ta4319 - collective for social science...

TA4319

Determinants and Drivers of Poverty Reduction and ADB’s Contribution in Rural Pakistan

Governance Impediments to Pro-Poor Change in Pakistan

Ali Cheema

© 2007 Asian Development BankPakistan Resident MissionAll rights reserved.

First printing February 2007

The views expressed in this book are those of the author and do not necessarily reflect the views andpolicies of the Asian Development Bank or its Board of Governors or the governments they represent.

The Asian Development Bank does not guarantee the accuracy of the data included in this publicationand accepts no responsibility for any consequence of their use.

Use of the term “country” does not imply any judgment by the author or the Asian Development Bankas to the legal or other status of any territorial entity.

Asian Development BankPakistan Resident MissionOPF Building, Shahrah-e-JamhuriyatG-5/2, Islamabad, P.O. Box 1863PakistanTel +92 51 282 5011–16Fax +92 51 282 3324, 227 [email protected]


iii

Preface

The Asian Development Bank (ADB) and Pakistan have been development partners for almost39 years. Assistance has averaged close to $1 billion a year since 2004 supporting good governance, sus-tained pro-poor growth, inclusive social development, and targeted investment projects to reducepoverty and generate employment.

ADB has an active research agenda on economic and social development issues. A technicalassistance titled ‘Determinants and Drivers of Poverty Reduction and ADB’s Contribution in RuralPakistan,’ was approved by ADB in March 2004. It supported a detailed study of both chronic andtransitory rural poverty and sought to identify appropriate policy and implementation measures to pro-mote sustained poverty reduction. It was financed on a grant basis from the Poverty ReductionCooperation Fund (PRF). The PRSP Secretariat of the Ministry of Finance, Government of Pakistan,was the Executing Agency of the technical assistance.

The thematic papers presented here are interim outputs of this project. They are based primarily onfieldwork carried out by the authors and their research teams in diverse regions of rural Pakistanbetween June 2005 and March 2006. The fieldwork included three types of data collection: (a) qualita-tive research at the village level, (b) extended household survey of selected villages, and (c) cross-villagerapid surveys in selected districts.

The innovative “drivers of change” approach to poverty reduction focuses on long-term structural driv-ers and determinants as well as on impediments to pro-poor change. Poverty-reduction correlates withhigher economic growth, more equal distribution of assets and opportunity, higher real wages, bettersocial indicators, improved governance, better access to resources, level playing field in markets, socio-economic mobility, and pro-poor delivery of public goods and services.

The thematic papers are being made available to support and enrich the debate on poverty reductionand economic growth. Comments and feedback received are appreciated.

Peter L. FedonCountry DirectorADB Pakistan Resident Mission

Abbreviations

ADB— Asian Development Bank

CCB— community citizen board

DFID— UK Department for InternationalDevelopment

LUMS— Lahore University of ManagementSciences

NWFP— North West Frontier Province

PRM— Pakistan Resident Mission

SC— school council

UNDP— United Nations DevelopmentProgramme

VNC— village and neighborhood council

NOTES

The analysis in this paper is up-to-date untilJanuary 2006 when the study was completed.

GLOSSARY

barani rain-fed (normally with reference toagricultural land)

deh (in Sindh) the lowest-defined geo-graphic unit functioning as a rev-enue village

goth (in Sindh) hamlet: a consolidatedgeographic settlement that is nor-mally but not necessarily smallerthan a revenue village

kissan farmer

mauza (in Punjab and NWFP) the lowestdefined geographic unit functioningas a revenue village

naib nazim deputy mayor: convenor of localgovernment council; takes overfrom nazim when he or she is away

nazim mayor: elected head of local gov-ernment

pind (in Punjab) hamlet: a consolidatedgeographic settlement that is nor-mally but not necessarily smallerthan a revenue village

taluka (in Sindh) administrative unit oflocal government, equivalent to asubdistrict

tehsil administrative unit of local govern-ment, between a district and union


iv


v

Contents

Preface

Abbreviations

1 Introduction 1

1.1 Social Gaps and Poverty 1

1.2 Governance Failures and Social Gaps 2

1.3 Objectives and Questions of the Study 3

1.3.1 Questions 3

1.3.2 Research Design 3

1.3.3 Contribution of the Study 4

1.3.4 Structure of the Study 5

2 Conceptualizing Governance Failures: A Framework and Empirical Strategy 6

2.1 World Development Report (2004) Governance Framework 6

2.2 Empirical Strategy to Measure Anti-Poor Accountability Failures 8

2.2.1 Focus of Empirical Strategy 8

2.2.2 Variables and Empirical Methodology 8

3 Devolution and the Long and Short Routes of Accountability 12

3.1 Empowering the Union 12

3.2 Engendering User Power and Citizens’ Participation 16

4 Data and Fieldwork Methodology 17

5 Analyzing the Long Route of Accountability in Pakistan 19

5.1 Barriers to Entry in the Electoral Sphere for the Poor and Women 19

5.1.1 Voting and Gender 19

5.1.2 Voting, Women, and the Poor 21

5.1.3 Male Voting and the Poor 22

5.2 Dependent Voters and Pro-Poor Accountability 23

5.2.1 Voting Typology Table 24


vi

5.2.2 Village Correlations 26

5.2.3 Conclusion 28

5.3 Do the Poor Have Equal Access to Union Nazims and Councilors? 29

5.3.1 Estimating Access 30

5.3.2 Extent of Anti-Poor Bias 31

5.4 Patterns of Bias in the Targeting of Localized Public Goods 31

5.4.1 Why is Measuring the Targeting Bias Insightful? 31

5.4.2 Methodology and Data 32

6 Analyzing the Short Route of Accountability in Pakistan 35

6.1 Awareness of and Participation in CCBs 36

6.2 Awareness of SCs 37

6.3 The Political Accountability Index 37

7 Conclusion 39

7.1 Long Route of Accountability 39

7.1.1 Electoral Sphere 39

7.1.2 Planning and Budgeting Spheres 39

7.1.3 Anti-Poor Targeting Bias in Service Delivery 40

7.2 Awareness of Citizen-Based Service Delivery and Oversight Bodies 40

7.3 Persisting Structural Constraints to Accountability 41

7.3.1 Degrees of Socioeconomic Inequality 41

7.3.2 Degree of Political Competition 43

7.4 Constraints on Voter Turnouts among the Poor 43

7.5 Is Literacy a Panacea? 44

7.6 Persisting Institutional and Information Gaps 45

References 47


vii

Tables

Table 1: Social Gaps and Poverty 1

Table 2: Drivers of/Impediments to Pro-Poor Accountability 10

Table 3: Male and Female Voting Turnout 19

Table 4: Female Voting Turnout by Village 20

Table 5: Incidence of Male and Female National Identity Card Possession 21

Table 6: Incidence of Female Identity Card Possession by Village 21

Table 7: Proportion of Male Adults who Voted 22

Table 8: Voting Typology Table 25

Table 9: Village Indicators 26

Table 10: Type of Voting and Potential Determinants 28

Table 11: Respondents Reporting Contact with Union Nazim/Naib Nazim to Convey

Needs in the Last 6 Months 30

Table 12: Respondents Reporting Contact with Union Councilors to Convey Needs

in the Last 6 Months 30

Table 13: Awareness of Citizen Community Boards 35

Table 14: Participation in Citizen Community Boards 36

Table 15: Awareness of School Councils 36

Table 16: Male Workers by Sector and Activity (Primary Occupation) 42

Figures

Figure 1: Analyzing Governance 6

Figure 2: Electoral Linkages 12

Figure 3: Interlinked Representation 13

Figure 4: Electoral and Budgetary Linkages 14

Figure 5: Participatory Budgeting in West Bengal 15

Appendix

Background Data 49

1 Introduction

1.1 Social Gaps and Poverty

There is growing appreciation in the aca-demic and policy literature that poverty is a mul-tifaceted phenomenon encompassing more thanincome and consumption deprivation (Dreze andSen 1996, and Gazdar 2005).1 Numerous studies inboth the international and Pakistani context showthat poverty is highly correlated with the acquisi-tion of social, cultural, and political assets andforms of capital that are essential determinants ofa citizen’s position on the poverty scale.2 The

important finding is that nonincome correlates ofpoverty are as, if not more, important as income

correlates of poverty.

The worrying but unsurprising findingfrom the Pakistani literature is that there is astrong association between poverty and socialgaps in Pakistan. The term “social gap” describesthe gap between Pakistan’s social developmentindicators relative to comparator countries withsimilar levels of income and similar rates ofgrowth. As indicated in Table 1, this is true forimportant social development indicators such asilliteracy, primary school enrolment, basic health


1

Table 1: Social Gaps and Poverty

Indicators Poor (%) Nonpoor (%)Overall Net Primary Enrolment

Total 36.6 59.3Male 43.0 66.0Female 30.2 52.3

Rural Net Primary EnrolmentTotal 33.4 53.6Male 41.7 61.7Female 25.0 44.8

Access to Drinking WaterPiped 18.7 28.2Open Source 24.0 18.6

Access to Drainage SystemUnderground 7.8 17.2Open Drain 29.8 34.3No Drain 52.4 48.5

Access to ToiletIn Home 40.9 61.0Outside Home 59.1 39.0

Access to Utilities (% of Population living in Households Connected to)Electricity 52.2 76.0Gas 10.9 22.9Telephone 7.4 22.2

Source:World Bank. 2004. World Development Report 2004, Making Services Work for Poor People. Washington, DC: World Bank. pp. 36–7.

1 See: Dreze, J., and A. Sen. 1996. Indian Development: Selected Regional Perspectives. New Delhi: Oxford University Press; and Gazdar, H. 2005. Scoping Paper.Revised draft submitted for Determinants and Drivers of Poverty Reduction and ADB’s Contribution in Rural Pakistan. Asian Development Bank (ADB),Islamabad.

2 See: Charlton, J., M. Cyan, Z. Hasnain, N. Manning, D. Porter, and H. Sharif. 2003. Devolution in Pakistan—Preparing for Service Delivery Improvements.World Bank, ADB, and UK Department for International Development (DFID); Dreze and Sen (1996) (footnote 1); Easterly, W. 2003. The PoliticalEconomy of Growth without Development: A Case Study of Pakistan. In In Search of Prosperity: Analytical Narratives on Economics Growth. Edited by D.Rodrick. New Jersey: Princeton University Press; Gazdar, H. 2002. A Qualitative Survey of Poverty in Rural Pakistan: Methodology, Data, and Main Findings .Karachi: Collective for Social Science Research; and World Bank. 2002. Pakistan Poverty Assessment. Poverty in Pakistan: Vulnerabilities, Social Gaps, and RuralDynamics. Washington, DC: World Bank.

indicators, and access to social and physical infra-structure (United Nations DevelopmentProgramme [UNDP] 2004 and World Bank 2002[footnote 2]).3 What is even more worrying is thatthese poverty-creating “social gaps” persist inPakistan in spite of the country’s relatively fasthistorical rate of growth, and continue to persisteven after controlling for differences in the mag-nitude of public spending on social services(Easterly 2003, footnote 2).

The multifaceted nature of poverty is alsohighlighted by studies of poverty profiles,4 whichidentify patterns of persistence of poverty interms of urbanization and in terms of regionaland household characteristics. The literature onPakistan shows that rural poverty tends to bemuch higher than urban poverty. It also high-lights considerable inter- and intra-provincial vari-ation in the persistence of poverty. Furthermore,results based on household survey data show thataccess to agricultural land and, in particular, landownership is a strong negative correlate of pover-ty as is levels of education, whereas levels ofdemographic dependency are positively related topoverty.

The poverty literature also suggests thatlevels of deprivation tend to be higher amongsocially excluded kinship groups, women, and themarginalized (Gazdar 2002 and World Bank 2002,footnote 2). That is, an individual’s gender andposition in the social hierarchy are as importantdeterminants of poverty as access to income andwealth. Furthermore, the literature shows thatlack of access to social services mirrors the find-ings regarding poverty profiles. Lack of access tosocial services is much worse for rural areas andtends to be higher among socially excludedgroups, women, and the marginalized. This sug-gests that there exists a vicious cycle betweensocial gaps and poverty—with poverty resulting inhigh levels of social deprivation and the conse-quent low levels of human development reinforc-ing poverty. In turn, this vicious cycle appears tomake poverty chronic for particular regions,

areas, gender groups, and socioeconomic groups.

A recognition that social gaps areextremely important correlates of poverty is auseful starting point for the design of a more com-prehensive set of poverty reduction policies andstrategies. However, a crucial issue from the poli-cymaking perspective is to identify processes anddrivers that help in the reduction of income andnonincome sources of poverty. In this perspec-tive, identifying a correlate of poverty is a startingpoint for the analysis of socioeconomic and gov-ernance impediments that drive the particular sta-tistical association in question. The task of policyis not only to identify which correlates of pover-ty to target but also to alter these impedimentswith targeted interventions to allow male andfemale citizens to break out of the vicious cycle ofpoverty.

1.2 Governance Failures and Social Gaps

The consensus in the literature is thatshortfalls in public spending do not provide theentire explanation for Pakistan’s social develop-ment gaps. Instead, the literature argues that amore robust explanation has to be sought in fail-ures of “governance” associated with public serv-ice providers, which include policymakers/politi-cians, bureaucrats, and frontline service providers(Charlton, et. al. 2003, footnote 2). The impor-tance of “governance impediments” as a variablethat explains Pakistan’s social gaps is best appreci-ated by the following findings from the literature:

w Pakistan’s social gaps persist relative to com-parator economies even after controlling forshortfalls in magnitudes of social spending(Easterly 2003, footnote 2).

w Pakistan’s social development has beenplagued by considerable inequality in socialdevelopment outcomes across not only poorand nonpoor households but also: (i) rural-urban areas, (ii) gender, (iii) class, (iv) socialgroups and, (v) in certain areas, even betweenvillages that lie in close proximity to one


2

3 United Nations Development Programme (UNDP). 2004. Human Development Report, Cultural Liberty in Today’s Diverse World. USA: Oxford University Press.4 For a comprehensive review of the literature, see Gazdar (2005) (footnote 1).

another (Cheema and Mohmand 2004;Cheema and Mohmand 2005; Easterly 2003[footnote 2]; and Keefer, Narayan, andVishwanath 2005).5

These findings suggest that Pakistan’ssocial gaps are rooted in political economy andgovernance failures. To the extent that theinequality of social provision outcomes reflectsthe failure of state targeting, it clearly revealspolitical and planning biases in state provisionacross different wealth, income, and social groups,and across different localities and jurisdictions.Similarly, the inequality of outcomes betweenpoor and nonpoor cohorts points to processes ofsociopolitical exclusion, which make the poorparticularly under-provided and this, in turn,appears to further reinforce an existence at themargins of poverty. It appears that “governance-related impediments” have historically muted theability of the poor, women, and marginalized cit-izens to hold governments and government func-tionaries accountable. This has, in turn, set incen-tives for governments and government func-tionaries to be unresponsive to the poor, women,and socially excluded and marginalized groups.This suggests that a viable poverty reduction strat-egy needs to analyze and overcome governance-related impediments to poverty.

1.3 Objectives and Questions of the Study

1.3.1 Questions

The main aim of this study is to analyzegovernance impediments that reinforce povertyin Pakistan. The study is motivated by two simplebut important questions:

(i) What constrains the poor, women, and mar-ginalized groups from holding the state andstate functionaries accountable?

(ii) What impediments to pro-poor accountabil-ity exist in the electoral and service delivery

spheres?

Answers to these questions are sought in the fol-lowing corollary questions:

(i) Do the poor, women, and marginalizedgroups have equal access to the electoralsphere or are there barriers to entry?

(ii) Do the poor, women, and marginalizedgroups have access to and participate in elec-toral coalitions that represent their interestsor are they dependent voters in electoralcoalitions that are led by and represent eliteand nonmarginalized groups?

(iii) Do the poor, women, and marginalizedgroups have equal access to local-level politi-cians for the purposes of conveying develop-ment needs? If not, what factors explain thislack of access?

(iv) Are there systematic biases in the targetingof local public goods provision in favor ofsocially and economically dominant groups?

(v) Is there equal awareness of and participationin citizen-based participatory organizationsthat are linked to service delivery?

By addressing these questions this studyaims to identify and analyze:

(i) important determinants of anti-poor“accountability failures” in Pakistan; and

(ii) processes that underlie the persistence ofthese determinants in Pakistan.

1.3.2 Research Design

The study addresses these questions in thecontext of Pakistan’s recent devolution reforms.The study, however, is not designed to assess theimpact of devolution reforms on poverty-reduc-tion outcomes. This is because, as pointed out byCharlton, et. al. (2003), the impact of devolution

Introduction

3

5 See: Cheema, A., and S. Mohmand. 2004. Provision Responses to Devolved Service Delivery—Case Evidence from Jaranwala Tehsil. Manuscript. LahoreUniversity of Management Sciences (LUMS)-McGill Social Enterprise Development Programme. LUMS, Lahore; Cheema, A., and S. Mohmand. 2005. ThePolitical Economy of Devolved Provision: Equity-based Targeting or Elite Capture—Case Evidence from 2 Pakistani Unions. Manuscript. LUMS, Lahore;and Keefer, P., A. Narayan, and T. Vishwanath. 2005. The Political Economy of Decentralization in Pakistan. In Decentralization in Developing Countries:A Comparative Perspective. Edited by P. Bardhan and D. Mookherjee. The MIT Press.

on poverty outcomes is difficult to assess at pres-ent because local government politicians andbureaucrats still do not have autonomous man-dates and finances (footnote 2). Furthermore, con-siderable overlap of jurisdictions continue to existbetween governments at different tiers, whichmakes it next to impossible to estimate and assessthe impact devolution reforms are having onpoverty-reduction outcomes.

