201301(2) arikun vid

8/19/2019 201301(2) arikun vid

1/44

Working Paper in Economics and BusinessVolume III No.1/2013

Varieties of Governance of Public Goods Delivery in Indonesia:The Case of Roads after Decentralization and Local Democratization

Ari KuncoroVid Adrison

Ifa Isfandiarni

January 2013

Institute of Economic and Social ResearchFaculty of Economics, University of Indonesia

8/19/2019 201301(2) arikun vid

2/44

Working Paper in Economics and BusinessChief Editor: Suahasil Nazara

Editors: Djoni Hartono, Beta Y. Gitaharie, Femmy Roeslan, Riatu M. QibthiyyahSetting: Rus’an Nasrudin

Copyright c2013, Department of EconomicsDepartment of Economics Building 2nd FloorDepokWest Java, Indonesia 16424Telp. 021-78886252Email: [email protected]: http://econ.fe.ui.ac.id/workingpage

http://econ.fe.ui.ac.id/workingpagehttp://econ.fe.ui.ac.id/workingpage

8/19/2019 201301(2) arikun vid

3/44

Contents

Contents 3

List of Tables 4

List of Figures 5

1 Introduction 1

2 Country and Sector Context 2

3 Conceptual Framework and Literature Review 3

4 Methodology 54.1 Field Interview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54.2 Theoretical Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64.3 Empirical Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

5 Data 10

6 Results 116.1 Field Interview in 50 sampled Districts . . . . . . . . . . . . . . . . . . . . . . . . . 116.2 Econometric Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

7 Referrences 23

8/19/2019 201301(2) arikun vid

4/44

List of Tables

1 Pre to Post Local Democratization (2003-2006): Paved Road Regressions . . . . . 152 Pre to Post Local Democratization (2003-2006): Passable Road Regressions . . . . 17A2.1 Data Sources for the Main Variables . . . . . . . . . . . . . . . . . . . . . . . . . . 27A2.2 List of Instruments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27A2.3 Classification Districts Based on Road Quality and Governance Compared to Prior

2005 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28A2.4 Descriptive Statistic for the Main Variables . . . . . . . . . . . . . . . . . . . . . . 29A2.5 Pre to Post Local Democratization (2003-2006): Passable Road and Bribes . . . . 30A2.6 Pre to Post Local Democratization (2003-2006): Paved Roads and Bribes . . . . . 31A2.7 Voters Preference: Probability of incumbent to win reelection . . . . . . . . . . . . 32A2.8 Before and After Decentralization (1999-2003): Paved Roads . . . . . . . . . . . . 33A2.9 Before and After Pre-Democratization (1999-2003): Passable Roads . . . . . . . . 34A2.10–Determinants of Bribes (2003-2006) . . . . . . . . . . . . . . . . . . . . . . . . . . 35A2.11–Tax Ratio: Before and After Decentralization (1999-2003) . . . . . . . . . . . . . 36

8/19/2019 201301(2) arikun vid

5/44

List of Figures

1 Figure 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 Figure 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 Figure 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74 Figure 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85 Figure 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86 Graph 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127 Graph 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138 Graph A4.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379 Graph A4.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

8/19/2019 201301(2) arikun vid

6/44

Varieties of Governance of Public Goods Delivery in Indonesia:

The Case of Roads after Decentralization and Local Democratization

Ari Kuncoroa, Vid Adrisonb, Ifa Isfandiarnic

a Institute of Economic and Social Research University of Indonesia bInstitute of Economic and Social Research University of Indonesia cInstitute of Economic and Social Research University of Indonesia

Abstract

For Indonesia the fall of Suharto in 1998 brought dramatic changes in the political landscape. Itsignified the beginning of transformation from an authoritarian regime towards a more democraticsociety where the distribution power is more devolved. In this respect Indonesia follow what iscalled a big-bang approach in the transition. First in the agenda of socio-economic reform isdecentralization which changes the relationship between the center and local governments. Next inthe countrys reform was democratization at the local government level. In this paper we examinehow decentralization and democratization affect governance at the local government level. Ourparticular interest is to assess the impact of decentralization and local democratization on thequality of road. We find that although the temptation to produce lower quality roads is high thecombination of democratization, free media, an ad-hoc anti-corruption agency and well educatedbureaucrats keep the quality of roads reasonably good while the corruption is held in check.

JEL Classifications:D73, H73, O17

Keywords: Governance, Decentralization, Democratization, Infrastructure Provision

1. Introduction

For Indonesia the fall of Suharto in 1998brought dramatic changes in the political land-scape. It signified the beginning of transfor-mation from an authoritarian regime towards

a more democratic society where the distribu-tion power is more devolved. In this respectIndonesia follow what is called a big-bang ap-proach in the transition. First in the agendaof socio-economic reform is decentralizationwhich changes the relationship between the

This research is apart of Global Research Project(GRP), generous funding from Global DevelopmentNetwork (GDN) is greatly acknowledged.

Email address: [email protected] (Ari Kuncoro)

center and local governments (Ranis [1994])1.Next in the countrys reform was democratiza-tion at the local government level. Starting in2004, a district head has been elected directlyby the people through a general election. Po-tentially this would give a district head more

freedom from the local assembly2.The broad definition of governance encom-

1This reform was also prompted by the demand fromregional governments, particularly those of resource-rich regions; to have more authorities in managing theirown affairs now resurfaces. Responding to this the leg-islative passed the decentralization law in 1999, whichwas subsequently enacted in 2001.

2Before 2001, a district head (regency or municipal-ity) was a part of the central government hierarchy (bu-reaucracy).But the regional decentralization law in 2001

1

8/19/2019 201301(2) arikun vid

7/44

Kuncoro et al./Varieties of Governance of Public Goods Delivery in Indonesia... 2

passes state and political institutions, theirchecks and balances, the rule of law and the

delivery system for the provision of publicservices. Effective governance includes vari-ous components such as accountability, trans-parency and effective bureaucracy, electoralcompetition and political checks and balances.The combination of these elements and themechanism implemented vary across countriesand across regions within a country. Thiswould make the case of studying inter-districtvarieties of governance in public service deliv-ery with a focus on roads in Indonesia is par-

ticularly interesting. Most recent initiatives of decentralization in Indonesia can be seen fromthe perspective of bringing policy makers androad providers closer to citizen, while local de-mocratization afterward is aimed to strengthenthe basic component of accountability.

2. Country and Sector Context

From the start the decentralization reform inIndonesia have been hampered by two major

constraints that could render the process inef-fective. First, after years of the politicizationby the Suharto regime, the civil performance isplagued by poor morale and low performance.At the local government level, the corruptionof the central government grants had been tol-erated as a way to compensate for low incomes.This problem was exacerbated by the lower ca-pacity of the local bureaucracy.

The implementation of democracy at the lo-cal government level is not without problem.

One urgent issue in the new democratic In-donesia is political party funding. Much of theparties money is thought to come from corruptgovernment officials. Corrupt money also oils

abolished the hierarchy b etween the center, provincesand districts. Instead, district heads will be elected bylocal (district) parliaments (legislative). While provin-cial governors will continue to represent the center ineach respective region, district head (regent or mayor)will no longer act as representative of the center andwill be accountable only to local legislative.

other parts of the political machinery. Thereare middlemen (calo) who use their money to

influence members of parliament making deci-sions about filling an important public office, abudget allocation, a fast-track draft of a law,and so on. So in practice, local democratiza-tion may not reduce corruption. At the locallevel, in this new political setting, corruptionalso takes in a different form. Any elected dis-trict heads would be potentially the prisonerof the respective local legislative will, since bylaw he or she will be subjected to censor andimpeachment by it. To get his or her pro-

gram passed by the local legislative, individ-ually or as a group, the local legislative mem-bers must be courted - a polite word for bribed.The bribes could range from money appropri-ated for traveling abroad to awarding lucrativepublic works to certain firms that have colludedwith local legislative members.

For political candidates the entry into politicrequires a huge sum of money which comes nei-ther from the state nor parties. To be able toqualify into an election of a public office for ex-

ample district head or bupati , a candidate mustbe endorsed by a political party or a coalitionof parties, and it does not come cheap. If a can-didate happens to be an incumbent then he orshe can use the budget appropriation or moreprecisely misappropriation to secure money forpolitical campaign. Like in the above case oneway is to sell concession on public work likeroad construction or road maintenance to thehighest bidder who in turn is required to paykickback. For a new entrant the situation is not

much different. He or she can ask special inter-ests like construction and business associationsto sponsor the candidacy in exchange for futureconcession in public work projects.

