13
Corruption, Manufacturing Plant Growth, and the Asian Paradox: Indonesian Evidence VIRGINIE VIAL Euromed Management, GRIDS, Marseille, France and JULIEN HANOTEAU * Euromed Management, GRIDS, Marseille, France GEM – Sciences Po, France Summary. Using panel data from the Indonesian manufacturing industry during the Suharto era (1975–95), we assess the impact of plant-level corruption on output and productivity growth. In support of the grease the wheelshypothesis and the view of an Asian paradox, we find that corruption, measured as bribes and indirect tax payments, has a positive and statistically significant effect on indi- vidual plant growth. This effect persists over the entire period, which suggests improvements in the efficacy of the bribe system and a strengthening of the long-term contract between firms and the government. Ó 2009 Elsevier Ltd. All rights reserved. Key words — corruption, taxation, plant growth and productivity, Asia, Indonesia 1. INTRODUCTION During the second half of the 20th century, economic growth in Southeast Asia was rapid compared with that of other regions of the world, and the various development mod- els in that region display some common characteristics, includ- ing strong government intervention and relative openness to foreign trade, investment, and global finance (Wade, 1990). Rapid growth rates in the 1980s and the 1990s suggested the region as potential example for other developing nations and attracted attention, as well as criticism, from some economists (Krugman, 1994). Observers soon began questioning the sus- tainability of such growth, arguing that it actually resulted from factor accumulation rather than total factor productivity (TFP) growth. When the financial crisis of 1997 hit, all Southeast Asian development models came under acute scrutiny, and a series of investigations noted potential causes of the breakdown (Claessens, Djankov, Fan, & Lang, 1998a, 1998b, 1998c, 1998d; Claessens, Djankov, & Lang, 1998; Pomerleano, 1999; Prowse, 1998; Rajan & Zingales, 1998). Rajan and Zin- gales (1998, p. 559) in particular underline that Southeast Asian economies relationship-based systems work well when contracts are poorly enforced and capital scarce.However, when arm’s-length investors provide capital, they prefer short-term contracts to limit the risk associated with a misal- location of capital. When these arm’s-length investors pull out, the system collapses. Most studies suggest cronyism and corruption as the cement that holds the relationship-based sys- tem together but also recognize the limitations of such a sys- tem when financial markets open, which usually precipitates a confidence crisis. We consider the case of Indonesia, which displayed high economic growth rates from the 1970s onward but suffered greatly from the 1997 crisis, partly because of the crony capitalism that spread under the rule of General Suharto (1965–98). During 1975–95, aggregate yearly output growth for the manufacturing sector reached 9.30%, and TFP growth estimates ranged from 2.70% (Aswicahyono, 1998) to 2.80% (Timmer, 1999) to 3.54% (Vial, 2006). At the micro-level, man- ufacturing plants’ average annual output growth rate was 4.60%, and average labor productivity growth was 2.40%. These characteristics epitomize the Asian paradox: Cronyism and corruption are prevalent, but they do not necessarily ham- per business (Rajan & Zingales, 1998). To investigate this hypothesis using unique, micro-level data on corruption, we begin with a review of literature on the con- troversial effects of corruption on firms’ performance. We then describe Indonesia’s corruption and performance during the period 1975–95. We present the data, the empirical model, and then the estimation results, and the last section concludes. 2. CORRUPTION AND PERFORMANCE: CONTRADICTING EFFECTS Corruption is generally considered as detrimental to both investment and growth at the macro-level (for empirical evi- dence, see Mauro, 1995; Me ´on & Sekkat, 2005; Wei, 2000; for theoretical arguments, see Bardhan, 1997; Rose-Ackerman, 1999). This consensus about the negative effects of corruption has however been challenged, especially in the case of Asian economies. Kaufmann and Wei (1999, p. 10) posit Asian excep- tionalism, such that corruption has been part of the Asian cul- ture for a long time and does not seem to hamper business there.Furthermore, Rock and Bonnett (2004) perform a com- * We thank Roland Bel, participants at the 2008 EPCS meeting, partic- ipants at the International Schumpeter Society Conference 2008, and five anonymous referees for their helpful comments and suggestions. Final revision accepted: November 6, 2009. World Development Vol. 38, No. 5, pp. 693–705, 2010 Ó 2009 Elsevier Ltd. All rights reserved 0305-750X/$ - see front matter www.elsevier.com/locate/worlddev doi:10.1016/j.worlddev.2009.11.022 693

Corruption, Manufacturing Plant Growth, and the Asian Paradox: Indonesian Evidence

Embed Size (px)

Citation preview

Page 1: Corruption, Manufacturing Plant Growth, and the Asian Paradox: Indonesian Evidence

World Development Vol. 38, No. 5, pp. 693–705, 2010� 2009 Elsevier Ltd. All rights reserved

0305-750X/$ - see front matter

www.elsevier.com/locate/worlddevdoi:10.1016/j.worlddev.2009.11.022

Corruption, Manufacturing Plant Growth, and the Asian Paradox:

Indonesian Evidence

VIRGINIE VIALEuromed Management, GRIDS, Marseille, France

and

JULIEN HANOTEAU *

Euromed Management, GRIDS, Marseille, FranceGEM – Sciences Po, France

Summary. — Using panel data from the Indonesian manufacturing industry during the Suharto era (1975–95), we assess the impact ofplant-level corruption on output and productivity growth. In support of the “grease the wheels” hypothesis and the view of an Asianparadox, we find that corruption, measured as bribes and indirect tax payments, has a positive and statistically significant effect on indi-vidual plant growth. This effect persists over the entire period, which suggests improvements in the efficacy of the bribe system and astrengthening of the long-term contract between firms and the government.� 2009 Elsevier Ltd. All rights reserved.

Key words — corruption, taxation, plant growth and productivity, Asia, Indonesia

* We thank Roland Bel, participants at the 2008 EPCS meeting, partic-

ipants at the International Schumpeter Society Conference 2008, and five

anonymous referees for their helpful comments and suggestions. Finalrevision accepted: November 6, 2009.

1. INTRODUCTION

During the second half of the 20th century, economicgrowth in Southeast Asia was rapid compared with that ofother regions of the world, and the various development mod-els in that region display some common characteristics, includ-ing strong government intervention and relative openness toforeign trade, investment, and global finance (Wade, 1990).Rapid growth rates in the 1980s and the 1990s suggested theregion as potential example for other developing nations andattracted attention, as well as criticism, from some economists(Krugman, 1994). Observers soon began questioning the sus-tainability of such growth, arguing that it actually resultedfrom factor accumulation rather than total factor productivity(TFP) growth.

When the financial crisis of 1997 hit, all Southeast Asiandevelopment models came under acute scrutiny, and a seriesof investigations noted potential causes of the breakdown(Claessens, Djankov, Fan, & Lang, 1998a, 1998b, 1998c,1998d; Claessens, Djankov, & Lang, 1998; Pomerleano,1999; Prowse, 1998; Rajan & Zingales, 1998). Rajan and Zin-gales (1998, p. 559) in particular underline that SoutheastAsian economies “relationship-based systems work well whencontracts are poorly enforced and capital scarce.” However,when arm’s-length investors provide capital, they prefershort-term contracts to limit the risk associated with a misal-location of capital. When these arm’s-length investors pullout, the system collapses. Most studies suggest cronyism andcorruption as the cement that holds the relationship-based sys-tem together but also recognize the limitations of such a sys-tem when financial markets open, which usually precipitatesa confidence crisis.

We consider the case of Indonesia, which displayed higheconomic growth rates from the 1970s onward but sufferedgreatly from the 1997 crisis, partly because of the cronycapitalism that spread under the rule of General Suharto

693

(1965–98). During 1975–95, aggregate yearly output growthfor the manufacturing sector reached 9.30%, and TFP growthestimates ranged from 2.70% (Aswicahyono, 1998) to 2.80%(Timmer, 1999) to 3.54% (Vial, 2006). At the micro-level, man-ufacturing plants’ average annual output growth rate was4.60%, and average labor productivity growth was 2.40%.These characteristics epitomize the Asian paradox: Cronyismand corruption are prevalent, but they do not necessarily ham-per business (Rajan & Zingales, 1998).

To investigate this hypothesis using unique, micro-level dataon corruption, we begin with a review of literature on the con-troversial effects of corruption on firms’ performance. We thendescribe Indonesia’s corruption and performance during theperiod 1975–95. We present the data, the empirical model,and then the estimation results, and the last section concludes.

2. CORRUPTION AND PERFORMANCE:CONTRADICTING EFFECTS

Corruption is generally considered as detrimental to bothinvestment and growth at the macro-level (for empirical evi-dence, see Mauro, 1995; Meon & Sekkat, 2005; Wei, 2000;for theoretical arguments, see Bardhan, 1997; Rose-Ackerman,1999). This consensus about the negative effects of corruptionhas however been challenged, especially in the case of Asianeconomies. Kaufmann and Wei (1999, p. 10) posit Asian excep-tionalism, such that “corruption has been part of the Asian cul-ture for a long time and does not seem to hamper businessthere.” Furthermore, Rock and Bonnett (2004) perform a com-

Page 2: Corruption, Manufacturing Plant Growth, and the Asian Paradox: Indonesian Evidence

694 WORLD DEVELOPMENT

parative analysis of corruption, growth, and investment andfind a positive and statistically significant relationship betweengrowth and corruption for newly industrialized, large, EasternAsian countries, such as China and Indonesia.