Instead, the study uses an indirect designto analyze the effect that recent governancereforms in Pakistan are having on pro-pooraccountability. The study is designed to analyzethe interface between the poor, women, and mar-ginalized citizens and local-level politicians, andbetween these citizen cohorts and citizen-basedservice delivery bodies that have been instituted asa result of the recent reforms. With regard tothese interfaces the study asks whether the rela-tive access of the poor, women, and marginalizedgroups to the local-level electoral sphere, to localgovernment politicians, to planning and budget-ing institutions, and to citizen-based service deliv-ery bodies is in line with nonpoor, male, and non-marginalized groups in the new governance envi-ronment.

This is a key question to ask because animportant logic underlying the devolutionreforms is to create space for participatory andinclusive citizen-based governance, and thereforeit is essential to analyze and ask:

w have adequate institutional spaces been createdto allow participation by the poor, women,and marginalized citizens in governance?

w whether the poor, women, and marginalizedcitizens are able to capture the new governancespaces that have opened up at the local level?and

w what variables and impediments constrain thepoor, women, and marginalized groups fromcapturing these new governance spaces?

Answering these questions will help assessthe degree to which new governance spaces havebecome available and the extent to which thesenew governance spaces are inclusive of the poor,women, and the marginalized, which is a neces-sary precondition for heightened pro-pooraccountability. Answering these questions willalso help in the design of second-generation poli-cies that strengthen devolution as a pro-poorreform.

1.3.3 Contribution of the Study

This study makes an important contribu-tion to the policy literature and debate on pover-ty because it directly estimates for the first timethe relative degree of access that poor citizencohorts have to (i) the electoral sphere, (ii) citizen-based service delivery bodies, and (iii) electedlocal-level representatives. The main empiricalestimate uses a simple measure to quantify pro-poor accountability, that is, how well poor citi-zens are faring relative to the nonpoor in terms ofthese three accountability variables. As men-tioned in Gazdar (2007),6 rigorous quantitativemeasures and techniques are used to preciselydefine poverty scores and the poor cohort. Thestudy does not stop at quantifying the relativeaccess that the poor have, it also uses rigorousqualitative fieldwork to interpret quantitativeresults, thereby unraveling the impediments thatthe poor, women, and marginalized groups face ingovernance relationships with local-level electedrepresentatives and when participating in citizen-based service delivery bodies (Gazdar 2005 [foot-note 1], pp. 1–3).

The study hypothesizes that, in additionto poverty impediments to pro-poor accountabil-ity, it can arise from exogenous factors such as (i)the degree of socioeconomic inequality embeddedin social structures, (ii) historical resource endow-ments, (iii) agro-ecological conditions, (iv) degreesof patriarchy, and (v) existing physical infrastruc-ture; as well as from endogenous factors such as (i)property rights over land, (ii) access to land


4

6 Gazdar, H. 2007. Rural Economy and Livelihoods in Pakistan. Thematic paper prepared under TA4319, Determinants and Drivers of Poverty Reductionand ADB’s Contribution in Rural Pakistan. ADB, Islamabad.

through tenancy, (iii) possibilities of accessing thelabor market through migration, (iv) differencesin literacy and schooling attainments, and (v)dependence on landlord credit, etc. In addition toestimating the correlation between being poorand having access to different spheres of account-ability, the study also estimates the effect each ofthe exogenous and endogenous drivers mentionedabove has on pro-poor accountability in Pakistan.

1.3.4 Structure of the Study

Section 2 of this paper introduces a con-ceptual framework that is used to analyze gover-nance failures for the purpose of this study. Thisis followed by a description of the empirical strat-egy and research design. It also introduces themain questions that will be tested and analyzed inorder to get a handle on the nature and causes ofpro-poor governance failures in Pakistan. Section3 describes the changes in the governance frame-work that have been brought about by Pakistan’srecent devolution reforms. It highlights the needto interpret many of the findings of this study inthe context of devolution. This section also sharp-ens many of the hypotheses that are introduced inthe previous section. Section 4 describes thenature of the data used and introduces the readerto the fieldwork methodology. Sections 5 and 6present the results of how well pro-poor gover-nance is working in Pakistan and identify theimpediments that continue to constrain the poor,women, and illiterate and socially marginalizedcitizens from holding the state accountable.Section 7 summarizes the findings of the studyand makes some initial recommendations regard-ing the drivers of pro-poor governance inPakistan.

Introduction

5

2. Conceptualizing GovernanceFailures: A Framework andEmpirical Strategy

2.1 World Development Report (2004)Governance Framework

The World Bank’s (2004) W o r l dDevelopment Report7 has developed a rigorousframework that provides a useful starting pointfor an analysis of the weaknesses and strengths ofgovernance relationships in specific contexts. Theframework analyses the “governance relation-ship” in terms of relationships of accountabilitybetween four sets of actors that include (i) citizensand clients, (ii) politicians and policymakers, (iii)organizational providers, and (iv) frontline profes-sionals. A schematic description of an ideal gover-nance relationship (Figure 1) is as follows:

An ideal governance relationship is onewhere citizens choose representatives that form agovernment, are delegated the task of public serv-ice delivery (delegation), and have adequate

finances and budgetary control at their disposal(financial control). Citizens, in turn, have at theirdisposal electoral and other legal mechanisms thatallow them to evaluate executive performance(information) and enable them to reward orpenalize politicians’ behavior (enforceability).The World Bank (2004) classifies this as the “longroute of accountability” (Figure 1)—this capturesmore than the exercise of “voice” in the electoralsphere, although this, too, is an extremely impor-tant component. Besides electoral voice, the longroute includes policy-directed actions and process-es such as constitutional and judicial activism,media oversight, and lobbying.

Evidence from around the world suggeststhat citizens “are combining electoral accountabil-ity and participation with what would traditional-ly have been considered the official accountabilityactivities of the state” (World Bank 2004 [footnote

7], pp. 85–86). These initiatives increase accounta-bility at various levels by strengthening citizenvoice in and citizen oversight of government poli-cies, projects and budgeting, and planning mecha-


6

Figure 1: Analyzing Governance

Source: World Bank. 2004. World Development Report 2004, Making Services Work for Poor People. Washington, DC: World Bank.

7 World Bank. 2004. World Development Report 2004, Making Services Work for Poor People. Washington, DC: World Bank.

nisms. Accountability is expected to be strength-ened because these initiatives are supposed to givecitizens not only greater “voice” but make themmore informed about the actions of government,thus making concerted action among citizens eas-ier (World Bank 2004 [footnote 7], pp. 86–89).However, it is important to recognize the follow-ing in the long route of accountability:

w Citizens hold politicians accountable by par-ticipating and competing in political and elec-toral process through formal (political partiesor associations) and informal coalitions (castegroups, religious networks, and clan groups).Competition between these coalitions definesthe collective objectives for public policy andpublic service delivery as each coalitionattempts to influence politicians’ actions intheir favor.

w The strength of the accountability relationshipbetween politicians and the poor, women, andmarginalized citizens depends on (i) whetherthese citizens have equal access to the electoralsphere, and (ii) whether they are able to cre-ate/participate in effective electoral coalitionsthat represent their interests.

w The ability of these citizens to hold politiciansaccountable to collective objectives depends onthe efficacy of mechanisms of information gen-eration and enforcement. This, in turn,depends on the existence of mechanisms thatgive citizens a greater participatory and over-sight role in the formulation and execution ofgovernment policies, projects and budgeting,and planning mechanisms.

Informed citizen participation in the elec-toral sphere is, however, not sufficient to enforceaccountability. Accountability also depends uponthe compact between politicians/policymakersand organizational providers. For collective objec-tives to be translated into outcomes, policymakershave to delegate adequate powers, authorities, andresources to organizational service providers. Inaddition, they have to monitor the actions oforganizational service providers and design com-

patible incentives for service providers to deliveron collective outcomes. However, it is importantto recognize that effective and informed citizenparticipation in the electoral sphere will set incen-tives for politicians to design and enforce incen-tive compatible compacts. In this sense, the sec-ond rung of the accountability relationship isendogenous to the first one.

Moving down the accountability chain,organizational service providers have to createcompatible incentives for frontline providers andmanage their actions in a way that ensures thedelivery of collectively mandated outcomes. Theframework acknowledges that it may not be pos-sible for policymakers/politicians to specify allactions in the compact nor may it be possible forall information to be revealed between citizensand politicians. Given these constraints, welfarecan be enhanced by

w having citizens directly reveal informationabout their demands for services as clients,

w having them monitor and reward the actionsof frontline providers, and

w allowing citizens to deliver services in collabo-ration with public service providers and toengage with these providers in a monitoringand oversight role.

This is classified as the short route ofaccountability, which envisages that frontlineproviders can be made directly accountable toclients by passing decisions and authorities direct-ly to citizens and communities.

The key difference between the long andshort routes of accountability as described here isthe different relationship between citizens, front-line service providers, and executive and politicalrepresentatives. The latter describes the relation-ship between citizens as clients and potential serv-ice providers, and the executive as the public serv-ice provider. In the long route, the emphasis is onthe accountability of political representatives to cit-izens as electoral candidates and as formulators ofgovernment policy, projects, budgets, and plans.


7

2.2 Empirical Strategy to Measure Anti-Poor Accountability Failures

2.2.1 Focus of Empirical Strategy

Governance failures are defined as devia-tions from this ideal governance relationship andcan occur along any of the routes taken by theaccountability relationship described above. Theemphasis of the study is on failures that occuralong the first route of the governance relation-ship, i.e., along the long route. There are two rea-sons for this focus. First, as already mentioned,with the recent reforms, there is considerable lackof clarity with regard to mandates, control overfinances, and jurisdictional authority between dif-ferent tiers of governments. This makes itextremely difficult to analyze failures along thesecond and third routes of the accountability rela-tionship, especially as the institutional structureitself is in transition.

Similarly, issues of institutionalizationcontinue to plague the newly instituted citizen-based service delivery and oversight bodies, whichmakes it extremely hard to evaluate the weakness-es in the short route of accountability. However,with regard to the short route, this study asks amore limited question, i.e., to what degree are thepoor, women, and marginalized and sociallyexcluded citizens aware of these service deliveryand oversight bodies and how does the level ofawareness in these citizen types compare withother citizen types. Although, this is a much morerestricted question, it gives us important insightsinto the process of institutionalization of these cit-izen-based bodies and how involved the poor,women, and marginalized and socially excludedcitizens are in this process. It is also an importantquestion to ask because local governments havecompleted their first electoral cycle at the time ofthis study.

Second, and more importantly, this studytakes a micro-village and household-level perspec-tive, which implies that the focus of the analysis isalong routes where the citizen directly interfaces

with governance structures, and not on weakness-es in administrative compacts, structures, and pro-cedures. Given this focus, an important caveatmust be kept in mind when interpreting theresults: how well a functioning long route ofaccountability might in and of itself fail to trans-late into pro-poor social delivery outcomes if it isoffset by weaknesses in administrative compacts,structures, and procedures.

With this in mind, it is, nonetheless,important to recognize that a well-functioninglong route of accountability for the poor will go along way in redressing accountability failures inPakistan’s governance framework. Weaknesses inadministrative compacts, structures, and proce-dures in the context of devolution have been rig-orously analyzed by Charlton, et. al. (2005) study(footnote 2). The current study should be treatedas a companion study that analyses the weakness-es in pro-poor accountability along the routes ofinterface between politicians and serviceproviders and the poor, women, and marginalizedcitizens.

2.2.2 Variables and Empirical

Methodology

Dependent Variables

The study uses three dependent variablesof access as proxy measures for pro-poor account-ability:

(i) access of different citizen types8 to the elec-toral sphere,

(ii) access of different citizen types to electedlocal-level representatives involved in plan-ning and budgetary needs assessment, and

(iii)access of different citizen types to citizen-based service delivery bodies.

These variables are first-order governancemeasures that capture the long and short routes ofaccountability that allow different citizen cohorts


8

8 This study is concerned with the poor, women, and socially excluded and marginalized citizens.

to hold government, politicians, and serviceproviders accountable through direct participa-tion in citizen bodies, in planning and budgetingmechanisms, and through direct participation inthe electoral sphere.

In addition, for a limited sample of vil-lages, this study uses a fourth dependent variableto estimate the bias in the targeting of localizedpublic good provision between elite and nonelitehouseholds by union nazims (mayors)/naib naz -ims (deputy mayors). Estimating biases in the tar-geting of localized public goods at the householdlevel allows important insights into the nature ofthe long route of pro-poor accountability at theunion council level. The emergence of anti-poortargeting biases will reveal that distortions existnot only in electoral accountability but also in thepro-poor nature of budgeting and planning mech-anisms. The dependent variable used to estimatethese biases is the post-devolution change in theprovision of sanitation and soling (street paving)by local governments at the household level. Thisvariable is generated by using pre-2001 baselinedata on provision and comparing it to a currentprovision line. Household data for this exercisecame from the village census instrument used bythe present study as well as the Cheema andMohmand (2005) survey instrument (footnote 5).

The main empirical strategy of the studyis to analyze whether a positive correlation existsbetween being poor and the four dependent vari-ables defined above. This provides a direct meas-ure of pro-poor accountability. This empiricalstrategy is an important contribution to thePakistani literature on governance as there is nostudy that directly estimates and analyzes the cor-relation between the poverty status of a house-hold and its access to the electoral sphere and tomechanisms of planning and service delivery.This correlation is important from a policy view-point because it is imperative to analyze whether“the poor” as a category are systematically dis-

criminated against in participatory bodies, withinthe electoral sphere, in terms of access to local-level politicians, in planning and budgeting mech-anisms, and in terms of the targeting of schemes.A negative and significant correlation between

governance variables and being a poor household9

should alert the policymaker to the immediacy ofdesigning specific policies that are aimed atincreasing the stake that the poor have in gover-nance mechanisms.

Independent Variables

This study, however, goes beyond thissimple correlation as it recognizes that poverty isa multifaceted phenomenon and that there exist anumber of exogenous and endogenous micro-drivers of pro-poor accountability. Table 2 liststhe more important exogenous and endogenous

micro-drivers of pro-poor accountability. 10

In addition to analyzing the effect ofpoverty on governance measures, the study esti-mates the effect these other exogenous andendogenous drivers have on the four dependentvariables that are being used as proxy measures toestimate the strength of the long and short routesof accountability. It is important to recognize thatthe micro-drivers/impediments listed in Table 2can contribute to the weakening of the accounta-bility relationship between citizens and govern-ment/politicians/service providers. This is likelyto occur when these impediments (i) raise entrybarriers for the poor, women, and the marginal-ized to participate in electoral, political, and plan-ning processes; (ii) increase the costs of collectiveaction for the poor, women, and socially exclud-ed citizens; and (iii) impede the formation of hor-izontal class-based coalitions that represent thepoor, women, and the marginalized.1 1

Furthermore, anti-poor governance failures arelikely if certain impediments raise the cost ofaccessing information on executive performance.Information constraints are more likely to be

Conceptualizing Governance Failures: A Framework and Empirical Strategy

9

9 For estimates of household-level poverty scores, rankings, and indices, see Gazdar (2007) (footnote 6).

10 For a wider discussion around the drivers of pro-poor change, see the references in Gazdar (2005) (footnote 1).

11 For the importance of class-based coalitions as a means to fight poverty in the Indian case, see Kohli, A. 1987. The State and Poverty in India: The Politics of

Reform. Cambridge University Press.

binding if the poor lack easy access to informationeither because (i) they lack access to education or(ii) they lack access to networks that in turn haveaccess to critical information on a politician’s per-formance as well as budgeting and planning.12

The literature suggests that for the longroute of accountability to work well, the poor,women, and socially excluded citizens must be

able to:

w create political and electoral coalitions thatrepresent their interests;

w freely participate in dominant coalitions thatare representative of their needs and to voicetheir demands in these coalitions;

w exercise an independent vote and voice irre-spective of whether they form class-basedcoalitions or cross-class representative coali-tions;

w have access to information on politicians’behavior; and

w participate in budgeting and planning deci-sions.

Similarly, the impediments listed in Table2 are likely to weaken the short route of account-ability because of (i) weaknesses in selection andrepresentation in client/user organizations; (ii)lack of awareness of and information on theseorganizations; and (iii) because the poor, women,and socially excluded households find it difficultto provide, access/interpret information on theperformance of frontline providers.

The literature suggests that for the shortroute of accountability to work well, the poor,women, and socially excluded citizens must beable to:

w provide and access information on client bod-ies and the performance of service providers;

w participate and be represented in user/clientbodies; and

w have access to and representation in client bod-ies that have effective mandates, mechanismsof enforceability, and access to finance.

These reasons highlight the need to esti-mate the direct effect that the impediments listedin Table 2 have on the study’s measures of


10

Table 2: Drivers of/Impediments to Pro-Poor Accountability

Driver Exogenous Endogenous

Micro Socioeconomic inequality in social PovertyDrivers structures Property rights over land

Historical resource endowments Access to land through tenancyAgro-ecological conditions Access to income through the labor marketPatriarchy Access to information through educationExisting social and physical Access to the labor market through migrationinfrastructure Dependence on landlords for credit/homesteadMembership of a dominant Electoral competition between patrons/politikinship group cal parties

Existence of pro-poor political parties/electoralcoalitions

Source: Author’s fieldwork.

12 See: Bardhan, P., and D. Mookherjee. 2005. Introductory chapter to Decentralization in Developing Countries: A Comparative Perspective. Edited by P.