In this scheme, the quality of work would de-pend on how much left in the hand of providersafter paying some kickback as well as their in-terest to obtain profits. The amount of kick-back as percentage of the project value in turnwill depend on the candidate preference for

8/19/2019 201301(2) arikun vid

8/44


personal versus public gains weighed againstthe prospect of punishment from voters in the

next election. In the Indonesian context, lack-ing strong consumer groups voting is perhapsthe only possible tool for citizen to punishelected politicians for bad public service de-livery (Henderson and Kuncoro [2011]). Thispunishment mechanism requires the presenceof reasonably good information provided byfree news media.

The purpose of this study is to examinehow decentralization and democratization af-fect governance at the local government level.

Our particular interest is to assess the impactof decentralization and local democratizationon the quality of road. For this, our approachis to combine a field work of interviewing realplayers in the real world with economic andeconometric modeling. The field work is de-signed as a fact findings mission. To gain a bet-ter insight of the problems, the primary datagathered in the field work are used for the con-ceptual modeling of the relationship betweendecentralization, local democratization and lo-

cal governance. The study found that despitethe initial worrisome of the deterioration of road quality, the combination of free media,an ad-hoc anti-corruption agency and well ed-ucated and better trained bureaucrats keep thequality of roads reasonably good in the atmo-sphere of decentralization and democratizationwhile the corruption is held in check.

3. Conceptual Framework and Litera-ture Review

The situation mentioned above is modeled asa principal-agent relationship where the votersare principal who demand good quality of in-frastructure specifically roads with newly ac-quired power to vote out incompetent and/orcorrupt district heads. The district heads andtheir associated bureaucracies are agents whoare much freer now from the center domina-tion after decentralization, whose task is to de-liver infrastructure with good quality but at

the same time under the current law they can-not build or maintain roads by themselves and

also under democratization they need moneyto finance political activities including reelec-tion campaign. In the middle are constructionfirms which bid to get contracts to build roadsatisfying a certain standard of quality. Thesefirms may subject to extortion from districtheads bureaucracy. Altogether, this approachis therefore akin to the long-route accountabil-ity.

From the theoretical standpoint, how de-centralization affects governance is ambiguous.

The first view suggests that decentralizationleads to greater fragmentation of governmentdecision making power. The breakdown of co-ordination among bureaucrats may lead to ex-cess rent extraction and poor public service de-livery (Shleifer and Vishny [1993]). In line withthis argument, in the case of corruption, Treis-man (2000) found that federalist country havehigher rate of corruption due to the problemof overgrazing among different levels of gov-ernment entities over the same targets. Fur-

thermore, Prudhome (1995) and Tanzi (1996)argued that the low capacity of local bureau-crats in delivering public goods and servicesmay prevent the realization of benefits fromdecentralization. The opposite view on theother hand suggests that decentralization is ex-pected to have moderating impact on corrup-tion and would bring better quality of pub-lic service, by increasing the role of the lo-cal government and accentuating the forcesof inter-jurisdictional and political competition

(Tiebout [1956], Huther and Shah [1998]), Fis-man and Gatti [2002]).

The theoretical work of Persson andTabellini (2000) provides an insight into howpolitical accountability may affect bureaucratsperformances. With democratization, a gov-ernment agent is responsible for a certain taskthat is specific to the respective jurisdiction,thus, logically a more direct accountabilityshould improve the agents performance, and

8/19/2019 201301(2) arikun vid

9/44


hence a better public service delivery is ex-pected. The opposite view of this matter, on

the other hand, argues that decentralizationwould also create multiple tiers of governments.Potentially it could have the opposing effect toweaken accountability, since voters would havemore troubles differentiating, those who weresuccessful, from those who were not (Fismanand Gatti [2002]).

In the case of rent-seeking activities althoughthe regime shift to local democratization is ex-pected to reduce corruption by increasing therole of the local government and strengthening

the forces of inter-jurisdictional and politicalcompetition, it is also recognized that democ-racy may have limited impact on corruption(Rasmusen and Ramseyer [1994]) and the tran-sition process is difficult given the role of his-tory, culture and expectations (Tirole [1996],and Andvig and Moene [1990]). One particu-lar worrisome situation is the increase of polit-ical transaction cost for lobbying which divertbudget money from its productive use (North[1990]).

The existing literature on the impact of decentralization/democratization in the devel-oping countries focuses mostly on its im-pact on infrastructure or capital spendingpattern (Faguet, 2004; Estache and Sinha,1995; Vergne 2009), while few other studiesfocus on the impact of decentralization onthe efficiency of public infrastructure provi-sion (Chavis, 2010), availability and quality of public infrastructure (Chowdhury et al, 2009).Burgess et. al (2012) examined the role of localpolitical cycle in Indonesia in affecting defor-estation. Local politicians who have to financepolitical activities may not enforce forest policyas they are benefiting from illegal logging.

Faguet (2004) investigates the effect of de-centralization on the responsiveness of publicinvestment in Bolivia. There are two mainquestions of his research; (1) whether or notthe pattern of public sector investments havechanged since the implementation of decentral-

ization in 1994, and (2) if so, do indicators of need determine the change? In order to answer

the research questions, public investments in 10sectors were analyzed for 1987-1996 period. Hefound that after decentralization, public invest-ments increased in five sectors, namely educa-tion, water and sanitation, water management,agriculture and urban. And these changes arestrongly associated with local needs. However,investment pattern did not change for indus-try and tourism, energy and communicationand transport. Another important finding of his research is that the poorest localities put

higher priorities on human capital investmentand basic social service (water) rather thanother sectors, such as agriculture and trans-portation/communication.

A study by Estache and Sinha (1995) ana-lyzes the infrastructure spending in both de-veloped and developing countries, using paneldata set from 1970 to 1992. The study departsfrom the fact that many countries have decen-tralized the provision of public infrastructure

to sub-national governments. The effect on in-frastructure spending is ambiguous. It will in-crease if the sub-national government put in-frastructure in a higher priority than the cen-tral government, or if it is less efficient at de-livering services, or if they give up the bene-fits economic scale to gain more autonomy. Onthe other hand, the spending will decrease if the local government put infrastructure in alower priority, or if they are more cost effective.Their study finds that degree of decentraliza-

tion measured by the ratio of sub national ex-penditure to total expenditure, and the ratio of own source revenue to total revenue does af-fect the level of infrastructure spending. Theyfound that the elasticity of total expenditurewith respect to both measures of decentraliza-tion is approximately 0.3 for developing coun-tries, and 0.2 for industrialized country. How-ever, the elasticity of sub national spendingon infrastructure is much higher in developingcountries than industrialized countries (1.1-1.4

8/19/2019 201301(2) arikun vid

10/44


as opposed to 0.7-0.9).

Vergne (2009) tries to relate the composi-

tion of expenditure with election in 42 devel-oping countries. His study departs from Ro-goff (1990) theoretical model which predictsthat incumbents have the incentives to spendmore towards more visible (current) expendi-ture rather, away from capital expenditure. Inhis empirical strategy, expenditure is classifiedinto three categories; (1) current expenditure,(2) capital expenditure, and (3) infrastructureexpenditure. The study finds that election doeschange the composition of expenditure, where

current expenditure increases, capital expendi-ture and infrastructure expenditure decreases.In one specification, this study also takes intoaccount the decentralization, measured by theshare of sub national expenditure in total gov-ernment expenditure. Although the result in-dicates that the distortion is smaller with de-centralization, the effect is statistically insignif-icant.

A recent study conducted by Chavis (2010)analyzes the effect of decentralization on theefficiency of public good provision at the vil-lage level in Indonesia. The study investigatesthe impact of local competition for block grantunder Kecamatan (Subdistrict) DevelopmentProgram (KDP). The efficiency public goodprovision is measured using two approaches;cost of road construction per square meter, andthe amount of missing materials. The find-ing suggests that competition and local partic-ipation increases efficiency of public good pro-vision. A 10% increase in competition leadsto a 1.8% reduction in per unit cost of road.The study also finds that efficiency is increasedwhere local participation is high.