On a more theoretical basis, Leff (1964), Huntington (1968),and Leys (1965) argue corruption can be beneficial from a sec-ond-best perspective. That is, corruption, though generallydetrimental to macroeconomic growth, can increase overallefficiency in the presence of deeper distortions, such as a rigidand overcentralized administration, excessive regulatory barri-ers, or poorly efficient and weakly competent bureaucracies.

The most popular explanation of this effect involves the“grease the wheels” hypothesis, which suggests that bribes,or “grease,” reduce long administrative delays and enableprogression through slow-moving queues for public services(Lui, 1985). 1 Several cross-country studies provide empiricalsupport for this hypothesis. For example, Dreher and Gasseb-ner (2007) analyze the impact of regulation on entrepreneur-ship, and show that the number of procedures and minimumsize of capital required to start a business have detrimental ef-fects on firm entry. However, corruption reduces these nega-tive impacts of regulation. Meon and Weill (forthcoming)find that, over the period 1994–97, corruption reduces macro-economic productivity in countries with effective institutionsbut raises it in countries with poor institutions.

The “efficient grease” idea rests on the crucial assumptionthat red tape and regulatory burdens are exogenous (see Bard-han, 1997; Kaufmann & Wei, 1999). Shleifer and Vishny(1993) challenge this assumption by suggesting bureaucratsand government officials have discretionary power over redtape and can customize it, in proportion to firms’ ability topay, so as to extract bribes that support their own interest. 2

This hold-up effect “sands the wheels” (Meon & Sekkat2005), creates administrative delays, that can then be reducedby the payment of bribes. Therefore, rather than improvingefficiency, corruption may add distortions and increase overallcosts. Meon and Sekkat (2005) offer empirical support for thisclaim in another cross-country study over the period 1970–98,that shows corruption is detrimental to investment andgrowth, a negative effect amplified for countries whose gover-nance is of the lowest quality.

Several theoretical arguments support both opposite views.If exogenous, corruption could have negative or positive im-pacts on economic performance, whereas if it is endogenous,it should have only a negative impact. We attempt to comple-ment this discussion with the argument that endogenous cor-ruption may have positive effects on plant performance ifpredatory officials embrace a long-term strategy. That is, cor-rupt officials may choose to extort money from the most suc-cessful companies, but if those officials remain in power forlong periods of time, they cannot run companies out of busi-ness. Instead, they extract a sustainable rent, which involvescreating a trusting relationship between the company andthe official; through this relationship, the company may evenobtain services that improve its performance in a second-bestsetting. 3 In this paper, we take a micro-level perspective of thedebate about the grease the wheels hypothesis. We posit thatthe accuracy of the efficient grease assumption represents anempirical question (Fisman & Svensson, 2007) that dependson the complex mechanisms at play and various explanatoryfactors linked to each country’s particular economic, political,and social settings.

Micro-level empirical studies of the economic consequencesof corruption are rare, mainly due to data limitations. Excep-tions are Kaufman and Wei (1999), McArthur and Teal(2002), and Fisman and Svensson (2007).

In a cross-country and firm-level study, Kaufman and Wei(1999) find that corruption explains both management timewasted with bureaucrats and regulatory burdens on firms.The authors assume the level of corruption is exogenous andinvestigate a direct, noncircular relationship between corrup-tion and red tape. They conclude that firms that face morebribe demands are also likely to spend more management timewith bureaucrats, which invalidates the efficient greasehypothesis. However, as they note, their results need to beinterpreted with caution, for two reasons. First, both corrup-tion (amount of bribes paid) and red tape (regulatory burden)should be considered endogenous, in that they may be influ-enced by the same factors: bureaucrats’ willingness to holdup firms and receive bribes. Second, Kaufmann and Wei usefirm-level data (based on the Global Competitiveness Reportindex, 1996 and 1997) to measure wasted management timebut country-wide data about corruption and regulatory bur-dens, which they derive from a country perception question-naire. Moreover, their data are primarily qualitative (indices).

Using similar data, based on the Africa Competitiveness Re-port, McArthur and Teal (2002) study the simple linear rela-tionship between corruption and productivity for a sampleof 505 firms from 27 African economies. They find a signifi-cant and negative effect of corruption, as measured country-wide, on productivity.

Fisman and Svensson (2007) examine the efficient greaseassumption with a sample of 243 firms in Uganda from 14industries, located in five different areas during 1995–97.Firms’ bribe payment rates affect their output growth nega-tively. The measure of firms’ bribe payments is a self-reporteditem. The authors consider this variable endogenous andinstrument it with an industry-location bribe average. Theyalso show that the rate of tax payment reduces output growthbut is less damaging than bribe payments.

We attempt to contribute to this empirical literature byexamining the effects of bribes, measured at the plant level,using two different data series on almost 200,000 plant-yearobservations in the Indonesian manufacturing industry during1975–1995.

3. CORRUPTION IN INDONESIA

Corruption has a long tradition in Indonesia and is emblem-atic of the era of former President Suharto. For example, in1980–83, Indonesia scored just 1.5 on the 0–10 corruption in-dex compiled by Business International (now The EconomistIntelligence Unit), ranking among the most corrupt countriesin the world (Mauro, 1995). During Suharto’s rule (1965–98), corruption served two main objectives. First, it contrib-uted to the personal enrichment of Suharto, his family, andhis allies. Bribes channeled through a system of cronies, as wellas through yayasans, the term for foundations that receivebribe payments from firms and then use those funds to financenew companies controlled by Suharto’s affiliates (McLeod,2000; Robertson-Snape, 1999). As Schwarz (2004, pp. 40–41)explains: “As the New Order progressed, so did the art ofpatronage. Revenues collected from Suharto’s close businessassociates in sectors such as oil, construction and agro-busi-ness—often washed through non-profit foundations—have en-abled Suharto to expand the distribution of patronage topotential critics in political, religious and social circles.”

In this system of cronies, or crony capitalism, Suharto andhis cronies exerted direct control over most (if not all) of thelargest firms in Indonesia. These controlled firms receivedadvantages granted by the government (McLeod, 2000) and

Page 3: Corruption, Manufacturing Plant Growth, and the Asian Paradox: Indonesian Evidence

CORRUPTION, MANUFACTURING PLANT GROWTH, AND THE ASIAN PARADOX 695

retransferred the rents to the cronies through various mecha-nisms, including manipulated prices in contracts, free grantsof shares in a company, cash envelopes, high salary positions,advantages in kind, and so on.

The yayasans are legal entities that collect funds to financesocial, religious, educational, or humanitarian activities andprograms. But according to The Economist (1993), “in addi-tion to their charitable work, yayasans act as giant slushfunds, dispensing patronage and cornering lucrative con-tracts.” Their books are not audited, and they do not paytax; because Suharto and his relatives controlled them, theirfunds got deployed for purposes known only to these cronies.Companies accounted for their donations to yayasans as giftsand charities, though Behrman and Deolalikar (1989) suggestthey clearly represented rewards for favors, such as preferen-tial tariffs, import quotas, and tax benefits from governmentagencies. In this sense, they are bribes to top officials in ex-change for favors. Robison (1986, p. 232) provides recurrentevidence of such payments and confirms yayasans are partof the corrupt system, such as P.T. Bogasari, a large Indone-sian flour mill for which “The articles of association stipulatedthat, in effect, 26% of profits be set aside for ‘charitable’ foun-dations including Mrs. Suharto’s Yayasan Harapan Kita andKostrad’s Yayasan Dharma Putra.” Another example is that“The articles of incorporation of Karana [an Indonesiandeep-see shipping line company] include a dispersal of divi-dends to various foundations, including Mrs Suharto’s Yaya-san Kartika Jaya.” (Robison, 1986, p. 260). Suharto also usedcorruption as an instrument to assert and maintain his power.Because corruption spread across all levels of the state and lo-cal administration, and with civil servants’ wages at a lowlevel, 4 most bureaucrats had an incentive and were encour-aged to participate in the corrupt system and sustain the re-gime.

Corruption in Indonesia might be described as a steeplyascending pyramid, with Suharto on the top (Robertson-Snape, 1999), above the delegation of power through a systemof “franchisees” (McLeod, 2000). Bureaucrats, both from gov-ernmental agencies and at the local level (i.e., districts, kabu-paten), had the discretionary power to set red tape andvarious taxes to pressure firms and influence their supply ofbribes. McLeod (2000) and Henderson and Kuncoro (2006)explain that in return, firms used bribes as grease money to re-duce taxes, administrative delays, red tape, and harassment bycivil servants.

McLeod (2000, p. 24) further argues that “the [Indonesian]bureaucracy has proven adept at creating countless regula-tions that require some kind of bureaucratic action before pri-vate sector firms can carry out their normal business activities:the issue of licenses, approvals, certificates, permits and soon, . . . most people take it for granted that it will be necessaryfor them to offer some ‘grease money’ if they are not to beblocked by the bureaucracy in whatever they are trying to do.”

In addition, firms expect real privileges, such as exclusiveimport licenses, tax exemptions, public contracts, and rightsto exploit natural resources. In this context, payments madeto obtain various licenses represent another proxy for corrup-tion, because firms without connections lack any access tothese various licenses.

