong> GP <strong>in</strong> <strong>the</strong> previous six months. For all outcomes <strong>for</strong> whichit was possible, we collected <strong>the</strong> same <strong>in</strong><strong>for</strong>mation at <strong>the</strong> GP-level and at <strong>the</strong> village-level. <strong>The</strong>village-level <strong>in</strong><strong>for</strong>mation is likely to be more reliable, because it is not provided by <strong>the</strong> Pradhan,and because it was easy <strong>for</strong> villagers to recall <strong>in</strong>vestments made <strong>in</strong> <strong>the</strong>ir village <strong>in</strong> <strong>the</strong> previous twoyears. However, <strong>the</strong> <strong>in</strong><strong>for</strong>mation given by <strong>the</strong> GP head refers to <strong>in</strong>vestment <strong>in</strong> <strong>the</strong> entire GP, and isthus free from sampl<strong>in</strong>g error. <strong>The</strong>re<strong>for</strong>e, when an outcome is available at both levels, we per<strong>for</strong>m<strong>the</strong> analysis separately <strong>for</strong> both and compare <strong>the</strong> results.Between August and December <strong>of</strong> 2002, we collected <strong>the</strong> same village-level data (<strong>the</strong>re was noPradhan <strong>in</strong>terview) <strong>in</strong> 100 hamlets <strong>in</strong> Udaipur, <strong>Raj</strong>asthan, chosen randomly <strong>in</strong> a subset <strong>of</strong> villagescoveredbyalocalNGO,SevaMandir. 10 <strong>The</strong> reference period <strong>for</strong> ask<strong>in</strong>g about <strong>in</strong>vestment wasalso two years, 2000-2002. In <strong>Raj</strong>asthan, <strong>the</strong>re was no regularly elected Panchayat system until1995. Table 2 displays <strong>the</strong> characteristics <strong>of</strong> villages reserved <strong>for</strong> women and unreserved <strong>in</strong> our9 <strong>The</strong> standard errors are omitted from <strong>the</strong> table <strong>for</strong> clarity, but none <strong>of</strong> <strong>the</strong> differences are significant at <strong>the</strong> 95%confidence level.10 <strong>Raj</strong>asthani villages are much more spread out than West Bengali villages (a <strong>Raj</strong>asthani village covers an area onaverage 10 times bigger than a West Bengali village) and are much less densely populated. <strong>The</strong>y are made <strong>of</strong> a series<strong>of</strong> <strong>in</strong>dependent “hamlets”, which are not adm<strong>in</strong>istrative entities but function as <strong>in</strong>dependent villages. Our sampl<strong>in</strong>gunit is <strong>the</strong> hamlet: We first sampled 100 villages (with probability <strong>of</strong> selection weighted by village size) and <strong>the</strong>n onehamlet per village (aga<strong>in</strong>, <strong>the</strong> probability <strong>of</strong> selection was weighted by village size).6
sample. 11 Udaipur is a much poorer district than Birbhum. It is located <strong>in</strong> an extremely aridarea with little irrigation and has male and female literacy rates <strong>of</strong> 27.5% and 5.5% respectively.Because <strong>the</strong> villages are bigger, <strong>the</strong>y are more likely to have a middle school, a health facility anda road connection, compared to villages <strong>in</strong> West Bengal. As <strong>in</strong> <strong>Raj</strong>asthan, we see no differencebetween <strong>the</strong> characteristics <strong>of</strong> reserved and unreserved villages be<strong>for</strong>e <strong>the</strong> reservation policy wasimplemented.4 F<strong>in</strong>d<strong>in</strong>gsIn <strong>the</strong> absence <strong>of</strong> reservation, if we found that different public goods <strong>in</strong>vestments are undertaken<strong>in</strong> GPs that elect women and SCs, compared to GPs that do not, it would be very difficult to<strong>in</strong>terpret <strong>the</strong>se results, s<strong>in</strong>ce <strong>the</strong> few places that elect women or SCs are presumably very differentfrom places that do not elect women. For example, places that are dom<strong>in</strong>ated by SCs may elect aSC Pradhan and also <strong>in</strong>vest more <strong>in</strong> <strong>the</strong> SC hamlet, but this would not imply that if we constra<strong>in</strong>random places to elect a SC Pradhans, <strong>the</strong> Pradhan will have any power to implement <strong>the</strong> policyhe chooses. By contrast, <strong>in</strong> this case, because <strong>the</strong> reserved seats were randomly assigned, we cannow compare <strong>the</strong> outcome <strong>in</strong> GPs where <strong>the</strong> position <strong>of</strong> Pradhan is reserved <strong>for</strong> a woman or a SCto those where it is not reserved, and be confident that any difference reflects only <strong>the</strong> impact <strong>of</strong><strong>the</strong> reservation policy.4.1 <strong>Reservation</strong> <strong>for</strong> Women4.1.1 Effects on <strong>the</strong> Political Participation <strong>of</strong> WomenTable 3 displays <strong>the</strong> effect <strong>of</strong> hav<strong>in</strong>g a woman Pradhan on <strong>the</strong> political participation <strong>of</strong> women.Columns (1) and (2) display <strong>the</strong> average participation rates <strong>in</strong> reserved and unreserved GPs, respectively.Column (3) displays <strong>the</strong> difference. Differences that are significant at <strong>the</strong> 95% confidence<strong>in</strong>terval, us<strong>in</strong>g standard significance tests, are <strong>in</strong> bold. In West Bengal, <strong>the</strong> percentage <strong>of</strong> womenamong participants <strong>in</strong> <strong>the</strong> Gram Samsad is significantly higher when <strong>the</strong> Pradhan is a woman(<strong>in</strong>creas<strong>in</strong>g from 6.9% to 9.9%). S<strong>in</strong>ce reservation does not affect <strong>the</strong> percentage <strong>of</strong> eligible votersattend<strong>in</strong>g <strong>the</strong> Gram Samsad, this corresponds to a net <strong>in</strong>crease <strong>in</strong> <strong>the</strong> participation <strong>of</strong> women, anda decl<strong>in</strong>e <strong>in</strong> <strong>the</strong> participation <strong>of</strong> men. This is consistent with <strong>the</strong> idea that political communicationis <strong>in</strong>fluenced by <strong>the</strong> fact that citizens and leaders are <strong>of</strong> <strong>the</strong> same sex. Women <strong>in</strong> villages withreserved Pradhans are twice as likely to have addressed a request or a compla<strong>in</strong>t to <strong>the</strong> GP Pradhan<strong>in</strong> <strong>the</strong> previous six months, and this difference is significant. 12 <strong>The</strong>factthat<strong>the</strong>Pradhanisawoman <strong>the</strong>re<strong>for</strong>e significantly <strong>in</strong>creases <strong>the</strong> <strong>in</strong>volvement <strong>of</strong> women <strong>in</strong> <strong>the</strong> affairs <strong>of</strong> <strong>the</strong> GP <strong>in</strong> WestBengal.11 For Udaipur, we could not obta<strong>in</strong> <strong>the</strong> data necessary to match villages to Panchayat <strong>in</strong> <strong>the</strong> entire district.12 In <strong>the</strong> subsample <strong>of</strong> villages <strong>in</strong> which we conducted follow-up surveys, we also asked whe<strong>the</strong>r men had broughtup any issue <strong>in</strong> <strong>the</strong> previous six months. In all cases but one (a reserved GP), <strong>the</strong>y had.7