Another study on Indonesian case is con-ducted by Chowdhury et al (2009), which ex-amines the impact of decentralization on thequality of three types of public infrastructure;village road, availability of primary and juniorhigh school, and availability of basic health fa-cility. They use three waves of village poten-

tial survey (PODES) of 1996, 2000, and 2006to construct changes in quality pre and post

decentralization. The change in public infras-tructure quality between 1996 and 2000 refersto pre-decentralization era, while the changein public infrastructure quality between 2000and 2006 reflects after decentralization is in-troduced. Estimation is performed using or-dered probit. Their finding shows a significantdependency of road, public schools and health-care facilities on local income and resources,in both the pre and post decentralization pe-riods. Their results imply that poor villages

are less likely to see improvement in the qual-ity of public infrastructure or provision of newpublic infrastructure. Comparing the parame-ter estimates between pre- and post- decentral-ization, they conclude that the dependency of local public infrastructure on the local incomeand resources declined substantially3.

Our study differs from Chowdhury et al(2009) and Chavis (2010) in the sense that weconsider the new political reality faced by dis-trict government after decentralization and de-

mocratization. While their studies are purelyempirical, we conceptually model the decisionthat must be made by a district head with re-spect to road construction taking into accountall political as well as economic constraints.Our study takes into account the election cy-cle in a district which influences district headdecision on roads provision. In that sense ourstudy is parallel to Burgess et.al (2012) whoexamine the impact of local political cycle ondeforestation in Indonesia.

4. Methodology

4.1. Field Interview

The primary interest to conduct a field in-terview is to examine the anatomy of the roadconstruction and maintenance business at the

3Note that they do not perform statistical tests onthe difference of the parameter between pre- and post-decentralization.

8/19/2019 201301(2) arikun vid

11/44


district level. The main goal is to collect infor-mation on construction and maintenance bud-

gets per kilometer and their execution espe-cially time for project negotiation, constructiontime and the tolerance for construction timeextension, actual road construction and main-tenance costs.

The structured questionnaire was fielded torespondents in the road construction business.If possible the questionnaire was seeking infor-mation as far back as 1999 to allow for cap-turing the impact of decentralization in 2001and local democratization in 2004. The respon-

dents were a representative construction firmsor its association, a representative of trans-portation firms and local public work officesfrom 50 districts in Java. For the constructionbusiness we found in the field that the associ-ation basically serves as a cartel. They decidethe order of queuing in receiving local govern-ment contract among a small circle of firms.In turn they may sub-contract smaller portionof work to firms outside the circle. It is alsoplanned to ask question about the percentage

of kickback from the association. Since somequestions like bribe or kickback may involvesensitive matter, interviews were conducted asa conversation between friends.

The field survey was completed in 120 days(around 4 months). We employed our exten-sive network with local universities throughoutJava. They have been involved in LPEM sur-vey since 2001. LPEM led the field survey, as-sisted by local university, thus, the team con-sisted of people from both LPEM and localuniversities. On average, the interview with re-spondents took 3 days and some were arrangedsimultaneously in survey areas due to time con-straints. Also, the local team was responsibleto collect secondary data at district level.

The choice of geographical coverage - focus-ing on the main island of Java, is based onthree considerations. First, location decisionsfor the Indonesian businesses are focused onJava which holds about 80% of economic ac-

tivities in Indonesia. Second, the distance be-tween localities in Java is not too far, which

permits to examine whether districts learningfrom each other with respect to governance(Manski [1993]). Third, all districts in Javacan be accessible by land transportation.

4.2. Theoretical Model

The following conceptual model is based onstylized facts we gathered in the fieldwork.The results of field interview will be pre-sented later. The impact of decentralization-democratization is ambiguous. One undis-

puted commonality obtained from the fieldwork is that after decentralization a districtholds monopoly of the issuance of contractsin public work projects. There may be sev-eral viable construction firms in a district butnot many. In order to be eligible for biddingprocess they must join association or cartel ap-proved by a district government which requiresthem to pay entrant or membership fee or sim-ply bribes. The winner of the bidding processis usually the most dominant firm in the cartel

while others will serve as sub-contractors4. If the strength of each firm in the cartel is aboutthe same then the winner of the project will berotated among the most powerful firms whilelosers of both powerful and less powerful firmsmay act as sub-contractors.

Based on this, we model the above situationas interaction between two agents. Construc-tion firms are the bribe suppliers or kickbackswhich need government contracts to survive.Meanwhile, district heads after democratiza-

tion have to show to the people that they canprovide good quality roads otherwise they willbe reelected. So district heads to be crediblemust possess some sort of monitoring system toensure the provision of reasonably good roads.

4In many instances the so-called dominant firm isa pure-rent seeker without production capacity. Thefirm only sells concessions to tru construction compabtfor fees. The fees are then used to finance politicalactivities.

8/19/2019 201301(2) arikun vid

12/44


Figure 1: Figure 1

At the same time they are also extract moneyfrom budget in the form of inflated budgetitems, as well as soliciting bribes from firms inorder to finance their political activities. Thedetailed theoretical model is presented in theannex A1.

Figure 1 through Figure 3 illustrates theworking of the model. In the case 1 dis-trict heads apparatus control the road qualitystrictly or z=1. This case is depicted in Figure1. The transition from the indirect election of a district head to a direct election by the peo-ple has resulted in a more complex democraticprocess at the local which would require morefunding requirement for political activities oran increase in m.

From point A, democratization has resultedin the increase of money needed for campaign,while at the same time it results in the increaseof perceived minimum quality of roads to winthe election from h* to h as well as the need toreduce budget corruption from θ∗ to θ (pointB). The overall result is to lower the utilityfrom bureaucrats to bureaucrats . The point B isthe first best situation where there is no bribe.

The second case in Figure 2 is similar butthe quality control is relaxed allowing the firm

Figure 2: Figure 2

Figure 3: Figure 3

to cheat on road quality (0¡z¡1). In order torecoup some profit losses, the firm now movessouthwestward from point C0 to C. The qual-ity of roads h will be lower than h but stillhigher than the pre-democratization level h*.The firm profit is now higher than π’.

In Figure 3 quality control is completely re-laxed or z=0. The firm will move from C 0 toC passing the point A. The quality of roads islower than the pre-democratization era. The

firm profit is the highest in this case. In thissituation there will be an inverse relationshipbetween the road quality h and bribe b.

To summarize the relationship implied byFigure 1 through 3, the first order conditionfrom the district head optimization which re-sults in a reaction function for him is graphedin the h and θ space as bureaucrats given othertings constant, while firm is derived from thefirm optimization. In the vertical axis is thequality of road h. In the horizontal axis is the

8/19/2019 201301(2) arikun vid

13/44


Figure 4: Figure 4

Figure 5: Figure 5

fraction of the full value of project defrauded

by the local bureaucrats or θ. Figure 4 is thecounterpart of Figure 1 where a new form of democratization has full impact on road qual-ity. The increase of bribes has no impact onroad quality since the quality is strictly moni-tored or z=1. In Figure 5 the quality controlis not as strict as before so the constructionfirm is able to compensate some of profit lossdue to higher bribes or 0¡z¡1. Since bribes bis determined recursively after a district headchooses an optimal level of budget corruption

θ what the firm can do facing increasing bribesis o lower road quality from h to h. The dis-tance between h and h represents the amountof cheating. Comparing to the starting pointA, the overall effect is however still positive. Inthe final case z=0 which means there is no qual-ity control so the increase in bribes will promptthe firm to lower road quality enormously suchthat the cheating effect totally dominate thepositive effect of democratization. The roadquality is worse than before democratization.

4.3. Empirical Strategy

Since we have well-behaved functions the

first-order condition from the the district headand the firm optimization can be solved for thesemi reduced form of h and θ.

h = h(θ, A−1,m,q ,H ) (1)

θ = θ(h,A−1,m,q ,H ) (2)

From the firm optimization the road qualityproduced by firm is given by

h = h(b(θ), θ , ci.H,z) (3)

While given θ, bribes that must be paid byfirms is

b = b[(θ(h,A−1m,q,H )] (4)

Substituting (4) into (3) we get the semi-reduced form

h = h(θ, A−1m,q,c,.H,z) (5)

The past accumulated asset A-1 is unobserved

so it is assumed that which it influenced byGDP per capita which also serve as the basefrom which the corrupt officials can extractrents. The full value of projects H is assumedto depend on the fiscal situation in the respec-tive district which is represented by local taxratio to district GDP.

The money for political campaign m is un-observable what we know is that since 2004in which the direct election by the people waslaunched the money needed by a candidate for

political campaign have increased significantlyso we will use a time dummy separating yearsafter 2004 from before it. Beside this to con-trol for political factors we include local po-litical competition, the extent of political par-ties presence (the number of sub-district po-litical parties headquarters).The probability of getting caught by the anti corruption agency(KPK) is measured by the proportion of house-holds exposed to the media in this case havingTV sets.

