Can Mandated Political Representation Increase Policy Influence for Disadvantaged Minorities? Theory and Evidence from India By RoHiNi PANDE'" A basic premise of representative democracy is ihat all those subject to policy should have a voice in its making. However, policies enacted by electorally accountable governments often fail to refiect the interests of disadvantaged minor- ities. This paper exploits the institutional features of political reservation, as practiced in Indian states, to examine the role of mandated political representation in providing disadvantaged groups infiuence over policy-making. 1findthai political reservation has increased transfers to groups which benefit from the mandate. This finding also suggests that complete policy commitment may be absent in democra- cies, as is found in this case. (JEL D72, D78, HI 1. H50) There are strong moral and economic argu- ments suggesting that it is in the interest of society to improve the economic standing of historically disadvantaged minority groups. In democracies, the use of legislative policy to bring about such improvements remains contin- gent on legislator behavior, and arguably, a significant barrier to the introduction of such policies is the political underrepresentation of individuals belonging to minority groups who * Department of Economics, Yale University, P. 0, Box 208269, New Haven. CT 06520 (e-mail: rohini,pande@yale, edu). This paper is based on Chapter 2 of my Ph.D, ihesis (LSE 1999), I am indebted lo Tim Besley for help, advice and, above all, enthusiasm for the subject. I thank an anonymous refen;e, Esther Duflo, Dom Lcggetf, and Neel Mukheijee for very helpful :ind detailed comnienls, I have also benefited from cotnmenfs iuid discussions wifh Abhijit Banerjee, Robin Bur- gess, Sieve Coate. Lena Ediund. Maiiree.sh Ghauik. Torsfen Persson. Debraj Ray. Ken Shepsle. and numerous seminar participants. Financial siipptirt fR)m LSE-STICERD. Overseas Research Students Awaid. Royal Economic Society, and Win- gale Founciaiion is gratefully acknowledged. Any errors re- main mine. ' Hisforieally disadvanfaged grotips arc commonly de- fined as groups which have been systematically excluded from institulions imd culfural pracfices fhat provide skills and resources. An important moral argument for directing public policy at such groups is that historical discrimination againsi a group should nol be allowed to perpetuate itself and inhibit the groups' right to well-being. Moreover, sueh policies may enhance efficiency by improving the talent allocaiion across differeni occupations (Harry Holzer and David Neumark. 2000). i might vote in their own interest.^ Both sets of arguments are particularly compelling in the case of India, where the hierarchical caste sys- tem has contributed to the economic deprivation of those bom into lower castes. At indepen- dence, the Indian State committed to use puhlic policy to end caste-based discrimination, and to improve the economic status of disadvantaged groups. A centerpiece of this endeavor has been the implementation of the constitutional man- date which ensures the presence of legislators belonging to minority groups in state and na- tional legislatures. This paper examines the im- pact of this mandated political representation on policy outcomes in India at the state level. While many countries have experimented with mandates which seek to increase minority representation in the political process, the In- dian experiment remains, by far, the most rad- ical (Bernard Grofman and Arend Lijphart, 1986). Prior to every state election, specified jurisdictions are declared reserved for disadvan- taged castes and for tribes. Only members of the group which benefit from reservation can stand for election. However, the entire electorate votes over the set of candidates. The effect of the mandate is to alter legislator identity without ^ Cross-country evidence documents the fact that mem- bers of minority groups are less likely to get selected as candidates by parties, and are therefore underrepresented in the legislature (Wilma Rule and Joseph Zimmerman. 1994), 1132
Transcript
Can Mandated Political Representation Increase PolicyInfluence for Disadvantaged Minorities?
Theory and Evidence from India
By RoHiNi PANDE'"
A basic premise of representative democracy is ihat all those subject to policyshould have a voice in its making. However, policies enacted by electorallyaccountable governments often fail to refiect the interests of disadvantaged minor-ities. This paper exploits the institutional features of political reservation, aspracticed in Indian states, to examine the role of mandated political representationin providing disadvantaged groups infiuence over policy-making. 1 find thai politicalreservation has increased transfers to groups which benefit from the mandate. Thisfinding also suggests that complete policy commitment may be absent in democra-cies, as is found in this case. (JEL D72, D78, HI 1. H50)
There are strong moral and economic argu-ments suggesting that it is in the interest ofsociety to improve the economic standing ofhistorically disadvantaged minority groups. Indemocracies, the use of legislative policy tobring about such improvements remains contin-gent on legislator behavior, and arguably, asignificant barrier to the introduction of suchpolicies is the political underrepresentation ofindividuals belonging to minority groups who
* Department of Economics, Yale University, P. 0, Box208269, New Haven. CT 06520 (e-mail: rohini,pande@yale,edu). This paper is based on Chapter 2 of my Ph.D, ihesis (LSE1999), I am indebted lo Tim Besley for help, advice and, aboveall, enthusiasm for the subject. I thank an anonymous refen;e,Esther Duflo, Dom Lcggetf, and Neel Mukheijee for veryhelpful :ind detailed comnienls, I have also benefited fromcotnmenfs iuid discussions wifh Abhijit Banerjee, Robin Bur-gess, Sieve Coate. Lena Ediund. Maiiree.sh Ghauik. TorsfenPersson. Debraj Ray. Ken Shepsle. and numerous seminarparticipants. Financial siipptirt fR)m LSE-STICERD. OverseasResearch Students Awaid. Royal Economic Society, and Win-gale Founciaiion is gratefully acknowledged. Any errors re-main mine.
' Hisforieally disadvanfaged grotips arc commonly de-fined as groups which have been systematically excludedfrom institulions imd culfural pracfices fhat provide skillsand resources. An important moral argument for directingpublic policy at such groups is that historical discriminationagainsi a group should nol be allowed to perpetuate itselfand inhibit the groups' right to well-being. Moreover, suehpolicies may enhance efficiency by improving the talentallocaiion across differeni occupations (Harry Holzer andDavid Neumark. 2000). i
might vote in their own interest.^ Both sets ofarguments are particularly compelling in thecase of India, where the hierarchical caste sys-tem has contributed to the economic deprivationof those bom into lower castes. At indepen-dence, the Indian State committed to use puhlicpolicy to end caste-based discrimination, and toimprove the economic status of disadvantagedgroups. A centerpiece of this endeavor has beenthe implementation of the constitutional man-date which ensures the presence of legislatorsbelonging to minority groups in state and na-tional legislatures. This paper examines the im-pact of this mandated political representation onpolicy outcomes in India at the state level.
While many countries have experimentedwith mandates which seek to increase minorityrepresentation in the political process, the In-dian experiment remains, by far, the most rad-ical (Bernard Grofman and Arend Lijphart,1986). Prior to every state election, specifiedjurisdictions are declared reserved for disadvan-taged castes and for tribes. Only members of thegroup which benefit from reservation can standfor election. However, the entire electoratevotes over the set of candidates. The effect ofthe mandate is to alter legislator identity without
^ Cross-country evidence documents the fact that mem-bers of minority groups are less likely to get selected ascandidates by parties, and are therefore underrepresented inthe legislature (Wilma Rule and Joseph Zimmerman. 1994),
1132
VOL 93 NO. 4 PANDE: POLITICAL RESERVATION AND POLICY OUTCOMES 1133
affecting voter identity. The placing of require-ments on candidate identity in reserved jurisdic-tions directly increases the political representationafforded to minority groups in the legislature.Political reservation has had a profound effecton the Indian political landscape—a quarter ofall legislators in India, at both the national andstate level, come from reserved jurisdictions.
The use of mandates to enhance minoritypolitical representation is predicated on the as-sumption that legislative capture by nonminor-ity individuals adversely affects the policyinterests of minority groups, and that partiescannot fully control candidate behavior afterelections (for if they could, candidate identitywould be irrelevant to the policy process). Thisassumption is, however, invalid if a party's pre-ferred policy is independent of its candidates'identity, and parties and voters can ensure thatcandidate behavior after elections is guided bythe commitments they made beforehand—astandard assumption in many political economymodels (Anthony Downs, 1957). An analysis ofIhe impact of political reservation on policyprovides a direct empirical test of thisassumption.
In the first part of the paper, 1 develop asimple model of political competition to showthat the effectiveness of political reservation inaltering policy depends on the nature of thecontract between the electorate and the elected.Two important elements in this contract arewhether a party can commit its candidates topolicies and how legislators resolve policy dif-ferences within the legislature. In situationswhere candidate entry is mediated by politicalparties with policy preferences which are inde-pendent of their candidates identity, changes inlegislator identity brought about by reservationcan only affect policy in the absence of fullpolicy commitment. Moreover, such changesmay not be significant unless every legislatorhas voice in the policy-making process.
