Corruption, Economic Growth, and Income Inequality: Evidence from Ten Countries in Asia

Transcription

1 Corrupton, Economc Growth, and Income Inequalty: Evdence from Ten Countres n Asa Chung-Ju Huang Abstract Ths study utlzes the panel vector error correcton model (PVECM) to examne the relatonshp among corrupton, economc growth, and ncome nequalty experenced wthn ten Asan countres over the 1995 to 2010 perod. Accordng to the emprcal results, we do not support the common percepton that corrupton decreases economc growth. On the contrary, we found that corrupton ncreases economc growth. Meanwhle, an ncrease n economc growth wll cause an ncrease n ncome nequalty, although the effect s nsgnfcant. Smlarly, an ncrease n ncome nequalty wll cause an ncrease n economc growth but a decrease n corrupton, although the effect s also nsgnfcant. Keywords Corrupton, economc growth, ncome nequalty, panel vector error correcton model C I. INTRODUCTION ORRUPTION s a global commonalty. The corrupton percepton ndex (CPI) has been publshed annually by Transparency Internatonal (TI) snce 1995 and has been wdely credted wth puttng the ssue of corrupton on the nternatonal polcy agenda. The CPI ranks almost 200 countres on a scale of zero to 10, wth zero ndcatng hgh levels of corrupton and 10 ndcatng low levels. The CPI generally defnes corrupton as the msuse of publc power for prvate beneft. In the 2011 CPI, Tawan ranks 32 nd wth a score of 6.1; comparng to the Asan countres, behnd Sngapore (rankng 5 th wth a score of 9.2), Hong Kong (rankng 12 th wth a score of 8.4) and Japan (rankng 14th wth a score of 8.0), but ahead of South Korea, Malaysa, Chna, Thaland, Indonesa, Vetnam, and Phlppnes. The relatonshp between corrupton and economc growth has been a popular topc and has been examned n several emprcal studes. Jan [1] argues that three types of corrupton phenomena mght occur n a democratc naton. The three types of corrupton are grand corrupton nvolvng corrupton among hgh level executves n government, legslatve corrupton nvolvng corrupton among representatves of the general publc, and bureaucratc corrupton nvolvng corrupton among government offcals and staff. No matter what type of corrupton, corrupton hurts economc development and causes resources msallocaton and economc neffcences. Corrupton mght decrease a country's compettveness, cause a decrease n economc growth, crowdng out government spendng n educaton and health, and an ncrease ncome nequalty, and dstort the market mechansm and resource allocaton [2], [3]. Chung-Ju Huang s wth the Department of Publc Fnance, Feng Cha Unversty, Tachung, Tawan. (e-mal: cjhuang@fcu.edu.tw) The author s grateful to the Natonal Scence Councl, Tawan for fnancal support through project no. NSC H Mauro s cross-country emprcal study shows that severe corrupton sgnfcantly deters nvestment and economc growth [4]. Brunett et al. [5] and Brunett and Weder [6] fnd that the mpact of corrupton on nvestment s negatve. Monte and Papagn [7] fnds that corrupton n Italy muncpalty not only drectly lmts the average labor ncome, but also decreases prvate nvestments, whch n turn, decreases the effcency of publc nvestment expendtures and slows down economc growth. Neeman et al. [8] suggests that corrupton has a negatve mpact on a country s economc growth whch s determned by the openness of the specfc country. Svensson [9] re-vst of Mauro s study [4] smlarly supports the fndng that the mpact of corrupton on economc growth s negatve. Gymah-Brempong and de Camacho [10] shows that corrupton has a negatve mpact on economc growth and there are sgnfcant countres-specfc effects. Corrupton s negatve mpact s most sgnfcant among Afrcan countres and least sgnfcant among Asan and OECD countres. Although most of the studes support that the mpact of corrupton on economc growth s negatve, some scholars beleve that the mpact of corrupton on economc growth s postve because corrupton contrbutes to economc development on the grounds that bureaucratc corrupton can mprove the admnstratve effcency