Bardhan and D. Mookherjee. The MIT Press; Besley, T., R. Pande, L. Rahman, and V. Rao. 2001. The Politics of Public Goods Provision: Evidence from

Indian local governments. Unpublished manuscript. London School of Economics and Political Science, London; Dreze and Sen (1996) (footnote 1); and

Kohli (1987) (footnote 11).

accountability. An analysis of the constraintsimposed by micro-drivers/impediments otherthan the poverty status of a household providesfurther information for a government on addi-tional tools/instruments of intervention that itcould use to strengthen the long and short routesof accountability, which, as mentioned earlier, areimportant prerequisites for strengthening the gov-ernance relationship between citizens and thestate.

In order to test for the effects of thesedrivers, the study primarily uses the followingcontrol variables.

w Land ownership. This variable simply meas-ures the quantity of land owned by a house-hold. Gazdar (2007) shows that the quantity ofland owned is an important negative correlateof poverty (footnote 6).

w Dominant Kinship Group Status. This vari-able simply identifies whether a householdbelongs to a kinship group that was historical-ly recognized as the dominant kinship groupin the village. It is important to note that thestudy is making no assumptions about the cur-rent socioeconomic status of the kinshipgroup, which makes this a purely exogenousvariable. In fact, Gazdar (2007) not onlydescribes the village-level hierarchy of kinshipgroups, he also documents cases where histori-cally dominant kinship groups continue todominate, as well as cases where socially andeconomically mobile groups have started tocontest the dominance of these groups at thevillage level (footnote 6).

w Proportion of literate adults (male andfemale) in a household. Again, Gazdar (2007)shows that literacy is an important correlate ofpoverty and that literacy attainment tends tovary considerably between dominant and non-dominant kinship groups (footnote 6).However, he also documents cases of mobilitythat reveal a break in the one-to-one corre-spondence between social status and literacy.Literacy is also thought to be important

because it is an asset that gives households theability to absorb and process crucial politicaland administrative information.

w Village fixed effects. The study includes vil-lage fixed effects because Gazdar (2007) showsthat considerable variation exists in the samplebetween village-level socioeconomic structuresand between historical patterns of socioeco-nomic mobility (footnote 6). It is, therefore,important to control for inter-village variationwhen estimating the effect of independent vari-ables.

w Other variables. The study controls for anumber of other variables that capture theexogenous and endogenous micro-drivers men-tioned above, such as occupation, tenancy sta-tus, and female ownership of land, etc.

Conceptualizing Governance Failures: A Framework and Empirical Strategy

11

3 Devolution and the Long andShort Routes of Accountability

Pakistan’s recent devolution reforms haveinstituted a number of governance changes thatare expected to strengthen the direct long andshort routes of accountability for citizens at thegrassroots level. This section briefly describesthese important reforms. It is important to under-stand the nature of these reforms because devolu-tion provides the context for the new accountabil-ity framework between citizens and the state inPakistan. However, in line with the study’sresearch design, this document only describes theimportant changes that have occurred at the directinterface between state and citizenry.

3.1 Empowering the Union

By far the most significant change has

been the strengthening of the union as a founda-tion of the local state. The union council is theonly directly elected tier of a more empoweredlocal government system (Figure 2). Nazims, naibnazims, and councilors are directly elected at theunion level, which is a multimember ward con-sisting of an agglomeration of villages. Althoughthe multi-ward nature of union elections meansthat there is no village-level reservation in theunion councils, seats are, nonetheless, reserved forwomen, minorities, laborers, and peasants.

Reforms at the union level are expected tostrengthen the long route of accountability in anumber of ways:

(i) The introduction of elections is expected toexpose the old local-level unelected patron toelectoral competition and, thus, create a directlink of accountability between these patronsand the union-level citizenry.


1213 A tehsil is an administrative unit of local government, between a district and union.

Figure 2: Electoral Linkages


(ii) The mechanism of reservation is expected toincrease accountability to the marginalizedand powerless through the mechanism ofdirect representation.

(iii) Not only does political bargaining over devel-opment funds now occur between levels (theunion and the district/tehsil13) that are closerto citizens, all the actors involved in this bar-gaining are either directly or indirectly elect-ed and, therefore, accountable to their voters(Figures 2 and 3).

Even more importantly, at least in theo-ry, union councilors and the union executive arenow in a position to hold a much more empow-ered, higher-tier, local government executiveaccountable because of the manner in which theunion has been integrated into the higher tiers oflocal government. First, union nazims, naib naz-ims, and councilors constitute the electoral col-lege for the district/tehsil nazims, naib nazims,and councilors elected on reserved seats at thesehigher levels (Figure 2). Second, union nazims areex-officio members of the district council and

union naib nazims are ex-officio members of thetehsil council (Figure 3). The political and elec-toral integration of the union into higher tiers ofgovernment allows union nazims and naib nazimsto play a critical role in the allocation of funds andprojects across villages and social groups in theirrespective unions for two reasons.

While union nazims/naib nazims cannow monitor the spatial allocation of the dis-trict/tehsil budgets directly in their role as dis-trict/tehsil councilors, the tehsil administration

and district government do not currently have thecapacity to “monitor” and “verify” the use ofdevelopment funds for localized public goods(such as sanitation and soling) by union represen-tatives at the local level. This gives union nazimsand naib nazims considerable “authority” and“autonomy” over the allocation of funds and proj-ects for localized public goods within the union.In fact, the current practice of dividing claimsover the district and tehsil development budgetsbetween union nazims and union naib nazims,respectively, is similar to the way in which mem-


13

Figure 3: Interlinked Representation


ber of national/provincial assembly grants work(Figure 4).

Under the current reform, an “institu-tional gap” remains in that neither citizens norindividual mauzas/dehs or pinds/goths participatedirectly in budgeting, development planning, andthe selection of schemes.14 The direct participationof mauzas/dehs and pinds/goths was to have been

made operational through the enactment of vil-lage and neighborhood councils (VNCs), which,to date, have not been made active in spite of thefact that the Local Government Ordinance (2001)provides for their existence. This “institutionalgap” means that citizens can only exercise voiceover satisfaction with local government budgets,

plans, and schemes at the time that union nazims,naib nazims, and councilors are to be elected(Figure 4). Apart from this, whatever contact indi-vidual citizens have with union, tehsil, and dis-trict representatives to convey their developmen-tal needs is “informal” and “ad hoc” and thisaccess is likely to be asymmetric between differentcitizen types. Furthermore, this input is not bind-ing on a government.

After the elections, citizens have little orno direct oversight role or participation in the for-mulation and execution of union, tehsil, and dis-trict government budgets and plans.Furthermore, given that unions are multi-villageelectoral wards, it is possible for whole villages toend up having no representatives in the union


14

14 A revenue village, also called a mauza (in Punjab and NWFP) or deh (in Sindh), is the lowest defined geographic unit with a jurisdiction list number for the

purposes of revenue. A revenue village may be populated or depopulated. A hamlet/village, also called a goth (in Sindh) or pind (in Punjab), is a consolidat-

ed geographic settlement of citizens that is known to them as a village. A hamlet/village may be the same as a revenue village or there may be more than one

hamlet in a revenue village. As a rule of thumb, a hamlet/village is always populated. The other interesting difference is that, while a revenue village has

administratively defined physical boundaries, the boundaries of a hamlet/village are invariably defined by its inhabitants.

Figure 4: Electoral and Budgetary Linkages


council and, as a result, no say over the budgetaryand planning process for a particular electoralcycle. This is in contrast to experiments of directcitizen participation in West Bengal and Kerala inIndia, and Porte Algre in Brazil, where citizenshave a direct role in budget formulation and over-sight (Figure 5) (Ghatak and Ghatak 2002,15 andWorld Bank 2004 [footnote 7], pp. 78–94).Citizens can exercise some decision-making con-trol over the citizen community board (CCB)component of the budget, but there is no enforce-able requirement that the formulation of CCBsmust involve or represent the wider village-levelcitizenry.

These reforms have clearly empoweredunion nazims and naib nazims in comparisonwith the pre-devolution system. However, towhat extent this will result in the more equitable

and efficient management of development fundsdepends largely on how accountable union naz-ims and naib nazims are to the union-level citizen-ry. The “institutional gap” in budgetary and plan-ning processes means that citizens can only con-vey their developmental needs informally andthat the process is likely to be ad hoc and need notallow equal access to all citizens. Furthermore, itis only at the time of elections that all citizens,irrespective of their type, have the opportunity tovoice their satisfaction over the policies, plans,and budgets formulated by different local govern-ments. Therefore, if direct elections at the unionlevel increase the accountability of these represen-

tatives, one would expect to see increased equityand efficiency in the use of development funds atthis level.

Devolution and the Long and Short Routes of Accountability

15

15 Ghatak, M., and M. Ghatak. 2002. Recent Reforms in the Panchayat System in West Bengal: Toward Greater Participatory Governance? Economic and

Political Weekly of India 37 (1).

Figure 5: Participatory Budgeting in West Bengal

Source:Ghatak, M., and M. Ghatak. 2002. Recent Reforms in the Panchayat System in West Bengal: Toward Greater Participatory Governance? Economic andPolitical Weekly of India . 5 January.

However, if the union remains prone toelite capture, the expected results could show thepersistence of inefficient spending and spatial andsocial inequity in the allocation of projects.Which effect holds is really an empirical questionthat this study partly attempts to address by ask-ing (i) whether the poor, women, and marginal-ized groups face any barriers to entry in accessingthe electoral sphere; (ii) to what extent thesegroups’ voting decisions are based on influenceand patronage; (iii) how easy it is for these groupsto access union-level nazims/naib nazims andcouncilors in order to convey their developmen-tal needs; and (iv) whether one can estimate pat-terns of bias in the targeting of localized publicgoods between elite and nonelite groups.

3.2 Engendering User Power and Citizens’Participation

The local government reforms have pro-vided for institutional arrangements that fostercitizen participation in service delivery and over-

sight, such as CCBs,16 VNCs, and school councils(SCs). CCBs and VNCs have been enacted toenable citizens to directly participate in servicedelivery. Oversight bodies, such as SCs, have beencreated to empower citizens as “users” by givingthem the authority to monitor and supervise localservice providers. It is hoped that these bodies willincrease the efficiency of project selection as wellas the accountability of government serviceproviders to the needs of citizens.

These new bodies are expected tostrengthen the short route of accountability forthe poor, women, and marginalized groups.However, a minimum assumption for this to hap-pen is that citizens belonging to these groups mustbe aware of these institutional arrangements andthere should be evidence suggesting that they havean equal chance of participating in these bodiesrelative to nonpoor, male, and elite groups. Aspointed out earlier, the main emphasis of this

study is on the long route of accountability and itonly asks this far more limited question withregard to the short route of accountability.However, to reiterate this is an important ques-tion as it provides insights into the process ofinstitutionalization of citizen-based user bodiesduring the first electoral cycle of the newlyempowered local governments.

This analysis is particularly pertinentbecause the field surveys for this study were con-ducted at the end of the first cycle of the devolu-tion reforms, which was a good time to assess theperformance of these bodies. In addition, thestudy asks whether the awareness of and participa-tion in these bodies by the poor, women, andmarginalized groups varies with (i) the type of vil-lage-level social structures, (ii) the degree ofinequality of asset holding between groups, and(iii) the degree of socioeconomic exclusion at thevillage level.


16

16 Qadir (2005) points out that CCBs are expected to play two roles: (i) give citizens a direct role in service delivery, and (ii) mobilize citizens to hold their rep-

resentatives accountable. In this sense CCBs exist at the border of both the long and short routes of accountability. See: Qadir, A. 2005. Straddling State and

Society: Devolution and Pakistan’s Ambiguous Associations. Preparatory analytical paper prepared for ADB workshop. ADB, Islamabad.

4 Data and FieldworkMethodology

Data for this study was collected in sevendistricts of Pakistan, using three main primarydata-gathering instruments to collate the data.Districts were selected to represent Pakistan’s dif-ferent agro-ecological and socioeconomic regions.Four of the seven sample districts—(i) Toba TekSingh in central Punjab, (ii) Muzaffargarh insouthern Punjab, (iii) Mardan in the North WestFrontier Province (NWFP), and (iv) Sanghar incentral Sindh—were picked from among districtswhere the main source of livelihood is irrigatedcrop farming. Nine-tenths of Pakistan’s rural pop-ulation resides in villages that correspond to theseecological and socioeconomic conditions. The dis-trict of Chakwal in northern Punjab was pickedto capture issues of governance and poverty in abarani (rain-fed) agricultural zone. Upper Dir wassampled to represent the forestry areas in NWFPand Thatta to represent a coastal district on theIndus delta where the main source of livelihood ismarine fishing.

Intra-district sampling reflects the objec-tive of capturing features of interest within eachdistrict. In Muzaffargarh, villages from the Alipurand Jatoi tehsils were sampled to ensure that vil-lages from the riverine plains of the Punjab hadsome representation. Within Dir, villages weresampled from the mountainous regions of UpperDir to ensure that the sample was restricted to theforest-growing parts of the district. In order toincrease the sample of districts from the NWFPplains, villages were selected from both theMardan and Swabi districts. Within Thatta, thesample of villages was restricted to the coastal vil-lages. For Sanghar, the arid taluka (subdistrict) ofKhipro was dropped from the sample since theobjective was to represent villages reliant on irri-gated crop farming in Sindh. Finally, in the case ofToba Tek Singh and Chakwal, villages were ran-domly chosen from the entire district frame.

Having selected the district and subdis-

trict frame, a sample of 7% of revenue villages(footnote 14) was randomly selected in these areasfrom the 1998 Population Census village list(Government of Pakistan 2002).17 In selecting vil-lages, the study excluded very small (less than 300persons) and very large (over 3,000 persons) vil-lages from the 1998 Population Census village list.The final sample contains 61 revenue villages: 14in Toba Tek Singh, 11 in Chakwal, 7 inMuzaffargarh, 7 in Upper Dir, 7 inMardan/Swabi, 8 in Sanghar, and 7 in Thatta.Villages in Muzaffargarh, Sanghar, and Thattacontain numerous hamlets and, as a result, theoverall hamlet sample is over 100.

The first survey instrument employed thetechnique of rapid community surveys, whichwas used in all 61 sample revenue villages. Thesurvey collated information on infrastructure,social services, assets, land and tenancy, collectiveaction, and governance. In each of the seven dis-tricts, a typical village was selected for in-depthwork where a complete census was conductedusing three instruments. These seven villages arereferred to as census villages for the purpose ofthis study. The first instrument employed in-depth qualitative fieldwork around issues of labormarkets, land ownership patterns, functioning oflocal democracy, and collective action. Particularattention was paid to documenting historicaldynamics within and across village variations.

The scoping fieldwork provided informa-tion that was used to develop a village censusinstrument that, in turn, was used to collate datafrom all households in each of the seven villages.In villages comprising more than one settlement,the census was conducted in all the settlementsconstituting the revenue village. The census vil-lage questionnaire collated information on infra-structure provision, food consumption, assets,tenancy, patterns of wage labor, awareness of andparticipation in citizen-based participatory bod-ies, patterns and rationale of voting in local gov-ernment elections, and meetings attended withunion nazims/naib nazims and councilors to con-


1717 Government of Pakistan. 2002. 1998 Census Report of Pakistan . Islamabad: Population Census Organization.

vey development needs. In addition, informationon age, literacy, education, marital status, occupa-tion, possession of national identity card, and vot-ing was collated for each member of the house-hold.

Apart from these instruments, the study’sgovernance team collected qualitative data on gov-ernance and voting using key respondent andgroup interviews in the seven census villages aswell as the 61 sample villages. This qualitative dataprovides rich information on underlying struc-tures and processes, and has been used to interpretthe findings from the quantitative data.

Some of the analysis in the study com-bines household-level survey data from four rev-enue villages in Jaranwala tehsil in Faisalabad dis-t r i c t .1 8 These revenue villages were randomlyselected from a cluster of villages that lay within a20-km-radius of Khurrianwala town, Jaranwala’sfastest growing urban center. In this sense, theseare revenue villages with a high degree of expo-sure to urbanization. Within each revenue village,a sample of 30% of households was surveyed. The30% sample was randomly selected from a rev-enue village sampling frame that was stratified bykinship group. The survey instrument used byCheema and Mohmand (2005) (footnote 5) wasmore comprehensive than the instrument used toconduct a census in each of the seven census vil-lages. A similar survey methodology was used inboth the Jaranwala survey and the seven censusvillages.


18

18 This data was collated by the author along with Ms Shandana Mohmand for the purposes of research. For details, see Cheema and Mohmand (2005)

(footnote 5).

5 Analyzing the Long Route ofAccountability in Pakistan

The study analyses the long route ofaccountability by examining the following:

w Is electoral participation of the poor andwomen in the recent union council elections inline with the participation levels of the non-poor and male cohorts? The study also ana-lyzes whether the electoral participation of thepoor and women varies with exogenous andendogenous micro-drivers, such as the type ofvillage-level social structures, the degree ofinequality of asset holding between groups,and the degree of socioeconomic exclusion atthe village level.

w To what extent are these groups’ voting deci-sions based on influence and patronage? Thestudy analyzes the extent to which the decisionto vote dependently or independently is con-tingent upon (i) differences in socioeconomicstructures at the village level, (ii) the degrees ofpolitical competition within the area, (iii) thepresence of political party activity, and (iv) thedegree of political party competition.

w How easy is it for these groups to access union-level nazims/naib nazims and councilors inorder to convey their development needs?Again, the study analyzes the extent to whichaccess to union-level nazims/naib nazims iscontingent upon (i) differences in socioeco-nomic structures at the village level, (ii) the

degree of inequality of asset holding betweengroups, and (iii) the degree of socioeconomicexclusion at the village level.

w What patterns of bias exist in the targeting oflocalized public goods between elite andnonelite groups? The point of interest is toexamine how biases in targeting vary with (i)differences in socioeconomic structures at thevillage level, (ii) the degree of inequality ofasset holding between groups, and (iii) thedegree of socioeconomic exclusion at the vil-lage level.