In the remainder of the paper I use an Indianstate-level panel data set to examine whetherpolitical reservation for scheduled castes andscheduled tribes in state elections has affectedpolicy-making. A state-level analysis is appro-priate as India is a federal democracy, withstates enjoying substantial independent policy-making powers (Jean Dreze and Amartya Sen,1995), Moreover, the choice of affirmative ac-tion policies in favor of scheduled castes and
.scheduled tribes has been an important elementof policy activism by stale govemments (SunitaParikh, 1997).
The Indian constitution requires that the ex-tent of state-level political reservation enjoyedby a group reflect the group's population sharein the state. But, the extent of political reserva-tion can be revised only when new census pop-ulation estimates are received. Thus, while agroup's population share varies continuously,the proportion of jurisdictions reserved for itchanges with a lag—that is, only at the point ofelection, and after fresh census population esti-mates for the group are received. 1 exploit thisinstitutional feature of reservation to isolate itsimpact on policy outcomes. I use changes in theextent of political reservation, which are spe-cific to a given state, to identify how changes inthe group shajes of minority legislators affectpolicy outcomes. Since the response of politicalreservation to population changes is character-ized by a lag I can separately control for thevariable which causes changes in reservation—the census population shares of the two groups.This allows me to disentangle the effects ofchanges in the political representation affordedto a group on policy from those due to changesin its population share.
The main finding is that political reservationin Indian states has increased redistribution ofresources in favor of the groups which benefitfrom political reservation. Such increases havebeen accompanied by increases in overallspending and decreases in spending on educa-tion programs. I interpret these findings as evi-dence that reservation can enhance a group'sinfiuence on policy-making, and that legislatorsbelonging to minority groups have used thisinfluence to increase the incidence of targetedredistribution. Whether these changes in gov-emment spending away from general redistri-bution programs toward targeted programsimprove the well-being of either the minoritygroups or the polity at large remains an openquestion.
In conjunction with the theoretical argumentspresented in this paper, the findings suggest thatcomplete policy commitment may be absent indemocracies, as is found in this case. Thesefindings are consonant with recent politicaleconomy papers which assume that existing po-litical institutions cannot enforce full policycommitment (see, for example, Martin Osborne
1134 THE AMERICAN ECONOMIC REVIEW SEPTEMBER 2003
and Al Slivinski, 1996; Timothy Besley andStephen Coate, 1997). Such models of policy-making predict that increases in political repre-sentation afforded to a group will enhance itsinfluence on policy.'
My findings are also in line with recent em-pirical papers on the relationship between poli-tician identity and policy outcomes. Joseph Kaltand Mark Zupan (1984) and Steven D. Levitt(1996) show that a candidate's personal "ideol-ogy" is a key determinant of observed policyoutcomes. Besley and Anne Case (2000) andRaghabendra Chattopadhyay and Esther Duflo(2001) document significant differences be-tween the policies favored by male and femalepoliticians. Evidence on how a country's choiceof political institutions mediates the relationshipbetween legislator identity and policy outcomesis, however, limited. My paper addresses thisquestion in the context of a specific institufion—political reservation.
The remainder of the paper is organized asfollows. Section I sets out a simple model ofpolitical competition which identifies how par-ties' ability to commit their candidates to poli-cies affects the impact of political reservationon policy. Section II describes the institution ofpolitical reservation, as practiced in India andthe data set used in the empirical analysis. Sec-tion III presents the empirical findings, and Sec-tion IV is the conclusion.
I. Political Reservation and the Political Process
In this section I use a theoretical example toillustrate the link between my empirical analy-sis and the literature on political competition inrepresentative democracies. The objective is toshow that, in a world where individuals chooserepresentatives to select and implement policyand where parties mediate candidate entry, thepossibility of policy commitment on the part ofcandidates affects the impact of political reser-vation on policy in an empirically identifiablemanner.
Consider a large population of /V individualswho differ in their eaming potential. Each indi-vidual supplies one unit of labor, and depending
^ Thomas Husted and Lawrence Kenny (1997) and LenaEdiund and Pande (20(12). among others, provide empiricalevidence that an individual's group identity is correlatedwith their policy preference.s.
TABLE 1—PRKJ'ERRED REDisriuBuirvE POLICY, BY CASTEAND INCOME;
Group Preferred redistributive policy
Rich high castePoor high casteRich low caste
Poor low caste
/ = 0 and no redistributioni = 1 and general redistribution/ = 1 and targeted redistribution
ify V
else t = 0 and no redistributioni = 1 and targeted redistribution
on her/his eaming potential either earns v' andis rich or earns y'' and is poor, where v'' > / ' .Individuals also differ with respect to an unal-terable attribute, which 1 call caste—an individ-ual is born either a high iH) or a low (L) caste.Let A* denote the population share of individu-als who belong to caste c G iH, L) and eamincome v*. where k E (r, p). I assume low-caste citizens constitute a population minority(A/ < '/:). In addition, they are more likely to
k<l Kbe poor, i.e., rr > TT •
At AH
Individual income, if taxed, is taxed at rate /.Taxes can be redistributed in up to two ways.First, via a general transfer T to all individualsand second, via a targeted transfer 8 to low-caste individuals (that is, 5 = 0 for high-casteindividuals). The former redistributes incomefrom rich to poor individuals, and the latterfrom high- to low-caste individuals, I assumethat the selected redistribution policy must bebudget balancing. Individual utility is increas-ing with own posttax income, and is denoted u^.Formally, H?, = (I - / ) / + Tand u^ = (1 -r)v* + T + 8. Table 1 describes an individu-al's preferred redistributive policy, by incomeand caste. The redistributive preferences of richlow-caste individuals vary with the demo-graphic makeup of the population. These indi-viduals favor no redistribution to targetedredistribution if, and only if, the populationshare of poor low-caste individuals exceeds
^ = / - / ' •
Individuals elect legislators to select and im-plement the levels of general and targeted re-distribution. For expositional clarity I assume
•• When A > A* thf targeted transfer a rich low-casteindividual receive.s is less than the fax she pays.
VOL 93 NO. 4 PANDE: POLITICAL RESERVATION AND POLICY OUTCOMES 1135
the population is divided into Z jurisdictions,with one legislator elected per jurisdiction.^ Po-litical competition is mediated by two politicalparties. The political process has three stages.At stage 1, the two parties choose one candidateper jurisdiction. At stage 2, individuals choosefor whom to vote. In the final stage, electedlegislators choose the type and extent of redis-tribution. These stages are described in reverseorder.
The elected legislators select the type andextent of redistribution in the legislature, I as-sume that the policy influence of a legislator isproportional to, and increasing in, the group sizeof legislators which share her policy prefer-ences.*^ The Mathematical Appendix defines thepolicy determination rule.
In every jurisdiction a fraction a of the indi-viduals are rational voters, and the remaining(1 — a) fraction are noise voters. With two-party competition, sincere voting is rational.^Rational voters know whether or not parties cancommit their candidates to policies, and voteaccordingly. Such policy commitment, if feasi-ble, renders candidate identity irrelevant to thepolitical process. Therefore, with full policycommitment a rational voter directly conditionsher/his vote on the policies associated with acandidate. In the absence of such commitment acandidate will instead pursue her/his own pre-ferred policies in the legislature, and a rationalvoter will condition her/his vote on both thecandidate's party and group identity. In con-trast, a noise voter's choice of candidate isuncorrelated with the policies associated withthe candidate. The concept of noise voting cap-tures the idea that some individuals base theirvoting decisions on nonpolicy aspects of candi-date identity, such as a candidate's personal
' Qualitatively identical results also hold if we insteadconsider a single jurisdiction bul proportional representa-tion. In this environment political reservation would lakelhe form of a retjuirement thai party lists include low-caslecandidates, and fhat thiise candidates form a strict propor-tion of legislators.
''Barry Weingast 11979). among others, provides themicro-foundations for such a "universalistic" legislativebargaining procedure.
^ If possible, a voter will seek to move policy towardsher preferred outcome. Since she/lic can only affect theeleciorai outcome in the jurisdiction where she/he votes,she/he will vole for the candidate whose policies she pre-fers.
charisma. The presence of noise voters, by makingelection outcome,s probabilistic, also ensures theexistence of a voting equilibrium. The votingequilibrium is defined in the MathematicalAppendix.
Candidate selection is undertaken by two po-litical parties, indexed b> J G (/?, P). Partiesare ideologically differentiated on income—party R favors the rich, and party P the poor.Each party consists of individuals who share theparty's income preferences. That is, party R hasrich individuals as members, and parly P poorindividuals. Another way of saying this is thatparties function as brand names.^ Depending onparty membership costs, an individual will ei-ther join the party which favors her/his incomegroup or not join any political party.