In 1999 and 2000, the World Bank (2000) surveyed a sampleof Southeast Asian firms, including 100 Indonesian firms. Thesurvey aimed to determine the quality of the investment cli-mate and business environment. Of the 100 Indonesian firmssurveyed, 89 responded to a question related to taxes, andmore than two-thirds of them acknowledged high taxes hadbeen a major obstacle to their performance. Furthermore, 80

firms responded to a question about bribe payments and re-ported average payments equal to 3.5% of their total sales. Fi-nally, among the 93 firms that responded, nearly 70%indicated the service for which they paid the bribe had beenrendered as expected. If the services purchased through bribesget rendered, one question remains: Are payments of bribesfor rendered services actually harming growth?

4. EMPIRICAL METHODOLOGY AND DATADESCRIPTION

Our empirical model draws from Fisman and Svensson(2007), who provide a more detailed presentation. For thisstudy, we consider a linear relation between a plant’s out-put/productivity growth Dyijt;tþ1, between the periods t andt þ 1, and for bribe payments bijt at period t, such that

Dyijt;tþ1 ¼ a0 þ a1bijt þ a2xijt þ aij þ gijt; ð1Þ

where i indicates plant i in sector j observed during year t; xijt

is a vector of the time-variant variables; aij refers to the unob-served plant fixed effect; and gijt represents a zero-mean errorterm. Similar to Fisman and Svensson (2007), we introduceindirect taxes in the model only later.

Bribe payments should affect output growth and productiv-ity growth, though with a lag. This corresponds to the Indone-sian context of corruption during the Suharto era as discussedin Section 3. As the World Bank Environment Survey 2000shows, for example, Indonesian firms clearly pay bribes, butthe favorable treatments they gain in return (e.g., reducedtaxes, administrative delays, red tape, and harassment by civilservants, exclusive import licenses, right to exploit natural re-sources, public contracts) do not have an instantaneous effecton output and productivity growth. They rather perform overa subsequent and extended period of time. In the literature, therelationship between bribe payments and growth suggests thatbribe payments induce growth; for example, Kaufman andWei (1999) hold that firms choose to bribe to maximize theirafter-bribe profits. In their empirical study, Fisman and Gati(2006, p. 131) also assume bribes affect firms’ growth onlyafter a lag, such that they measure “expected growth rate insales in the subsequent three years.”

However, our preceding discussion also suggests that bribepayments may be endogenous to the firm, because bureaucratsmay have an incentive to customize bribes and extract higherrents from plants that indicate higher growth or growth poten-tial. In addition, firms may opt to follow a rent-seeking strat-egy, such that they pay bribes in exchange for economic favorsand preferential treatment from regulators and bureaucrats(Fisman & Svensson, 2007).

A potential reverse causality between bribes and plantgrowth thus may exist: Corrupt bureaucrats establish taxes,administrative delays, and red tape to extort bribes accordingto firms’ future ability to pay. Svensson (2003, p. 208) consid-ers corrupt bureaucrats rational and forward-looking in theirassessments of firms’ ability to pay, and “firms’ ability to pay,proxied with their current and expected future profitabil-ity . . . can explain a large part of the variation in bribes acrossgraft-reporting firms.” 5

Thus, we posit that bribe payments and plant output or pro-ductivity growth are jointly determined by some common fac-tors, which may be specific to an industry (e.g., underlyingtechnology) or location (e.g., rent extraction talents and cor-ruption of local bureaucrats). For Indonesia, Kuncoro(2004) and Henderson and Kuncoro (2006) confirm that firms’

Page 4: Corruption, Manufacturing Plant Growth, and the Asian Paradox: Indonesian Evidence

696 WORLD DEVELOPMENT

bribe payments depend on local factors, including local gov-ernments’ design of red tape aiming at collecting bribes.

Other factors are plant or firm specific. Kuncoro (2004) sug-gests that the amount of bribe payments in Indonesia dependson firms’ size, tax payments, and foreign ownership. Fismanand Svensson (2007) note however that plant- or firm-specificfactors that influence bribes and growth simultaneously maybe unobservable and thus likely to introduce omitted variablebias.

To solve this issue, Fisman and Svensson (2007) assume thatindustry-location-specific and plant-specific factors are uncor-related and use industry-location average bribe payments asan instrument of plant-level bribery. This grouped averagestechnique can solve a second empirical problem resulting fromthe possible measurement errors associated with bribery: Themost corrupt firms may be reluctant to report their bribe pay-ments, so using industry-location averages helps mitigateplant-level measurement errors.

As exposed previously, the bribes bijt are determined jointlyby geographic and sector-specific factors, as well as plant-spe-cific factors that also influence output or productivity growth,Dyijt;tþ1. In these conditions, bribes are endogenous to the firm,and our estimate of a1 using ordinary least squares (OLS), isbiased. Therefore, following Fisman and Svensson (2007),we assume that a plant’s bribe payments consist of two ele-ments:

bijt ¼ Bijt þ Bjt; ð2Þwhere Bijt corresponds to payments justified by reasons idio-syncratic to plant i at time t and Bjt is the share of paymentsexplained by factors specific to a particular industrial sectorin a particular geographical area (district, kabupaten) at timet. In a two-stage empirical procedure, we can use this indus-try-location average of bribe payments Bjt, along with otherexogenous variables xijt, as instruments to generate the fittedvalues for bijt:

bijt ¼ b0 þ b1Bjt þ b2xijt þ a0ij þ g0ijt: ð3Þ

Then, our empirical model becomes:

Dyijt;tþ1 ¼ a0 þ a1bijt þ a2xijt þ d75;...;95t75;...;95 þ aij þ gijt; ð4Þ

where bijt are the fitted values from the first-stage regressions.In addition to bribes, we use plants’ payments of indirect

taxes as an indicator of rent extraction by corrupt officials.Whereas Fisman and Svensson’s (2007) variable covers bothdirect and indirect taxes, we focus on indirect taxes, which ismore appropriate for the Indonesian context. 6 Indeed, Hen-derson and Kuncoro (2006) explain that the rent extortedfrom Indonesian firms by governments and bureaucrats chan-nels through indirect taxations, such as import duties, licenses,and levy fees, as well as indirect revenues in the form of bribes(see also McLeod, 2000). Therefore, indirect taxes and bribesare either complementary or alternative, from the perspectiveof corruption and rent extraction. Kuncoro (2004) confirmsthis by testing for the micro-level determinants of bribery.Over a sample of 1808 Indonesian firms for the year 2001,he finds that firms’ tax payments significantly explain theamount of bribes paid. He thus concludes that both variableshave the same determinants, namely, firm profitability and ci-vil servants’ willingness to extract bribes, and that “the moreprofitable the firm, the higher the bribes it will pay” (Kuncoro,2004, p. 336).

If we let sijt be the amount of indirect taxes paid by plant i attime t, we recognize it is likely to be endogenous if bureaucratsand governments can set indirect taxation to extract rents

from a particular plant. Therefore, we follow the same ap-proach and use the industry-location average of indirect taxpayments (T jt) to generate fitted values for sijt:

sijt ¼ b00 þ b01T jt þ b02xijt þ a00ij þ g00ijt: ð5Þ

The empirical model thus becomes:

Dyijt;tþ1 ¼ a0 þ a1bijt þ a01sijt þ a2xijt þ d75;...;95t75;...;95 þ a000ij þ g000ijt

ð6Þwith sijt equal to the fitted value of sijt.

Indonesian manufacturing data are available yearly and atthe plant level for the period 1975–95, so we use panel datatwo-stage instrumental variables with a fixed effect estimationtechnique, controlling for the time with year dummies. Be-cause we use a fixed effect model and plants do not shift theirindustry over the period, we do not need to include a dummythat controls for industries. 7

We estimate the model for the entire period, then comparethe results for three subperiods: 1975–80, or the oil boom per-iod; 1981–89, during deregulation; and 1990–95, the postdere-gulation and investment boom period. The first periodfeatured fast output growth, fuelled by oil revenues. The sec-ond one started with the oil crisis in 1981–83 and continuedwith a period of deregulation as a response to the crisis. Dur-ing the third period, Indonesia’s markets should have bene-fited from a marked increase in competition; theliberalization context seemingly should have affected boththe levels and the influence of firms’ payments of bribes andindirect taxes. Because we use population rather than sampledata, the number of observations grows over time as theindustrial sector develops in the form of more plant entriesthan exits. The increasing number of observations from onesubperiod to the other improves the generalizability of our re-sults, but it could increase or decrease the significance of thecoefficients of the explanatory variables. This evolution overtime should indicate both the effect of the explanatory vari-ables and the robustness of the results as the number of obser-vations increases.

We use a plant-level data set, Statistik Industri, that providesa census of all Indonesian manufacturing plants with 20 ormore employees over the period 1975–95. It comes from an an-nual survey conducted by the Indonesian Bureau of PublicStatistics (BPS). The questionnaire, administered at the na-tional level, is anonymous and detailed and covers the estab-lishments’ characteristics, output, input use, expenditures,ownership, and so on. 8

We measure output and labor productivity growth as the logdifferences between t and t þ 1. To measure bribes, we use avariable labeled “gifts, charities, donations,” which stems fromthe plants’ answers to a question about their “other expenses.”Our preceding discussion reveals that this item likely indicatesbribe payments, which mostly channel through charities(yayasans). Behrman and Deolalikar (1989) also use this vari-able as a proxy for plant-level corruption. Although it is a self-reported value, it should be relatively truthful, because theBPS survey is anonymous, with plants identified only by num-bers. We acknowledge though that it cannot include otherforms of bribery, such as commissions, contract shares, andoption contracts that fall below or above market prices.