5.1 Barriers to Entry in the ElectoralSphere for the Poor and Women

This section asks whether the barriers toentry in the electoral sphere are low for the poorand women, i.e., whether poor citizens andwomen are as likely to participate in the electoralsphere as males and nonpoor households. It alsoestimates whether other exogenous and endoge-nous micro-drivers constrain these citizens’ deci-sion to exercise their right to vote. As mentionedearlier, access to the electoral sphere is an essentialprecondition for a well-functioning long route ofaccountability.

5.1.1 Voting and Gender

Table 3 shows that, in the census vil-

lages,19 only approximately half the proportion ofadult women voted in comparison to adult men.Qualitative fieldwork reveals that this statisticreflects social structural constraints that define


19

19 The term “census villages” refers to the seven villages in which a complete census was conducted by the ADB TA 4319-PAK field teams.

Table 3: Male and Female Voting Turnout (%)

Voted Male Female Total

No 42.1 68.7 54.9Yes 57.8 31.2 45.0

Total 100.0 100.0 100.0

Note: Figures may not add up to total because of rounding.Source: Author’s fieldwork

behavioral roles for women, which do not includetheir active participation in the electoral sphere(Mohmand and Gazdar 2007). 20 This suggests thatthe electoral sphere is suitably gendered, whicheffectively curtails the role women play inPakistan’s politics. The concern is that a large partof Pakistan’s active political population does notparticipate in the act of voting. An even greaterconcern is that, given the current levels of femaleelectoral participation, it is highly unlikely thatgovernments (local, provincial, or national) willbe responsive to the needs of women.

Table 4 shows that the patterns of womenvoting across the census villages is highly correlat-

ed with the patterns of patriarchy documented inMohmand and Gazdar (2007) (footnote 20). Theresults show that women are far less likely to votein the NWFP districts and in Thatta, whereasthey are much more likely to vote in Chakwal.

As argued in Mohmand and Gazdar(2007), the low propensity of voting amongwomen in the Dir and Mardan villages is integral-ly linked to restricted patterns of physical andsocial mobility (footnote 20). Restricted patternsof physical and social mobility, in turn, appear tobe related to patterns of endogamy that exist to

preserve and perpetuate kinship groups(Mohmand and Gazdar 2007, footnote 20). In thecase of villages such as Chakwal, where there is amuch higher degree of social and physical mobili-ty, the study finds that a large majority of womenexercise their right to vote. Low levels of votingamong women can also be an outcome of the wayvoting booths are located. Interviews revealedthat women in the Thatta village chose not tovote in the 2005 local government elections—inspite of voting in previous elections—because nopolling booth had been set up in their village.

The pattern of national identity card pos-session is equally gendered and mirrors the male-

female divide found in voting patterns (Table 5).Interestingly, the relative incidence of identitycard possession is much higher among womenthan is the exercise of the right to vote.

This finding is pushed into stark reliefwhen one analyses the incidence of identity cardpossession at the village level. Table 6 shows thatthe lack of identity card possession does notexplain low voter turnout among women in vil-lages with extremely low female voting propor-tions. This suggests that lack of access to identitycards is not really an impediment to voting for


20

20 Mohmand, S., and H. Gazdar. 2007. Social Structures in Rural Pakistan. Thematic paper prepared under TA4319, Determinants and Drivers of Poverty

Reduction and ADB’s Contribution in Rural Pakistan. ADB, Islamabad.

Table 4: Female Voting Turnout by Village (%)

Village (in District) No Vote Vote Total

Chakwal 33.33 66.67 100.00Dir 97.27 2.73 100.00Mardan 77.14 22.86 100.00Muzaffargarh 66.15 33.85 100.00Sanghar 63.91 36.09 100.00Thatta 94.89 5.11 100.00Toba Tek Singh 51.62 48.38 100.00Total 68.73 31.27 100.00

Source: Author’s fieldwork

approximately 16% of adult, voting-age women.This finding reinforces the argument inMohmand and Gazdar (2007) that differences inpatterns of patriarchy are an important part of theexplanation for low levels of female votingturnout (footnote 20).

The important question to ask is whetherthere are systematic patterns in the type ofwomen who vote. Multivariate results suggestthat, within the group of voting women, literatewomen are 9% more likely to vote than illiteratewomen, although literate women are still almost20% less likely to vote compared to illiterate men(Appendix, Table A.1, columns 1–3). The correla-tion between literacy and women’s participationin the electoral sphere persists even after control-ling for village fixed effects. This suggests that lit-eracy does have a significant impact on the partic-ipation of women in the electoral sphere.

5.1.2 Voting, Women, and the Poor

The relationship between the poverty sta-tus of a household and voting turnout amongwomen still needs to be analyzed, given that therelationship between poverty and accountability

variables is the main question this study isdesigned to answer. Multivariate regressionresults suggest that the poverty status of a house-hold is negatively and significantly related towomen voting turnout at the household level(Appendix, Table A.1, columns 4–6). The resultsalso suggest that female literacy is a highly signifi -cant positive correlate of female voting.Interestingly enough, the results suggest that, con-trary to popular myth, women from dominantkinship groups are more likely to vote relative toothers. Given the low levels of overall electoralparticipation by women, this suggests that domi-nant kinship groups are cognizant of the impor-

Analyzing the Long Route of Accountability in Pakistan

21

Table 5: Incidence of Male and Female National Identity Card Possession (%)

Possession of Identity Card Male Female Total

No 25.35 52.53 38.46Yes 74.65 47.47 61.54Total 100.00 100.00 100.00


Table 6: Incidence of Female Identity Card Possession by Village (%)

Village (in District) No ID card ID card Total



tance of not wasting votes that are in their con-trol.

These results strongly suggest thatincreasing female literacy has a spillover benefit asit encourages women to participate in the elec-toral sphere. Therefore, if the aim of policymak-ers is to increase women’s participation in theelectoral sphere, a medium-term goal must be toincrease female literacy. However, the results alsosuggest that patterns of patriarchy embedded inthe social structure are exogenous impediments towomen’s participation in the electoral sphere.

5.1.3 Male Voting and the Poor

The data reveals a number of interestingfindings about male voting patterns. Table 7shows that the proportion of male voters is notmuch higher than 50%. The important question iswhether there are systematic biases in the types ofmales who vote and those who do not. In linewith the study’s empirical strategy, it starts byasking whether males from poor households aremore or less likely to vote than males from non-poor households. Simple averages suggest thatpoor male voters vote in roughly the same pro-portion as nonpoor male voters across most of thesample villages.

However, regression analysis shows that,after controlling for village fixed effects and for

other micro-drivers, poor households are, onaverage, less likely to vote than nonpoor house-holds (Appendix, Table A.2, columns 1–3). Thedifference in order of magnitude also does notappear to be trivial. This suggests that it is morethan at the margin that the poor are less likely toparticipate in the electoral sphere. Access to theelectoral sphere appears to be structured in a waythat discriminates against the poor. This result isfurther put into perspective by the finding thatthe decision to participate in voting at the house-hold level is positively and significantly correlatedwith the household belonging to a dominant kin-ship group and with the proportion of literatemale adults in a household. The latter effect is ofa significantly large order of magnitude.

The result regarding the dominant kin-ship group is interesting because, as described byGazdar (2007), some of these groups are no longerdominant in a social and economic sense in thecensus villages (footnote 6). The fact that the coef-ficient on the dominant kinship group dummy ispositive and significant after controlling for vil-lage fixed effects suggests that, irrespective of theircurrent position in specific villages, these groupsare likely to dominate the local-level electoralsphere.

Putting together these results suggeststhat members of a (i) poor, (ii) nondominant kin-


22

Table 7: Proportion of Male Adults who Voted(%)

Village (in District) Nonpoor Poor

Chakwal 47.19 58.22Dir 26.12 27.18Mardan 34.16 37.76Muzaffargarh 39.68 34.69Sanghar 33.06 30.19Thatta 57.54 41.39Toba Tek Singh 44.96 34.69Total 38.37 36.42


ship group, and (iii) illiterate household are farless likely to participate in the electoral sphereand exercise their right to vote. These are precise-ly the types of households the study considerssocially excluded and marginalized. The resultsregarding possession of identity cards mirror theresults for male voting (Appendix, Table A.2,columns 4–6).

Again from a policy perspective, if theaim is to increase participation in the electoralsphere as a medium-term goal, then increasing lit-eracy appears to be an important policy variable.In addition, it becomes necessary to support socialmobilization efforts that aim to catalyze participa-tion in the electoral sphere by poor householdsfrom nondominant kinship groups.

5.2 Dependent Voters and Pro-PoorAccountability

The previous section argues that women,the poor, and illiterate and nondominant kinshipgroup households face differential access to theelectoral sphere relative to male, nonpoor, and lit-erate and dominant kinship group households.The political economy literature, however, isreplete with examples that suggest that the act ofvoting may itself be circumscribed for groups thatare in relations of social and economic depend-ence, such as the landless, tenants, and sociallyexcluded groups.21 The argument is that the votingdecision of these groups is defined by relations ofdependence that constrain their exercise of choicewhen voting. In defining dependent voting, thestudy follows Alavi’s (1971) definition, “In mak-ing his (voting) choice, he (voter) must act withinthe constraints imposed on him by his situation(position) in the social structure of the village”(footnote 21, p. 112). For example, landless agri-cultural servants without homestead rights whoreside on a landlord’s land are more likely to votein line with the landlord’s expectations as com-pared to self-cultivators with secure homestead

rights. Therefore, it is important to analyzewhether or not relations of dependent voting existin the sample villages and which groups of votersare more prone to dependent voting.

The analysis of dependent voting was con-ducted in three parts and by combining quantita-tive and qualitative field research methodologies.The census questionnaire included a question onthe rationale behind a household’s decision tovote. The response options included:

(i) trust in the individual being voted for,

(ii) political party affiliation,

(iii) relationship with an influential person of thearea,

(iv) relationship with an influential member ofthe village,

(v) decision after consultation with family mem-bers, and

(vi) decision under pressure from a notable(s).

The study ran regressions with each ofthese options as the dependent variable and withthe following as independent variables: (i) pover-ty status of a household, (ii) membership of adominant kinship group, (iii) landownership, (iv)village fixed effects, and (v) other controls. Theregression estimates provide two important piecesof information.22 First, they allow one to identifywhich reason(s) for voting dominates in each ofthe census villages. Second, they allow one to seewhich reasons for voting are significantly corre-lated with a respondent belonging to a poorhousehold, a literate household, and a dominantkinship group household,23 holding village differ-ences in reasons for voting constant.

As the first step, each economically andstatistically significant regression coefficient(s)was entered into a table (footnote 22) against a vil-


23

21 See: Alavi, H. 1971. The Politics of Dependence: A Village in West Punjab. South Asian Review 4 (2); Cheema and Mohmand (2005) (footnote 5); and Kohli

(1987) (footnote 11).

22 The estimates are available from the author on request.

23 These were the only variables of significance in our regressions.

lage name—this allows us to see the important andsignificant regression coefficients at the villagelevel at a glance. This voting choice table (foot-note 22) creates a village-level matrix that docu-ments the significant reason(s) for voting found ineach village.

The second step in the analysis was tointerpret the meaning of each of the six votingchoice variables in each village using qualitativefieldwork techniques. Key respondent and smallgroup interviews were conducted to analyze thevillagers’ interpretation of each of the choice-to-vote variables in each village. The idea was tounderstand what “trust in an individual” means inthe Chakwal village as opposed to the Mardan vil-lage. Is the notion of trust one of free choice ordoes it reflect a relationship of dependence? Thequalitative interpretation of these variables wasthen entered into the voting choice table men-tioned earlier, which provides village-level snap-shots of households’ reasons for voting.

In addition, the voting choice tablereports significant regression coefficients on thefollowing independent variables: (i) poverty statusof a household, and (ii) membership of a domi-nant kinship group.24 These entries provide tabu-lar information on what drives the decision tovote among poor, literate, and dominant kinshipgroup households. A separate table is used todevelop a typology of dependent voting at the vil-lage level as well as to analyze the rationale behindthe voting decisions of poor, literate, and domi-nant kinship group households. This votingtypology table is presented in the next section andis the main basis for analyzing dependent voting.

The third step in the analysis is to com-pare the typology of dependent voting developedthrough the voting typology table with degrees ofsocioeconomic inequality at the village level. Thetypology of socioeconomic inequality at the vil-lage level was developed using the informationpresented in Gazdar (2007) (footnote 6). Thiscomparison allows us to analyze the manner in

which social structure types create relations ofdependence in the electoral sphere. This analysisis important because it allows us to identify struc-tural impediments to the long route of pro-pooraccountability.

5.2.1 Voting Typology Table

The voting typology table shows thatpoor households are far more likely to votebecause of pressure exerted on them by influentialpeople, and less likely to vote in consultation withinfluential village members or because of strongparty affiliations. These results strongly suggestthat poor households are more likely to act asdependent voters in comparison with nonpoorhouseholds. This suggests that there is a great like-lihood that the votes of poor households will bediscounted even if they are active in the electoralsphere. Table 8 also shows that the poor are morelikely to be de-linked from political parties rela-tive to nonpoor households and, in certain con-texts, this can reinforce dependence at the locallevel.

As opposed to this, dominant kinshipgroup households are more likely to vote in con-sultation with influential village members and lesslikely to vote because of external pressure. Asmentioned earlier, the study’s dominant kinshipgroup variable captures membership in a groupthat was historically recognized as the dominantgroup in the village. This variable makes noadjustment for the current socioeconomic statusof the household or group. Given this definition,these results are quite intriguing. Gazdar (2007)has documented evidence of considerable socioe-conomic mobility among nondominant kinshipgroups in Chakwal, Muzaffargarh, and Sanghar(footnote 6). One would expect this evidence ofsocioeconomic mobility among nondominantkinship groups to have weakened any strong andsignificant statistical association between being amember of a dominant kinship group and votingin consultation with the village influential.However, the study finds not only that the rela-


24

24 These were the only two variables of significance in the voting choice regressions.

tionship is positive, it is also statistically and eco-nomically significant. This suggests that the localelectoral sphere is quite likely dominated by thedominant kinship group in alliance with influen-tial member(s) of the village. However, an impor-tant question is whether we find variations in thetypology of voting choices across village typesand whether there is a correlation between socioe-conomic mobility and inequality at the villagelevel and the prevalence of dependent and influ-ence-based voting. This question is taken up in thenext section.

In order to contextualize this analysis,Table 9 provides a summary of village-levelsocioeconomic variables. Rows 1 and 2 of thetable draw up a typology of social and economicinequality at the village level. Given these indica-tors, dependent and influence-based voting is like-

ly to be low in Chakwal and Dir, but unambigu-ously high in Sanghar. If dependent voting is anoutcome of extreme levels of poverty, it is expect-ed to be high in Thatta and Muzaffargarh as well.Finally, dependent and influence-based voting isalso likely to be high in Toba Tek Singh andMardan if social inequality matters. In the nextsection, the study uses information from its qual-itative surveys, the voting typology table (Table8), and village indicators table (Table 9) to analyzewhether correlations exist between village-levelsocioeconomic inequality and dependent voting.


25

Table 8: Voting Typology Table

Statistically Significant Rationale for Voting by Village and Group

Chakwal Trust in area influential(s)

Dir Political party affiliation Trust in candidate

Mardan Influence of village Consultation with villageinfluential(s) influential(s) Consultation with family

Muzaffargarh Consultation with village Trust in area influential(s) influential(s)

Sanghar Political party affiliation Consultation with area and village influential(s)

Thatta Trust in area influential(s)

Toba Tek Singh Trust in village influential(s)

Poor Pressure of influential(s) Less likely to consult Less likely to vote village influential(s) because of political

party affiliation

Dominant Kinship More likely to vote in Less likely to vote becauseGroup consultation with village of the pressure of

influential(s) influential(s)


5.2.2 Village Correlations

Chakwal

Chakwal is a case of dependent votingwhere the entire village votes for an area landlordwho dominates national politics. Voting is classi-fied as dependent because the landlord controls amonopoly vote faction that dominates the entire

cluster of villages in the area and there are no out-side voting options available to villagers.However, villagers do not see this dependence ontheir landlord as an outcome of coercion, insteadit is seen as a way of signaling their loyalty to abenevolent patron. Therefore, voting in theChakwal village is typed as area-level, patron-client dependent voting.