8/19/2019 201301(2) arikun vid

14/44


The variable z represents probability thata cheating firm can be caught by the district

heads bureaucracy. To capture this we use thepercentage of district apparatus with collegedegree. We assume that cheating in road qual-ity is difficult to detect since it involves manytechnical matters. So a district staffed withbetter educated bureaucrats may be able to de-tect cheating better than others5.

We cannot observe directly the fraction of defrauded money out of the total budget θ.However we know the number of cases of bothexecutive and legislative member under inves-

tigation by local attorney general offices due tobudget corruption. So instead θ will be repre-sented by this variable. One caveat is that thisvariable may reflect more zeal of the anti cor-ruption office instead of more corruption so onehas to be careful when interpreting the results.

The cost of constructing roads will vary fromone district to another depending on topo-graphic, geographic and economic conditions.To represent the inter-district variability of construction costs we use the district GDP

deflator. In addition we also control for dis-trict physical characteristics which capture ge-ographic, topographic and climate condition of a district like flat or hilly area, coastal or not,earthquake area, flooding areas which eventu-ally determines the true cost of road mainte-nance.

The empirical counterpart of (5) is given by

h = α0 + α1θ + α2D2004 + (6)

α3θ.D2004 + ∆.G+

ΓY + Φ.P + u

In (6) θ is defrauded money from local bud-get, D2004 is the time dummy, G is a vectorof district physical characteristics containing

5According the procurement regulation it is impos-sible for bureaucrats to do their own design. Designingis the responsibility of construction firms. Certainlybureaucrats can set up local standard of quality basedon the guidance (minimum standard) from the nationalpublic work ministry.

dummy variables whether a district is in flat orhilly area, coastal or not, which determines the

true cost of road maintenance. The variableθ will be interacted with the time dummy toexamine the effect of the regime change in theelection system on h. In the meantime vectorY captures socio-economic condition of districtlike GDP per capita, ratio of local tax to GDPto represent the district fiscal capacity and lo-cal price or cost of living. The vector P cap-tures all political variables like local politicalcompetition measured by the Herfindahl indexof the parties share of local votes, the propor-

tion of households with TV and the share of households subscribing newspapers represent-ing the probability being caught by the anticorruption agency.

By construction, θ represented by the num-ber of cases of suspected budget corruption, isendogenous so one has to implement instru-mental variabel method (IV). For this a setof geographic variables mentioned above likecoastal or not, hilly or not, flooding area ornot and so on can potentially be used as instru-

ments. In addition a set of socio-demographicvariables like religious affiliation represented bythe number of Islamic schools, fraction of gov-ernment employment in the local populationand the fraction of population with high schooleducation or more will be used as instruments6.

According to the theoretical model, the co-efficient α1 is expected to be negative. A newform of democratization will make a districthead to pay more attention to the quality of public goods as this is one the prerequisite to

be reelected. In the process in the face of in-creasing need for political campaign, he hasto balance this with the desire to accumulatewealth. The first best situation is when thereis no bribe. In this case both α1 and α3 will bestrongly negative while α2 is significantly pos-itive and large. If both coefficients α1 and α3are positive or negative but insignificant thenthe differential effect of the budget corruption

6The exaustive list can be seen in Table 2.

8/19/2019 201301(2) arikun vid

15/44


before and after democratization in reducingroad quality is at best very little7.

The coefficient of GDP per capita is expectedto be positive indicating that richer districtswould have more potential to afford better roadquality. The same thing also applies to districttax ratio which captures the fiscal situation.The higher is the tax ratio the more likely theycan afford to build good quality roads.

Probability to win the relection Another interesting issue to examine is theelection itself. The probability of an incum-bent to win election (λ) is in the form of Probitequation given by

λ = λ(h,θ,m,G,Y,P ) (7)

The sign of the coefficient of h in (18) cannotbe determined in apriori. One one hand thedistrict head needs to extract money rent tofinance his campaign which lower his reputa-tion as well as may result in a lower qualityof roads. On the other hand he needs to showhis competence to the voters by providing good

quality of roads people But this would requirehim to extract less from the budget which un-der the absence of private contribution fromelsewhere may result in less money for relec-tion campaign. if he does too much then hemay not be releccted since there will be publicdisgruntlement on his performance.

For the budget corruption θ we use the num-ber of budget corruption cases involving boththe excutive and legislative branches in a dis-trict. If θ and h have the opposite signs then

there is a trade off between those two districthead must lower budget corruption to buildgood quality roads. If θ and h are both positivethen the district head bureaucracy may give

7It is also possible that α1 and α3 are strongly pos-itive. This case is probable but unlikely since after thedemocratization the bureaucrats would increase theirbudget corruption θ and use some of the money to sub-sidize the construction firm to produce road with betterquality than before the regime change.

consession (or subsidy) from corupted budgetto the firm in the form of lower bribes in order

to encourage a firm to build good quality road.Like before we cannot observe the political

funding for this as in the previous exercise weuse the year 2009 (the most recent election)dummy to signify the difference between beforeand after the introduction of the new lectionlaw allowing the district head and the membersof local legislative to be ellected directly by di-rect people vote. The positive sign suggeststhat incumbents are more likely to be electedafter 2004 than before.

The empirical version of (7) is given by

λ = β 0 + β 1h+ β 2θ + (8)

β 3D2004 + ∆G+ ΓY +

PhiP + u

5. Data

To estimate the model empirically we usea data set from the Statistical Bureau (BPS)

called PODES or village potential. We arelooking specifically for information related toroad quality. There are two pieces of infor-mation regarding the road quality that can beaggregated from the village to the district levelfrom PODES: fraction of passable roads andfraction of paved roads at the district level. Weargue that this information is credible since itis created from the responses of village headconcerning the roads passing through the vil-lage.

Fraction of paved roads in a district may alsorepresent quantity rather than quality since itcontains a political element. Roads have beenused to demonstrate competency by local po-litical candidates since the Suharto era. Sim-ply because it is free and the benefits can befelt by all, unlike services like hospital, schoolwhich are not free. Before the election to at-tract voters a district head may intensify theroad paving but the quality can only be ob-served later on. If it turns out that the roads

8/19/2019 201301(2) arikun vid

16/44


are impassable within 6 months later then con-struction firms may cheat on quality assuming

that the traffic volume is unchanged. This suitsour model that allows firms to cheat on qual-ity. So fraction of paved and passable roadsin a district will be our two main outcomes torepresent quality.

There are three election cycles in the periodof analysis in 1999, 2004 and 2009. In the 1999election, a district head was chosen by a locallegislative, while starting in 2004 he or she willbe chosen directly by the people. Our researchaim is to assess how these two different schemes

of democracy affect the quality of public goodsin this case roads. We assume 3 to 4 years lagfor all covariates allowing some adjustment totake place (see Table A2.1 in Appendix). Forinstruments we use variables farther back intime (see Table A2.2 in Appendix).

The PODES data sets in our possession arefor 1986, 1990, 1996, 2000, 2003 and 2006. Theroad quality in 2003 and 2006 will be our mainoutcomes overtime. We have measurement atthe sub-district level so for those two election

cycles altogether there are 2052 observation.Some of other covariates like the fraction of hilly and flooding areas are also measured atthe sub-district level. Other covariates like thenumber of budget corruption cases, the fractionof bureaucrat with college degrees, the numberof households with TV, the fraction of popula-tion exposed to newspaper, the tax ratio, GDPper capita, GDP deflator are calculated at thedistrict level. Table A2.1 summarizes our datasources.

6. Results

6.1. Field Interview in 50 sampled Districts

Governance/Institution Most of districts (23 out of 50) show a reduc-tion in bribe in 2011 compared to the valuein 2005. However, the role of local parlia-ments is increasing not only in budgeting pro-cess but also in planning and disbursement pri-ority. The latter is supposed to be the local

government authority. As consequence, manyroads built or preserved are not based on the

priority but on the constituency areas of localparliament members. This is in line with thegeneral trend in manufacturing sector (Graph1).

In general the strategy to join the projectawarding circle is for several contractors to es-tablish a cartel to win the bidding process. Inthis scheme each cartel member will have topay some amount to local government. As acartel, construction firms could negotiate theamount of bribe with the local government of-

ficials. This mechanism also allows other mem-bers to carry out the project as sub-contractorseven though only one firm is selected as a win-ner.