Each party seeks to maximize its averagemember's utility. Formally, party J's payoff is
where ^j is the share of low-caste members in aparty, I assume this share is independent of anysingle individual's party affiliation decision.Since the income identity of party candidates( Vft) is fixed, a party's entry decision reduces toa decision over the proportion of jurisdictions inwhich it fields low-caste candidates. 1 denotethis proportion as 77,
A pohtical equilibrium is a pair of party entrydecisions which constitute best responses. Ev-ery such equilibrium is associated with a prob-ability distribution over policy outcomes. Theprobability that the policies associated with theelection of a party's candidates are implementedequals the party's probability of electoralsuccess.
In this setting, a party's membership baseaffects its entry decision in two ways. First, aparty can only field members as candidates.Second, a party's payoff, and therefore its pre-ferred policies, vary with the caste compositionof its membership pool. Party membership ispotentially open to all individuals who share theparty's income identity. However, if. relative tohigh-caste individuals, low-caste individualsface higher party membership costs then low-caste
" James Snyder and Michael Ting (2(H12) and AnoukRiviere (2{K)0l, among others, develop political economymodels which focus on ihe role of parties as brand names.
1136 THE AMERICAN ECONOMIC REVIEW SEPTEMBER 2003
underrepresentation in political parties can re-sult.'' The idea that minority underrepresenta-tion in political parties can, in tum, cause theirunderrepresentation in the legislature is shownin the following result.
RESULT 1: If the proportion of low-castemembers in each party is below their populationshare then an equilibrium with no low-castecandidates and no targeted redistributionexists.
To see why the ahove result is true in mysetup, observe that low- and high-caste individ-uals in a party (potentially) differ in their pref-erence for targeted redistribution. Low-casteunderrepresentation in a party implies that, rel-ative to their population share, the party payofffunction gives "too high" a weight to the policypreferences of high-caste individuals. Result Itells us that in such a scenario the potentialelectoral gain for either party from deviatingaway from only fielding high-caste candidates(or equivalently, committing its candidates tozero-targeted redistribution) is strictly less thanthe utility loss it incurs from the increased like-lihood of targeted redistribution.
Result 1 demonstrates that the underrepresen-tation of individuals belonging to minoritygroups in political panies can adversely affecttheir chances of gaining representation in thelegislature. The introduction of political reser-vation in India was motivated by a desire toremove such legislative underrepresentation,and provide these groups with influence overpolicy. Political reservation, by requiring par-ties to field low-caste candidates in specifiedjurisdictions, ensures the presence of low-castelegislators. The demographic composition ofthe electorate is. however, unaffected by theintroduction of such a mandate. To explore theimpact of reservation I examine how policyoutcomes are altered by the imposition of res-ervation, relative to a scenario in which nolow-caste candidates are fielded. In line with theIndian experience. I assume that the fraction of
'' In general, individuals belonging to minority groupsremain underrepresented in the main national politicalparties—see Rule and Zimmerman (1994) for cross-countryevidence, and Pradeep Chhibber (1999) for evidence fromtndia. Possible reasons include discrimination by party elite,and financial costs of participating in pany activities.
jurisdictions reserved for low-caste individualsequals their population share.'" Result 2 cap-tures the idea that the impact of political reser-vation on policy depends on the ability ofparties to commit their candidates to their pre-fened policies.
RESULT 2: If parties can commit their candi-dates to policies then political re.servation doesnot affect policy outcomes. However, if suchcommitment is absent then, relative to an equi-librium with no low-caste candidates, politicalreservation increases the likelihood of targetedredistribution.
Political reservation forces parties to fieldlow-ca.ste candidates in a certain fraction ofjurisdictions. If possible, a party will commit itscandidates, both high and low caste, to its pre-ferred policies. A party's preferred policies de-pend on the caste composition of itsmembership pool and the demographic compo-sition of the electorate. Since neither are af-fected by political reservation a party'spreferred policies are invariant to the introduc-tion of reservation. Hence, with full policy com-mitment electoral and policy outcomes areunaffected by reservation.
Result 2 tells us that in the absence of fullpolicy commitment reservation increases thelikelihood of targeted redistributioti (relative toan equilibrium with no low-caste candidates).To see why this is true, remember that poorlow-caste individuals favor targeted redistribu-tion. Reservation causes party P to field poorlow-caste candidates who, if elected, will im-plement targeted redistribution. Since both par-ties enjoy a positive probability of winningreservation increases the likelihood of targetedredistribution. However, the magnitude of thisincrease will vary with the composition of theelectorate.
If the population share of poor low castes isbelow A*, then all low-caste individuals favortargeted redistribution. In this case, with reser-vation targeted redistribution will occur withcertainty. If, however, the population share ofpoor low-caste individuals exceeds A* then rich
'" Pande (1999) shows that in this environmem a partywill not field low-caste candidates in unreserved jurisdic-tions.
low-caste individuals, like their high-castecounterparts, oppose redistribution. Reservationwill, therefore, only alter the policies associatedwith party F candidate.s. Hence, with reserva-tion, the likelihood of targeted redistributionwill equal the probability of party P^s electoralsuccess.''
The analysis has assumed that a legislator'sinfluence on policy is increasing in the groupsize of legislators who share her/his prefer-ences. An alternative is to assume that policiesfavored by the majority group of legislators areimplemented. Clearly, under this assumption,political reservation will increase targeted redis-tribution only if it is implemented in a majorityof jurisdictions.
In sum. this model of political competitionaffords predictions on the conditions underwhich reservation will affect policies, and onwhich policies will be affected. Two key as-suinptions are that candidate entry is mediatedby poiiticai parties, and that a party's preferredpolicy does not vary with its candidates' iden-tity.'" Given these assumptions, the model dem-onstrates that a statistically significant linkbetween changes in the fraction of jurisdictionsreserved for a minority group and the extent ofredistribution targeted towards the same groupcan be interpreted as evidence that politicalparties cannot enforce policy commitment onpart of their candidates and that the bargainingprocedure adopted in the legislature allowsthese legislators influence over policy. In theremainder of this paper I use data from Indianstates to test these predictions.
II. Institutional Background and Data
Reservation of Jurisdictions in favor of sched-uled castes and scheduled tribes has ensured themrepresentation in Indian state legislatures (Lelah
" If the population share of poor low castes is above A*,then reservation can reduce party P's probability of elec-toral success (relative to the case where no parly lieidslow-caste candidates). In pariicuiar. if the extent of reser-
vation (TT) exceeds 7-^—^. then poor high-caste indi-y y
viduals will switch their vote from party P to R.'"State elections in India are party-based. In addition,
qualitative evidence demonstrates significant differences inthe membership pool of Indian political parties, and thatthese differences are rellected in the staled policy prefer-ences of the political parties (Chhibher. 1999).
Dushkin. 1972. Marc Galanter. 1979. and Ol-iver Mendelsohn and Marika Vicziany. 1998.among others, provide detailed evidence on thisissue). Quantitative evidence on how such rep-resentation has affected electoral and policyoutcomes is, however., lacking, and politicalcommentators remain divided on this issue.Some argue that party control of candidates'policy activism, and the structure of state legis-lative bargaining procedures imply political res-ervation has had little to no impact on policy.For instance, Upendra Baxi (1995) argues thatscheduled caste and scheduled tribe legislatorsneed to appeal both to upper-caste constituentsin reserved jurisdictions and to the primarilyupper-caste membership of party plenary com-mittees; as a result, they do not pursue theirpersonal policy preferences in the legislature.Others, such as Dushkin (1972). Barbara Joshi(1982), and Galanter (1984). claim that minor-ity legislators act en bloc, and have succeededin increasing transfers to their own group.These, they argue, include cabinet positions instate governments, educational .scholarships andreservations in higher educational institutions,and above all, government jobs.'^
In Section III I exploit the institutional fea-tures of political reservation and data on itspractice to provide evidence on the impact ofpolitical reservation on policy outcomes in In-dian states. In the remainder of this section Idescribe the institution of political reservationand the data set which will be used in theanalysis.
A. Political Reservation in Indian States
The 1950 Indian constitution mandates polit-ical reservation in favor of scheduled castes andscheduled tribes in every .state and national elec-tion. In addition, it explicitly directs state gov-ernments to use public policy to improve theeconomic well-being of these two groups.
'^Dushkin (1972) cites as instances of such activismduring the 1967-1972 national parliamentary session thedefeat of the Congress party on the amendment, "'the opin-ion of ihe House (that) safeguards provided in the Consti-tution for the scheduled castes and tribes are not being fullyimplemented" due to bloc voting by minority legislators: theliberalization of job reservation policy (July 1968. 1970),increased flexibility in targeted educational subsidies(1969). and the hardening of the untouchability offenses act(1970).