26

Table 9: Village IndicatorsIndicator Chakwal Dir Mardan Muzaff

argarhSanghar Thatta Toba Tek

SinghSocial inequality(differ-ence in poverty head-counts between sociallyexcluded and dominantcastes)

Moreequal

Moreequal

U n e q u a l Moreequal

Highlyunequal

More equal Highlyunequal

Economic inequality Smallequal

Smallequal

Mediumunequal

M e d i u munequal

Largeunequal

Large unequal(agriculture high-ly unproductive)

Mediumunequal

Cross-village poverty(headcount)

Low Low Low High Medium-High

Very high Medium

Mean landownership inlandowning class (acres)

3.9 0.8 1.9 1.1 5.1 5.1 1.8

Standard deviation 5.2 1.3 6.0 2.1 31.4 22.0 3.1Max. land owned (acres) 30.2 8.5 30.0 15.0 500.0 200.0 30.0Cultivating others’ land(%)

31.8 2.8 34.7 17.9 66.2 3.3 12.7

Landless (%) 29.4 14.2 77.7 52.2 62.4 83.1 54.4Illiterate males (%) 40.0 42.0 16.0 52.0 58.0 65.0 17.0DFServants on others’land (%)

36.0 3.3 34.0 57.0 42.0 2.2 38.0

Illiterate females (%) 45.6 53.0 40.0 63.0 76.0 74.0 16.6Agricultural productivity Low Low High High High Low HighCross-village extremepoverty (headcount)

2.6 3.2 3.2 35.6 15.1 53.9 10.0

Cross-village middlepoverty (headcount)

16.7 15.1 25.5 65.7 32.1 79.3 26.3

Nonfarm earners (%) 3.5 13.0 13.0 14.0 9.0 41.0 11.0Casual labor (%) 15.1 15.0 22.0 31.0 9.0 55.0 17.1Migrant labor(international) (%)

0.0 31.4 10.7 0.2 0.0 0.0 2.0

Migrant labor(domestic)(%)

0.7 2.7 8.3 4.4 0.3 0.6 2.5


Interestingly, Chakwal’s village indicatorswould suggest a low dependent voting village, butthe existence of a higher-level monopoly vote blocdominated by an area influential recreates pat-terns of dependent voting in a village with other-wise low socioeconomic inequality and lowpoverty. This suggests that the existence of high-er-tier monopoly voting factions is associatedwith dependent voting irrespective of the under-lying socioeconomic conditions of the area.

Dir

Low levels of socioeconomic inequalityand poverty in Dir have provided the foundationfor the exercise of free and independent electoralchoice. Clusters of households vote together andswitching for and between candidates has been anorm in Dir. Competitive political party electoralactivity reinforces the practice of independentvoting. Political party affiliations are an impor-tant reason for supporting a candidate. However,we were repeatedly told that the character of anindividual matters as much as the party the candi-date is affiliated with. Voting in Dir is typed asfamilial nondependent voting.

Mardan

Voting in the Mardan village is also typedas village–level, patron-client dependent voting. Amonopoly vote bloc of the dominant Mohmandkinship group prevails over village politics in spiteof the existence of competitive politics at the high-er tiers. Dependence on land and clan affiliationsare the main reasons for villagers’ electoral alle-giance to the Mohmand vote bloc.

Muzaffargarh

Low levels of socioeconomic inequality inMuzaffargarh have created preconditions of inde-pendent voting as no one landlord or faction dom-inates the village. Qualitative fieldwork suggeststhat households are fairly independent in theirvoting decisions and tend to switch their supportbetween the two village-level factions based ontheir own judgments about the benefits of back-ing particular candidates. Independent voting is

further catalyzed by the prevalence of competi-tive factions at the higher tiers. The combinationof low levels of socioeconomic inequality andcompetitive higher-tier politics has translated intopolitical mobility for nondominant kinshipgroups. Voting in Muzaffargarh is, therefore,typed as familial nondependent voting.

Sanghar

Given the levels of poverty and highdegree of socioeconomic inequality, one wouldexpect Sanghar to be a stark case of dependentvoting. However, two factors have reduced thetendency towards dependent voting here. First, asdescribed in Gazdar (2007), there have beenstrong signs of socioeconomic mobility amongmembers of some of the nondominant kinshipgroups (footnote 6). Opportunities and patternsof socioeconomic mobility have translated intopolitical independence because of the highlyactive presence of a national-level populist politi-cal party. The combination of socioeconomicchange and high-intensity political party competi-tion has translated into political mobility forsome members of some of the nondominant kin-ship groups. These groups have emerged as inde-pendent voters.

In parallel, however, there are other non-dominant kinship groups that remain tied to thelarge landlords’ land and are dependent on themeconomically. Members of these groups continueto vote as dependent client voters. In order to cap-ture these parallel realities, we have typed votingin Sanghar as competitive, patron-client, andparty-based semi-dependent voting.

Thatta

Unlike Dir and Muzaffargarh, low levelsof social inequality have not produced a tendencytowards independent voting in Thatta. However,like Chakwal, Thatta is a case of dependent votingwhere the entire village votes for a political influ-ential who dominates district and national poli-tics. Voting is classified as dependent because theinfluential controls a monopoly vote faction thatdominates the entire cluster of villages in the area


27

and there are no outside options available to vil-lagers. However, villagers do not see this depend-ence on their landlord as an outcome of coercion,instead it is seen as a signal of loyalty to a benev-olent patron. Therefore, voting in the Thatta vil-lage is typed as area–level, patron-client, depend-ent voting.

Toba Tek Singh

Toba Tek Singh is also a case of depend-ent voting. The politics of the village is dominat-ed by two vertically hierarchical dominant kin-ship group factions. Nonlandlord kinship groupsand the Musalli kinship group vote for their ownlandlord’s faction. Bonds of economic depend-ence based on unequal landownership and a high-ly unequal social hierarchy recreate dependencyin the electoral sphere. The bonds of economicdependency result in dependent voting in spite ofthe existence of competitive factions within thevillage and at the higher levels. Voting in the TobaTek Singh village is typed as village–level, patron-client, dependent voting.

5.2.3 Conclusion

Village case studies suggest the existenceof a weak correlation between socioeconomicinequality and the existence of dependent patron-client voting. The correlation is weakenedbecause political conditions are equally importantdeterminants of the existence of dependentpatron-client voting. The nature of the correla-tion is summarized in Table 10.

Table 10 reveals three prominent relation-ships between voting types and their determi-nants:

(i) Where socioeconomic inequality is highlyunequal and personality-based politics domi-nates, dependent village-level patron-clientvoting prevails. This relationship exists inboth Toba Tek Singh and Mardan. In thesecases, dependent voting persists in spite of theexistence of competitive higher tier factions.

(ii) The cases of Dir and Muzaffargarh suggestthat, where socioeconomic inequality is low


28

Table 10: Type of Voting and Potential DeterminantsItem Chakwal Dir Mardan Muzaffargarh Sanghar Thatta Toba Tek

SinghVoting type Dependent

area-levelpatron-client

Familialnonde-pendent

Dependentvillage-levelpatron-client

Familial non-dependent

Competitivepatron-clientand party-based semi-dependent

Dependentarea-levelpatron-client

Dependentvillage-levelpatron-client

Social

inequalityaMore equal More

equalUnequal More equal Highly

unequalMore equal Highly

unequal

Economicinequality

Small equal Smallequal

Mediumunequal

Mediumunequal

Large unequal Large unequal(agriculturehighly unpro-ductive)

Mediumunequal

Political partyactivism

Low High Low Medium High Low Low

Personality-based politics

High Medium High High Medium-High High High

Competitivehigh-tier factions

No Yes Yes Yes Yes No Yes

a Difference in poverty headcounts between socially excluded and dominant castes.


and competitive higher-tier factions domi-nate, we should expect independent voting toprevail.

(iii) However, low levels of socioeconomicinequality are not sufficient to ensure inde-pendent voting if it is combined with monop-oly higher-tier factions and personality-basedpolitics. This occurs in the cases of Chakwaland Thatta.

Sanghar represents an intermediate case and com-bines features of the three cases detailed above.

The analysis in this section suggests thatthe long route of pro-poor accountability will beweak for the following reasons:

w Poor households are more likely to act asdependent voters than nonpoor households,irrespective of the type of village. This suggeststhat there is a great likelihood that the votes ofpoor households will be discounted even ifthey are active in the electoral sphere.

w Dominant kinship groups are more likely tovote in consultation with village-level influen-tial(s). This suggests that the more localized thepolitics the more likely it will be dominated bya concert of dominant kinship groups and vil-lage-level influential(s).

w Voters in villages with high levels of socioeco-nomic inequality and prevalence of personali-ty-based politics are more likely to be depend-ent voters.

w Voters in villages that are dominated bymonopolist higher-tier factions are more likelyto be dependent voters.

As opposed to this, low levels of socioeco-nomic inequality, competitive higher-tier politics,and high degrees of local-level political partyactivism appear to be important preconditions forstrengthening the long route of pro-poor account-ability.

5.3 Do the Poor Have Equal Access toUnion Nazims and Councilors?

For the long route of accountability towork well, it is important that politicians are inregular contact with the poor, women, and mar-ginalized citizens to ascertain their developmentneeds. Formal decentralized mechanisms of par-ticipatory budgeting such as the Gram Sabhashave worked well in parts of India25 and appear tohave had a positive impact in terms of poor citi-zens’ participation in budgetary decision making.Unfortunately, mechanisms of participatory orcitizen-based budgeting have not been institutedin Pakistan at the union council or village-leveluntil now. As pointed out earlier, it was hopedthat VNCs would play this role but these councilswere not made operational during the first cycleof devolution. Direct village-level or citizen par-ticipation in budgeting and planning is an “institu-tional gap” in Pakistan’s current local governancesystem.

The current practice with regard to citi-zen-based budgeting is that union councils, as cit-izen representatives, have to approve develop-ment schemes for the union even if they arefinanced from the development budgets of the dis-trict or tehsil. Qualitative fieldwork suggests thatthis requirement is not strictly followed and, inpractice, the selection of schemes is done by theunion nazims/naib nazims depending on whetherthe funds for the scheme come from the district ortehsil. Furthermore, a part of the developmentbudget of these higher tiers of local government isallocated to general, minority, women, and kissan(farmer)/labor councilors directly and in thesecases there is no requirement that the union coun-cil approve the development scheme. The unioncouncil only appears to fulfill an oversight rolewith regard to its development budget. Since alarge part of the finances for developmentschemes come from the higher tiers of local gov-ernment, it is important to assess how the unionnazim/naib nazim and councilors actually assess


29

25 See: Besley, et. al. (2001) (footnote 12); and Chattopadhyay, R., and E. Duflo. 2003. The Impact of Reservation in the Panchayati Raj: Evidence From a

Nationwide Randomized Experiment. Typescript. Massachusetts Institute of Technology, Cambridge.

the ”needs” of the villagers. This is particularlyimportant with regard to schemes for localizedpublic goods, such as sanitation and soling, wherehigher-tier local governments really do not havesufficient information about the spatial distribu-tion and historic spread of these schemes at thevillage and household level because of a paucity ofinformation databases.

5.3.1 Estimating Access

Our qualitative fieldwork reveals that theprocess followed by the union nazim/naib nazimand councilors to assess the needs of villagers is

completely ad hoc and informal. The worry isthat, given the ad hoc and informal nature of theprocess, there may be a systemic anti-poor bias inaccess to the union nazim/naib nazim and coun-cilors for the purposes of conveying developmentneeds, resulting in the needs of the poor not beingcounted in a systematic fashion. To put it differ-ently, if the expected result holds, then it wouldimply that the needs of the poor are unlikely to beintegrated systematically into the planningprocess.

Given this context, it is critical to esti-


30

Table 11: Respondents Reporting Contact with Union Nazim/Naib Nazim toConvey Needs in the Last 6 Months(%)

Village (in District) No Yes Total


Naib nazim = deputy mayor, nazim = mayor.Source: Author’s fieldwork.

Table 12: Respondents Reporting Contact with Union Councilors to Convey Needsin the Last 6 Months(%)




mate the number of respondents reporting partic-ipation in informal meetings where they sharedinformation on development, infrastructure, andsocial service needs with union nazims/naib naz-ims and councilors. Table 11 shows that, withinthe pooled village data, only a small minority ofrespondents reported attending such informalmeetings. Chakwal, which is a relatively high-equality village in our sample (see previous sec-tion), is the only village where the reported inci-dence of these meetings tends to be high. Theresults for councilor meetings mirror the resultsfor the union nazim/naib nazim meetings (Table12). This suggests that the activity of needs plan-ning is far from participatory and inclusive and isconsiderably de-linked from the village communi-ty.

5.3.2 Extent of Anti-Poor Bias

A more serious concern is whether thereis a systemic anti-poor bias in the identity of thesmall minority of respondents reporting that theyhad attended such meetings with unionnazims/naib nazims and councilors. Multivariateresults (Appendix, Table A.3, columns 1–3) showthat there is indeed a systemic anti-poor bias inthe respondents who reported attending suchmeetings. The results suggest that the poor are lesslikely to report attending these meetings in com-parison with the nonpoor. The results also showthat the proportion of literate adults in a house-hold is an extremely strong and positive correlatefor attending these meetings. Interestingly,belonging to the dominant kinship group is notan advantage to respondents when it comes toattending these meetings. The results for coun-cilor meetings echo these results (Appendix,Table A.3, columns 4–6).

These findings suggest that the voices ofilliterate and poor households are systemically lesslikely to be counted in the assessment of needs inthe planning process. Unless there are systemicinterventions to formulate planning proceduresand needs assessment procedures that ensure that

the voices of these groups are counted, it is mostlikely that the needs of the most vulnerable willcount the least in the development planningprocess.

5.4 Patterns of Bias in the Targeting ofLocalized Public Goods

5.4.1 Why is Measuring the Targeting BiasInsightful?

Previous sections show that the poor, illit-erate male and female citizens, and citizens fromnondominant kinship groups are on average lesslikely to participate in the electoral sphere to exer-cise their right to vote. We have also seen that illit-erate and poor households are less likely to be partof the network around union nazims/naib nazimsand councilors and hence less likely to have theirneeds counted in the development process. This isthe result of an “institutional gap” that preventscitizens and villages from having a direct role inbudgetary and planning processes. Finally, Section5.2 has argued that the probability of a citizen act-ing as a dependent voter is contingent upon (i) theexistence of high degrees of socioeconomicinequality at the village level, (ii) the lack of polit-ical party activism, and (iii) monopolization of thepolitical space by an area influential. This suggeststhat the long route of accountability is weak forthe poor, women, and illiterate households, espe-cially in villages where these conditions pertain.This gives rise to the expectation that, in these cir-cumstances, we would expect the targeting oflocalized public goods to be biased in favor of elitegroups in villages where the three conditions men-tioned above hold.

This is more likely to be the case for local-ized public goods because the higher tiers of localgovernment lack the information databases thatcapture information on the allocation and exis-tence of these schemes, which gives considerablediscretion to union nazims/naib nazims inscheme targeting.26 Given that the long route ofaccountability appears to be weak at the union


31

26 This is unlikely to be the case with a school building/health facility whose location and existence is much easier to verify.

level, we would expect an anti-poor bias inscheme targeting. As pointed out earlier, the longroute of accountability is weak because of theanti-poor bias in the local-level electoral, plan-ning, and budgeting spheres.

Furthermore, qualitative field visitsrevealed that higher-tier staff rarely visited thecensus villages and when, and if, visits are made,no formal mapping is undertaken. This informa-tion gap, combined with the fact that a higher-tiercouncilor’s union nazims/naib nazims have con-trol over development funds and are not answer-able to citizens except at the time of elections,leaves a lot to their discretion. The involvementof union-level officials in scheme selection actual-ly brings village-level socioeconomic and politicalgroups into play in the bargaining game over theselection of localized public good schemes.Therefore, estimating the degree of nonelite biasin localized public good scheme targeting can pro-vide a neat measure of the effectiveness withwhich the long route of pro-poor accountabilityworks at the union level.

Estimating the degree of the targeting biasin the selection of localized public good schemesgives useful insights into how well citizen-basedaccountability works in the local governmentcontext. This is because direct citizen contactwith local government officials only happens atthe union level. Therefore, assessing how well thetargeting of a localized public good is workinggives important insights into how accountableunion councils are to the nonelite citizenry.Furthermore, heightened accountability at theunion council-level is a necessary condition forheightened pro-poor accountability at the highertiers of local government because of the indirectnature of the local government electoral system.

5.4.2 Methodology and Data

These reasons are good grounds for esti-mating the pro-elite targeting bias in the selectionof localized public goods at the union level. Thestudy uses census data from the villages of

Chakwal, Muzaffargarh, and Toba Tek Singh toestimate the targeting bias in the provision of sol-ing and sanitation schemes to nonelite house-holds. It was not possible to conduct this exercisefor the other villages because there was no varia-tion in the provision of these schemes within theremaining four villages. This is because either allor none of the households in these villages wereprovided with these schemes. Estimates of the tar-geting bias were, however, possible for theChakwal, Muzaffargarh, and Toba Tek Singh vil-lages because there is considerable variation in theprovision of soling and sanitation schemes sincethe enactment of elected local governments in2001. In order to increase the sample for this esti-mation, we have merged the data from our censusvillages with data from fieldwork conducted bythe author (with Ms. Shandana Mohmand) in fourvillages in the Faisalabad district.27 These villagesalso saw considerable variation in the provision ofsanitation and soling schemes after the enactmentof elected local governments in 2001.

Unfortunately, given the nature of thedata, the exercise could only be conducted for vil-lages in the Punjab province. As a result, the datais far more heavily weighted in favor of centralPunjab villages. Therefore, the results of this exer-cise, while insightful, must be kept in perspective.Another constraint of the data is that, for theFaisalabad district villages, the authors did notestimate poverty rankings of households (Cheemaand Mohmand 2005, footnote 5). Therefore, inorder to estimate the elite bias in provision target-ing, this study uses three variables to capture theelite status of a household: (i) land ownership, (ii)membership of a dominant kinship group, and(iii) literacy. Gazdar (2007) shows that these threevariables are extremely important correlates ofpoverty (footnote 6). Therefore, a significant esti-mate of bias on these measures provides impor-tant insights into the anti-poor nature of targetingin scheme selection at the union council level.