The variable for indirect taxes includes sales taxes, establish-ment licenses, building and land taxes, annual motor vehicletaxes (SWP3D), import duties, and custom fees, but not in-come and personal taxes (BPS, 1996).

We control for the presence of a large crony firm in eachindustry by including the five-digit crony dummy compiled

Page 5: Corruption, Manufacturing Plant Growth, and the Asian Paradox: Indonesian Evidence

CORRUPTION, MANUFACTURING PLANT GROWTH, AND THE ASIAN PARADOX 697

by Basri (2001), on the basis of qualitative data. A five-digitsector contains a crony if at least one of the establishmentsin that sector is well connected to Suharto’s family, that is,“businesspersons who incorporate Soeharto’s family into theirbusiness. This relationship could be in terms of a share, a po-sition, or setting up business together” (Basri, 2001, p. 17).The presence of a large crony firm in the industry may be det-rimental to other firms, because the crony firm captures all theavailable advantages, or beneficial through positive externali-ties such as the provision of infrastructure capabilities orfavorable policies for the sector as a whole.

We also include three plant-level control variables to ac-count for the size, type of ownership, and ratio of white collarto total workers’ wages. The size of plants, measured as the to-tal number of employees, should influence output and laborproductivity growth. That is, we expect larger plants to enjoylower output growth, based on the law of diminishing returnsto scale and the mean reversion hypothesis (De Wit, 2005).

The type of ownership might affect growth, in that foreignownership in particular alters the performance of plants indeveloping economies. Sjoholm (1999a), Sjoholm (1999b) con-firms this effect for Indonesian manufacturing plants. Wetherefore create a dummy variable for foreign ownership thatequals 1 if the plant’s capital share owned by foreigners is 50%or more. Considering the fixed effect set up of the model, thecoefficient on this dummy variable indicates whether a changethat leads to foreign ownership majority affects growth. Final-ly, the share of white collar wages to total wages may reflectthree overlapping components: managerial/support servicesintensity, overall workforce quality, and resources devotedto dealing with bureaucrats. We reasonably expect that mana-gerial/support services intensity and overall workforce qualityproduce positive effects on growth; however, the results maybe mixed, in that the third component could either boost orhamper growth, depending on whether resources spent to dealwith bureaucrats produce positive returns or just higher costs.

In the IV procedure, we instrument bribes and indirect taxesusing their industry-location averages, and we add two otherexcluded instruments in order to have two overidentifyingrestrictions. First, following Fisman and Svensson (2007), weuse an indicator of the quality of infrastructures. It is definedas the ratio of electricity sold over total electricity purchasedand produced (in KwH), which is not correlated with eitheroutput or labor productivity growth. In Indonesia, electricityis distributed by a state monopoly, PLN, and due to inefficien-cies and underinvestment, distribution is very uneven both in

Table 1. Summa

19

Output growth Mean 0std. 0

Labor productivity growth Mean 0std. 0

Bribes (ratio to value added) Mean 0std. 0

Indirect taxes (ratio to value added) Mean 0std. 0

Number of employees (log) Mean 4std. 1

White collar wages (ratio to total wages) Mean 0std. 0

Observations 22

Note: std. indicates standard deviation.

terms of quantity and quality. In order to dampen the effectsof a patchy electricity supply, establishments took on produc-ing their own electricity, partly for self-consumption, partlyfor resale, mostly to PLN itself (Kristov, 1995).

Second, we use an interaction term between one of the ex-cluded instruments and a control variable which Nichols(2007, p. 12) suggests as a valid procedure for addingexcluded instruments. 9 We take the interaction between theshare of electricity sold and the share of white collar workers’wages.

5. ARE BRIBES DETRIMENTAL TO PLANT GROWTH?

(a) Summary statistics

The Raw Statistik Industri data set counts 277,663 plant-year observations for the period 1975–95. 10 If we excludeplants that indicate bribe or indirect tax rates greater than100% or less than 0%, 261,872 observations remain. Finally,removing observations with missing data about output or la-bor productivity growth, number of workers, share of whitecollar wages, ownership type, or the sector crony dummy re-duces the number of observations to 220,867. Table 1 presentsthe summary statistics for these data.

As we show in Table 1, plants’ average output and laborproductivity growth rates both have been increasing over thestudy period. The trend and magnitude of these growth ratesare in line with results provided by previous studies (Berry,Rodriguez, & Sandee, 2001; Timmer, 1999). Output and laborproductivity growth rates appear slow during the oil boomera, then take off after the mid-1980s, or the period that corre-sponds with the liberalization of the economy and an invest-ment boom.

Specifically, during the oil boom period (1975–80), therespondent plants declared average bribe payments of 1.1%of their value added. This rate decreased during subsequentperiods, to averages of 0.9% and 0.6% in 1981–89 and1990–95, respectively. We attribute the drop in the bribe-to-value-added ratio to a significant increase in the number ofplants that report positive bribe payments, from 4,829 in1975 to 17,451 plants in 1995. This increase is mainly due tothe growing size of the manufacturing sector, which dilutedthe bribe remuneration system overall. To a lesser extent, itsuggests an increase in the share of plants that reported yearlybribes, from 78% in the first period to 83% in the last period.

ry statistics

All plants

75–95 1975–80 1981–89 1990–95

.057 0.011 0.073 0.062

.670 0.644 0.684 0.667

.038 0.012 0.039 0.047

.795 0.774 0.827 0.773

.008 0.011 0.009 0.006

.025 0.029 0.028 0.019

.040 0.054 0.045 0.031

.093 0.110 0.096 0.079

.117 3.879 4.042 4.275

.137 1.021 1.102 1.188

.218 0.201 0.219 0.222

.194 0.193 0.195 0.1940,867 38,328 84,788 97,751

Page 6: Corruption, Manufacturing Plant Growth, and the Asian Paradox: Indonesian Evidence

698 WORLD DEVELOPMENT

This dominant share of plants reporting positive bribe pay-ments throughout the period attests to the magnitude of thephenomenon and its widespread institutionalization. Finally,fast output and value added growth tends, mechanically, to re-duce the bribe-to-value-added ratio.

Plants declared that they paid many more indirect taxesthan bribes, and the average plant indirect taxes-to-value-added ratio reached 5.4% in the first period, 4.5% in the sec-ond, and 3.1% in the last period. We attribute this drop tothe same causes we noted for the drop in bribe rates; plantsreporting positive indirect tax payments constituted 85% ofall plants in the first period but 89% in the two subsequentperiods.

(b) Correlations

In Table 2, we present, for each pair of variables, the overalland within-correlation coefficients over the entire period andnote two critical findings. First, bribe and indirect tax ratesboth correlate positively with output growth, as well as withlabor productivity growth. Second, the bribe and indirect taxrates correlate positively and significantly with each other.Both results depart from those reported for Ugandan firms(Fisman & Svensson, 2007), for which the bribe and tax (directand indirect) rates both correlate negatively with outputgrowth and negatively with each other. However, our resultsare consistent with those of Kuncoro (2004), who explains thatin Indonesia, bribe and indirect tax payments share the samedeterminant, namely, firms’ profitability and civil servants’willingness to extract rents.

Also, appearing in a crony sector, as defined by Basri’s(2001) dummy variable, correlates positively to the bribe pay-ment ratio (share of value added) and negatively to the indi-rect taxes ratio during the first two subperiods. In the lastperiod, after liberalization, plants in crony sectors pay moreindirect taxes, and belonging to a crony sector has no influenceon bribe payments. Thus, belonging to a crony sector has apositive correlation with output and productivity growth onlyduring the oil boom period.

(c) Test for sample bias

We consider the potential for a sample selection bias that ex-cludes plants that never declare gifts, charities, or donations(our bribe proxy). Because we do not know whether they truly

Table 2. Correlation matrix (o

Outputgrowth

Labor productivitygrowth

Labor productivity growth Overall 0.6145*

Within 0.6124*

Bribes rate Overall 0.0397* 0.112*

Within 0.0523* 0.1320*

Indirect taxes rate Overall 0.0273* 0.0706*

Within 0.0354* 0.1017*

White collar wage ratio Overall 0.0049 �0.0030Within 0.0008 0.0019

Crony (dummy) Overall �0.0050 �0.0013Number of employees (log) Overall 0.0168* 0.0207*

Within 0.1886* 0.0526*

Foreign ownership dummy Overall 0.0152* 0.0022Within 0.0112* 0.0011

Note: n.a. means not applicable.* Significant at the 5% level.

do not pay bribes or simply are unwilling to answer, we distin-guish two groups of plants: those that never report positivebribes over their entire lifetime (8,499 observations) versusthose that report bribes at least once in their lifetimes(212,368 observations). Following Svensson (2003), we createa group dummy that equals 1 if the plants report positivebribes at least once and 0 otherwise. We regress this groupdummy on several plant characteristics, such as plant size(log of number of workers), indirect tax rate, and outputand labor productivity growth, while controlling for time withyear dummies. The results in Table 3 show that plants thatnever report any bribes are smaller but otherwise not signifi-cantly different than other plants in terms of their indirecttax rate or output and labor productivity growth.