1138 THE AMERICAN ECONOMIC REVIEW SEPTEMBER 2003
TABLH 2—LEGAL IDENTIRCATION OF SCHEDUI.ED CASTES .AND SCHEDULED TRIBES
Selection criteria for scheduled castes
1. Cannot be served by clean Brahman.s2. Cannot be served by the barbers, water-carriers, tailors, etc. who serve the caste Hindus3. Pollutes a high-caste Hindu by contact or by proximity4. Is one from whose hands a caste Hindu cannot take water5. Is debarred from using public amenities such as roads, ferries, wells, or schools6. Will not be treated as an equal by high-caste men of the same educational quai ill cation in ordinary sociaJ intercourse7. Is depressed on account of the occupation followed and, but for that, occupation would be subject to no social
disability
Selection criteria for scheduled tribes
1. Tribal origin • •..2. Primitive ways of life and habitation In remote and less accessible areas3. General backwardness in all respects
Note: The above criteria were the required basis for the selection of "sehedtiled caste" and "scheduled tribe" communities.as stated in the Constitutional (scheduled easte and scheduled tribe) orders of 1950.
TABLE 3—ECONOMIC CHARACTERI.STICS OF ScHEUUt-ED CASTES AND SCHEDLILEO TRIBES: 1991
Variable
Overall population shareWirliir>-i;r<>up fhanwtemlics:
Urban population shareLiteracy rateLabor force participation ralePercent labor force in the primary sectorPercent population below poverty line
Scheduled castes
16.4
18.737.43677.148,3
Scheduled tribes
7.9
7.329.6429052,0
Non-SCVST population
75.4
29.257.832.862.131.4
Notes: All numbers are from 1991 census, excepi poverty figures which are from ihe Indian National .Sample Survey(1993-1994). Planning Commission Estimates. The primary sector includes those employed in agricultural and alliedactivities. Within-group characteristics are reported as a percentage of the group population.
The constitutiona] (scheduled caste and sched-uled tribe) orders of 1950 established state-specific lists which identified the castes and tribesthat fall in the categories of scheduled castes andscheduled tribes respectively. The caste identifi-cation criteria of the 1931 census formed the basisfor the selection of scheduled castes, and a tribiilidentification criteria developed by a 1950 Parlia-mentary the basis for choosing scheduled tribes.Table 2 describes these criteria. The scheduledcaste and scheduled tribe lists have been revisedtwice—in 1956 to remove anomalies arising fromthe linguistic reorganization of states, and in 1976to remove witliin-state discrepancies in the iden-tification of certain castes and tribes as scheduledcastes and scheduled tribes respectively.
Scheduled castes make up roughly 16 percentof the Indian population, and scheduled tribesanother 8 percent. Relative to the rest of thepopulation, individuals belonging to these two
groups remain socially and economically disad-vantaged. The incidence of poverty in thesegroups is roughly one and a half times that inthe rest of the population (see Table 3). Theeconomic backwardness of scheduled castescan be directly traced to the caste system.'''^Members of scheduled castes were traditionallyassigned to menial occupations such as skinninganimal carcasses and removing human waste,and faced restrictions on asset ownership.''' In
'•* Roughly 85 percent of the Indian population is Hindu.Every Hindu belongs to a caste, and caste membership ishereditary. The genesis of the caste system is usually tracedto the Aryan invasion of India in approximately 1500 B.C.Caste groupings are. in general, endogamous. Tbe castesystem is hierarchical, with a ciLste's rank the primarydeterminant of il.s members' occupation.
' For instance. Manu Smriti. the definitive treatise oncaste system, decrees that the dwellings of low castes be
VOL 93 NO. 4 PANDE: POLITICAL RESERVATION AND POUCY OUTCOMES 1139
TABLE 4—THE TIMING AND REASONS POR RESERVATION CHANGES
Year of change Reason for change Cominis,sii>n responsible
1962196519671972. 1974. 19761977, I97S. 1980
Double member jurisdictions abolishedCreation of HaryanaRevi.sed in line with 1961 censusRevised in line with 1971 censusRevised in line with 1976 area restriction removal act
the case of scheduled tribes, their geographicisolation, combined with their dependence ontraditional agricultural practices for subsistence,has contributed to their poverty.
Article 332 of the Indian constitution pro-vides for political reservation in state elections.In a jurisdiction reserved for scheduled castes(scheduled tribes), only a scheduled caste(scheduled tribe) individual may stand for elec-tion. The entire electorate, however, participatesin choosing among candidates so qualified. Thearticle states that two independent national-level commissions will be responsible for im-plementing this mandate. The orders of thesecommissions will have the force of law, andcannot be questioned in court. These commis-sions are the Election Commission and the De-limitation Commission. The Election Commissionis a national-level body which oversees stateand national elections. The Delimitation Com-mission is responsible for redistricting. and isconstituted whenever new census population es-timates are announced.'" The article also re-quires that in selecting reserved jurisdictions,preference be given to jurisdictions with ahigher population share of the group in whosefavor reservation is being ptacticed, while en-suring a sufficient dispersal of reserved jurisdic-tions within the state.'^
outside the village, and their wealth be confined only todogs and donkeys. It states that a member of the upper castemay take possession of the property of a low caste wilhporlect impunity (Manu Smriti VIII:4I7. X:52).
'"Membership to the Delimitation Commission is re-stricted 10 a retired national court judge, a sitting state counjudge and the chief election commissioner.
' Scheduled castes are a populaliiin minority in everyJurisdiction, irrespective ol' its reservation slatus. Relative tononreserved jurisdictions the population shaie of scheduledcastes is. on average. 5-6 percentage points higher in re-served jurisdictions. In contra.sl, scheduled tribes are a pop-ulation majority in roughly half the jurisdictions reserved intheir favor (Galanter. 1984).
My empirical analysis exploits the diachronicvariation in the extent of political reservationenjoyed by a group in a state. The cause of suchvariation is defined by Section 3 of Article 332.Section 3 states that the jiroportion of jurisdic-tions reserved for sehedtiled castes (scheduledtribes) should equal, as nearly as possible, thepopulation share of scheduled caste (scheduledtribe) in the state. Moreover, the only permissi-ble basis for changes in the extent of reservationenjoyed by a group in a state is changes in thecensus estimates of the group's populationshare in that state.
Table 4 lists the years in which the proportionof jurisdictions reserved for a group changed,the stated reason for change, and the commis-sion responsible. As elections across states arenot synchronized, a single commission's recom-mendations are often implemented in multipleyears. Table 4 tells us that changes in the pro-portion of jurisdictions reserved for a groupwere always caused by changes in the censuspopulation estimate for the group. In every case,Ihe extent of change in reservation for a groupequaled the change in its census populationestimate. Changes in census population esti-mates for a group are. in turn, either caused bythe arrival of new census population estimatesor by a centrally mandated institutional changewhich altered the existing cetisus populationestimate for the group. These instituiionalchanges include the uational shift to singlemember jurisdictions in 1962. the creation ofthe state of Haryana in 1965, and the 1976national mandate which required that a caste ortribe which was identified as a scheduled casteor scheduled tribe in any part of the state be sodefined for the entire state. " Finally, due lo anational decision to disallow further changes in
'^Such within-state differences in the defmition of acaste/tribe as a scheduled caste/schedtiled tribe arose due tothe reorganization of state boundaries over time.
Political variablesSC reservationST reservationElection dummy
Deinofiraphic variablesSC census population shareSC current population shareST census population shareST current population shareCensus population density
Olher economic variablesState income per capita
Mean
153
•21.513.22.950.12
22.61
14.047.250.22
14.1914,527.157.27
205.46
1,036.22
Standarddeviation
(87.36)
(4.48)(2.19)(4,07)(0.45)
(10.39)
(5.36)(7.69)(0.41)
(6.47)(6.02)(7.57)(7.47)
(132.3)
(357.49)
Notes: The Data Appendix describes the construction andsources of variables. The data are for the 16 major Indianstales, and the period 1960-1992. For Haryana. which splitIrom Punjab in 1965, I use data for the period 1967-1992and Cor Jammu/Kashmir I use data for 1962-1992, Thisgives a sample size of 519: deviations from this are ac-eounled for by missing observations (on which, see the DataAppendix). SC and ST welfare spending is available po.st-1974.
the number of jurisdictions in a state the extentof reservation in place for a group in a state hasremained constant since 1980.
B. Data and Descriptive Statistics
The unit of observation in my analysis is theIndian state. 1 use a data set of the 16 majorIndian states which spans the period 1960-1992. These states account for over 95 percentof the Indian population. Table 5 provides de-scriptive statistics,
I measure the political re.sei vation afforded toa group in a state as the fraction of jurisdictionsreserved for that group in the state. The reser-vation variables for scheduled castes and sched-uled tribes are denoted as "SC reservation" and"ST reservation," respectively. In the sample,the average SC reservation is 13 percent, andthe average ST reservation is 7 percent. Threestates in the sample have no scheduled tribe
population, and therefore no ST reservation.The data .set includes two measures of a group'spopulation share. The first is the group's popu-lation share as measured by the census in theyear when reservation was determined, and isdenoted "SC/ST census population share." Thisvariable is used by the Election and Delimita-tion Commissions to determine the extent ofreservation for a group. The second measure isthe group's population share as measured in thecurrent year: this variable is denoted as "SC/STcurrent population share."