For districts experiencing an increase (19 outof 50) in bribes the winner of the bidding pro-cess is usually preselected. Various practicesto extort money such as making payment tothe local government as an information fee orbooking fees, or to the local party as a mem-bership fee, are at play. By paying the in-formation fee, the construction firms will getthe first information earlier so they will haveenough time to prepare good bidding docu-ments then be selected as a winner. Similarto the information fee, booking fee is paid bythe construction firms to obtain a guaranteefor upcoming projects. This is usually paid inthe early fiscal year. Unlike the two previousfees, the construction firms who pay the mem-bership fee will be included in a bidding ar-rangement. This mechanism is quite sophisti-cated as it needs a good cooperation among lo-cal parties, business associations and local gov-ernment officials. The transaction will then ap-pear as a normal procurement process althoughthe winner has actually been pre-selected.

Regarding to the practice to bribe local par-liament members, each contractor usually payssome money as a percentage i.e. 5% to 10%of total contract to the campaign team of in-cumbent in the local election. Besides making

8/19/2019 201301(2) arikun vid

17/44


Figure 6: Graph 1

Sources: Annual Manufacturing Surveys

payment to their political campaign teams, thecontractors could also pay directly to the localparliament members if the project areas hap-pen to locate where their voters live.

In some districts where the clan of districthead is dominant in both executive and leg-islative agencies, the bidder of constructionprojects is usually the one who has special con-nection with the clan of incumbent and or withthe local party. Interest groups such as busi-

ness associations are also part of their clique.In this typical district, the bribe tends to de-crease, the quality of road remains unchangedor improved, but all construction projects arecontrolled by their clique.

Most of construction firms prefer to pre-finance their project due to the complexityof administration process and higher unoffi-cial payment. The more frequent the con-tractor requests the payment, the more doc-uments should be prepared and the higher un-

official payments are paid to the local govern-ment. They usually borrow from local banks or

joint financing with other firms for about 3-6months, depending on the length of the project.

The impact of new regulation on procure-ment process implemented in January 2011 isunclear some respondents reveal positive im-pacts, others not. The frequent complaintsfrom construction firms on new procurementsystem are (i) expensive, (ii) more complicatedprocedure, and (iii) unreliable internet connec-

tion. Meanwhile, the positive impacts on newprocurement regulation are (i) less documents

required, (ii) less face to face meeting, and (iii)open to any firms. Despite those positive im-pacts, this new e-procurement system still canbe hacked by the internal procurement official.For instance, announcing an invitation to bidwhere the winner has actually already been de-cided in advance. In this scheme district offi-cials try to split up the projects into the smallerones so the grantee can be chosen without hav-ing to go to the official bidding process sinceby law small projects can be carried out based

on direct appointment by local government.During the project implementation, collu-

sion between construction firms and road su-pervisor from both local government and con-sultant still exists in some districts, particu-larly at the project hand over stage. For exam-ple, monthly certificate and assessment reportcould be manipulated by the consultant super-visors and or by the officials in order to speedup the disbursement process. Physical inspec-tion is not carried out randomly; instead, this

is carried out only at a certain part of road toavoid detection of cheating. This would makeroad durability shorter than it should be.Road Quality According to the respondents, in general, mostof districts have better road quality in 2009compared to the condition in 2005 (Graph 2).8

It is interesting to observe that the percentageof good roads actually higher in the era of di-rect election after 2004 compared to those in2001. This is in line with the travel time which

is faster in some districts due to new or goodroad quality, but slower in most of city areasdue to the congestion. All municipalities orcities in the sample have similar road condi-tion due to their small areas and flat contour.In addition, cities (municipalities) roads havebeen built since hundred years ago and trafficpassing through cities is well regulated due totheir relatively small areas of coverage.

8We do not have road data in 2010 and 2011

8/19/2019 201301(2) arikun vid

18/44


Figure 7: Graph 2

Source: Field Interview in 50 Districts

Unlike cities, districts (kabupaten) charac-teristics vary from one to another which affectstheir road quality. For instance, the existenceof industry and natural resources such as min-ing, forestry, and plantation requires inter-citytransportation which pass district roads Also,natural condition i.e. topography (mountainor flat), heavy rainfall, river mouth, etc. couldaffect the road quality in general.

One new development in this matter is that

the new law on traffic has been stipulated in2010 and effectively implemented in January2011. This law says that only the police couldinspect and fined the vehicles violating the law.Unfortunately, not all police are trained orequipped with adequate tools to do some mea-surements and assessment. So many heavy ve-hicles with high tonnage continue to pass dis-trict roads which are not designed to withstandheavy loads. Consequently roads in districtsdeteriorate quicker than their supposed stan-

dard specification.The percentage of heavily damaged roads

is slightly higher in 2009 compared to 2001(Graph 2). For this bad road category, thetravel time increases by two folds or triple,while the vehicle maintenance cost also risesby almost a double, and the time to replaceslow moving part is getting shorter. This typ-ical bad road will last only for 3 to 4 months.Road durability with a proper maintenance pe-riod is about one to two years for the hot-mix

type, and less than six month for the asphalttype. In general local governments have allo-

cated road maintenance in their annual bud-gets.9

Relationship between Road Quality and Gover-nance In this study, road quality is measured based onthe interview with local transportation associa-tions regarding the travel time, perception overthe road quality, and time to replace the slowmoving parts. Meanwhile, the governance is as-sessed by asking construction firms question onunofficial payment (bribe) paid to a respective

local government and or to local parliament,time spent to receive the first payment; andperception over the procedure to obtain con-tract, procurement and payment. All questionsare asked by comparing the current conditionto the prior 2005 conditions. The results aresummarized in Table A2.3.

From Table A2.3, almost half of district inthe sample have better road quality comparedto the 2005 condition. About one third of to-tal districts with better quality road are facing

an increase in bribe and very few are dealingwith unchanged bribe. It seems that the localgovernment together with the local parliament,in one hand, increase local budgets to financeroad maintenance to provide higher quality of road but on the other hand extorting higherbribes from the construction firms. Detaileddiscussion can be seen in Box2 Annex 6.

In most districts with better or unchangedroad quality and a decrease in bribe is oftenperceived by the media as a success story of the

anti corruption agency (KPK) in combatingcorruption (one example is Kabupaten Sidoarjoin East Java in Box 1 Annex 6). In realityit could be very different story. The respectiveconstruction firms do not reduce the qualityof road because there is a lower bribe paid to

9However, in some areas, some concrete roads arebuilt and preserved by the private sector budget whichis usually manufacturing companies located in indus-trial areas.

8/19/2019 201301(2) arikun vid

19/44


the local government or to the local parliament.Moreover, those typical districts may have high

own sources of revenue from local taxes andlevies so the government officials or parliamentmembers have no temptation to request higherbribe as those have already received higher in-come due to higher own source revenue. Theimplementation new procurement system (e-procurement) as well as the domination clan orcliques of district head and construction firmscould also contribute to this good road quality-low bribe situation.

For districts with bad road quality and a

decreasing or unchanged bribe, constructionfirms appear to pay lower or same amount of bribe but due to complicated local regulationto prevent corruption and cheating very fewfirms are willing to take risk to carry out theconstruction projects so the district roads mayfall into disrepair (a good example is Kabu-paten Temanggung in Box 3 Annex 5).

The worst situation occurred when districtshave bad road quality and while the bribe isincreasing. It seems that demand for bribe

has forced firms to sacrifice road quality justto maintain some decent profits (one exampleis Kabupaten Cirebon in West Java in Box 4Annex5).

The results of the above field interviewsuggest that the impact of decentralization-democratization on road quality is at best am-biguous for some it leads to the better while forothers it goes the other way. This observationis incorporated into the reformulated theoreti-cal model.

6.2. Econometric Results

Local Democratization Graph A4.1 shows the general picture of ourtwo main outcomes paved and passable roads.Overall it is suggested that the quality in termsof paved roads increases significantly after theintroduction of the direct election in 2004 Thisis accompanied by the decline of variabilityof the percentage of paved roads across dis-tricts (Graph A4.2). All districts tried to in-

crease the percentage of paved roads with thosethat lagged behind attempted to catch up with

those already ahead. There is a slight decreaseof the average of passable roads across sub-district, but the most obvious case is the jumpof standard of deviation from passable roads.So the quality in term of whether the roadsare passable for year around show more vari-ations in 2006 with some districts continue tolag behind. This graph provides useful infor-mation on the whole situation but it is hard todraw any meaningful conclusion about the firmcheating. This has to wait for the econometric

analysis.Table A2.4 in the annex gives the descrip-

tive statistic of all variables used as the depen-dent variables and the main covariates. Thesevariables vary significantly from one district toanother. For example for the tax ratio it variesenormously from around 0.5 percent to 70 per-cent of district GDP. Some districts have lowexposure to TV program while others almostall households have TV sets. There are also sig-nificant topographic and geographic variations

there are flat areas with no hills and while oth-ers are very mountainous. Some areas virtuallyhave no paved roads, while it is possible that inthe rainy season only about 20 percent of thetotal areas are passable to 4 wheels vehicle.