Plants that never report positive bribes only represent 3.85%of observations, and they tend to be relatively smaller. How-ever, density graphs of the plant size distribution in Figure 1show that they still follow distributions fairly similar to thoseassociated with plants that report at least one bribe.

We repeat the analysis with two alternative groups: thosereporting no bribes and those reporting positive bribes. Thistime, the group dummy equals 0 if a plant reports 0 bribesin year t, even if it reports positive bribes in another year. Thisapproach extends the number of observations that appear inthe potentially underreporting group. We obtain similar re-sults in both cases though (see Table 3, second column). Thus,we confirm that plants that report zero bribes are on averagesmaller, but this difference does not bias the sample, becausethey do not differ systematically on other characteristics, andthe plant size distributions remain fairly similar.

Our data include 41 districts (kabupaten) and 371 five-digitindustries, with a total of 6,203 industry-location clusters.We count 590 industry-locations with only one observation,1,406 industry-locations with two to five observations, 1,036industry-locations with 6–10 observations, 2,085 industry-locations with 11–50 observations, and 1,086 industry-loca-tions with more than 50 observations. The industry-locationclusters are numerous enough to offer great variation acrossall observations while also accounting for the industry-loca-tion effect. To increase the exogeneity of the instrument, wecompute the yearly bribe and indirect tax industry-locationaverages but exclude the value declared by the observed plant.We also remove all observations that belong to an industry-location cluster with fewer than three observations. Therefore,at a minimum, the industry-location average of bribes or indi-

verall and within), 1975–95

Bribes rate Indirect taxesrate

White collarwage ratio

Crony(dummy)

Number ofemployees

0.0803*

0.0852*

0.0134* 0.0504*

0.0065* 0.0134*

�0.0022 �0.0139* 0.0457*

0.0244* 0.1318 0.0325* 0.0366*

0.0387* 0.031* 0.0113* n.a.�0.091* 0.0652* 0.1480* �0.0253* 0.2155*

0.0038 0.0121* 0.0033 n.a. 0.0145*

Page 7: Corruption, Manufacturing Plant Growth, and the Asian Paradox: Indonesian Evidence

0.2

.4.6

.8kd

ensi

ty lo

g_nu

mbe

r_w

orke

rs

0 5 10 15x

bribe no bribe

Figure 1. Plant size density distributions: plants reporting bribes versus plants reporting no bribes.

Table 3. Comparison of plants/observations reporting bribes versus plants/observations reporting zero bribes

Dependent variable[number of observations]

Plants reportingbribes at least once

Observationsreporting bribes

Number of workers (log) 0.034 0.015[220,867] (0.020) (0.021)

[0.099] [0.459]Indirect taxes (ratio to value added) 0.001 0.002[220,867] (0.001) (0.001)

[0.628] [0.292]Output growth �0.002 �0.005[220,867] (0.009) (0.010)

[0.806] [0.579]Labor productivity growth 0.003 0.049[220,867] (0.009) (0.009)

[0.791] [0.000]

Notes: The second and third columns provide the coefficient estimates from OLS regressions on the variable dummy, which takes a value of 1 if positivecorruption data are reported and 0 otherwise; standard errors are in parentheses, and p-values are in brackets.

CORRUPTION, MANUFACTURING PLANT GROWTH, AND THE ASIAN PARADOX 699

rect taxes for a given plant equals the average of at least twoother plants in the industry-location cluster. We retain205,424 observations. After dropping any singletons—plantsfor which there is only one year of observation—we consider199,958 observations. With these last two steps, we can per-form industry-location, cluster-robust, standard error estima-tions.

(d) Estimation results

We run our model for the whole period and for the three dif-ferent subperiods to capture the evolution of the different ef-fects of the plants’ bribe and indirect tax rates, as well as thecrony sector dummy on output and productivity growth.However, as we noted previously, results estimated with moreobservations are more generalizable, and for our data, the sub-period with the most observations is the last period, 1990–95.Because we work with population rather than sample data,shifts in the coefficient size and significance across periodsshould indicate a shift in the effect of the explanatory vari-ables.

In Table 4, we present the results of OLS panel dataregressions with fixed effect. Over the entire period, as wellas in the three subperiods, both the bribe- and indirecttax-to-value-added ratios have strongly significant impactson both output and labor productivity growth, which sug-gests that the returns on investments in bribes and indirecttaxes are both positive. That is, companies receive the ser-vices they have paid for, and the removal of red tape andaccess to resources help them grow faster than companiesthat pay less or nothing in the form of bribes. Companiesthat pay larger bribes relative to the value added grow insize (output) and also manage to improve the productivityof their labor. However, operating in a crony sector has asignificant positive effect only on labor productivity growthduring the deregulation period.

The size of plants always reveals a significantly negative im-pact on output growth, consistent with the law of diminishingreturns. However, it has in every case a significantly positiveeffect on labor productivity growth, because larger plants tendto be more capital intensive and thus contribute positively tolabor productivity.

Page 8: Corruption, Manufacturing Plant Growth, and the Asian Paradox: Indonesian Evidence

700 WORLD DEVELOPMENT

Overall, the ratio of white collar wages to total wages doesnot have a significant effect on either output or labor produc-tivity growth, except for the 1975–80 period. During the oil

Table 5. IV estimates of bribes and indirect taxes effect on o

Variable Dependent variable = output growth

1975–95 1975–80 1981–89

Bribe payment (ratio to value added) 3.645*** �6.077 3.453

(0.008) (0.301) (0.108)

Indirect taxes (ratio to value added) 0.495** �0.177 0.893*

(0.012) (0.789) (0.052)

Number of employees (log) �0.288*** �0.365*** �0.408***

(0.000) (0.000) (0.000)

White collar wages (ratio to total

wages)

�0.020 0.203*** �0.0347

(0.187) (0.000) (0.218)

Foreign ownership �0.080*** 0.002 �0.0778*

(0.000) (0.975) (0.060)

Crony (dummy) 0.009 – 0.445

(0.234) – (0.433)

Year dummies Not reported Not reported Not reported

Constant 1.289*** 1.459*** 1.700***

(0.000) (0.000) (0.000)

Observations 199,958 34,945 78,784

Anderson–Rubin F-statistic (test of

joint significance of endogenous

regressors)

36.08 589.08 2.14

(0.000) (0.000) (0.073)

Kleibergen–Paap LM-statistic

(underidentification test)

33.79 11.82 49.63

(0.000) (0.008) (0.000)

Kleibergen–Paap rk Wald F-statistic

(weak identification test)a

196.24 26.13 8.83

Hansen J test of 2 overidentifying

restrictions, p-value

0.476 0.256 0.180

1st-stage partial R2/Shea’s partial R2

for bribes

0.003/0.003 0.001/0.001 0.003/0.003

1st-stage partial R2/Shea’s partial R2

for indirect taxes

0.009/0.009 0.005/0.005 0.005/0.005

Notes: p-values are in brackets. Standard errors are industry-location cluster-ra Stock Yogo (2005) weak identification test critical values are respectively 11.* Significant at 10% level.** Significant at 5% level.*** Significant at 1% level.

Table 4. OLS estimates of bribes and indirect taxes effect on

Variable Dependent variable = output growt

1975–95 1975–80 1981–89

Bribe payment(ratio to value added)

1.212*** 0.767*** 1.287***

(0.000) (0.000) (0.000)Indirect taxes(ratio to value added)

0.254*** 0.261*** 0.333***

(0.000) (0.000) (0.000)Number of employees (log) �0.329*** �0.386*** �0.471***

(0.000) (0.000) (0.000)White collar wages(ratio to total wages)

�0.0288 0.210*** �0.0330(0.112) (0.000) (0.367)

Foreign ownership �0.0923*** 0.0264 �0.0811*

(0.000) (0.643) (0.089)CRONY (dummy) 0.00912 – 0.802

(0.319) – (0.277)Year dummies Not reported Not reported Not reported NConstant 1.407*** 1.410*** 1.856***

(0.000) (0.000) (0.000)Observations 199,958 34,945 78,784Number of plants 27,246 7,676 17,490F-statistic 132.69 42.43 89.85

(0.000) (0.000) (0.000)

Note: p-values are in brackets. Standard errors are industry-location cluster-ro* Significant at 10% level.*** Significant at 1% level.

boom, managerial/support services intensity apparently pro-duced higher output and labor productivity growth rates.

utput growth and labor productivity growth, fixed effect

Dependent variable = labor productivity growth

1990–95 1975–95 1975–80 1981–89 1990–95

7.062* 6.113*** 6.678 6.015*** 7.772*

(0.066) (0.000) (0.322) (0.001) (0.071)

1.824** 1.255*** 1.457 1.946*** 4.046***

(0.015) (0.000) (0.112) (0.005) (0.000)

�0.428*** 0.100*** 0.302*** 0.142*** 0.146***

(0.000) (0.000) (0.000) (0.000) (0.000)

�0.056* �0.007 0.129** �0.021 �0.044

(0.058) (0.673) (0.011) (0.546) (0.211)

�0.167*** �0.048 �0.083 �0.084 �0.082

(0.000) (0.051) (0.178) (0.119) (0.104)

0.002 0.008 – 0.546 0.022

(0.934) (0.369) – (0.195) (0.490)