The Indian constitution provides for a federalstructure of govemment, with state govern-ments enjoying independent jurisdiction onmost types of social-sector spending. In addi-tion, the constitution explicitly allows state gov-ernments to target welfare programs towardsscheduled castes and scheduled tribes, In theanalysis I distinguish between two types of statepublic policies—-policies whose benefits are notrestricted to scheduled castes and/or scheduledtribes (henceforth, "general" policies), and pol-icies whose benefits are so restricted (hence-forth, "targeted" policies).
In the category of general policies, I firstconsider the size of the state govemment, asmeasured by log real per capita state govem-ment spending. Second, I consider the share ofa state's total spending going to education. Ed-ucation spending, on average, makes up 21 per-cent of an Indian state's budget, and is thesingle largest category of general social-sectorspending in every Indian state. The final generalpolicy I consider is an asset-based redistribu-tion policy—land reform. This variable is ofinterest, hoth because landlessness is highamong scheduled castes and scheduled tribesand because land reform has been a politicallycontentious issue in most Indian states. Landreform is measured by an indicator variablewhich takes the value one in years when a statepasses a land reform act, and the value zerootherwise.'*'
In the category of targeted policies, I considertwo measures of targeted spending. These arethe fraction of government spending devoted toscheduled caste and scheduled tribe welfare
' ' 'I use a measure of land refonn activism created byBesley and Robin Burgess (2000). who show that landrefonn had a significant negative effect on rural povertyacross Indian states.
programs, and are denoted as "SC welfarespending" and "ST welfare spending" respec-tively. The programs financed by such spendinginclude, among others, group housing projects,hostels for students belonging to these groups,and the provision of public goods in scheduledcaste and scheduled tribe hamlets. The averagestate spends between 3-4 percent of its budgeton each of these two categories of targetedspending, Einally, I consider job quotas which isthe fraction of state govemment jobs reservedfor scheduled castes and scheduled tribes. In thesample, the mean fraction of state govemmentjobs so reserved is 20 percent. Arguably, in-creases in the extent of job re,servation has beenthe most important political concession grantedto scheduled castes and .scheduled tribes inpostindependence India (Parikh, 1997, Mendel-sohn and Viczianv, 1998).^"
m . Results
Variation in the percentage of jurisdictionsreserved for a group in a state is attributable tochanges in the census population estimates forthe group. Such changes are caused hy the ar-rival of fresh population census estimates andnational institutional changes, as outlined in Ta-ble 4. Hence, reservation for a group is a non-linear function of the group's population in themost recent census. I exploit this feature ofpolitical reservation to identify its impact onpolicy,
A. Ba.sic Results
For the st\\ state at time t, I can write:
where y^ is a policy outcome, and R^ is a vectorwhose elements are SC reservation and ST res-ervation, respectively, a,, and j3, are state andyear fixed effects, and e,, is the state-level errorterm. The reservation coefficient y is the param-eter of primary interest. State effects control forthe influence of unobserved time-invariant state
characteristics on policy. Year effects controlfor the policy effects of national events whichaffect all states in a similar manner: they, how-ever, do not control for national events whichaffect different states differentially.
In this empirical specification identificationof the effect of political reservation on policyoutcomes is obtained from within-state varia-tion, i.e., state-specific changes in reservation.As discussed above, such changes are caused bychanges in a group's census population share.The main threat to the validity of this identifi-cation strategy is omitted effects of the factorswhich determine reservation, of which laggedpopulation effects of a group seem most likelyto be of concem. To guard against such omittedvariable bias I sequentially expand the set ofcovariates. I start by including the vector P ,*whose elements are SC and ST census popula-tion shares as right-hand-side variables. I thenadd as a covariate the two groups' current pop-ulation shares, denoted by the vector P^,. Fi-nally, I include as covariates three variableswhich are potentially correlated with scheduledcaste and scheduled tribe population shares in astate; this vector is denoted as X j. Here I in-clude state income per capita lagged by oneperiod as changes in a state's income may di-rectly affect groupwise fertility rates. In addi-tion, cross-state income differences may inducemigration, and thereby alter group populationshares." ' Since changes in a group's populationshare are likely to be correlated with changes inpopulation density I include a state's populationdensity, as measured when reservation was de-termined. The final element in this vector is anelection year dummy. Reservation changesonly occur in an election year; the electiondummy is to ensure that the reservation vari-ables do not simply pick up election year ef-fects. To summarize, the llnal specification is ofthe form:
^°Job quotas for different population groups w;is firstintroduced in India by the British on the basis of the 1922Miller repon. Parikh (1997J describes lhe evolution of jobreservation policy in India.
*' I use stale income lagged by one period as this vari-able is potentially endogenou.s The results are robust tolagging state income by dilferent periods, see Pande (1999).
1142 THE AMERICAN ECONOMIC REVIEW SEPTEMBER 2003
Tables 6 and 7 report the findings for gen-eral and targeted policies, respectively. Foreach policy I report four specifications, whereI sequentially expand the set of covariates.Columns (I)-(4) of Table 6 consider totalspending. Increases in ST reservation, but notSC reservation, raise total spending in a state.The estimated effect is robust to the inclusionof additional population and economic con-trols [columns (2)-(4)!. Columns (5)-(8) ex-amine the impact of reservation on educationspending, ST re.servation has a significantnegative impact on education spending. Thecolumn (8) estimate suggests that a l-percentincrease in ST reservation reduces educationspending by slightly under 0.4 percentagepoints. Given the high levels of illiteracyamong scheduled tribes, the finding thai STlegislators choose not to prioritize educationspending is striking. SC reservation is alsonegatively correlated with education spend-ing; this relationship, however, is statisticallyinsignificant. Finally, in columns (9)-(12) weobserve that increases in the number of leg-islators belonging to .scheduled castes andscheduled tribes does not affect the likelihoodof land reform legislation.
I next consider policies whose benefits areexplicitly targeted towards scheduled castesand/or scheduled tribes. The results in Table7 suggest a significant relationship betweenreservation and these policies. Moreover, theimpact of SC reservation and ST reservationon policy differs. The results in columns ( I ) -(4) reveal a positive correlation between SCreservation and job quotas. The effect is largeand significant, and robust to the inclusion ofpopulation and other controls. To give someidea of magnitudes, the estimates in column(4) tells us that a 1-percent rise in SC reser-vation is associated with a 0.6-percent in-crease in job quotas. In contrast, SCreservation is unrelated to the level of SCwelfare spending [columns (5)-(8)], and STwelfare spending [columns (9)-(l2)]. Thesefindings stand in sharp contrast to those forST reservation. Columns (l)-(3) show thatincreases in ST reservation do not signifi-cantly affect job quotas; there is weak evi-dence of a negative effect in column (4).However, ST re.servation has a significantpositive effect on ST welfare spending. Theestimates in column (12) suggest that a one
point increase in ST reservation increases theshare of total state spending devoted to STwelfare programs by 0.8 percentage poinEs. ^
The finding that SC reservation increases jobquotas while ST reservation increases spendingon ST welfare programs is consistent with dif-ferences in scheduled caste and scheduled tribegroup characteristics. Relative to scheduledtribes, scheduled caste individuals are bothmore educated and geographically more dis-persed. Hence, their relative retums from indi-vidual-specific policies, such as job quotas, arehigher. In contrast, relative to scheduled castes,the benefits to scheduled tribe individuals fromgeographically localized welfare programs suchas housing schemes are greater. The finding thatST, but not SC, reservation increases targetedspending also helps us make sense of the findingin Table 6 that increases in ST reservation re-duce education spending, and raise overallspending. Taken together, these results suggestthat some of the observed increases in targetedredistribution have come at the expense of gen-eral redistribution.
Finally, Tables 6 and 7 reveal interestingdifferences in the relationship that SC and STcurrent population share variables bear to thepolicy outcomes. Increases in SC current pop-ulation shares are associated with increases injob quotas and reductions in ST welfare spend-ing. In contrast, increa.ses in ST current popu-lation share are negatively correlated with mostpolicy outcomes. While the potentially endoge-nous nature of these population variables pre-vents a causal interpretation, these findings areconsistent with the fact that the political activ-ism of members of these two groups differssignificantly. While scheduled castes haveemerged as an important political bloc in post-Independence India, scheduled tribes remain,by and large, politically marginalized.'^^
B. Robu.stness
The empirical analysis exploits state-specificvariation in political reservation to examine its
" F-tests reject the null that SC and ST reservation havesame impact on job quotas and ST welfare spending,
'•* These findings are in line with Abhijit Banerjec andRohini Soinanathan (2001), They find that (ndian districtswith a higher scheduled tribe population gel tower publicgoods. This, however, is nol the case with scheduled castes.