of bureaucracy and reduce transacton tme cost. Leff [11], Bayley [12], and Huntngton [13] suggest that n some cases, brbery of certan decson-makers can reduce the ncompleteness of laws and regulatons and admnstratve rgd adverse effects to promote economc effcency. Lu [14] suggests that corrupton may smplfy the admnstratve procedures. Kltgaard [15] and Acemoglu and Verder [16] use the theoretcal model to demonstrate that consderng the cost of combatng corrupton, under the condton of a country's output maxmzaton, the optmal level of corrupton may be low, but t does exst. Colombatto [17] fnds that n some developng countres, corrupton has a postve mpact on economc growth. Tresman s cross-country emprcal study shows that corrupton has not sgnfcant mpact on economc growth [18]. As to the relatonshp between corrupton and ncome nequalty, Gupta et al. [19], L et al. [20], Hendrks and Muthoo [21], Jan [1], and Johnston [22] thnk corrupton wll ncrease the level of ncome nequalty. Corrupton has changed the dstrbuton of socal welfare spendng and wll beneft the rch people [23], [24]. A large number of emprcal studes have attempted to explore the relatonshp between ncome nequalty and economc growth, such as Persson and Tabelln [25], Psacharopoulose et al. [26], Barror [27], Janvry and Sadoulet [28], Alfranca et al. [29], Jomo [30], Rcardo [31], Samanta and Heyse [32] etc. Whle most studes explore how OECD countres, European countres, Latn Amerca, or Amercas have experenced rapd economc growth accompaned wth ncreasng economc nequalty, there are few studes that focus on Asan countres. Ths study focuses on the Asan countres. 1141

2 Over the years of 1995 to 2010, the Asan countres/regons wth the CPI score of 6 or more are only Sngapore, Hong Kong and Japan. Unversalsm argues that ncreased corrupton s negatvely correlated wth economc growth. Corrupton prevents the economc development. Therefore, the countres all over the world are devoted to ant-corrupton. In recent years, economc growth n Asa has rapdly ncreased, besdes some countres have both serous corrupton and rapd economc growth. Is unversalsm argument approprate for the Asan countres? Does corrupton hurt economc development or corrupton ncrease the economc growth? In ths study, ASEAN+3 (excludng Brune, Camboda, Laos, and Myanmar) and Tawan are selected as the man countres of nterest for ths emprcal study. We use panel data from ten Asan countres over the perod of and adopt the panel vector error correcton model to examne whether corrupton ncreases ncome nequalty and then reduces economc growth. II. ECONOMETRIC METHODOLOGY AND DATA A. Panel Unt Root Tests A varety of procedures for the analyss of unt roots n a panel context have been developed n an attempt to combne nformaton from the tme-seres dmenson wth that of the cross-sectonal dmenson. Gven that many nterestng fndngs nvolve relatvely short tme-seres dmensons, and that conventonal unt root tests turn out low power results when appled to sngle tme seres the well-known low power of conventonal unt root tests when appled to a sngle tme seres, four panel unt root tests whch are Levn, Ln and Chu (LLC) test, Im, Pesaran, and Shn (IPS) test and PP-Fsher are employed n ths study. One of the popular panel unt root tests s that of Levn, Ln, and Chu [33]. Ther test s based on analyss of the equaton as shown below: k y, t = α + βy, t 1 + γ t + θj y, t j + ε, t j= 1 (1) Where s the frst dfference operator, ε, t s a whte nose 2 dsturbance wth a varance ofσ, t =1,2,,T represents tme perods, and =1,2, 10 ndexes cross-sectonal regons. Ths model allows for two-way fxed effects ( α and θ ) and unt-specfc tme trends. LLC test nvolved the null hypothess H : β 0 for all aganst the alternatve H : = β < 0 for = 0 all, wth auxlary assumptons under the null also requred about the coeffcents relatng to the determnstc components. The Im, Pesaran, and Shn [34] test extends the LLC framework to allow for heterogenety n the value of under the alternatve hypothess. IPS relaxed the assumpton of dentcal