Not reported Not reported Not reported Not reported Not reported

1.806*** �0.309*** �1.257*** �0.559*** �0.735***

(0.000) (0.000) (0.000) (0.000) (0.000)

86,229 199,958 34,945 78,784 86,229

5.28 14.72 40.82 5.38 8.19

(0.000) (0.000) (0.000) (0.000) (0.000)

14.91 33.79 11.82 49.63 14.91

(0.002) (0.000) (0.008) (0.000) (0.002)

3.94 196.24 26.13 8.83 3.94

0.416 0.203 0.319 0.148 0.329

0.003/0.003 0.003/0.003 0.001/0.001 0.003/0.003 0.003/0.003

0.003/0.003 0.009/0.009 0.005/0.005 0.005/0.005 0.003/0.003

obust.04 and 7.56 for 5% and 10% maximal IV relative bias.

output growth and labor productivity growth, fixed effect

h Dependent variable = labor productivity growth

1990–95 1975–95 1975–80 1981–89 1990–95

1.742*** 4.252*** 4.278*** 4.510*** 5.572***

(0.000) (0.000) (0.000) (0.000) (0.000)0.353*** 1.056*** 1.441*** 1.351*** 1.279***

(0.000) (0.000) (0.000) (0.000) (0.000)�0.508*** 0.133*** 0.390*** 0.176*** 0.177***

(0.000) (0.000) (0.000) (0.000) (0.000)�0.0640* 0.00166 0.171*** �0.00856 �0.0212(0.077) (0.938) (0.010) (0.843) (0.625)�0.166*** �0.0482 �0.0978 �0.0595 �0.0606

(0.003) (0.121) (0.211) (0.393) (0.336)�0.00192 0.00721 – 0.772* 0.0131

(0.956) (0.475) – (0.079) (0.745)ot reported Not reported Not reported Not reported Not reported2.211*** �0.591*** �1.626*** �0.845*** �0.802***

(0.000) (0.000) (0.000) (0.000) (0.000)86,229 199,958 34,945 78,784 86,22919,694 27,246 7,676 17,490 19,694121.64 93.09 51.60 91.84 51.01(0.000) (0.000) (0.000) (0.000) (0.000)

bust.

Page 9: Corruption, Manufacturing Plant Growth, and the Asian Paradox: Indonesian Evidence

CORRUPTION, MANUFACTURING PLANT GROWTH, AND THE ASIAN PARADOX 701

A move toward foreign ownership does not significantly af-fect labor productivity growth. However, plants for which for-eign ownership becomes prevalent grow more slowly in termsof their output, especially in the last period. This provides astrong signal that during this period, corruption became apowerful means to compete in the market that even joint ven-tures could not supersede.

The results of instrumental variable (IV) regressions, inwhich we instrument for bribes and indirect taxes using theindustry-location average technique in Table 5, confirm andreinforce the previous results, according to the significantand large coefficients on instrumented variables, except for1975–80. During the oil boom, bribe rates influenced neitheroutput nor labor productivity growth rates, but indirect taxrates had a positive effect, significant at a 11% level, only onlabor productivity growth rates. This period also was charac-terized by very slow growth rates, so the results should notseem surprising. However, managerial/support services inten-sity plays a significant and positive role for output and laborproductivity growth.

As a robustness check, we remove potential severe outliersfrom the plant population, that is, all observations with bribeand/or tax rates in the top 1% of their respective yearly distri-butions, which equates to average bribe rates greater than8.8% and indirect tax rates greater than 49.2%. We also re-move observations with output and/or labor productivitygrowth rates in the top and bottom 1% of their respectiveyearly distributions. This procedure removes approximately2.03% of the total observations. We further note that all theoutliers we removed because of their extreme values in thebribe and/or indirect tax rates also display extreme values interms of output and/or labor productivity growth. Thus, wefind a strong correlation between bribe rates and growth ratesand between indirect tax rates and growth rates. When we ex-clude extreme, though not necessarily untrustworthy, values,our results should be less significant. There is no theoretical

Table 6. OLS estimates of bribes and indirect taxes effect on output g

Variable Dependent variable = output grow

1975–95 1975–80 1981–89

Bribe payment (ratio to valueadded)

3.621*** 1.760*** 3.954***

(0.000) (0.000) (0.000)Indirect taxes (ratio to valueadded)

0.311*** 0.316*** 0.422***

(0.000) (0.000) (0.000)Number of employees (log) �0.325*** �0.382*** �0.466***

(0.000) (0.000) (0.000)White collar wages (ratio tototal wages)

�0.037** 0.207*** �0.042(0.044) (0.000) (0.265)

Foreign ownership �0.084*** 0.030 �0.062(0.001) (0.609) (0.186)

Crony (dummy) 0.009 – 0.855(0.315) – (0.323)

Year dummies Not reported Not reported Not reported NConstant 1.366*** 1.387*** 1.880***

(0.000) (0.000) (0.000)Observations 195,890 33,579 75,913Number of plants 27,074 7,629 17,413F-statistic 132.46 41.63 88.42

(0.000) (0.000) (0.000)

Note: p-values are in brackets. Standard errors are industry-location cluster-ro* Significant at 10% level.** Significant at 5% level.*** Significant at 1% level.

reason to remove these observations, but this procedure pro-vides a good robustness check of our previous results.

Results for this reduced population, as we show in Tables 6and 7, confirm the magnitude and significance of the coeffi-cients for bribe and indirect tax rates except for the coefficienton bribe rate over the deregulation period. We note that this isalso the case for the other explanatory variables. The results inTables 6 and 7 thus reinforce our previous findings.

In Tables 5 and 7, we also present a series of tests of the rel-evance of the instruments. The Hansen J-test of overidentify-ing restrictions has p-values that always lead us to considerour instruments as valid.

In order to conduct the weak identification test, in the pres-ence of several endogenous regressors (see Stock & Yogo,2005) and as standard errors are industry-location cluster-ro-bust (see Baum, Schaffer, & Stillman 2007), we use the Klei-bergen–Paap Wald F-statistic. In all cases except one, it issuperior to the Stock and Yogo’s (2005) critical values at 5%and 10% of maximal IV relative bias (two endogenous regres-sors and four excluded instruments), and we reject the nullhypothesis that our excluded instruments are weak. For theperiod 1990-1995, in presence of outliers, the low value ofthe Kleibergen–Paap rk Wald F-statistic suggests a potentialweak-instrument bias. However, once we exclude outliers thatrepresent potential measurement errors (Table 7), we reject thehypothesis of instruments’ weakness.

The relevance of instruments is confirmed by the first stagepartial R2 and Shea’s partial R2, which are always equal orvery close.

The Kleibergen–Paap LM-statistic is consistently large en-ough so that we can reject the null hypothesis of underidentif-ication. The Anderson–Rubin F-test has always a p-value thatis sufficiently low, and we fail to reject the joint insignificanceof endogenous regressors (bribes and indirect taxes), exceptfor the output growth regression in 1981-89 when removingoutliers.

rowth and labor productivity growth, fixed effect, outliers excluded

th Dependent variable = labor productivity growth

1990–95 1975–95 1975–80 1981–89 1990–95

9.534*** 10.388*** 8.524*** 11.488*** 21.356***

(0.000) (0.000) (0.000) (0.000) (0.000)0.497*** 1.185*** 1.698*** 1.577*** 1.378***

(0.000) (0.000) (0.000) (0.000) (0.000)�0.503*** 0.140*** 0.390*** 0.191*** 0.184***

(0.000) (0.000) (0.000) (0.000) (0.000)�0.086** �0.023 0.159** �0.041 �0.053(0.021) (0.281) (0.016) (0.348) (0.231)�0.168*** �0.048 �0.113 0.058 �0.069

(0.003) (0.129) (0.159) (0.422) (0.272)�0.003 0.007 – 0.968* 0.005(0.931) (0.481) – (0.054) (0.911)

ot reported Not reported Not reported Not reported Not reported2.137*** �0.676*** �1.649*** �0.858*** �0.899***

(0.000) (0.000) (0.000) (0.000) (0.000)83,477 195,890 33,579 75,913 83,47719,558 27,074 7,629 17,413 19,558135.68 123.15 62.21 109.15 104.94(0.000) (0.000) (0.000) (0.000) (0.000)

bust.