•—I r*A '— O C O •—' C ^ - i/ '—I ^ ^ "^O O w O O C ^ O C O O
d d d d d d c o
rN fN rJu-i tN Tt 00— — lO p
d d d d
— o — c r i o
o o o o o o o c c c o o
O OO O O O O w C O O w O
d d d d d d d d d d
d d d d d d d d
O5;Z d
in in ro m
p o pd d d d
•2 •§ a a
U
^ ts Xu a pc- —b ^"3 -a 3O < Z
2 cy ^ ^ ^ E3 d ^ S 3 3T3 > c oj c-a
~ a. G- £, S
•5 5^-3S - UJ
c: E
^ i<- 8-5 c
U.2 ™"3 D.
X 2
2 -
5
c
and
CO
u
CO
x :
pwo
3Xr.C3 u
coucOJ
2'^,
cOJ
• a
ia E 3w S 5
a.a
o<ua..5
and
6
fro
V
0
to
0
ulat
i
• au3
£O. OJ U
. a ^ - " ^ ^ : : i^ ^ ^aj O ^^ [/; v; : / ;•« y •— » » *OJ a o * *
5-^ S
1144 THE AMERICAN ECONOMIC REVIEW SEPTEMBER 2003
• ^ - ^
I
dddddddddddd
C O O o aof, 00 — O ' t i ^ o ^ . c i T i o o r - f ^' T r . j p r ^ — ^—ITT—-ir i fN
d d — d d d d d d d d d
— r-1inro
dd
d d d d
lOoo-^r^-^ce — r^if. m —r^iOs(NoiC!>Oi^ooovr j \om_ _ - 3 p p p p '
dddddddddddd
IZ o ri
Z c
Z o n
Opr-idd
t O i
c — q q p p p — — —d d d d d d d d d d
c o c o n r ^ fo — c — ' p p q pd d d d d d d d Z c
Z o
[ N OCTOiC — r n t N - p s q - \C — v q r n
d d d d d d d d d d d d
CTvoc — i n — - r o r J O M t N i n - i -^ — r - ^ c N p p s o - ^ — ^qr- idddddddddddd
^ p
dddddddd
o ov 2;Z d U-,
o a 2:Z d <r;
Z d m
u 3 c
o < z
'S'-
^ y
--a -a
2 i- ^ s' ^ -3 'S '" "".
"" 2 ^ M >;
^ .!2
2 a
VOL 93 NO- 4 PANDE: POLITICAL RESERVATION AND POUCY OUTCOMES 1145
effects on policy outcomes. In this empiri-cal setup key robustness checks are the sen-sitivity of the results to the introduction ofadditional controls for variables that may bein the function determining reservation. Asdiscussed, changes in reservation are attribut-able to changes in the groups" populationshares, as reported by the census.""* Table8 examines the robustness of the findings toincluding additional population controls ascovariates. For expositional ease, I restrict theanalysis to the subset of policies which havebeen shown to be affected by politicalreservation.
All regressions reported in Table 8 includethe SC and ST reservation variables, the SCand ST census and current poptjiation sharevariables, and the controls for state income,population density, and election year as co-variates. Reservation for a group is a nonlin-ear function of the group's lagged (census)population share. It is. therefore, relevant tocheck that the observed results are robust tothe inclusion of nonlinear and lagged popula-tion share variables. In column (1). I includequadratic controls for SC and ST census pop-ulation shares as additional covariates. SCand ST reservation variables do not appear tobe proxying for the nonlinear effects of thecensus population variables. In the case ofeducation spending, the inclusion of quadraticcontrols reduces the statistical significance ofthe estimated effect of ST reservation. How-ever, the economic magnitude of this effectremains unchanged. In all other cases, theimpact of reservation on policy is unaffectedby this change in specification (see Panel Athrough Panel D). In column (2), I include theone- and two-period lagged values of SC and
'"* The main reason for changes in ihe population sharesof these groups has been differential fertility rales. Between1961 and 1971. relative to the general population, popula-tion growth was slightly lower among scheduled caste andscheduled tribe populalions. This was mainly driven byhigher infant mortality rates among these two groups. How-ever, by the end of the 196O's, infant mortality rates hadconverged across groups, and since 1971 the Indian censuscon.sistenily reports higher fertility, iind population growthrates among these groups. This iinding is also corrohoraledby other surveys^—for instance, the Indian Demographic andHealth survey for 1993 reports all-India total fertility ratesamong scheduled castes, scheduled tribes, and ihe rest of thepopulation as 3.15. .1.06. and 2.60. respectively.
ST current population shares as covariates.Once again, the size and significance of theestimated relationship between SC and STreservation and policy outcomes is largelyunchanged. In column (3). I include a state-specific piecewise linear trend. For any givenstate, this trend variable increases by incre-ments of one in years in which reservation fora group changed in a stale. This trend variablecan be viewed as partially controlling for anyomitted variables which also change at thepoint when reservation for a group changes. Ifind no significant change in the estimatedrelationship between reservation and policyoutcomes.
As a final check, I examine whether the re-sults are robust to restricting the sample for eachstate to five-year periods that are centeredaround an election in which the proportion ofjurisdictions reserved for a group changed. Theidea is to check whether the discontinuouschanges in political reservation which occur inan election year are associated with subsequentpolicy changes. The results are presented incolumn (4). Reductions in the sample implygreater imprecision in the estimated relationshipbetween reservation and policy outcomes, i.e.,the standard errors tend to be larger. However,both the size and significance of the estimatedeffects are robust to the reduction in samplesize. The only exception is total spending—inthis case, the effect of ST reservation is sta-tistically insignificant (though the economicmagnitude of the estimated effect remains com-parable to earlier specifications). This specifica-tion increases our confidence that omittedvariable bias is not driving the observed rela-tionship between political reservation and pol-icy outcomes—there is less reason to believethat the impact of omitted population variableson policy outcomes would follow a similar dis-continuous pattem.
In Pande (1999) 1 repoit additional robustnesschecks. I show that the results are robust toincluding an array of contemporaneous stateeconomic, demographic, and political variables.I also check that the results are robust to includ-ing controls for interstate migration—the con-cern being that policy-induced migration mayunderlie observed populaiion. and therefore res-ervation, changes. The results are also robust tousing a two-stage least-square procedure whereI use SC and ST census population shares to
1146 ..1 , THE AMERICAN ECONOMIC REVIEW , J , SEPTEMBER 2003
TABLE 8—POLITICAL RESERVATION AND POUCY OUTCOMES: ROBUSTNESS CHECKS
Nonlinear censuspopulation controls
(1)
PANEL A: Dependent variable: Total .spendingSC reservation
Notes: Robust standard errors are reponed in parentheses. A!l regressions include (i) state and year fixed effects, (ii) stateincome per capita lagged one period, population density and election year dummy, and (iii) SC/ST census population shareand SC/ST current population share as controls. Panel A includes as covariates SC/ST census population shares squared/100;panel B includes SC/ST one- and two-period lagged cuirent population shares. Panel C includes a state-specific trend whichincreases by units of one in years in which reservation changes. The data are for the 16 main states, and the period 1960-1992,For Haryana. which split from Punjab in 1965. the data spans I967-1992. and for Jammu-Kashmir 1962-1992. This gives519 observations. Deviations are accounted for by tiiissing data (on which, .see the Data Appendix), Panel D regressionsrestrict the sample for each stale to data for two years prior to an election in which the proportion reserved jtirisdictionschanged, the election year and two subsequent years. The number of observations is 187. except for ST spending for whichil is 82.
* Significant at lhe 10-percent level,** Significant at the 5-percent level.
•"** Significant at the l-percent level. ' " • ' ' "
instrument for SC and ST current populationshare. "•'
In summary, the regressions reported in Ta-ble 8 control, in different ways, for argumentsof the function which determines the extent ofreservation enjoyed by a group in an Indianstate. In every case. I continue to find a sig-nificant relationship between reservation andpolicy outcomes. Increases in both SC and STreservation are associated with increases intargeted redistribution. In addition, increasesin ST reservation lower education spendingwhile raising overall government spending.
^ This specification checks for possible measurementerror hias in the regression induced by the use of interpo-lated population data for inter-cen.sus years.
Taken together, the results in this section sug-gest that changes in legislator identity in Indiahave exerted a significant influence on state-level policies in a way that is consistent with amodel of political competition in which partieshave policy preferences, but cannot committheir candidates to policies. .