The estimation results for two outcomes forroads fraction of paved and passable are pre-sented in Table 1 and Table 2. We use twoestimation methods. First is a plain OLS withrobust standard error labeled as model I (col-

umn 1 and 3). To account for the endogeneityof the number of corruption cases θ we employthe IV-GMM (instrumental variable generalmethod of moment) method. The list of poten-tial instruments can be seen in Table A2.2. Inthe paved road regressions (Table 1), the im-pact of changing the election system from anindirect one through local legislative in electinga district head to a direct system election is toincrease the fraction of paved road. The coef-ficient of time dummy is positive and strongly

8/19/2019 201301(2) arikun vid

20/44


Table 1: Pre to Post Local Democratization (2003-2006): Paved Road Regressions

Fraction Paved Road OLS IV-GMM

I II I II

Number of budget Corruption cases -0.129 -0.002 -0.057 0.158[-1.90]* [-0.24] [-0.28] [2.08]**

Dummy 2006 0.166 0.195 0.149 0.626[8.26]*** [7.51]*** [7.44]*** [2.84]**

Dummy 2006 X # of budget Corruption cases -0.021 -0.357[-1.65]* [-2.15]**

Fraction of households with TV 0.074 0.076 -0.018 0.014[1.76]* [1.80]* [-0.42] [0.27]

Log GDP per capita 0.053 0.051 0.085 0.06[3.86]*** [3.73]*** [6.56]*** [3.12]***

Tax ratio 0.561 0.525 0.787 0.196[2.24]** [2.09]** [3.14]** [0.45]

District price indicator (GDP deflator) -0.244 -0.024 -0.031 -0.02[-2.75]** [-2.75]** [-2.89]** [-1.48]

Fraction of hilly areas in a district -0.198 -0.198 -0.216 -0.212[-7.59]*** [-7.60]*** [-7.99]*** [-6.91]***

Fraction of areas prone to flooding -0.292 -0.285 -0.307 -0.199

[-5.66]*** [-5.57]*** [-5.91]*** [-2.80]**Herfindahl Index of political parties -0.024 -0.031 -0.027 -0.161

[-0.43] [-0.52] [-0.44] [-1.77]*Fraction of population subscribing newspaper 0.316 0.323 0.314 0.426

[7.20]*** [7.24]*** [6.97]*** [5.47]***Number of Observation 2052 2052 2052 2052F-value 52.74** 48.52**Wald-chisq 620.05** 424.52**P-value 0.88 0.09

Notes:1. Figures in parentheses are t-ratio2. *** Significant at 1% level

3.** Significant at 5% level4. * Significant at 10% level5. Instruments are number of voters in the previous election, percentage of government employees in 1990s,average distance to district capital from a village, percentage of population with high school education in1990s, bribe in manufacturing sector 3 years back and fraction of population subscribing newspaper.

8/19/2019 201301(2) arikun vid

21/44


significant in all model specifications and esti-mation methods.

We manage to get a negative relationship be-tween the fraction of areas with paved roadsand budget corruption. The less is the corrup-tion the higher is the road quality. Howeverwhat is more important is the differential ef-fects on paved roads after the introduction of direct election. For this purpose we introducean interactive term between time dummy andthe number of corruption cases (model II) inboth OLS and IV-GMM estimation methods.The theoretical framework predicts that this

new form of democracy would increase the roadquality due to lower budget corruption. The in-teractive term is negative and significant giv-ing support to this prediction. The coefficientof interactive term is more negative and sta-tistically stronger if we take into account theendogeneity of budget corruption.

The highest is the fraction of householdswith TV the better is the road quality. Sothe public exposure will force the district tobuild better roads which is only possible if hereduces corruption. This variable however doesnot survive the application of IV-GMM proce-dure. To begin with, its effect has never beenvery strong in the OLS works (significant at the10 percent level, so it is not too surprising thatthe effects will disappear if the endogeneity istaken care of.

Richer areas appear to have greater abilityto build better quality roads. The coefficientof logarithm of GDP per capita is positive and

strongly significant in all specifications. HigherGDP per capita serves as a base for local taxcollection. Though not always the case, dis-tricts with higher GDP per capita would havemore potential to have higher tax ratio. Poorerdistricts in the meantime would have to relymore on the central government transfer. Thecoefficient of district tax ratio is positive andsignificant at the 5 percent level stressing theimportance of government resources from tax-ation.

It appears that the construction of pavedroad is also sensitive to local cost. The higher

the price deflator the lower is the fraction of areas with paved roads. In 3 out of 4 specifica-tions, the coefficient is negative and significantat the 5 percent level. Related to cost as well asto the level of technical difficulties is the con-tour of the areas. Hilly regions are less likely tohave paved roads since it will cost more. Thecoefficient of the fraction of mountainous ar-eas in a district is strongly significant at the1 percent level for all specifications. Movingto the next covariate, areas frequently visited

by flooding also tend to have lower fraction of paved roads.

The local politic appears to have little in-fluence on the decision to build or to improveroad. The local politics is represented byHerfindahl index calculated from the share of votes of all parties. The variable is significantonly in IV-GMM with an interactive term thevariable so the results have to be interpretedwith a caution. The coefficient is negative andsignificant at the 10 percent level. The more

concentrated are the votes in few political par-ties the less likely the areas to be paved. So lo-cal political competition has positive effect onroad quality and may work to the advantage of local people.

Finally the fraction of local population ex-posed to newspaper or subscribing newspaperserves as an early warning to a district head forany public discontent especially with respect toroads. The coefficient is positive and signifi-cant at the 1 percent level. The existence of

free media has positive impact on the qualityof roads.

Table 2 presents the results of the passableroad regressions. The regressions are intendedto detect the cheating of road quality mostlyon the part of construction firms.

The fraction of areas of paved roads in a dis-trict passable by 4 wheelers year around is cre-ated from the eye witness account of villageheads for district roads passing through their

8/19/2019 201301(2) arikun vid

22/44


Table 2: Pre to Post Local Democratization (2003-2006): Passable Road Regressions

Fraction Paved Road OLS IV-GMM

I II I II

Number of budget Corruption cases -0.004 0.001 -0.003 -0.02[-1.56] [-0.73] [-1.45] [-1.60]

Dummy 2006 -0.011 -0.004 -0.011 -0.042[-2.85]*** [-1.08] [-2.98]*** [-1.97]**

Dummy 2006 X # of budget Corruption cases -0.004 0.026[-1.08] [1.54]

Fraction of households with TV 0.002 0.003 -0.01 -0.01[0.25] [0.32] [-1.20] [-0.89]

Log GDP per capita 0.0002 -0.00001 0.001 0.003[0.05] [0.00] [0.32] [0.84]

Tax ratio 0.16 0.154 0.13 0.15[3.54]*** [3.37]*** [3.30]*** [3.41]***

District price indicator (GDP deflator) 0.003 0.003 0.001 0.005[1.20] [1.21] [0.48] [1.60]

Fraction of hilly areas in a district -0.011 -0.012 -0.011 -0.014[-1.83]* [-1.85]* [-1.97]** [-2.47]**

Fraction of areas prone to flooding -0.032 -0.03 -0.034 -0.036

[-2.34]** [-2.26]** [-2.96]*** [-2.64]***Herfindahl Index of political parties -0.002 -0.003 0.001 -0.01

[-0.19] [-0.25] [0.13] [-0.93]Fraction of bureaucrats with college degree 0.04 0.04 0.038 0.022

[4.79]*** [4.57]*** [3.18]*** [1.85]*Number of Observation 2052 2052 2052 2052F-value 6.72** 6.18**Wald-chisq 66.88*** 50.97**P-value 0.21 0.32

Notes:1. Figures in parentheses are t-ratio2. *** Significant at 1% level

3.** Significant at 5% level4. * Significant at 10% level5. Instruments are number of voters in the previous election, percentage of government employees in 1990s,average distance to district capital from a village, percentage of population with high school education in1990s, bribe in manufacturing sector 3 years back and fraction of population subscribing newspaper.

8/19/2019 201301(2) arikun vid

23/44


villages. This variable is somewhat more reli-able than the data produced by district offices

which may strategize to exaggerate or to dis-count the cases depending on their motives, inorder to attract funding from the central gov-ernment.