Page 10: Corruption, Manufacturing Plant Growth, and the Asian Paradox: Indonesian Evidence

Table 7. IV estimates of bribes and indirect taxes effect on output growth and labor productivity growth, fixed effect, outliers excluded

Variable Dependent variable = output growth Dependent variable = labor productivity growth

1975–95 1975–80 1981–89 1990–95 1975–95 1975–80 1981–89 1990–95

Bribe payment (ratio to valueadded)

2.655*** �7.057 1.356 6.112* 5.535*** 4.417 5.205*** 7.599*

(0.005) (0.187) (0.306) (0.088) (0.000) (0.457) (0.003) (0.083)Indirect taxes (ratio to valueadded)

0.451*** �0.247 0.750** 1.038 1.153*** 1.283 1.430*** 3.424***

(0.007) (0.684) (0.046) (0.105) (0.000) (0.118) (0.009) (0.001)Number of employees (log) �0.211*** �0.329*** �0.321*** �0.301*** 0.117*** 0.262*** 0.162*** 0.187***

(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)White collar wages (ratio tototal wages)

0.011 0.183*** 0.006 �0.007 0.001 0.132*** 0.005 �0.034(0.375) (0.000) (0.783) (0.751) (0.967) (0.003) (0.877) (0.255)

Foreign ownership �0.023 0.0002 �0.029 �0.060** �0.022 �0.074 �0.014 �0.042(0.147) (0.995) (0.360) (0.036) (0.285) (0.154) (0.757) (0.300)

Crony (dummy) 0.008 – 0.015 �0.011 0.008 – �0.027 0.007(0.255) – (0.967) (0.643) (0.273) – (0.923) (0.792)

Year dummies Not reported Not reported Not reported Not reported Not reported Not reported Not reported Not reportedConstant 0.868*** 1.373*** 1.355*** 1.251*** �0.553*** �1.139*** �0.628*** �0.920***

(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)Observations 195,890 33,579 75,913 83,477 195,890 33,579 75,913 83,477

Anderson–Rubin Wald F-statistic (test of jointsignificance of endogenousregressors)

28.46 652.50 1.59 3.63 16.04 57.78 5.11 7.90(0.000) (0.000) (0.173) (0.006) (0.000) (0.000) (0.000) (0.000)

Kleibergen–Paap LM-

statistic (underidentificationtest)

30.90 9.82 45.79 38.35 30.90 9.82 45.79 38.35(0.000) (0.020) (0.000) (0.000) (0.000) (0.020) (0.000) (0.000)

Kleibergen–Paap Wald F-statistic (weak identificationtest)a

199.63 29.47 9.25 9.92 199.63 29.47 9.25 9.92

Hansen J test of 2overidentifying restrictions,p-value

0.504 0.271 0.502 0.453 0.171 0.332 0.356 0.172

1st-stage partial R2/Shea’spartial R2 for bribes

0.003/0.003 0.001/0.001 0.004/0.004 0.003/0.003 0.003/0.003 0.001/0.001 0.004/0.004 0.003/0.003

1st-stage partial R2/Shea’spartial R2 for indirect taxes

0.009/0.009 0.005/0.005 0.005/0.005 0.003/0.003 0.009/0.009 0.005/0.005 0.005/0.005 0.003/0.003

Notes: p-values are in brackets. Standard errors are industry-location cluster-robust.a Stock Yogo (2005) weak identification test critical values are respectively 11.04 and 7.56 for 5% and 10% maximal IV relative bias.* Significant at 10% level.** Significant at 5% level.*** Significant at 1% level.

Table 8. Contribution of bribes and indirect taxes to output and labor productivity growth

Period Average contribution to output growth (%) Contribution to output growthfor plants in the top 10% of bribesand indirect taxes distribution (%)

Average output growth (%)

Bribe payments Indirect taxes Crony dummy Bribe payments Indirect taxes

1975–80 n.s. n.s. n.s. n.s. n.s. 1.08%1981–89 3.11 3.98 n.s. 6.79 10.88 7.31%1990–95 4.31 1.11 n.s. 9.31 2.40 6.23%1975–95 2.91 2.00 n.s. 6.41 5.47 5.75%

Average contribution to labor productivity growth(%)

Contribution to laborproductivity for plants in the top10% of bribes and indirect taxes

distribution (%)

Average labor productivity growth (%)

Bribe payments Indirect taxes Crony dummy Bribe payments Indirect taxes

1975–80 n.s. 7.93 n.s. n.s. 20.93 1.23%1981–89 5.41 8.68 0.55 11.82 23.72 3.93%1990–95 4.74 12.68 n.s. 10.24 32.77 4.67%1975–95 4.88 5.07 n.s. 10.74 13.88 3.79%

Note: n.s. indicates not significant.

702 WORLD DEVELOPMENT

Page 11: Corruption, Manufacturing Plant Growth, and the Asian Paradox: Indonesian Evidence

CORRUPTION, MANUFACTURING PLANT GROWTH, AND THE ASIAN PARADOX 703

We next calculate the contribution of bribes and indirecttaxes to output and productivity growth. From Table 5, we rec-ognize that the coefficient of the bribe rate using the IV estima-tion for 1975–1995 is 3.645, but what does this value mean interms of additional percentage points of output growth? Theanswer differs for each plant, because each plant differs in itsbribe payments. However, we can estimate how average briberate alters the average output growth. For 1975–1995, the aver-age bribe rate equals 0.80% (Table 1), and when we multiplythat value by the estimated coefficient of the bribe rate, we ob-tain an average contribution of the bribe rate of 2.91 percent-age points—a significant portion of the average plant outputgrowth (5.75%). In Table 8, we list the contributions of theaverage bribe and indirect tax rates, as well as the crony sectordummy, to output and labor productivity growth (column 1),and report the average growth rates (column 3). In column(2), we estimate the contribution of bribes and taxes for plantsthat are in the top 10% of the bribe and tax rates distributions.

The average contribution of bribe payments to outputgrowth increases from 0 in the first period to 4.31 percentagepoints in the last period, and the average contribution of indi-rect taxes reaches a peak during the deregulation period. In thesecond period, the contribution of indirect taxes is greaterthan that of the bribe, and in the third period, bribe contribu-tion takes the lead. Over the entire period, average bribe ratescontribute to about half the average output growth, and aver-age indirect tax rates contribute to less than half of thisgrowth. For plants that pay bribe in the top 10% of the distri-bution, the contribution outperforms the average growth rateby 0.66 percentage points. The contribution of indirect taxrates almost matches the average growth rate.

The contribution of bribes and indirect taxes to labor pro-ductivity growth figures are even more striking. The effectsof indirect taxes increase over the entire period, and the effectsof bribe payments reach a peak during deregulation. That is,in 1975-1995, plants paying average bribe or indirect tax ratesoutperform the average labor productivity growth rate. In thesame period, plants paying bribe or indirect tax rates in the top10% of the distribution multiply the average labor productiv-ity growth rate by 2.8 and 3.7, respectively.

6. CONCLUSION

Using unique panel data of Indonesian manufacturing duringthe Suharto era, we test the effects of corruption on plant output

and labor productivity growth and find a positive relation.Using plant-level panel data estimations and controlling forthe potential endogeneity of bribe payments through theirindustry-location average (Fisman & Svensson, 2007) and aproxy for infrastructure quality, we find that plants displayinga higher bribe-to-value added ratio enjoy significantly higheroutput and productivity growth. The effect of the ratio of indi-rect taxes to value added, another proxy for corruption, is sim-ilar in the scope and significance of its effect on laborproductivity growth, but it is less dramatic in its scope for out-put growth.

These findings stand in clear opposition to the results forAfrican countries (Fisman & Svensson, 2007; McArthur &Teal, 2002), for which corruption always has a negative impacton firm growth, perhaps because of their political instabilityand short or one-off relationships between firms and officials.In contrast, our results support the efficient grease hypothesis,from the perspective of individual plants. Indonesian firmsthat pay bribes to yayasans are better able to overcome redtape and barriers to doing business.

We do not intend to argue, however, that corruption is agood thing for a country, even Indonesia, which suffers fromthe combination of high levels of corruption, relative politicalstability, and fast economic growth. We highlight our distinc-tion between the micro- and macro-level perspectives. Somefirms may benefit individually from corruption at the expenseof others, but the practice remains fundamentally a negativedistortion for the entire economy and a source of unproduc-tive activities (Bhagwati, 1982).

In turn, we imagine at least two avenues for further re-search. A study of the dynamics of the redistribution ofmarket shares across companies could assess the aggregateeffects of micro-corruption. In addition, a counterfactual ap-proach could offer information about the potential outputand productivity growth rates that might have been achievedin a purely competitive, rather than corrupt, system.Researchers might try to compare Indonesian plants withsimilar plants from a less corrupt country in support of suchan analysis.

Yet our results still offer important policy implications:On average, it pays for firms individually to take partin a corrupt system. Programs attempting to fight corrup-tion therefore must take this feature into account if theywant to convince entrepreneurs and managers to give uptheir corrupt practices and establish alternative incentivesystems.

NOTES

1. Bardhan (1997) and Meon and Sekkat (2005) offer detailed presen-tations of the “grease the wheels” assumption and its underpinningmechanisms, as well as the general effects of corruption on efficiency andgrowth.

2. Banerjee (1997) formalizes this point by examining situations in whichbureaucrats create red tape to discriminate across private agents.

3. Ventelou (2002) addresses the idea that in the long run, it is in theinterest of corrupt politicians to moderate their rent extraction and run thecountry more efficiently, as a condition for increasing their collective (i.e.,party, dynasty) reputation and remaining in office. Olson (2000) distin-guishes between stationary and roving bandits.

4. By 1990, wages only covered one-third of an official’s household needs(Robertson-Snape 1999, p. 590).

5. Svensson (2003) assumes that corrupt bureaucrats are rational andmaximize expected profit. He considers a bargaining model in whichhigher expected future profits weaken the firm’s bargaining position,because exiting the market becomes relatively more costly. As a result, thecorrupt bureaucrat can demand more graft.

6. Fisman and Svensson (2007) use a measure of tax payments byUgandan firms that includes both direct and indirect taxes; it comes fromthe World Bank Ugandan Industrial Enterprise Survey (see Svensson,2001).

7. We also use industry-location bribe and tax averages as instrumentalvariables and employ the cluster-robust estimation technique.