frst-order autoregressve coeffcents of the LLC test and developed a panel-based unt root test that allows β to be dfferent across regons under the alternatve hypothess. The null and alternatve hypotheses are defned as: H : β 0 ; A β = 0 H : β < 0, for some. The IPS test s based on the mean A group approach. IPS demonstrated that ther test has better fnte sample performance than that of LLC. The Fsher-ADF test proposed by Maddala and Wu [35] and the Fsher-PP test proposed by Cho [36] assume an ndvdual unt root process and compute probabltes by usng an asymptotc Ch-square dstrbuton. The advantage of the Fsher test s that unlke the IPS test, t does not requre a balanced panel. Addtonally, the Fsher test allows the use of dfferent lag lengths n the ndvdual ADF regresson and can also be carred out for any unt root test derved. One dsadvantage of the Fsher test s that the p-values have to be derved va Monte Carlo smulaton. B. Panel Contegraton Tests Pedron [37], [38] developed a number of statstcs based on the resduals of the Engle and Granger [39] study. Pedron s (2004) panel contegraton procedure allows for consderable heterogenety n the panel, snce heterogeneous slope coeffcents, fxed effects and ndvdual specfc determnstc trends are all permtted [40]. By so dong, Pedron had developed seven panel contegraton statstcs for varyng ntercepts and varyng slopes. Four of the statstcs, known as the pooled panel contegraton statstcs, are wthn-dmenson based statstcs, whle the remanng three, known as the group mean panel contegraton statstcs, are between-dmenson based. Pedron [37] argued that for cases wth longer tme spans (such as the number of observaton s greater than 100), the sample sze dstorton tends to mnmal, whle retanng a very hgh testng power across all seven statstcs. However, for shorter panels, alternatve statstcs appeared to yeld conflctng evdence. Pedron [37] showed that n terms of testng power, the group-adf statstc has the best performance n general, followed by the panel-adf. The panel-varance and group-ρ statstcs performed poorly n comparson. C. Panel Vector Error Correcton Model If varables n the emprcal model are nonstatonary and contegrated, we can use the panel vector error correcton model (PVECM) to characterze both long run equlbrum relatonshps and short run dynamc adjustment processes between economc growth and other varables. The PVECM s a restrcted panel vector autoregresson (PVAR) model wth a contegraton bult nto ts specfcaton. The contegraton term s known as the correcton term snce devatons from the long-run equlbrum are corrected gradually through a seres of partal short run adjustments. The PVECM s shown as follows: P X t = C + k X, t k + λec, t 1 + u, t k= 1 β (2) where X s the vector of varables ncludng GDP, CPI, Gn, G, FTD, and HC; represents the panel dentty or cross-country dentfer; k represents the lag length; whlst P represents the optmal lag length selected n accordance wth the Schwarz Crteron (SC); EC represents the error correcton terms; λ and 1142

3 u represent the speed of adjustment to long run equlbrum and the statstcal nose, respectvely. GDP denotes economc growth measured by the real GDP. CPI represents the corrupton measured by the corrupton percepton ndex; the greater the CPI, the lower the corrupton. Gn standards for ncome nequalty measured by Gn coeffcents. G s the real government expendtures. FTD denotes the degree of dependence on foregn trade. Human captal (HC) s measured as the secondary educaton enrollment rates. D. Data Annual data nvolvng ten Asan countres (ncludng Chna, Indonesa, Japan, South Korea, Malaysa, Phlppnes, Sngapore, Tawan, Thaland and Vetnam) from 1995 to 2010 was used n the analyss. Varables GDP and G are expressed at constant 2005 prces and denomnated n U.S. dollar. Data on GDP s obtaned from the Internatonal Monetary Fund (IMF). CPI s obtaned from the Transparency Internatonal. Vetnam s CPI n 1995 and 1996 s not avalable. We adopt SPSS procedures to handle mssng data and obtan the predcted data n SPSS data