32

27 For details of data and methodology, see Section 4 of this paper, and Cheema and Mohmand (2005) (footnote 5).

There are three further empirical reasonswhy analysis of targeting bias in the provision oflocalized public goods provides useful insightsinto the strength of pro-poor accountability in thenewly devolved system at the union council level.First, in all the Punjab villages, soling and sanita-tion was communicated as an essential need bykey respondents during the qualitative part of thefieldwork. Second, nearly all functional sanitationand soling was provided through governmentschemes. Third, the qualitative fieldwork revealedthat the union nazim and naib nazims of these vil-lages were pivotal players in the selection of theseschemes, and in all these villages, funds were allo-cated for the provision of these schemes. That is,in the parlance of the World Bank (2004) frame-work, delegation and financial control exists inthe case of this provision (footnote 7), and there-fore analyzing the targeting bias provides cleaninsights into which citizen group’s voice countsmore.

The main questions addressed by this sec-tion is (i) whether empowering the union councilhas reduced the elite bias in service provision,which is suggested to be a hallmark of the prede-volution governance structure,28 and (ii) whetherthis has reduced inequitable provision betweenhouseholds belonging to marginalized and elitegroups. Our main empirical question is whetherbeing (i) a landowner, (ii) a member of a dominantkinship group, or (iii) relatively literate gives ahousehold a higher likelihood of receiving newprovision of post-local government reforms.

The direction of targeting bias is estimat-ed by comparing the change in pre- and post-devo-lution provision of schemes to nonelite groupswith the change in pre- and post-devolution pro-vision in favor of elite groups. The regression,which is discussed in detail in Table A.4, columns1–2 (Appendix), calculates the difference in thesechanges holding pre-devolution provision to eliteand nonelite groups as constant. In simple terms,

this exercise allows us to estimate the targetingbias by comparing the change in number ofschemes received by nonelite groups in a village tothat received by elite groups, while holding theinitial provision differences between these groupsconstant. The study attempts to measure whether,as a result of the reforms, nonelite households areable to catch up with elite households in schemeallocation. It is possible to calculate measures ofchange because both the field surveys, i.e., theFaisalabad survey and the census conducted aspart of TA-4319, not only capture data on theprovision of a scheme but also the date on whichthe scheme was provided.

This data not only provides a picture ofcurrent provision, it also allows us to build a base-line of pre-devolution provision. An importantcontribution of this methodology is that the out-come variable, drain and soling provision beforeand after 2001, allows us to measure the impact ofreforms on the equity of public service provisionacross social groups. By estimating the post-reform change in targeting biases towards differ-ent socioeconomic groups, we can get a good han-dle on how well the long route of “accountabili-ty” works for the poor at the union council level.

Table A.4 (columns 1–2, Appendix) sug-gest that there has been an 8% increase in the pro-vision of these schemes since the local govern-ments were enacted.29 The coefficients of the vil-lage dummies reveal that these were underprovid-ed villages pre-reforms, which suggests that theunion council reforms has some impact in termsof increasing the magnitudes of provision to pre-viously underprovided villages. However, thesecoefficients also reveal that not all villages havebenefited equally from the increase in provision.However, columns 1–2 also show that theincrease in provision is heavily biased towardsmembers of the dominant kinship group andhouseholds with higher levels of literacy.30 In asense, this is an unsurprising result given the find-


33

28 See Easterly (2003) and Gazdar (2002) (footnote 2).

29 Estimated from the coefficient of the variable POST.

30 Estimated from the coefficients of (POST*DOM) and (POST*LITERACY). Household literacy is measured as the proportion of literate adults in a house-

hold.

ings in the last three sections. This result confirmsthat the long route of accountability does notwork well for the poor and for marginalized citi-zens at least in the Punjab villages.

Table A.5 (columns 1–2, Appendix) testswhether having a politician resident31 in the vil-lage matters in obtaining a higher share in theallocation of schemes. The results show that it isassociated with a significant increase in post-2001provision.32 This indicates that spatial biases mayalso exist in post-reform provision. Again, thisresult is unsurprising given the ad hoc and infor-mal nature of the contact between village citizengroups and union-level politicians with regard toplanning. The ad hoc nature of the contact weak-ens the long route of accountability and makes iteasier for union-level politicians to target provi-sion to their own villages and constituencies.Furthermore, it appears that, at least, in the cen-tral Punjab case, literacy acts as an asset and beingpart of a dominant kinship group gives consider-able advantage in local-level politics. Again, itappears that belonging to a group that has a histo-ry of dominance gives its members considerableleverage in the political sphere.

This analysis provides some preliminaryevidence that the weaknesses in the electoral,planning, and budgetary spheres result in the anti-poor targeting of local public goods. In a sense,this result is unsurprising given that the poor areon average less likely to vote and that the infor-mal and ad hoc manner in which the budgetaryand planning process is structured results in thepoor having far lesser access to local politiciansfor the purpose of conveying their developmentneeds.


34

31 The resident politician could be a union-level or higher-tier politician.

32 Estimated from the coefficient (POST*NAZIM).

6 Analyzing the Short Route ofAccountability in Pakistan

The Local Government Ordinance 2001and the wider devolution reforms have instituteda number of citizen-centered bodies to overseepublic service delivery and catalyze direct citizenparticipation in the delivery of public goods andservices. CCBs are envisaged as spontaneous andvoluntary associations of citizens that can directlyaggregate and express citizen needs and allow citi-zens to participate in the delivery of public goodsand services (Qadir 2005 [footnote 15], pg. 9).

Similarly, SCs empower citizens as “stake-holders” by giving them the authority to monitorand supervise local service providers. These newbodies are expected to strengthen the short routeof accountability for the poor, women, and mar-ginalized groups. Accountability is expected toincrease because these bodies involve local govern-ments and citizens in a symbiotic relationship thatplaces the onus of managing and monitoring thedelivery of public services on both parties (Qadir2005 [footnote 15], pg. 9). Given the importanceof these bodies in the new governance framework,it is important to assess the degree to which theystrengthen the short route of accountability.

Measuring the impact of these bodies on

the short route accountability is difficult at pres-ent because issues of delegation, financial control,and regulation have curtailed their true empower-ment and this has affected their functioning duringthe first cycle of devolution.3 3 This study uses anindirect method to assess the potential of these bod-ies as a means to act as organizations that strengthenthe short route of pro-poor accountability. We sim-ply ask whether citizens belonging to poor and illit-erate households and nondominant kinship groupsare “aware” of the existence of CCBs and SCs. Ourinterest is in estimating whether these sociallyexcluded groups are less likely to be aware of thesebodies compared to nonmarginalized and nonpoorhouseholds. A finding that suggests that this is thecase reveals a systemic bias in “information flows”between poor and nonpoor households, whichcould emanate from the socioeconomic structureand/or the manner in which the selection and start-up mechanisms of these bodies have been designed.

We also ask whether poor citizens have anequal chance of participating in these bodies rela-tive to nonpoor groups. In addition, it is alsoimportant to ask whether poor citizens’ aware-ness of and participation in these bodies varieswith (i) the type of village-level social structure,(ii) the degree of inequality in asset-holdingamong groups, and (iii) the degree of socioeco-nomic exclusion at the village level.


35

33 See Qadir (2005) for a useful summary (footnote 16).

Table 13: Awareness of Citizen Community Boards (%)




6.1 Awareness of and Participation in CCBs

Tables 13 and 14 show that both the levelof awareness and level of participation in CCBs isextremely low. This is a consistent finding acrossvillages. This data is in line with the popular viewthat citizen awareness of and participation inCCBs is extremely low.

Multivariate regression results (Appendix,Table A.6, columns 1–3) suggest that being pooris not associated with a heightened lack of aware-ness of CCBs.34 In a sense, this is expected giventhe low levels of overall awareness of CCBs.However, the results also suggest that the propor-

tion of literate adults in a household is a signifi-cant positive correlate of CCB awareness. Thecoefficient remains significant after we have con-trolled for village fixed effects and for other con-trol variables. This implies that literacy as an assetmatters for catalyzing citizen awareness of CCBs.It also suggests that even after the first cycle oflocal governments, CCB awareness among illiter-ate households was nonexistent and that animportant body of citizen participation had failedto function in an inclusive mechanism thatengaged marginalized households.


36

Table 14: Participation in Citizen Community Boards(%)




Table 15: Awareness of School Councils(%)




34 The coefficient on poor becomes insignificant once we put in the controls.

6.2 Awareness of SCs

While the level of citizen awareness ofSCs is higher than the level of awareness of CCBs(Table 15), the overall level of citizen awareness ofthe former is also poor in absolute terms. Amajority of respondents reported lack of aware-ness of SCs. This is an interesting result given thatSCs were operational in nearly all the census vil-lages, yet Dir was the only village where a signifi-cant number of people were aware of SCs.

Multivariate regression results (Appendix,Table A.6, columns 4–7), however, report consid-erable inequity in awareness of SCs among citizencohorts. Poor households are significantly lesslikely to have heard of SCs even after controllingfor village effects. As opposed to this, being amember of the dominant kinship group has a sig-nificant positive association with SC awareness.Regression results35 also suggest that being a par-ent does not imply a higher degree of awareness ofSCs. This suggests that information regarding theexistence of SCs is not widespread at the villagelevel and that poor households and citizens fromnondominant kinship groups are excluded fromthe flow of this information. Part of the reasonfor this is that there is a strong, significant, andpositive association between household-level maleliteracy and SC awareness. It is more than likelythat poor households and households from non-dominant kinship groups are also more likely tobe illiterate.

Interestingly, the proportion of literateadult women in a household is insignificantlyassociated with awareness of SCs. This is an inter-esting result because one would expect SCs ingirls’ schools to attempt to involve educatedwomen from the village. The upshot is that SCshave not emerged as institutions that are inclusiveof the poor and marginalized. This points to theneed for designing selection mechanisms and con-sultation mechanisms at the SC level that activelyengage citizens who are poor and belong to non-dominant marginalized social groups and illiterate

households.

6.3 The Political Accountability Index

To address the question of whether poorcitizens are able to hold governments accountablealong multiple dimensions, we construct a politi-cal accountability index, which, in turn, involvesconstructing two separate indices. The first index,“political accountability 1,” is constructed by tak-ing an adjusted sum of scores of six separate sur-vey items: (i) meetings with nazim, (ii) meetingswith councilors, (iii) proportion of females whopossess a national identity card, (iv) proportion ofmales who possess a national identity card, (v)proportion of females who voted, and (vi) propor-tion of males who voted. An individual house-hold’s score on this index reflects the effect ofdiverse modes through which citizens can holdgovernments accountable through the long route.The principal-component-analysis method wasused to initially compute group scores for the sixinitial classes of variable identified above. Oncethese scores had been computed, the principal-component-analysis was used to compute a com-posite accountability score across the six groups.The higher this score, the greater an individualhousehold’s overall ability to hold the govern-ment accountable through the long route.

We use this index to address the basicquestion this study asks, i.e., are the poor,women, and socially excluded citizens as able tohold governments accountable as nonpoor, male,and elite groups? Table A.7, column 1 (Appendix)shows that poverty is negatively and significantlyrelated to political accountability through thelong route. Furthermore, the orders of magnitudeappear to be quite large. Column 3 of the tableshows that this relationship holds even after con-trolling for village fixed effects and other variablesof interest. In addition, we find that (i) being fromthe dominant kinship group and (ii) male andfemale literacy have a highly significant positivecorrelation with political accountability. This isin line with our earlier findings that dominant

Analyzing the Short Route of Accountability in Pakistan

37

35 Not reported in the Appendix but available from the author on request.

and literate groups are better able to hold govern-ments accountable through the long route whilepoor households are unable to effectively holdgovernments accountable.

We also calculate a second index of politi-cal accountability, “political accountability 2,”which takes into account awareness of CCBs andSCs along with the above six variables. The resultsusing this index are the same as the resultsobtained from using the first index. The resultssuggest the existence of a strong anti-poor, anti-illiterate, and anti-illiterate women accountabilitybias in Pakistan’s current governance structure.


38

7 Conclusion

7.1 Long Route of Accountability

7.1.1 Electoral Sphere

The main focus of this study was to exam-ine how well the long route of accountabilityworks in Pakistan. The results show that electoralparticipation in Pakistan is low and segmentedalong lines of gender, poverty, literacy, and castehierarchy. The study shows that current levels offemale electoral participation are low and, giventhese levels, it is unlikely that governments willbe responsive to the needs of women. We alsofind that patterns of voting across villages are cor-related with patterns of patriarchy documented inMohmand and Gazdar (2007) (footnote 20).

One interesting finding is that, control-ling for other factors, poor male and female vot-ers tend to vote far less than nonpoor male andfemale voters. It appears that the electoral sphereis suitably truncated by a household’s poverty sta-tus. This is in stark contrast with the case of India,where findings show that the “poor in India tendto vote more than the middle class or the rich, vil-lages more than cities, and lower castes more thanupper castes” (Keefer and Khemani 2004, andYadav 2000).36 This raises the important questionas to why political elites are unable/reluctant tomobilize the votes of the poor, which constitute asignificant demographic population. Some pre-liminary explanations for this paradox are givenbelow in Section 7.3 explaining the persistence ofstructural constraints.

The study also finds that literacy is animportant positive correlate of voting turnoutsboth among male and female voters. However,the evidence from Gazdar (2007) (footnote 6) andMohmand and Gazdar (2007) (footnote 20) sug-gests that literacy is endogenous to poverty.37 Thissuggests that poverty and literacy appear to have

a self-reinforcing cycle, which suggests that theconstraints to voting are much more structuralthan appears at first glance. This interpretation isstrengthened when we recognize that female andmale members of a dominant kinship group havemuch higher voting turnouts compared to othersin their village. The dominant kinship groupresult is interesting because, as pointed out byGazdar (2007), the variable used to capture domi-nant kinship groups is the group that was histori-cally recognized as the dominant group (footnote6). This variable makes no assumption about thecurrent socioeconomic status of the kinshipgroup. The fact that historical groups continue todominate the electoral sphere irrespective of theircurrent status points to the persistence of struc-tural constraints. Again, this raises more ques-tions than it answers. An attempted explanationfor the persistence of structural constraints isgiven below.

7.1.2 Planning and Budgeting Spheres

The study shows that, under the currentreform, an “institutional gap” remains in thatindividual mauzas/dehs and pinds/goths, as wellas citizens do not participate directly in budget-ing, development planning, and the selection ofschemes. The enactment of VNCs was supposedto bridge this “institutional gap;” however, thesebodies have not been made operational eventhough the recently promulgated local govern-ments have entered their second electoral cycle.The “institutional gap” means that the processthrough which citizens convey their developmentneeds to their union nazim, naib nazim, and coun-cilors remains “informal” and “ad hoc” and is notbinding on the agents of the state.

Furthermore, citizens can only voicetheir “levels of satisfaction” concerning local gov-ernment budgets, plans, and schemes ex post atthe time of election of their local representatives.No formal space for citizen-state engagement on


39

36 See: Keefer, P., and S. Khemani. 2004. Why Do the Poor Receive Poor Services? Economic and Political Weekly, 28 February; and Yadav, Y. 2000.

Understanding the Second Democratic Upsurge. In Transforming India: Social and Political Dynamics of Democracy. Edited by F. Frankel. India: Oxford

University Press.

37 We would like to thank Dr. Aliya Khan and Mr. Richard Martini for pointing this out to us.

planning and budgeting is available in the currentinstitutional framework. This is in contrast toexperiments with direct citizen participation inWest Bengal and Kerala in India and Porte Algrein Brazil, where citizens play a direct role in budg-et formulation and oversight. Citizens can exer-cise some decision-making control over the CCBcomponent of the budget, but there is no enforce-able requirement that the formulation of CCBsmust involve or represent the wider village-levelcitizenry.

Another key finding of the study is thatcitizens’ access to “informal” and “ad hoc” meet-ings at which they can convey their developmentneeds to union nazims, naib nazims, and coun-cilors, is not symmetric between different citizentypes. The study shows that a small minority ofrespondents report attending these informal meet-ings. More importantly, there is a systemic anti-poor bias in respondents reporting informal con-tact. Furthermore, the study also shows that theproportion of literate adults in a household is astatistically strong correlate of attendance to theseinformal meetings. This suggests that the voice ofilliterate and poor households is systemically lesslikely to be counted in the assessment of needs inthe planning process.

7.1.3 Anti-Poor Targeting Bias in ServiceDelivery

The study documents considerable biasessince the local government reforms in the target-ing of local public goods (such as sanitation andsoling) in central Punjab38 between householdswithin the same village. For these villages, thestudy shows that, while there has been a signifi-cant increase in provision, this increase has beenbiased in favor of dominant kinship group house-holds with higher levels of literacy. The study alsofinds that having a politician resident in the vil-lage matters for obtaining a higher share in theallocation of schemes, which suggests that spatialbiases may also exist in post-reform provision. Animportant caveat that needs to be kept in mind

when interpreting these results is that this analy-sis was conducted for a small number of villages inPunjab, albeit the household sample was verylarge.

These results are unsurprising given thead hoc and informal nature of the contactbetween village citizen groups and union-levelpoliticians on issues of planning as well as themanner in which the electoral sphere is struc-tured. The results show that these weaknesses inthe long route of accountability make it easier forunion-level politicians to target provision to theirown villages and constituencies. What is impor-tant is that weaknesses in the long route ofaccountability have real outcome effects in thesample villages. Furthermore, it appears that, atleast in the central Punjab case, literacy and beingpart of a dominant kinship group gives consider-able advantage in local-level politics.