IV. Discussion
A number of countries, including the UnitedStates, have experimented with mandates thatseek to enhance minority representation in thelegislature. However, most of these experimentsstop short of directly changing legislator iden-tity. For instance, in the I98()"s. U.S. courtssucceeded in increasing African-American rep-resentation in the legislature by requiring states
VOL 93 NO. 4 P.ANDE: POLITICAL RESERVATION AND POLICY OUTCOMES 1147
to alter the composition of the electorate insome jurisdictions in favor of this group(Charles Cameron et al.. 1996; Marvin Overbyand Kenneth Cosgrove. 1996). However, in thiscase both voter and legislator identity were al-tered. As a result, isolating the effects ofchanges in legislator identity on policy hasproven difficult. The nature of the political res-ervation mandate in India allows us to makeprogress on this issue.
To isolate the effect of legislator identityon policy outcomes I use time lags in theresponse of political reservation to populationchanges. Such time lags arise because thefraction of jurisdictions reserved for a groupis adjusted to take account of changes in agroup's population share only when new pop-ulation census estimates arrive. Using decen-nial census estimates for electoral purposes isa commonly adopted practice in most democ-racies. This suggests that il may be possible touse variants of the identification strategyadopted in this paper to provide empiricalevidence on other features of the politicalprocess. For instance, in many countries ju-risdiction boundaries are redrawn after everydecennial census to ensure equal populationsin each jurisdiction. The lime lags betweenactual population growth across jurisdictionsand the equalization of population across ju-risdictions may be used to isolate the politicalconsequences of legislative redi.stricting.
This paper provides evidence that increasedpolitical representation for disadvantaged mi-norities can allow them greater influence onpolicy-making. I show that pohtical reservationfor scheduled castes and scheduled tribes inIndia has, in part, been responsible for the ob-served rise in targeted redistribution towardsthese groups over the lasi half-century. Thesefindings also shed light on tho functioning of thepolitical process in India. In general, politicalreservation can affect policy outcomes onlywhen parties cannot enforce policy commitmenton the part of their candidates, and even then itseffect remains sensitive to the decision-makingprocedures adopted in the legislature.
Tt would, however, be premature to viewthis paper's findings as suggesting that polit-ical reservation is a welfare-enhancing policy.To start with, neither the theoretical nor theempirical analysis has examined ihe implica-tions of political reservation for candidate
quality. It is wholly feasible that the restric-tions on candidate entry which are requiredby a policy of political reservation adverselyaffect the quality of candidates. Such restric-tions may also reduce the ability of voters(and parties) lo punish candidates who engagein wasteful redistribution or corrupt politicalpractices. Policies chosen by minority legis-lators may also be more likely to reflect thepolicy preferences of special interest groups.For instance, minority individuals may bemore easily intimidated by members of themajority social group. Alternatively, theymay be more willing to curry favors alonggroup identity lines. Finally, a candidate'ssocial identity as a primary prerequisite maylead to individuals belonging to nonminoritygroups becoming more disengaged from, anddisaffected by. the political process. All thesefactors go to suggest that the extent to whichenhanced political representation for a grouptranslates into improved welfare outcomes formembers of these groups, and the polity atlarge, remains an open and important questionfor future research. Clearly, the findings inthis paper suggest that one important elementof any such research agenda must involveunderstanding the relative successes of tar-geted and broad-based redistribution in af-fecting the welfare of such groups.
MATHEMATICAL APPENDIX
The Appendix is structured as follows. First,the voting equilibrium and the decision-makingmle in the legislature are described. Second, theproof for Result 1 is provided.
Voting Equilibrium.^-ln every jurisdiction afraction a of the voters ai'e rational, and a frac-tion (1 — a) are noise voters. A fraction /3 ofthese noise voters vote for party P. where (5 isa random variable with support [0. I] and cu-mulative distribution function G(j3). The func-tion is symmetric, such thai G(/3) ^ 1 —G( I - |3) for alt /3. That is. noise voters areunbiased. The electoral outcome depends onrational voters' voting choices, and the draw of/3. Let e denote the difference between the num-ber of voters who favor party P, and those whofavor party R. The party P candidate wins if
ae (\ - a) - a)(\ -
1148 THE AMERICAN ECONOMIC REVIEW SEPTEMBER 2003
or,
ae
In probability terms a party P candidate wins inevery jurisdiction with probability (f)(e), where
1 -a(|)(e) = 0 if e ^ ; (f>(a) = 1 if e ^
1 - a; and
aea • • \2 2(1 - a)f
otherwise. I assume that the fraction of noise vot-ers in a jurisdiction is sufficiently high, so that
1 «A_ -H A - A, < — — . Under this assump-
-' • ation, a party which only attracts a single demo-graphic group's vote will enjoy a positiveprobability of winning.
Legislative Procedure.—Idenxicd] jurisdic-tions imply that the same electoral outcomeoccurs in every jurisdiction. Hence, all legisla-tors share the same party identity. If possible, aparty will commit its candidates to the partypreferred policy. Hence, with full policy com-mitment the policy favored by the winningparty will be implemented by its legislators (allof whom will have been committed to thispolicy).
In the absence of full policy commitmentlegislative conflict may arise. Let S ,. and 7j,, bethe levels of targeted and nontargeted redistri-bution favored by a caste c legislator belongingto party J. I assume that the policy influence ofa legislator is increasing in the group size oflegislators who share her/his policy preferences.Here, this group size is given by the number oflegislators with the same income and caste iden-tity. Given that income identity is fixed within aparty, the legislative policy determination mleis defined as:
T*j=(l-
I assume
and5*=
= 0, and > 0.
RESULT 1: / / the proportion of low-castemembers in each party is below their populationshare, then an equilibrium with no low-castecandidates and no targeted redistributionexists.
PROOF:This proof considers the case where the low-
caste party membership share is less than theirpopulation share, i.e., ^j < A . Tf policy com-mitment is feasible then prior to an electioneach party will commit its candidaies to itspreferred policies. Since candidate identity isirrelevant to the political process, ir . = 7T^ =0 constitutes a pair of best responses.
Now consider the case where full policycommitment is infeasible. If TT^ = 7r = 0,then party P will be associated with generalredistribution and party R with no redistribu-tion. Rational voters will vote for the partywhich represents her/his income group, andparty P will win with probability cf)(A'' - A'').In equilibrium no targeted redistribution willoccur. I now show that Trp = TT = 0 con-stitute a pair of best responses for the twoparties.
(i) If A'' > A* then rich low-caste individualsfavor no redistribution. Hence, irrespectiveof caste identity, all party R candidates willfavor no redistribution, ir, = 0 constitutesa best response. In contrast, low- and high-caste members of party P will favor differ-ent policies. If party P deviates to TTp > 0,then election of its candidates causes bothtargeted and general redistribution (see leg-islative procedure definition). Such a devi-ation, however, cannot improve party P'spayoff. First, as rich citizens do not favorredistribution this deviation can not in-crease its probability of electoral success.Second, under the assumption $j < A thisdeviation lowers the party's expected pay-off. Hence TT = 0 is a dominant strategyfor party P.
(ii) If A'' < A* then rich low-caste individ-uals favor targeted redistribution. In thiscase given TT/^ = 0. TT , = 0 is a bestresponse. Consider a deviation by party Pto TTp > 0. Such a deviation leaves thevoting decisions of rich high-caste andpoor iow-caste individuals unaffected. Ttcauses rich low-caste individuals to votefor party P if
VOL 93 NO. 4 PANDE: POLITICAL RESERVATION AND POUCY OUTCOMES 1149
Poor high-caste citizens continue voting forparty P if
These two conditions are jointly satisfiedif A^ > A' . In this case, the deviationmaximizes party F's payoff if
+ (I -
where X =out gives:
-I- ^. Solving this
, - AJ
This inequality cannot hold if the share oflow-caste members in a party is less thanits population share, i.e., ^p < A . For, inthis case, the left-hand side is negativewhile the right-hand side is positive. Itfollows that given Tr' = 0. TT'' = 0 is abest response. A symmetric argument es-tablishes that TT' = 0 constitutes a bestresponse to TT'" = 0.
DATA APPENDIX
The data sst used in this paper builds on anIndian state-level data set which was collated byBerk Ozler et al. (1996). and updated by Besleyand Burgess (2000). The data set spans 1960-1992 and includes the 16 major Indian states.These states are Andhra Pradesh. Assam. Bihar.Gujarat. Har)'ana. Jammu-Kashmir, Kamataka.Kerala. Madhya Pradesh, Maharashtra, Orissa,Punjab. Rajasthan. Tamil Nadu. Uttar Pradesh,and West Bengal. Jammu-Kashmir enters thedata set in 1962. and Haryana in 1967."^
Political data come from Shankar Bose and
• Haryana was created in 1965. by splitting up Punjab.Poiiticai data for Harvana exists from !967.
V. B. Singh (1987) and Election Commissionstate election reports. The variable "SC (ST)reservation" is the proportion of jurisdictions ina state reserved for scheduled castes (scheduledtribes). The election dummy takes a value onein the year of a state election, and zerootherwise.