transformatons. Data on Gn s obtaned from World Development Indcators (WDI) databank, the Standardze World Income Inequalty Database (SWIID), Human Development Report (HDI), and each country s Bureau of Statstcs. Data on FTD s obtaned from WDI and Tawan s Bureau of Statstcs. Data on HC s obtaned from WDI data bank, the World Economc Forum, and the Mnstry of Educaton n Tawan and Chna. III. EMPIRICAL RESULTS The panel unt root tests were conducted usng four technques: LLC, IPS ADF-Fsher, and PP-Fsher. The results are reported n Table I. For varables GDP, Gn, G and FTD, we were unable to reject the null hypothess of unt root at the 1% level of sgnfcance accordng to four panel unt root tests. Thus, GDP, Gn, G and FTD are nonstatonary. Accordng to the four panel unt root tests except LLC test, we were unable to reject the null hypothess of unt root for varables CPI and HC at the 1% level of sgnfcance. It means Gn and HC are statonary accordng to LLC test but are nonstatonary accordng to the other three tests. Gn and HC are treated as nonstatonary varables. Thus, we observed that all varables are nonsatonary. Fnally, the frst dfferences of all varables were found to be statonary, although s not reflected n the fndngs below. To determne whether a contegraton relatonshp exsts, the recently developed methodology proposed by Pedron [38] s employed. It employs four panel statstcs and three group panel statstcs to test the null hypothess of contegraton. Table II presents Pedrono s test for potental contegraton among the followng varables: GDP, CPI, Gn, G, FTD, and HC. The panel contegraton results show that among the seven panel statstcs, the null hypothess of no contegraton s rejected by the panel PP, panel ADF, group PP, and group ADF test statstcs at the 1% level of sgnfcance. Therefore, we thnk that there s a contegraton relatonshp among varables. TABLE I PANEL UNIT ROOT TEST RESULTS Varable LLC IPS ADF-Fsher PP-Fsher GDP CPI Gn G FTD HC (0.9995) *** (0.0046) * (0.0850) (0.0782) *** (0.0097) (0.9999) (0.1526) (0.5813) (0.5842) (0.2174) (0.9945) (0.1910) (0.6088) (0.6224) ** (0.0165) (0.9945) ** (0.0306) (0.3915) (0.8297) ** (0.0363) Notes: *, **, and *** ndcate sgnfcance at the 10%, 5%, and 1% level, respectvely. Numbers n parentheses are p-values. TABLE II PANEL COINTEGRATION TEST RESULTS (WITH GDP AS DEPENDENT VARIABLE) Statstc p-value Panel v Panel ρ Panel PP *** Panel ADF *** Group ρ Group PP *** Group ADF *** Note: *, **, and *** ndcate sgnfcance at the 10%, 5%, and 1% level, respectvely. Due to the fact that the varables ncluded n the model are nonstatonary and contegrated, PVECM s adopted n ths study. The results of PVECM are reported n TABLE III. In ths study, we focus on the relatonshp among corrupton, economc growth, and ncome nequalty n ten Asan countres. The nteractve effects of other varables are mportant and computed n TABLE III. However, the focus of ths study s around the mpacts of economc growth (GDP), corrupton (CPI), and ncome nequalty (Gn). Indept. Varable EC t-1 TABLE III PVECM RESULTS Dependent Varable GDP t CPI t Gn t G t FTD t HC t 0.154*** (0.102) (0.118) 0.062*** (0.0003) (0.619) (0.647) 1143

4 GDP t-1 CPI t-1 Gn t-1 G t-1 FTD t-1 HC t-1 Constant (0.254) *** (0.005) (0.929) 0.874** (0.020) (0.464) (0.179) 34.98*** (0.009) 0.001** (0.031) * (0.096) (0.547) (0.699) (0.447) (0.178) (0.396) (0.907) (0.240) (0.136) (0.230) (0.453) (0.399) * (0.089) (0.752) -44.4*** (0.257) 0.520*** (0.0002) (0.347) (0.576) 13.22*** (0.007) (0.425) (0.658) (0.467) (0.244) -0.27*** (0.002) ** (0.045) (0.373) (0.910) (0.109) (0.752) (0.716) (0.307) 0.278*** 0.668*** Notes: ***, **, and * ndcate sgnfcance at the 1%, 5%, and 10% levels, respectvely. The optmal lag length s 1 selected by SC. The p-values are gven n parentheses. In Table III, the error correcton term has sgnfcantly postve coeffcents on economc growth equaton, ndcatng that prevous dsequlbrum n the economc growth wll be corrected n the current perod. In the economc growth equaton wth the dependent varable GDP