7.2 Awareness of Citizen-Based ServiceDelivery and Oversight Bodies

With regard to the short route of account-ability, the study asks a much more limited set ofquestions: (i) to what degree are the poor, women,and marginalized and socially excluded citizens“aware” of the newly constituted citizen-basedservice delivery and oversight bodies; and (ii) howdoes the level of awareness in these citizen typescompare with other citizen types. The study asksthese limited questions because citizen-based serv-ice delivery and oversight bodies such as CCBsand SCs are in the process of being institutional-ized and at this point it is difficult to conduct animpact assessment of them. However, our con-tention is that this more restricted set of questionsgives important insights into the process of insti-tutionalization of these citizen-based bodies andthe extent to which the poor, women, and mar-ginalized and socially excluded citizens areinvolved in the process. As pointed out earlier,this is an important set of questions to ask becauselocal governments had completed their first elec-


40

38 A similar analysis could not be done for other provinces because of the low variation in our outcome variable. This is because villages were either

universally provided local public goods prereforms, or universally not provided any local public goods post-reforms.

toral cycle at the time of this study and answersfrom this set of questions will give us an insightinto the likelihood of the poor, women, and illit-erate households accessing these citizen-basedbodies within the current system.

The study shows that the level of aware-ness of CCBs and SCs is extremely low, and thatthere are systemic biases in lack of awarenessbetween different citizen groups. In the case ofCCBs, we find that literacy is the most significantcorrelate of CCB awareness. In the case of SCs,poor and illiterate households and women are onaverage less likely to be aware that such bodiesexist. As opposed to this, we find that being amember of the dominant kinship group of the vil-lage has a significant positive association withawareness of SCs. This, again, suggests that struc-tural constraints might impair the access that thepoor, women, and illiterate citizens have to citi-zen-based bodies unless measures are taken tomake access to these bodies more representativeand participatory.

7.3 Persisting Structural Constraints toAccountability

7.3.1 Degrees of Socioeconomic Inequality

Village case studies and empirical analysispoint towards the persistence of structural con-straints that continue to inhibit the long route ofaccountability in Pakistan’s governance system.39

In what follows, we measure the strength of thelong route of accountability by the degree towhich citizens can exercise independent choiceover their decision to vote. Analysis reveals threesalient relationships between socioeconomic andpolitical conditions and accountability:

(i) Where socioeconomic inequality is low andelectoral politics competitive, we find theprevalence of independent voting, which wewould expect to strengthen the long route ofaccountability for the poor.

(ii) High levels of socioeconomic inequality when

combined with personality-based politicsresults in dependent village-level patron-clientvoting, which appears to weaken accountabil-ity.

(iii) However, low levels of socioeconomicinequality are not sufficient to ensure inde-pendent voting if it is combined with monop-oly higher-tier factions and personality-basedpolitics.

We also find that the low propensity tovote among women is integrally linked to theirrestricted patterns of physical and social mobility.This, in turn, appears to be related to patterns ofendogamy that exist to preserve and perpetuatekinship groups (Mohmand and Gazdar 2007, foot-note 20).

This suggests that there is considerablevariation at the local level between levels ofsocioeconomic inequality across villages from dif-ferent districts. It appears that the socioeconomicstructure in some villages, such as ourMuzaffargarh and Dir villages, has evolved in thedirection of greater socioeconomic equality,whereas in villages from other districts we findthat high levels of socioeconomic inequality con-tinue to persist, suggesting the persistence ofstructural constraints. In particular, dependentvoting appears to be related to (i) the degree oflandlessness, (ii) the difference in poverty head-counts between socially excluded and dominantcastes, and (iii) dependence on others’ land as aservant.

Evidence of nondependent voting isfound in the cases of Dir and Muzaffargarh. Thesecases give interesting insights into the type of con-ditions that are needed to break the dependentvoting cycle. The three most salient features ofthese villages that put them apart from the restare: (i) low levels of social equality as measured bythe difference in poverty headcounts betweensocially excluded and dominant kinship groups,(ii) active political party activity, and (iii) strongpolitical competition. The political conditions

Conclusion

41

39 This section draws heavily on Section 5.2 of this paper.

prevailing in these villages are a result of the struc-tural features discussed below (Section 7.3.2) aswell as the history of local-level political activism.Low levels of social equality are partly the resultof low levels of inequality in land ownership.However, an important difference between thesevillages and others is the access its socially exclud-ed and nondominant castes have to nonfarm activ-ities (Table 16).

Table 16 shows that the main source ofnonfarm employment in Dir is the internationallabor market and government employment. InMuzaffargarh, the main source of nonfarmemployment is the domestic labor market.Furthermore, in both Dir and Muzaffargarh, non-farm labor market opportunities are available tonondominant kinship groups and these opportu-nities tend to play an equalizing role in the extentof poverty between elite and nonelite groups. Inthe Muzaffargarh village, this is done by redistrib-uting poverty but, in both cases holding political

conditions constant, access to nonfarm labor mar-ket opportunities reduces the extent of dependentvoting¸ which is likely to lead to strengthenedpro-poor accountability. This suggests thatdependent voting is more likely to occur in agrar-ian transitions that decrease the relative access toland for nondominant kinship groups without acommensurate compensation of income throughnonfarm activities

This analysis suggests that strengtheningaccountability would require either:

w increasing nondominant kinship groups’access to land assets,

w diversifying economic opportunities for non-dominant kinship groups by opening up thepossibility of entering nonfarm activities, or

w ensuring a reduction in the poverty differentialbetween dominant and nondominant kinship groups.


42

Table 16: Male Workers by Sector and Activity (Primary Occupation)(%)Sector/Activity Chakwal Dir Mardan Muzaffargarh Sanghar Toba Tek Singh Thatta TotalEconomically activeproportion of refer-ence age group

88.5 74.3 77.1 83.6 90.9 78.6 86.7 83.3

Self-cultivator 45.9 23.5 11.9 25.1 22.3 22.8 1.6 19.7Tenant cultivator 2.7 0.4 20.9 6.9 45.6 4.2 2.0 15.2Livestock owner 2.1 2.2 6.3 0.9 6.6 2.2 2.8 3.5Casual labor 15.1 15 22.1 30.9 8.7 17.1 55.4 23.9Skilledworker/mechanic

1.4 0.4 0.0 3.5 0.1 2.4 16.3 4.0

Farm servant 0.0 0.0 0.0 0.5 1.7 8.2 0.0 2.4Nonfarm servant 0.0 0.9 0.4 1.4 0.7 0.9 2.2 1.1Government job 18.5 14.6 5.1 0.7 3.1 12 1.8 6.4Shop 2.1 4.0 2.0 4.4 2.4 5.2 11.3 4.9Transport 0.7 0.0 7.5 0.9 2.0 2.1 0.2 1.8Factory worker 0.0 0.9 2.0 4.8 0.6 13.5 0.2 4.2Labor abroad 0.0 31.4 10.7 0.2 0.0 1.9 0.0 3.8Labor in big city 0.7 2.7 8.3 4.4 0.3 2.5 0.6 2.4Other 11.0 3.5 2.8 15.4 5.7 4.9 5.8 6.9Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0Source: Author’s fieldwork

7.3.2 Degree of Political Competition

We also find that low levels of socioeco-nomic inequality reduce dependent voting in onlythose areas where politics has evolved to becomecompetitive. In areas where politics is personal-ized and dominated by oligarchies and monopo-lies, we find the persistence of dependent votingirrespective of the level of socioeconomic inequal-ity. In addition, we find that, in these areas, poorhouseholds are far less likely to vote because ofpolitical party affiliation and far more likely tovote because of the pressure of a village influen-tial. Again, it appears that structural constraints inthe political sphere prevent the poor from exercis-ing their choice to act as independent voters.These structural constraints include the absenceof well-organized political parties that can mobi-lize the poor and reduce their dependence on thevillage influential in the electoral sphere.Strengthening the long route of accountabilitywould require strengthening the presence of polit-ical collectivities capable of mobilizing the poorin the electoral sphere.

Elections on a nonparty basis appear tostrengthen personality-based politics and rein-force precisely the type of politics that strength-ens dependent voting. Instead of limiting electoralcompetition, the governance structure shouldencourage competition by enabling the entry oforganized political parties at the local level thathave the incentive to mobilize poor voters andcan affect this mobilization by reducing the trans-action and financial costs of electoral mobilizationamong poor voters. In the absence of these collec-tivities, the poor appear to be faced with consider-able financial and transaction costs that reducetheir incentive to exercise their choice as inde-pendent voters. However, there should be pres-sure on political parties to democratize and estab-lish grassroots representation, internal accounta-bility, and electoral mechanisms to increase the

leverage of the poor vis-à-vis dominant grouppoliticians.

7.4 Constraints on Voter Turnouts amongthe Poor

An important finding of this study is thatpoor voters tend to vote less, controlling for otherfactors, and are more likely to vote as dependentvoters. This is an important finding and needsexplanation.40 In the Pakistani context, the stan-dard explanation is that it reflects voters’ lack offaith in the democratic process because of thecountry’s weak democratic tradition. However,this does not account for why the poor have sys-tematically low voting turnouts since lack of faithin the democratic process should affect the poorand nonpoor equally. The standard explanationfor this finding in other contexts is that povertyincreases the opportunity cost of investing timeand resources in the electoral sphere (Krishna2003 and Young 2000).41

This still begs the question why politi-cians do not have the necessary incentive toorganize this potentially large vote bank. Oneexplanation for this is that politicians find it diffi-cult to make credible promises to poor voters onthe provision of broadly targeted public goods orredistributive private transfers.42 As a result, thepolicy platforms of different politicians becomeirrelevant to elections. The lack of credibility ofpromises implies that the poor will participate lessactively in the electoral sphere because theyexpect lower outputs from their representatives.In addition, evidence from India seems to suggestthat politicians’ promises are credible to a narrowsegment of client voters with whom they havelong-term personal relationships, while other vot-ers, especially the poor, tend to be discounted inthe electoral game (Keefer and Khemani 2004,

Conclusion

43

40 We would like to thank Dr. G. M. Arif for raising this question.

41 See: Krishna, A. K. 2003. Poor People’s Participation in Democracy at the Local Level. Working Paper Series SAN03-04, Terry Sanford Institute of Public

Policy, Duke University, North Carolina, p. 3; and Young, I. M. 2000. Inclusion and Democracy. USA: Oxford University Press.

42 This argument draws on Keefer and Khemani (2004) who focus on public goods provision (footnote 36). The reasoning holds equally well for redistributive

private transfers such as land reforms, homestead land schemes, or the efficient and equitable functioning of social safety nets.

footnote 36). This provides an explanation for ourfinding that, in the survey villages, local publicgoods tended to target dominant kinship grouphouseholds and not the landless and sociallyexcluded.

In addition, lack of coordination amongvoters makes it even more difficult for politiciansto credibly commit to certain promises to a broadsegment of the electorate and for excluded votersto effectively threaten politicians with electoraldefeat. We have seen that, in Pakistan, highdegrees of socioeconomic inequality and theabsence of sufficient political competition make itextremely difficult for the poor to enter the elec-toral sphere as independent voters. This bluntsthe credible electoral threat they exercise andmutes the credibility of politicians’ promises toprovide broadly targeted public goods. This lackof credibility reinforces their incentive to targetthose clients with whom they have personal link-ages, because their promises are credible to thisnarrow segment.

Finally, the particular system of local elec-toral representation in Pakistan does not createstrong incentives for politicians at the district ortehsil/taluka levels to directly invest in votermobilization. These higher-level politicians—whomight enjoy economies of scale in political invest-ment due to the size of their constituencies—areelected indirectly by an electoral college. Theirprime focus, therefore, is likely to be the unioncouncil level-politician and not the individualvoter. This indirect form of representation tendsto weaken the political-economic rationale fororganized and stable political parties, because itundervalues the potential for economies of scalein political transactions. One of the functions ofpolitical parties—and a clear advantage that theyenjoy over individual dominant group politi-cians—is to internalize the economies of scale.

Our analysis suggests that political clien-telism in Pakistan appears to have been strength-ened by (i) the absence of effective political partycompetition with well-defined voter bases andcampaign platforms, (ii) by asymmetric relations

of socioeconomic dependence among poor andnonpoor citizens, and (iii) by the persistence ofconsiderable socioeconomic inequality betweendifferent social groups. A switch from indirect todirect elections for higher levels of local govern-ment (tehsil and district) is likely to increase theincentive for politicians to invest in the mobiliza-tion of poor voters. This might be a possibledirection of reform if the pro-poor potential oflocal democracy is to be realized more fully. Theparticipation of poor voters is likely to increase ifpoliticians are able to make more credible prom-ises with respect to public goods provision orredistributive private transfers. Some issues ofcredibility are likely to be addressed over time asthe system becomes more institutionalized andinitial confusion about the division of powers andresponsibilities is resolved.

7.5 Is Literacy a Panacea?

This study has emphasized that, in addi-tion to socioeconomic equality and political com-petition, literacy is a strong positive correlate of(i) electoral accountability, (ii) access to the plan-ning mechanism and public services, and (iii)awareness of the newly constituted citizen-basedservice delivery and oversight bodies. However, acareful analysis reveals that literacy is not apanacea for weak accountability. Gazdar (2007)shows that literacy outcomes are much worseamong the poor, precisely the group of voters thatare dependent and less active (footnote 6). Thissuggests that literacy itself is a differentiated assetand that high levels of inequality in literacy out-comes represent an equilibrium that is endoge-nous to poor governance and to socioeconomicinequality. It is, therefore, unlikely that literacywill be the variable that strengthens the ability ofthe poor to hold governments accountable. Theequilibrium represents a vicious cycle where poorcitizens who need to acquire education as an assetwith which to enhance their ability to compete inthe electoral sphere are least able to exercise“voice” in the electoral sphere, precisely becausethey lack access to education and literacy.Furthermore, the local governments that need toeducate the poor most are likely to be the very


44

governments least responsive to the poor becauseof their lack of educational assets. This suggeststhat removing the structural constraints men-tioned above may be a necessary condition foraltering the low accountability equilibrium.

However, additional measures thatincrease access to literacy and education for poorpeople are likely to have positive governanceeffects. It needs to be reiterated that, in a politicaleconomy framework, it is unclear whether gov-ernments that are irresponsive to poor and illiter-ate citizens will have the incentive to providethem with education and literacy, especially in theabsence of changes that reduce socioeconomicinequality and increase political competition atthe local level. If the situation can indeed be char-acterized as a “low-level equilibrium” in whichsocial, economic, and political structures rein-force one another, it is important to think ofmutually reinforcing reforms in the political,social, and economic spheres. Our findings sug-gest that electoral changes in the absence of otherpro-active reforms may not be leading drivers ofpro-poor change. The analysis reported herereveals that historically dominant groups havetended to retain their influence in the politicalsphere even after having lost their economic pre-eminence in a number of cases.

7.6 Persisting Institutional andInformation Gaps43

The study finds that the ad hoc and infor-mal nature of budgetary needs assessment andplanning results in avenues of elite capture at theunion level. This is because of an “institutionalgap” in the budgetary and planning processes thatdo not allow for any direct participation by vil-lages and village citizens. VNCs, that were to pro-vide direct participation, remain dormant, and theonly mechanisms available to citizens for voicingtheir satisfaction with budgetary choices is eitherinformal contact or at the time of elections. Wehave already argued that, with weaknesses in thelong route of accountability, elections are going

to provide a blunt instrument with which thepoor, women, and socially excluded can hold elec-toral representatives accountable.

Institutionalizing budgetary mechanismsthat allow for citizen participation is an impor-tant precondition to give the poor, women, andsocially excluded citizens a voice in budgetarychoices. For these measures to strengthen pro-poor accountability, they must be accompaniedby other measures that strengthen electoral com-petition and reduce the difference in access to keyassets between poor and nonpoor citizens. Thereis, nonetheless, a possibility of “elite capture” inparticipatory budgetary mechanisms in areas thatare plagued by high levels of socioeconomicinequality and the absence of political competi-tion. In these areas, it is extremely important tostrengthen planning systems in tehsil and districtgovernments and to institute performance-basedgrants for union councils, which are tied to equi-ty indicators that aim to give union councilsincentive to make pro-poor allocations. Theseperformance-based grants will require the institu-tionalization of a mapping exercise that, usingGIS technology, can document the availability offacilities at the village and union council level.

The mapping exercise will provide thenecessary information on differences in access tofacilities within and between villages. Union-levelreport cards can be generated and reported at thelevel of union, village, and lane. In addition, disag-gregated budgetary information must be providedat the union, village, and lane level. This mappinginformation could provide the foundation formaking performance-based grants operational forunion councils. In addition, this informationcould be made public at district, tehsil, and unioncouncil meetings and through local nongovern-ment organizations at village-level meetings. Theaim of this information would be to make citizensaware of how poor services are and which servic-es are doing worse than others, but also how poorservices are in their union, village, and lane rela-tive to others. An added benefit of this interven-

Conclusion

45

43 This section draws on Keefer and Khemani (2004) (footnote 36).

tion would be that it would allow voters to bench-mark the performance of their union-level repre-sentatives relative to others both in terms ofacquiring budgets from higher tiers of local gov-ernment and in establishing equity of distributionwithin union councils.


46

References

Alavi, H. 1971. The Politics ofDependence: A Village in West Punjab. SouthAsian Review 4 (2).

Bardhan, P., and D. Mookherjee. 2005.Introductory chapter to Decentralization inDeveloping Countries: A Comparative Perspective.Edited by P. Bardhan and D. Mookherjee. TheMIT Press.