Statistically the results for the passableroads are weaker than the paved road case.None of the number of budget corruption casesis significant. Meanwhile all 4 coefficients of time dummy are negative and three of themare significant at the 5 percent level. So thereis some indication that there the durability of

roads are declining significantly at least in someareas, the situation which is not clearly indi-cated in the summary statistic (Table A2.4).Finally local politics have nothing to do withroad durability.

The fraction of household with TV is not sig-nificant indicating that the problem of qualitycheating is hard to detect by the general publicor the press which do not have technical compe-tency to do the job. From physical appearanceroads may look fine but they may be washed

away in the rainy season or at least would bemarked by big potholes.

The detection of cheating may be performedmore effectively by competent specialists ortechnicians or at least by bureaucrats of certaintypes who can deter cheating before it starts ei-ther through better monitoring management orhigher technical competency. For this purposewe include the percentage of bureaucrats withcollege degree as a proxy for those who may do

better job in the road quality monitoring. Theresult shows that districts with better educatedbureaucrats are less likely to have problem withroad impassability especially during the rainyseason. The coefficient of the fraction of localbureaucrats with college degree is positive andstrongly significant at the 1 percent level.

Related to the road durability is the abilityof districts to do regular maintenance whichneeds to be financed from tax revenues. Thereis no difference between rich and poorer dis-

tricts with respect to whether roads are pass-able or not but those with higher ratio of lo-

cal tax to GDP face less problems with roadimpassability. The coefficient of tax ratio ispositive and strongly significant in all specifi-cations. Higher tax ratio could also mean thatpotentially there would be more resources to adistrict which imply they could do everythingbetter.

The task of maintaining roads year aroundappears to be harder in mountainous regionsas well in districts vulnerable to frequent flood-ing. Flooding is the worst enemy to road

maintenance. The coefficients of for frequentlyflooded areas are always negative and signifi-cant at the 5 percent level in all specifications.

Comparing results from Table 1 and Table 2,it suggests that an introduction of direct elec-tion would make district heads to improve theroad quality in terms of the increase of pavedroad areas but perhaps at the expense of somedecline in durability. In the theoretical partwe assert that the decline of quality (in termsof durability) may result because construction

firms have to pay higher bribes to compensatefor the district head loss of corruption incomesfrom defrauding local budgets. The above re-sults give a hint but not a direct result thatthis may be the case.

To be able to confirm the results we replacethe number of corruption cases with the trendof bribes in a district. The bribe data from thefield interview have too many missing valueswhile the timeframe is too short. It may begood enough for analytical narrative but cer-

tainly not sufficient for any econometric exer-cises. To overcome this problem we use thetrend of bribes in manufacturing sector in adistrict. We assume that bribes of all sectorsin a district move in a similar fashion sincein the end all of those must be paid to thesame corrupt district heads bureaucracy in adistrict.10 The bribe in the manufacturing sec-

10In other words the relationship θ(b) is substitutedin to the semi-reduced form in equation (15).

8/19/2019 201301(2) arikun vid

24/44


tor in Indonesia can be calculated for every dis-trict since we know firms location though the

firm identity is censored. The regressions re-sults are presented in Table A2.5 and TableA2.6 in Annex 2.

This does not mean there is no cheating byconstruction firms since on average there is aslight decline in durability. Also the year 2006dummy is negative and significant which meansalthough there are more paved roads in 2006as indicated in the paved road regression, theroads are less durable which may indicate lowerquality or cheating on the part of construc-

tion firm. As the theoretical model indicatedand also corroborated by a field interview, thismay stem from the increase in bribes whichforce firms to cheat to preserve profits. It maynot happen in general scale rather only in somecertain districts whose district heads are reallypure rent seeker and hence affects the overalldistrict bureaucracy attitude. For good dis-tricts which are the majority in our sample,the temptation to increase bribes and to cheaton quality may be held in check by local free

media and monitoring system. So the situationis more in line with Figure 2 where the positiveeffects from the new election system still persistdespite the continuation of some cheating.

In the era of direct election system, districtheads need to build better quality roads moreto satisfy potential voters while at the sametime a more vigorous public scrutiny wouldforce them to lower corruption from local bud-get. The risk of extorting more bribes fromfirms is also high. As our field interview indi-cated, this however will not eliminate bribingaltogether since the results suggest the nega-tive relationship between bribes and the roaddurability still persists. So what has been hap-pening after 2004 all interested parties will justpreserve the old bribing system quietly not at-tracting public attention.

The burden of the new form of electionregime may fall on construction firms since dis-trict heads bureaucracy has monopoly power to

award project grants. At the same time theyneed to build better quality roads to attract

voters as well as to signal their competence.This will cost firms more which would squeezeprofits but construction firms may have verylittle choices except to continue the business aslong as the profits are comparable to the op-portunity costs elsewhere. To check this wererun the model in Table A2.5 and replacingthe fraction of passable roads with paved roads.The results are presented in Table A2.6.

None of the coefficient of bribe is significant.Bribes have no effect on road quality. Mean-

while, the time dummy is still positive and sta-tistically strong, which means a big positive ef-fect on the number of areas with paved roadafter the introduction of a direct election sys-tem. In model II of OLS version the interac-tive term between bribes and time dummy isnegative and significant suggesting that after2004 the increase of bribes tend to lower roadquality but the effect does not survive whenthe IV-GMM procedure is applied. So at bestour interpretation is that some cheatings have

indeed taken place perhaps almost everywherebut not in a sufficient magnitude that wouldmake the positive effect of democratization onroad quality disappears altogether.Probability to win reelection There is a question whether road quality asthe single most important variable in the vot-ers mind is really valid outcome. There maybe some other public goods that are equally ormore important. To answer this question weestimate empirically the probability to win re-

election equation (7). The dependent variableis a dichotomous one having the value of one if the incumbent won the election and zero oth-erwise.11

In the regression equation beside a measureof road quality we also incorporate other publicgoods such village markets, village bus termi-

11We attempted to instrument both the road qual-ity h and defrauded budget Θ but the model does notconverge.

8/19/2019 201301(2) arikun vid

25/44


nal, village health facility and access to districtwater company which serves the public either

through piped in water, central village watertower or water truck. In another experimentwe pair road quality and the number of bud-get corruption cases to examine which issue isreally more important in the election. The re-sults are presented in Table A2.7 in Annex2.

None of other public goods is important inthe reelection of incumbent. For the voters tohave passable roads year around is more impor-tant than just having roads paved. The frac-

tion of areas with passable roads is positivelysignificant at the 5 percent level, while it is notsignificant for paved roads. The probability of an incumbent to be reelected is higher if heor she did a good job in maintaining the roaddurability in the district and not simply justmaking new or paving roads. The number of budget corruption cases has never been signif-icant in any of the above specifications.

The time dummy for 2009 is positive andstrongly significant suggesting that more in-cumbent won the reelection in 2009. In 2004many incumbents were the product of the oldindirect election system in which campaigningdirectly to the voter had never been important.At the time of transition in 2004, they were stillgrasping on issues that might attract voters.There was also a public sentiment against in-sider of which outsiders were perceived as beingcleaner or less corrupt (Henderson and Kun-coro [2011]). In 2009 however, the incumbentswere cleverer having learned their lesson. Theuse of road quality as a political instrument isnot new. To mobilize popular support from ru-ral areas Suharto had used among others roadconstruction as income generating public workas well as to smooth people and goods move-ment in the countryside (Kuncoro and Resosu-darmo [2006]).

Only few of other covariates are significantwhich may be a sign of good thing for the vot-ers rationality. This suggests that inter district

variations in socio-economic, demographic andgeographic do not matter very much. Local

governments are expected to deliver somethingreal and durable. The above results also sug-gest that the voters preferences are more or lessuniform across districts favoring road qualityover other things.

Decentralization Results Table A2.8 summarizes the econometric resultsfor the paved roads before and after decentral-ization in 2001 but local democratization hasyet to take place in 2004. So decentralizationhas not yet fully accompanied by full account-

ability to local people. The difference betweenmodel I and model II is the inclusion of frac-tion of population exposed to newspaper in thelater and not in the former. The purpose isto examine whether the existence of local freemedia in this case newspaper provides a suffi-cient counterbalance for the potential lack of accountability on the part of local governmentsince local people still could not vote out in-competent or corrupt local government.