t, the estmated coeffcent of CPI t-1 s sgnfcantly negatve. Ths ndcates that corrupton has a sgnfcantly postve mpact on economc growth because a decrease n the CPI score means an ncrease n corrupton. In addton, the estmated coeffcents of G t-1 are sgnfcantly postve. Ths ndcates that government expendture has a sgnfcantly postve mpact on economc growth. Thus, the ten Asan countres could use expansonary fscal polcy to ncrease economc growth. However, the estmated coeffcents of Gn t-1 are nsgnfcantly postve. Ths ndcates that an ncrease n ncome nequalty wll cause an ncrease n economc growth, although the effect s nsgnfcant. In the corrupton equaton, the estmated coeffcent of GDP t-1 s sgnfcantly postve. Ths mples that an ncrease n economc growth wll cause a country to mprove the degree of corrupton and ncrease the CPI score. In the ncome nequalty equaton, the estmated coeffcent of GDP t-1 s nsgnfcantly postve and the estmated coeffcent of CPI t-1 s nsgnfcantly negatve. Thus, an ncrease n economc growth or corrupton wll cause an ncrease n ncome nequalty, although the effect s also nsgnfcant. IV. CONCLUSION In recent years, several Asan countres have experenced rapd economc growth accompaned wth ncreasng economc nequalty. Whle the relatonshp between corrupton and economc growth has been a popular research topc, ths study examnes the relatonshp among corrupton, economc growth, and ncome nequalty experenced wthn ten Asan countres, utlzes the PVECM on data coverng the perod 1990 to 2008, and takes nto consderaton varables ncludng economc growth, corrupton, ncome nequalty, government expendture, foregn trade dependency, and human captal. The emprcal results show that corrupton s mpact on economc growth s sgnfcantly postve, ndcatng that corrupton causes an ncrease n economc growth. For Asan countres, corrupton may smplfy the admnstratve procedures, mprove the admnstratve effcency of bureaucracy, and reduce transacton tme cost. Ths may be the reason why the mpact of corrupton s postve. Thus, we do not support the common percepton that corrupton decreases economc growth. Addtonally, the mpact of government expendture on economc growth s also sgnfcantly postve, mplyng that a government could use expansonary fscal polcy to ncrease economc growth. Meanwhle, the mpacts of economc growth on corrupton and ncome nequalty are sgnfcantly negatve and nsgnfcantly postve, respectvely. These relatonshps suggests that an ncrease n economc growth leads to decreased corrupton, whle an ncrease n economc growth leads to ncreased ncome nequalty, although the effect s nsgnfcant n the latter case. Furthermore, ncome nequalty also appears to have an nsgnfcant postve effect on economc growth and an nsgnfcant negatve effect on corrupton, ndcatng that an ncrease n ncome nequalty wll lead to ncreased economc growth and decreased corrupton, although the effects are once agan nsgnfcant. REFERENCES [1] A. K. Jan, Corrupton: a revew, Journal of Economc Surveys, vol. 15, pp , [2] V. Tanz and H. Davood, Corrupton, Publc Investment and Growth, IMF workng Paper No. WP/97/139, [3] S. Rose-Ackerman (1999), Corrupton and Government. Causes, Consequences, and Reform, Cambrdge Unversty Press, pp. 2-4, [4] P. Mauro, Corrupton and growth, The Quarterly Journal of Economcs, vol. 110, pp , [5] A. Brunett, G. Ksunko, and B. Weder (1997), Insttutonal obstacles to dong busness: regon-by regon results from a worldwde survey of the prvate sector, World Bank PRD Workng Paper, no. 1759, [6] A. Brunett and B. Weder, Explanng Corrupton, Unversty of Saarland and Unversty of Basel, [7] A. D. Monte and E. Papagn (2001), Publc expendture, corrupton, and economc growth: the case of Italy, European Journal of Poltcal Economy, vol.17, no. 1, pp. 1-16, [8] Z. Neeman, M. D. Paserman, and A. Smhon (2004), Corrupton and openness centre for ratonalty and nteractve decson theory, Hebrew Unversty, Jerusalem, Dscusson Paper Seres, no. dp353, [9] J. Svensson, Eght questons about corrupton, Journal of Economc Perspectves, vol. 19, no. 3, pp , [10] Gymah-Brempong, K. and S. M. de Camacho, Corrupton, growth, and ncome dstrbuton: are there regonal dfferences? Economcs of Governance, vol. 7, pp , [11] N. H. Leff, Economc development through bureaucratc corrupton, Amercan Behavoral Scentst, vol. 8, pp , [12] D. H. Bayley, The effects of corrupton n a developng naton, Western Poltcal Quarterly, vol. 19, no. 4, pp , [13] S. P. Huntngton, Poltcal Order n Changng Socetes, New Haven: Yale Unversty Press, [14] F. T. Lu, An equlbrum queung model of brbery, Journal of Poltcal Economy, vol. 93, no. 4, pp , [15] R. Kltgaard, Controllng Corrupton, Unversty of Calforna Press, [16] D. Acemoglu and T. Verder, Property rghts, corrupton and the allocaton of talent: a general equlbrum approach, The Economc Journal, vol. 108, no. 450, pp ,

5 [17] E. Colombatto, Why s corrupton tolerated? Revew of Austran Economcs, vol. 16, no. 4, pp , [18] D. Tresman, What have we learned about the causes of corrupton from ten years of cross-natonal emprcal research? Annual Revew of Poltcal Scence, vol. 10, pp , [19] S. Gupta, H. Davood, and R. Alnso-Terme, Does corrupton affect ncome dstrbuton? Economc Governance, vol. 3, no.1, pp , [20] H. L. L, C. Xu, and H. Zou, Corrupton, ncome dstrbuton and growth, Economc Poltcal, vol. 12, no. 2, pp , [21] Hendrks, J. M. Keen, and A. Muthoo, Corrupton extorton and evason, Unversty of Exeter Department of Economcs Dscusson Paper, no.98/09, [22] M. Johnston, Corrupton, nequalty, and change, n Peter M. Ward (ed.) Corrupton, Development and Inequalty: Soft Touch or Hard Graf, London: Routledge, [23] S. Gupta, L. de Mello, and R. Sharon, Corrupton and mltary spendng, IMF workng Paper, no. WP/00/23, [24] D. Tresman, What have we learned about the causes of corrupton from ten years of cross-natonal emprcal research? Annual Revew of Poltcal Scence, vol. 10, pp , [25] T. Persson and G. Tabelln, Is nequalty harmful for growth? The Amercan Economc Revew, vol. 84, no. 3, pp , [26] G. Psacharopoulos, Poverty and ncome nequalty n Latn Amerca durng the 1980s, Revew of Income and Wealth, vol. 41, no. 3, , [27] R. Barro, Democracy and growth, Journal of Development Economcs, vol. 1, no. 1, pp. 1-28, [28] A. Janvry and E. Sadoulet, Growth, poverty, and nequalty n Latn Amerca: a causal analyss, Revew of Income and Wealth, vol. 46, no. 3, pp , [29] O. Alfranca and M. A. Galndo, Publc expendture, ncome dstrbuton, and growth n OECD countres, Internatonal Journal of Producton Economcs, vol. 9, pp , [30] K. S. Jomo, Growth wth equty n East Asa? Unted Natons, Department of Economcs and Socal Affars, vol. 33, pp. 1-52, [31] A. C. Rcardo, Revstng the relatonshp between ncome nequalty and economc growth, Lecturas de Economía, vol. 64, no , [32] S.K. Samanta and A. Heyse, Income nequalty and economc growth n developng countres: an emprcal analyss, Indan Journal of Economcs, vol. 5, no. 2, pp , [33] A. Levn, C. F. Ln, and C. S. J. Chu, Unt root test n panel data: asymptotc and fnte-sample propertes, Journal of Econometrcs, vol. 108, pp. 1-24, [34] K. S. Im, M. H. Pesaran, and Y. Shn, Testng for unt roots n heterogeneous panels, Journal of Econometrcs, vol. 115, pp , [35] G. S. Maddala and S. Wu, A comparatve study of unt root tests wth panel data and a new smple test, Oxford Bulletn of Economcs and Statstcs, vol 61, pp , [36] I. Cho, Unt root test for panel data, Journal of Internatonal Money and Fnance, vol. 20, pp , [37] P. Pedron, Panel Ccontegraton; asymptotc and fnte sample propertes of pooled tme seres tests, wth an applcaton to the PPP hypothess: new results, Workng Paper, Indana Unversty, [38] P. Pedron, Crtcal values for contegraton tests n heterogeneous wth multple regressors, Oxford Bulletn of Economcs and Statstcs, vol. 61, pp , [39] R Engle and C. Granger, Contegraton and error correcton: representaton, estmaton and testng, Econometrca, vol. 55, pp , [40] Pedron, P. (2004), Panel contegraton: asymptotc and fnte sample propertes of pooled tme seres tests wth an applcaton to the PPP hypothess, Econometrc Theory, vol. 20, pp ,