Besley, T., R. Pande, L. Rahman, and V.Rao. 2001. The Politics of Public GoodsProvision: Evidence from Indian local govern-ments. Unpublished manuscript. London Schoolof Economics and Political Science, London.

———. 2003. The politics of public goodsprovision: Evidence from Indian local govern-ment. Mimeo. London School of Economics andPolitical Science, London. Forthcoming inJournal of the European Economic Association.

Charlton, J., M. Cyan, Z. Hasnain, N.Manning, D. Porter, and H. Sharif. 2003.Devolution in Pakistan—Preparing for ServiceDelivery Improvements. World Bank, AsianDevelopment Bank (ADB), and UK Departmentfor International Development (DFID).

Cheema, A., and S. Mohmand. 2004.Provision Responses to Devolved ServiceDelivery—Case Evidence from Jaranwala Tehsil.Manuscript. Lahore University of ManagementSciences (LUMS)-McGill Social EnterpriseDevelopment Programme. LUMS, Lahore.

———. 2005. The Political Economy ofDevolved Provision: Equity-based Targeting orElite Capture—Case Evidence from 2 PakistaniUnions. Manuscript. LUMS, Lahore.

Chattopadhyay, R., and E. Duflo. 2003.The Impact of Reservation in the Panchayati Raj:Evidence From a Nationwide RandomizedExperiment. Typescript. Massachusetts Instituteof Technology, Cambridge.

Dreze, J., and A. Sen. 1996. I n d i a nDevelopment: Selected Regional Perspectives. NewDelhi: Oxford University Press.

Easterly, W. 2003. The Political Economyof Growth without Development: A Case Studyof Pakistan. In In Search of Prosperity: AnalyticalNarratives on Economic Growth. Edited by D.Rodrick. New Jersey: Princeton University Press.

Gazdar, H. 2002. A Qualitative Survey ofPoverty in Rural Pakistan: Methodology, Data, andMain Findings. Karachi: Collective for SocialScience Research.

———. 2005. Scoping Paper. Revised draftsubmitted for Determinants and Drivers ofPoverty Reduction and ADB’s Contribution inRural Pakistan. ADB, Islamabad.

———. 2007. Rural Economy andLivelihoods in Pakistan. Thematic paper preparedunder TA4319, Determinants and Drivers ofPoverty Reduction and ADB’s Contribution inRural Pakistan. ADB, Islamabad.

Ghatak, M., and M. Ghatak. 2002. RecentReforms in the Panchayat System in West Bengal:Toward Greater Participatory Governance?Economic and Political Weekly. 5 January.

Government of Pakistan. 2001. 1 9 9 8Census Report of Pakistan. Islamabad: PopulationCensus Organization.

Keefer, P., A. Narayan, and T.Vishwanath. 2005. The Political Economy ofDecentralization in Pakistan. In Decentralizationin Developing Countries: A ComparativeP e r s p e c t i v e. Edited by P. Bardhan and D.Mookherjee. The MIT Press.

Keefer, P., and S. Khemani. 2004. WhyDo the Poor Receive Poor Services? Economic andPolitical Weekly, 28 February.

Kohli, A. 1987. The State and Poverty inIndia: The Politics of Reform . CambridgeUniversity Press.


47

Krishna, A. K. 2003. Poor People’sParticipation in Democracy at the Local Level.Working Paper Series SAN03-04, Terry SanfordInstitute of Public Policy, Duke University,North Carolina.

Mohmand, S., and H. Gazdar. 2007.Social Structures in Rural Pakistan. Thematicpaper prepared under TA4319, Determinants andDrivers of Poverty Reduction and ADB’sContribution in Rural Pakistan. ADB, Islamabad.

Qadir, A. 2005. Straddling State andSociety: Devolution and Pakistan’s AmbiguousAssociations. Preparatory analytical paper pre-pared for ADB workshop. ADB, Islamabad.

United Nations DevelopmentProgramme (UNDP). 2004. Human DevelopmentReport, Cultural Liberty in Today’s Diverse World.USA: Oxford University Press.

World Bank. 2002. Pakistan PovertyAssessment. Poverty in Pakistan: Vulnerabilities,Social Gaps, and Rural Dynamics. Washington,DC: World Bank.

———. 2004. World Development Report2004, Making Services Work for Poor People.Washington, DC: World Bank.

Yadav, Y. 2000. Understanding theSecond Democratic Upsurge. In T r a n s f o r m i n gIndia: Social and Political Dynamics of Democracy.Edited by F. Frankel. India: Oxford UniversityPress.

Young, I. M. 2000. Inclusion andDemocracy. USA: Oxford University Press.


48

APPENDIX: BACKGROUNDDATA


49

Table A.1: Multivariate Regressions for Dependent Variable

Variable Voted Voted Voted Females Females Females 1 2 3 4 5 6

Intercept 0.577a 0.794a 0.578a 0.331a 0.561a 0.494a

(0.008) (0.027) (0.012) (0.011) (0.030) (0.045)

Poor -0.083a -0.040b -0.036c

(0.018) (0.018) (0.021)

Female -0.265a -0.265a -0.302a

(0.012) (0.012) (0.016)Literacy -0.001

(0.016)

Dir -0.427a -0.505a -0.429a

(0.032) (0.045) (0.046)

Mardan -0.249a -0.384a -0.249a

(0.032) (0.042) (0.043)

Muzaffargarh -0.190a -0.232a -0.133a

(0.030) (0.036) (0.038)

Sanghar -0.257a -0.260a -0.156a

(0.029) (0.034) (0.038)

Thatta -0.277a -0.439a -0.333a

(0.030) (0.040) (0.049)

Toba Tek Singh -0.113a -0.014a -0.071b

(0.012) (0.034) (0.035)

Literacy (Female)c 0.091a 0.099a

(0.024) (0.030)

Dom 0.0498c

(0.019)R-Squared 0.006 0.090 0.103 0.020 0.200 0.256Observations 1,656 1,650 1,648 1,186 1,186 839Village Dummies No Yes Yes No Yes YesControls No No Yes No No Yes

Notes: Dependent variables: (i) “voted” = adults who voted in 2005 union council elections, (ii) “females” = proportion of females in household who voted.Independent variables: (i) “poor” = household below the poverty line, (ii) “literacy” = proportion of literate adults in household, (iii) “dom” = member ofdominant kinship group, (iv) controls include: occupation, tenancy, migration status of household, landownership, headcount, and female landownership.

a 1% level of significance.b 5% level of significance.c 10% level of significance.



50

Table A.2: Multivariate Regressions for Dependent Variable

Variable Males Males Males Identity Card Identity Identity Card Card

1 2 3 4 5 6


(0.010) (0.032) (0.044) (0.010) (0.032) (0.048)

Poor -0.007 -0.043b -0.062a -0.056a -0.043a -0.055a

(0.015) (0.017) (0.019) (0.014) (0.016) (0.019)

Dir -0.218a -0.213a -0.022 -0.008(0.040) (0.041) (0.039) (0.042)

Mardan -0.109a -0.048 -0.166a -0.126a

(0.042) (0.041) (0.041) (0.042)

Muzaffargarh -0.068c -0.050 -0.146a -0.119a

(0.037) (0.037) (0.037) (0.038)

Sanghar -0.149a -0.083b -0.155a -0.010a

(0.035) (0.039) (0.035) (0.039)

Thatta 0.014 0.122a -0.083b 0.019(0.038) (0.048) (0.038) (0.046)

Toba Tek Singh -0.045 -0.044 -0.081b -0.083b

(0.035) (0.035) (0.035) (0.038)

Literacy (Male)c 0.139a 0.206a

(0.031) (0.031)

Dom 0.037b 0.0420b

(.0184) (0.019)R-Squared 0.0001 0.057 0.057 0.009 0.040 0.172Observations 1,479 1,479 1,585 1,583 1,583 1,058Village Dummies No Yes Yes No Yes YesControls No No Yes No No Yes

Notes: Dependent variables: (i) “males” = proportion of males in household who voted, (ii) “identity card” = proportion of males in household with iden-tity cards. Independent variables: (i) poor = household below the poverty line, (ii) literacy = proportion of literate adults in household, (iii) dom = mem-ber of dominant kinship group, (iv) controls include: occupation, tenancy, migration status of household, landownership, headcount, and female landown-ership.



APPENDIX: BACKGROUND DATA

51

Table A.3: Multivariate Regressions

Variable Nazim Nazim Nazim Councilor Councilor Councilor1 2 3 4 5 6


(0.010) (0.034) (0.057) (0.01) (0.033) (0.054)

Poor -0.067a -0.076a -0.053b -0.068a -0.080a -0.058b

(0.016) (0.017) (0.024) (0.015) (0.017) (0.023)

Dir -0.223a -0.314a -0.043 -0.095c

(0.042) (0.053) (0.041) (0.050)

Mardan -0.244a -0.342a -0.128a -0.180a

(0.044) (0.053) (0.043) ((0.051)

Muzaffargarh -0.244a -0.313a -0.094b -0.107b

(0.039) (0.049) (0.038) (0.046)

Sanghar -0.269a -0.374a -0.110a -0.202a

(0.037) (0.049) (0.0364) (0.047)

Thatta -0.168a -0.248a -0.034 -0.055(0.040) (0.059) (0.039) (0.056)

Toba Tek Singh -0.250a -0.386a -0.135a -0.184a

(0.037) (0.046) (0.036) (0.043)

Literacy (Male)c 0.077b 0.075b

(0.038) 0.037R-Squared 0.011 0.049 0.068 0.012 0.029 0.038Observations 1,648 1,647 1,081 1,636 1,635 1,073Village Dummies No Yes Yes No Yes YesControls No No Yes No No Yes

Nazim = mayor.

Notes: Dependent variables: (i) “nazim” = meeting with union nazim to convey development needs, (ii) “councilor” = meeting with union councilor to con-vey development needs. Independent variables: (i) poor = household below the poverty line, (ii) literacy = proportion of literate adults in household, (iii)con-trols include: occupation, tenancy, migration status of household, landownership, headcount, female landownership, and membership of dominant kinshipgroup.




52


Variable Provision of Sanitation/Soling Provision of Sanitation/Soling1 2

Intercept 0.016 0.147b

(0.040) (0.032)

Post 2001 0.077b 0.015c

(0.350) (0.130)

Land 0.004c 0.005b

(0.002) (0.002)

Post (Land)c 0.001 0.001(0.003) (0.003)

Dom 0.014 0.044(0.028) (0.032)

Post (Dom)c 0.105a 0.071c

(0.04) (0.035)Literacy 0.007 0.010

(0.005) (0.006)

Post (Literacy)c 0.306a 0.377b

(0.050) (0.153)

Faisalabad 1 0.422a

(0.049)

Faisalabad 2 0.124b

(0.049)

Faisalabad 3 0.181a

(0.048)

Faisalabad 4 0.205a

(0.049)Muzaffargarh -0.035

(0.039)

Toba Tek Singh 0.221a

(0.038)R-Squared 0.237 0.190Observations 2,102 2,102Village Dummies Yes NoControls Yes Yes

Notes: Independent variables: (i) post 2001 = provision of sanitation/soling post-2001.

a 1% level of significance.

b 5% level of significance.

c 10% level of significance.



53


Variable Provision of Sanitation/Soling Provision of Sanitation/Soling1 2

Intercept 0.126a 0.03(0.032) (0.041)

Post 2001 0.014 0.402(0.131) (0.036)

Land 0.005b 0.004b

(0.002) (0.002)

Post (Land)c 0.001 0.001(0.003) (0.003)

Dom 0.044 0.023(0.032) (0.028)

Post (Dom)c 0.071c 0.083b

(0.035) (0.04)Literacy 0.010 0.006

(0.006) (0.008)

Post (Literacy)c 0.377b 0.329a

(0.153) (0.05)

Nazim 0.105 0.326a

(0.092) (0.054)

Post (Nazim)c 0.198c 0.193a

(0.094) (0.049)

Jaranwala 2 0.125b

(0.049)

Jaranwala 3 0.182a

(0.048)

Jaranwala 4 -0.218a

(0.044)Muzaffargarh -0.0339

(0.039)

Toba Tek Singh 0.220a

(0.038)R-Squared 0.19 0.24Observations 2,102 2,102Village Dummies No YesControls Yes Yes

Notes: Independent variables: (i) post 2001 = provision of sanitation/soling post-2001.






54


Variable CCBs CCBs CCBs SCs SCs SCs SCs1 2 3 4 5 6 7

Intercept 0.034a 0.028c 0.007 0.141a 0.197a 0.082b 0.097b

(0.005) (0.016) (0.024) (0.009) (0.03) (0.044) (0.053)

Poor -0.026a -0.016c -0.165 -0.111a -0.074a -0.058a -0.067a

(0.007) (0.008) (0.011) (0.014) (0.015) (0.019) (0.023)

Literacy 0.035c 0.105a

(0.019) (0.035)

Dir 0.073a 0.080a 0.136a 0.119a 0.072(0.02) (0.024) (0.037) (0.043) (0.05)

Mardan -0.023 -0.026 -0.141a -0.146a -0.194a

(0.021) (0.024) (0.038) (0.044) (0.051)

Muzaffargarh -0.002 0.002 -0.127a -0.104a -0.136a

(0.018) (0.022) (0.034) (0.039) (0.046)

Sanghar -0.016 -0.008 -0.137a -0.104a -0.141a

(0.018) (0.021) (0.033) (0.039) (0.046)

Thatta -0.011 -0.011 -0.086b 0.001 -0.017(0.019) (0.025) (0.035) (0.046) (0.055)

Toba Tek Singh 0.011 0.003 -0.037 -0.053 -0.092b

(0.017) (0.02) (0.032) (0.037) (0.043)Land -0.0002

(0.0005)

Literacy (Male)c 0.078b

(0.036)

Literacy (Female)c 0.053(0.033)

Dom 0.065a 0.0762a

(0.019) (0.022)R-Squared 0.008 0.035 0.036 0.037 0.111 0.114 0.124Observations 1,656 1,650 1,311 1,652 1,648 1,309 1,309Village Dummies No Yes Yes No Yes Yes YesControls No No Yes No No Yes Yes

CCB = community citizen board, SC = school council.

Notes: Dependent variables: (i) “CCBs” = households aware of CCBs, (ii) “SCs” = households aware of SCs. Independent variables: (i) poor = householdbelow the poverty line, (ii) literacy = proportion of literate adults in household, (iii) dom = member of dominant kinship group, (iv) controls include: occu-pation, tenancy, migration status of household, landownership, headcount, and female landownership.




55


Variable Political Accountability1 Political Accountability1 Political Accountability11 2 3

Intercept -0.229a 0.526a 0.486b

(0.056) (0.154) (0.221)

Poor -0.343a -0.207b -0.314a

(0.090) (0.098) (0.105)

Land 0.004b

(0.002)

Dom 0.381a

(0.094)

Dir -1.250a -1.137a

(0.251) (0.225)

Mardan -1.309a -0.795a

(0.220) (0.215)

Muzaffargarh -.875a -0.664a

(0.188) (0.182)

Sanghar -1.011a -0.461b

(0.180) (0.191)

Thatta -1.143a -0.259(0.206) (0.236)

Toba Tek Singh -0.525a -0.475a

(0.171) (0.163)

Literacy (Female)c 0.251c

(0.151)

Literacy (Male)c 0.575a

(0.171)R-Squared 0.010 0.007 0.278Observations 1,000 1,000 741Village Dummies No Yes YesControls No No Yes

Notes: Dependent variables: “political accountability 1” = index of political accountability generated from meetings with nazim (mayor), meetings withcouncilor, proportion of females with identity card, proportion of males with identity card, proportion of females who voted, and proportion of males whovoted. This variable has been generated for adults only. Independent variables: (i) poor = household below the poverty line, (ii) literacy = proportion of lit-erate adults in household, (iii) dom = member of dominant kinship group, (iv) controls include: occupation, tenancy, migration status of household,landownership, headcount, and female landownership.






56


Variable Political Accountability 2 Political Accountability 2 Political Accountability 21 2 3

Intercept -0.213a 0.602a 0.480a

(0.057) (0.155) (0.223)

Poor -0.394a -0.248b -0.345a

(0.09) (0.098) (0.106)

Land 0.004b

(0.002)

Dom 0.423a

(0.095)

Dir -1.187a -1.112a

(0.251) (0.226)

Mardan -1.139a -0.849a

(0.222) (0.214)

Muzaffargarh -0.965a -0.750a

(0.188) (0.183)

Sanghar -1.110a -0.566a

(0.180) (0.192)

Thatta -1.207a -0.281b

(0.207) (0.238)

Toba Tek Singh -0.586a -0.558a

(0.172) (0.164)

Literacy (Male)c 0.628a

(0.172)

Literacy (Female)c 0.282c

(0.153)R-Squared 0.02 0.08 0.29Observations 998 998 739Village Dummies No Yes YesControls No No Yes

Notes: Dependent variables: “political accountability 2” = index of political accountability generated from meetings with nazim (mayor), meetings withcouncilor, proportion of females with identity card, proportion of males with identity card, proportion of females who voted, proportion of males whovoted, awareness of citizen community boards, and awareness of school councils. This variable has been generated for adults only. Independent variables:(i) poor = household below the poverty line, (ii) literacy = proportion of literate adults in household, (iii) dom = member of dominant kinship group,

(iv) controls include: occupation, tenancy, migration status of household, landownership, headcount, and female landownership.





Asian Development BankPakistan Resident Mission (PRM)OPF Building, Shahrah-e-Jamhuriyat,G-5/2, Islamabad.Tel: (92-51) 2825011-16Fax:(92-51) 2823324, 2274718email: [email protected]

ta4319 - collective for social science...

Documents