Looking at glance the impact of introduc-ing fraction of population exposed to news-paper is to weaken the impact of bribes onfraction of paved roads in a district. Oneplausible explanation is that in the 2001-2003periods although the time line for local de-mocratization had not been announced, thelogic of decentralization-local democratizationsequencing was already a public knowledge sodistrict heads and people might form their ex-pectation and behavior accordingly. In the re-

gression with newspaper variable, the inclusionof newspaper variable (model II) weakens thebribe effects on paved roads. In the regres-sions without newspaper variable we managedto find a significant negative relationship be-tween fraction of paved roads and bribes. Butthese effects disappear if newspaper variableis added into the equation. The coefficient of newspaper variable itself is positive and highlysignificant. In the regression without newspa-per variable the TV ownership is not strong

8/19/2019 201301(2) arikun vid

26/44


enough to weaken bribe effects. It suggeststhat public exposure to newspaper may provide

strong deterrent against corrupt behavior.The coefficient of yearly dummy for 2003 is

significant and negative for regression withoutnewspaper variable suggesting that quantita-tively the road situation deteriorated after de-centralization went into effect. But control-ling for newspaper the negative effects becomeinsignificant basically telling the same storyas above. The tax ratio is positive but in-significant which is understandable since al-most all aspects of local public finance were

still very much controlled by the central gov-ernment prior to 2001. Other covariates con-trolling for topographical contour of the areasmore or less behave in the same way as inthe local democratization regressions. DistrictGDP deflator as a proxy for cost show that lo-cal costs, does not impose any constraint onroad construction since it is either insignificantor positive but weakly significant.

The same sort of exercises with respect to

passable roads is presented in Table A2.9.The coefficients of bribes are negative and sig-nificant in 3 out of 4 specifications. The addi-tion of newspaper in the set of covariates doesnot weaken this relationship. Again it suggeststhat cheating on the part of construction firmis hard to detect people will only realize aboutthe defect perhaps some months after the roadsare in place. The newspaper variable itself ispositive and highly significant so more passableroads that are not washed away in the rainy

season will be available to the public. The ex-istence of free media in the form of newspa-per coverage is essential to keep pressure ondistrict heads to maintain roads in good con-dition. The TV ownership remains significantand positive at least in 2 specifications. Thereis no change in fraction of passable road be-fore and after decentralization - year dummyand its interaction with bribes are basically in-significant except in one case where the yeardummy is negative and weakly significant.

Turning to other covariates, none of GDPper capita variable is significant which suggests

cheating in road quality is nothing to do withrich or poor districts. Richer districts could af-ford to have better road quality as shown bythe paved road regression but this effect disap-pear in the passable road regressions. All dis-trict GDP deflator which is used as a proxy of construction cost have wrong positive sign butthis may only reflect that construction costs inricher areas are more expensive. Despite this,all other control variables for construction costssuch as fraction of hilly, prone to flooding have

the right negative sign an significant at one per-cent level.

Determinants of Bribes In Table A2.10 in Annex 2, we present sev-eral results of the first stage regressions of firmbribe and number of budget corruption cases.In each respective dependent variable we in-corporate the average bribes/number of budgetcorruption cases in contiguous districts to ex-amine whether there is a demonstration effector social learning from adjacent areas.

Between 2003 and 2006, bribes have de-creased as shown by the negative and signifi-cant coefficient of the year 2006 dummy. In thebribes regressions average distant of villages todistrict capital approximates the size of a dis-trict. The coefficient is positive and significantat 1 percent level. The larges is a district thehigher is the bribes which suggest monitoringand enforcement problem. The larger is thesize of government employee in the local workforce the higher is the bribes reflecting the scaleeffect of potential extorters. The fraction of manufacturing in district GDP which suggestthat the formal sector serves a base for bribecollection. The coefficient of average bribes incontiguous districts is not significant indicatingthere is no learning effect for corrupt behav-ior. So it is possible that for a clean districtto locate side by side or surrounded by corruptdistricts and relatively unaffected.

Unlike in the bribe regressions the year 2006

8/19/2019 201301(2) arikun vid

27/44


dummy in the budget corruption regressions isnot significant. The average distant to district

is positive and highly significant. There is nochange between 2003 with respect to the num-ber of corruption cases involving district bud-get. Coastal districts appear to be less cor-rupt. In one model, the percentage of coastalvillages in a district is negatively significant atthe 10 percent level. The Herfindahl index of political parties is negative and significant atthe 10 percent level. This could mean thatthe increase of one or few parties dominancein a district strengthens the ruling parties lob-

bying power against local attorney general of-fice or local anti corruption agency (KPK) todismiss budget corruption cases. Again the av-erage number of corruption cases in contiguousdistricts is not significant so there is no sociallearning effect.

Determinants of Tax RatioIn our empirical estimation tax ratio is treatedas exogenous. Actually it can be endogenousinfluenced by political and socio-economic vari-ables. We investigate this matter and the re-

sults are presented in Table A2.11 in Annex2. Between 1999 and 2003 (before and afterdecentralization) the district level tax ratio inIndonesia has increased as shown by a positiveand highly significant the year 2003 dummy.The average distant to district capital is neg-ative and significant reflecting collection andadministrative problems as a district becomeslarger in geographical size. The percentageof government employee in the work force ispositive and significant which show the poten-

tial increase of tax collection capacity with thesize of government sector. The relationship be-tween district GDP per capita and tax ratio isnegative which indicate that richer areas do notpay their dues. The problem of tax evasion isperhaps more rampant in richer areas which ishelped by the collusion between tax payers andtax collectors.

Policy Implication Building on the momentum of decentralization

and democratization Indonesia now has under-gone another experiment in democracy namely

the change of election system from indirect todirect election by the people. For politicians(district heads) and their associated bureau-crats alike the transition is not easy. With adirect election system, politicians have to ad-dress the need of voters, while at the same timehave to find the way to finance the political ma-chinery. There is a little doubt that the cost of political activities has increased significantly.In the Suharto era this had been run by thecentralized political system. With the depar-

ture of Suharto the system has become frag-mented with no clear hierarchy of power.

One urgent issue in this new form of democ-racy democratic Indonesia is political partyfunding. A reform on political activities fund-ing is overdue. Much of the parties moneyis thought to come from the selling public of-fices, government project, influencing budgetallocation. In this study we focus on the gover-nance aspect of the granting road constructionprojects to private contractors. In the era of

direct election system, district heads will haveto build better quality roads more to satisfypotential voters while at the same time a morevigorous public scrutiny would force them tolower corruption from local budget. To com-pensate the loss of corruption money they mayincrease bribes from awarding contracts to pri-vate contractors. The findings however suggestthat there is no apparent dramatic increasewith respect to bribes though it is still the wayto do business in the road construction busi-

ness. Voters punishment may not be enoughto deter district heads from increasing bribes.Other important factors must also exist. Inthis case the other factors are the existence of local free media and the ad-hoc anti corruptionagency (KPK) which holds corruption in check.

There are two types of corruption in roadprojects, budget mark-up and collusion be-tween providers and district government. Tominimize budget corruption the establishment

8/19/2019 201301(2) arikun vid

28/44


of procurement standard for road constructionthat takes into account district heterogeneity

is necessary. In the early budgeting processthis should be verified by a district audit bodywhich is independent from a district govern-ment to prevent budget mark-up. Despite thefact that invitation to bid is an open process,formation of cartel or pre-selection of winneris prevalent almost everywhere. Ad hoc anticorruption agency should also focus on the bid-ding process not just after the projects are com-pleted in order to prevent collusion as early aspossible. Once a project is finished a quality

check should be done by an independent thirdparty instead of by local government agency.Standard qualification for bureaucrats for tech-nical department in charge of quality monitor-ing must be high political appointment of rent-seeker agent should be avoided as it would beself-damaging to district government credibil-ity.

Producing high quality roads as demandedby voters are not cheap while most of the con-tracts are fixed. Taking into account also the

bribes that have to be paid to , this would haveeaten into firms profit. So the temptation toproduce lower quality roads is high. Districtheads after knowing this would increase qualitymonitoring since their jobs in the eyes of voterswould depend on the road quality. To summa-rize, the combination of democratization, freemedia, an ad-hoc anti-corruption agency andwell educated bureaucrats holds corruption incheck and even may still maintain decent qual-ity of roads.

7. Referrences

[1] Andvig, J.C. and K.). Moene, (1990) How Cor-ruption May Corrupt, Journal of Economic and Behavioral Organization , vol 13, p. 63-76.

[2] Austen-Smith, D., (1987), ’Interest Groups, Cam-paign Contributions and P

Documents

201301(2) arikun vid