8. The data do not offer the means to identify plants that belong tocorporate groups during the period.

Page 12: Corruption, Manufacturing Plant Growth, and the Asian Paradox: Indonesian Evidence

704 WORLD DEVELOPMENT

9. If the equation is exactly identified, and the excluded instruments trulyexogenous, there is always a feasible overidentification test using anaugmented set of excluded instruments. Therefore, adding an interactionterm that is the product of a truly exogenous excluded instruments and acontrol variable, is a better strategy than looking for additional, and likelyweak excluded instruments (Nichols, 2007, p. 12).

10. We use the Raw Statistik data set, compiled since 1975, and updatethe output and labor force figures with the Backcast Statistik Industri,which was compiled in 1996.

REFERENCES

Aswicahyono, H. (1998). Total factor productivity growth in Indonesianmanufacturing. Unpublished doctoral dissertation, Australian NationalUniversity, Canberra.

Bardhan, P. (1997). Corruption and development: A review of issues.Journal of Economic Literature, 35, 1320–1346.

Banerjee, A. (1997). A theory of misgovernance. Quarterly Journal ofEconomics, 62(4), 1289–1332.

Basri, M. C. (2001). The Political economy of manufacturing protection inIndonesia, 1975–1995. Unpublished doctoral thesis, Australian Na-tional University, Canberra.

Baum, C., Schaffer, M., & Stillman, S. (2007). Enhanced routines forinstrumental variables/GMM estimation and testing. Boston CollegeWorking Paper No. 667.

Behrman, J. R., & Deolalikar, A. B. (1989). Of the fittest? Duration ofsurvival of manufacturing establishments in a developing country.Journal of Industrial Economics, 38(2), 215–226.

Berry, A., Rodriguez, E., & Sandee, H. (2001). Small and mediumenterprise dynamics in Indonesia. Bulletin of Indonesian EconomicStudies, 37(3), 363–384.

Bhagwati, J. (1982). Directly unproductive, profit-seeking (DUP) activi-ties. Journal of Political Economy, 90(5), 988–1002.

Biro Pusat Statistik (1996). Annual Manufacturing Survey Questionnaire,1996. Available at : <http://www.rand.org/labor/bps.data/data-docpdf/survei%20industri/si96.pdf>.

Claessens, S., Djankov, S., Fan, J. P., & Lang, L. H. (1998a). Expropri-ation of minority shareholders: Evidence from East Asia. World Bank,Working Paper No. 2088.

Claessens, S., Djankov, S., Fan, J. P., & Lang, L. H. (1998b). East Asiancorporates: Growth, financing, and risks over the last decade.Malaysian Journal of Economics, 35(1–2), 137–156.

Claessens, S., Djankov, S., Fan, J. P., & Lang, L. H. (1998c). Diversi-fication and efficiency of investment by East Asian corporations.World Bank, Working Paper No. 2033.

Claessens, S., Djankov, S., Fan, J. P., & Lang, L. H. (1998d). Corporatediversification in East Asia: The role of ultimate ownership and groupaffiliation. World Bank, Working Paper No. 2089.

Claessens, S., Djankov, S., & Lang, L. H. (1998). Who controls East Asiancorporations. World Bank, Working Paper No. 2054.

De Wit, G. (2005). Firm size distributions: An overview of steady-statedistributions resulting from firm dynamics models. InternationalJournal of Industrial Organization, 23, 423–450.

Dreher, A., & Gassebner, M. (2007). Greasing the wheels of entrepre-neurship? The impact of regulations and corruption on firm entry.Cesifo Working Paper No. 2013.

The Economist (1993). Now for the hard part, April 17, Vol. 327(7807).Fisman, R., & Gati, R. (2006). Bargaining for bribes: The role of

institutions. In S. Rose-Ackerman (Ed.), International handbook on theeconomics of corruption (pp. 127–139). Cheltenham: Edward Elgar.

Fisman, R., & Svensson, J. (2007). Are corruption and taxation reallyharmful to growth?. Journal of Development Economics, 83, 63–75.

Henderson, J. V., & Kuncoro, A. (2006). Corruption in Indonesia.Unpublished manuscript.

Huntington, S. P. (1968). Political order in changing societies. New Haven,CT: Yale University Press.

Kaufmann, D. & Wei, S. J. (1999). Does grease money speed up the wheelsof commerce? NBER Working Paper No. 7093.

Kristov, L. (1995). The price of electricity in Indonesia. Bulletin ofIndonesian Economic Studies, 31(3), 73–101.

Krugman, P. (1994). The myth of Asia’s miracle. Foreign Affairs, 73(6),62–78.

Kuncoro, A. (2004). Bribery in Indonesia: Some evidences from micro-level data. Bulletin of Indonesian Economic Studies, 40(3), 329–354.

Leff, N. (1964). Economic development through bureaucratic corruption.The American Behavioural Scientist, 8(2), 8–14.

Leys, C. (1965). What is the problem about corruption? Journal of ModernAfrican Studies, 3, 215–230, Reprinted in A.J. Heidenheimer, M.Johnston, & V.T. LeVine (Eds.) (1989), Political corruption: Ahandbook (pp. 51–66). Oxford: Transaction Books.

Lui, F. (1985). An equilibrium queuing model of bribery. Journal ofPolitical Economy, 93(4), 760–781.

Mauro, P. (1995). Corruption and growth. Quarterly Journal of Econom-ics, 110(3), 681–712.

McArthur, J., & Teal, F. (2002). Corruption and firm performance inAfrica. CSAE working paper, WPS/2002.10, Oxford University.

McLeod, R. (2000). Soeharto Indonesia: A better class of corruption. TheIndonesian Quarterly, 28(1), 6–27.

Meon, P.-G., & Sekkat, K. (2005). Does corruption grease or sand thewheels of growth?. Public Choice, 122(1–2), 69–97.

Meon, P.-G., & Weill, L. (forthcoming). Is corruption an efficient grease?World Development.

Nichols, A. (2007). Causal inference with observational data. StataJournal, 7(4), 507–541.

Olson, M. (2000). Power and prosperity: Outgrowing communist andcapitalist Dictatorships. New York: Basic Books.

Pomerleano, M. (1999). The East Asia crisis and corporate finance: Theuntold micro story. World Bank, Policy Research Working PaperSeries no. 1990.

Prowse, S. (1998). Corporate governance. Emerging issues and lessons fromEast Asia. World Bank, mimeo.

Rajan, R. G., & Zingales, L. (1998). Financial dependence and growth.American Economic Review, 88(3), 559–586.

Robertson-Snape, F. (1999). Corruption, collusion and nepotism inIndonesia. Third World Quarterly, 20(3), 589–602.

Robison, R. (1986). Indonesia: The rise of capital. London: Allen andUnwin.

Rock, M., & Bonnett, H. (2004). The comparative politics of corruption:Accounting for the East Asian paradox in empirical studies of cor-ruption, growth and investment. World Development, 32(6), 999–1017.

Rose-Ackerman, S. (1999). Corruption and government: Causes, conse-quences, and reform. Cambridge: Cambridge University Press.

Schwarz, A. (2004). A nation in waiting: Indonesia’s search for stability.Singapore Talisman Publishing.

Shleifer, A., & Vishny, R. W. (1993). Corruption. Quarterly Journal ofEconomics, 108, 599–617.

Sjoholm, F. (1999a). Exports, imports and productivity: Results fromIndonesian establishment data. World Development, 27(4), 705–715.

Sjoholm, F. (1999b). Technology gap, competition and spillovers fromdirect foreign investment: Evidence from establishment data. Journalof Development Studies, 36(1), 53–73.

Stock, J. H., & Yogo, M. (2005). Testing for weak instruments in linear IVregression. In D. W. Andrews, & J. H. Stock (Eds.), Identification andinference for econometric models: Essays in honor of Thomas Rothen-berg (pp. 80–108). Cambridge: Cambridge University Press.

Svensson, J. (2001). The cost of doing business: Firms’ experience withcorruption. In R. Reinikka, & P. Pollier (Eds.), Uganda’s Recovery:The Role of Farms, Firms and Government (pp. 319–342). WashingtonDC: World Bank.

Svensson, J. (2003). Who must pay bribes and how much. QuarterlyJournal of Economics, 118(1), 207–230.

Timmer, M. (1999). Indonesia’s ascent on the technology ladder: Capitalstock and total factor productivity in Indonesian manufacturing,1975–1995. Bulletin of Indonesian Economic Studies, 35(1), 75–97.

Ventelou, B. (2002). Corruption in a model of growth: Political reputa-tion, competition and shocks. Public Choice, 110(1–2), 23–40.

Page 13: Corruption, Manufacturing Plant Growth, and the Asian Paradox: Indonesian Evidence

CORRUPTION, MANUFACTURING PLANT GROWTH, AND THE ASIAN PARADOX 705

Vial, V. (2006). New estimates of total factor productivity growth inIndonesian manufacturing. Bulletin of Indonesian Economic Studies,42(3), 353–365.

Wade, R. (1990). Governing the market. Princeton, NJ: PrincetonUniversity Press.

Wei, S. (2000). How taxing is corruption on international investors.Review of Economics and Statistics, 82(1), 1–11.

World Bank (2000). World Bank economic survey. Available at: <http://info.worldbank.org/governance/wbes/>.

Available online at www.sciencedirect.com