Population data are from the decennial In-dian censuses. 1951 through 1991 (Census ofIndia, Registrar General). Tn accordance withthe scheduled castes and scheduled tribes orders(Amendment) Act. 1976 fresh census estimatesof scheduled caste and scheduled tribe popula-tion were issued in 1977. Data have been inter-polated for inter-censal years. The variable "SC(ST) census population share" is the scheduledcaste (scheduled tribe) population share as re-ported by the census at the time when reserva-tion was determined. This variable is updated toreflect new census estimates for a state in theyear of the first election held in the state afterthe new census estimates have been declaredand the Delimitation Commission has met. Thevariable is held constant between two such elec-tions. The SC (ST) current population share isthe interpolated SC (ST) population share fromthe census, as measured in the current year.Population density is the ratio of interpolatedtotal population data from the census as mea-sured when reservation was detennined in thestate divided by total land area of the state, asreported in the Census Atlas. The variable isupdated for a state in the year of the first elec-tion held in the state after the new census esti-mates have been declared and the DelimitationCommission has met.
Public finance variables: The general ex-penditure variables are from the Reserve Bankof India Report on Currency and Finance, andthe targeted expenditure variables are from theannual Ministry of Welfare handbook. The vari-able descriptions are as below:
1. Total spending is the log total state expendi-ture during the budget year expressed in realper capita terms.
2. Education spending is the share of total stateexpenditure going to elementary, secondary,university and higher, technical, and adulteducation. 1972 data are missing for Biharand Gujarat, and 1972, 1973 data for Hary-ana and Tamil Nadu.
3. SC welfare spending is the share of total
1150 THE AMERICAN ECONOMIC REVIEW SEPTEMBER 2003
State expenditure going to scheduled castewelfare programs. This variable exists from1975. 1981 data are missing for Jammu-Kash-mir. and 1984 data for all states except AndhraPradesh. Assam, and Jammu-Kashmir.
4. ST welfare spending is the share of totalstate expenditure going to scheduled tribewelfare programs. 1974 data are missing forMaliarashtra, Tamil Nadu, and UttarPradesh, 1975 data for Uttar Pradesh, and1986 data for Andhra Pradesh and UttarPradesh.
Land Reform is a dummy variable whichequals one for an Indian state in the year inwhich the state legislature passed a land reformlegislation, and is zero otherwise. This variablewas created and used in Besley and Burgess(2000).-^
Job quota is the proportion of state govem-ment jobs which are reserved for scheduledcastes and scheduled tribes. The data source isthe annual scheduled caste and scheduled tribecommissioner's reports.
State income is the log of the real per capitastate income. The data source is: Estimates ofState Domestic Product, published by Ministryof Planning, Govemment of India.
Price deflators: The "Consumer Price Indexlor Agricultural Laborers'* (CPIAL) and Con-sumer Price index for Industrial Workers'(CPIIW) are used to deflate all monetary vari-ables. The reference period for the deflator is1973_1974. The deflators are obtained fromGovemment of India publications (Indian LaborHandbook, the Indian Labor Joumal, the IndianLabor Gazette, and Reserve Bank of India Re-port on Currency and Finance).
REFERENCES
Banerjee, Abhijit and Somanathan, Rohini."Caste, Community and Collective Action:The Political Economy of Public Good Pro-vision in India." Mimeo, Massachusetts Insti-tute of Technology. 2001.
" Four types of reforms were considered: tenancy re-form, abolition of iniermediaries. ceiling.s on land holdings,and consolidation of land plots.
Baxi, Upendra. Crisis and change in contemporaryIndia. New Delhi: Sage Publications, 1995.
Besley, Timothy and Burgess, Robin. "Land Re-fonn. Poverty Reduclion. and Growth: Evi-dence from India." Quarterly Journal ofEconomics. May 2000. 115(2), pp. 389-430.
Besley, Timothy and Case, Anne. "Unnatural Ex-periments? Estimating the Incidence of En-dogenous Policies." Economic Journal,November 2000. 7/0(467). pp. F672-94.
Besley, Timothy and Coate, Stephen. 'An Eco-nomic Model of Representative Democracy."Quarterly Journal of Economics, February1997. II2(\), pp. 85-114.
Bose, Shankar and Singh, V. B. State elections inIndia: Data handbook on Vidhan Sabha elec-tions 1952-85. New Delhi: Sage Publica-tions. 1987.
Cameron, Charles; Epstein, David andO'Halloran, Sharyn. "Do Majority-MinorityDistricts Maximize Black Representation inCongress?" American Political Science Re-view, December 1996. 90(4). pp. 794-812.
Chattopadhyay, Raghabendra and Dutio, Esther."Women as Policy-Makers: Evidence froman India-Wide Randomized Experiment."National Bureau of Economic Research(Cambridge. MA) Working Paper No. 8615,December 2001.
Chhibber, Pradeep. Democracy without associ-ations: Transformation of the party systemand social cleavages in India. Ann Arbor,MI: University of Michigan Press. 1999.
Downs, Anthony. An economic theory of democ-racy. New York: HarperCollins, 1957.
Dreze, Jean and Sen, Amartya. India: Economicdevelopment and social opportunity. NewDelhi: Oxford University Press. 1995.
Dushkin. Lelah. "Scheduled Caste Politics." in J.Michael Maliar. eds.. The untouchable incontemporary India. Tucson, AZ: Universityof Arizona Press. 1972, pp. 167-200.
Edinnd, Lena and Pande, Rohini. "Why HaveWomen Become Left-Wing? The PoliticalGender Gap and the Decline in Marriage."Quarterly Journal of Economics. August2002. 7/7(3). pp. 917-61.
Galanter, Marc. "Compensatory Discriminationin Political Reservation: A Preliminary As-sessment of India's Thirty-Yeiu" Experiencewith Reserved Seats in Legislature." Eco-nomic and Political Weekly, February 1979,14.
VOL 93 NO. 4 PANDE: POUTICAL RESERVATION AND POUCY OUTCOMES
. _ ^ . Competing equalities: Law and thebackward classes in India. Berkeley. CA:University of California Press, 1984.
Grofman, Bernard and Lijphart, Arend. Elec-toral laws ami their political con.sequences.New York: Agathon Press. 1986.
Holzer, Harr; ' and Neumark, David. "AssessingAffimiative Action." Journal of Economic Lit-erature, September 2(XX). JS(3). pp. 483-568.
Hnsted, Thomas and Kenny, Lawrence. "The Ef-fect of the Expansion of the Voting Franchiseon the Size of the Govemment." Journal ofPolitical Economv, February 1997. 705(1),pp. 54-82.
,Toshi, Barbara. Democracy in search of equal-ity: Untouchable politics and Indian socialchange. New Delhi: Hindustan PublishingCompany, 1982.
KaU, Joseph and Znpan, Mark. "Capture andIdeology in the Economic Theory of Poli-tics." American Economic Review, June1984, 74(3). pp. 279-300.
Levitt, Steven D. "How Do Senators Vote? Dis-entangling the Role of Voter Preferences.Party Affiliation and Senate Ideology."American Economic Review. June 1996.H6(3), pp. 425-41.
Mendelsohn. Oliver and Vicziany, Marika. Theuntouchables: Subordination, poverty andthe state in modern India. Cambridge: Cam-bridge University Press. 1998.
Oshorne, Martin and Slivinski, Al. "A Model ofPolitical Competition with Citizen-Candidates."
Quarterly Joumal of Economics, February1996. //7(1), pp. 65-96.
Overhy. Marvin and Cosgrove. Kenneth. "Unin-tended Consequences? Racial Redistrictingand the Representation of Minority Inter-ests." Journal of Politics, May 1996. 58(2),pp. 540-50.
Ozler, Berk; Datt, Gaurav and Ravallion, Martin."A Database on Poverty and Growth in In-dia." Mimeo, World Bank, 1996.
Pande, Rohini. "Minority Representation andPolicy Choice: The Significance of LegislatorIdentity." Development Economics Discus-sion Paper Series, STICERD (London) Dis-cussion Paper No. 16, June 1999.
Parikh, Sunita. Politics of preference. Ann Ar-bor, MI: University of Michigan Press. 1997.
Riviere, Anouk. "Citizen Candidacy. Party For-mation and Duverger's Law." Royal Hollo-way Discussion Paper Series (London)Discussion Paper No. dpeOO/l, 2000.
Rule, Wilma and Zimmerman, Joseph. Electoral.'iy.^tems in comparative perspective: Theirimpact on women and minorities. Westport.CT: Greenwood Press, 1994.
Snyder, James and Ting. Michael. "An Informa-tional Rationale for Political Parties." Amer-ican Journal of Political Science, January2002. 46(1). pp. 90-110,
Weingast, Barry. "A Rational Choice Perspec-tive on Congressional Norms." AmericanJoumal of Political Science, May 1979,2J(2), pp. 245-62.
