Investigating the interaction effect of democracy and economic freedom on corruption: a cross-country quantile regression analysis

Transcription

1 Investgatng the nteracton effect of democracy and economc freedom on corrupton: a cross-country quantle regresson analyss Author Saha, Shraban, Su, Jen-Je Publshed 2012 Journal Ttle Economc Analyss and Polcy Copyrght Statement The Author(s) The attached fle s reproduced here n accordance wth the copyrght polcy of the publsher. For nformaton about ths journal please refer to the journal s webste or contact the authors. Downloaded from Lnk to publshed verson Grffth Research Onlne

2 Economc Analyss & Polcy, Vol. 42 No. 3, december 2012 Investgatng the Interacton Effect of Democracy and Economc Freedom on Corrupton: A Cross-Country Quantle Regresson Analyss Shraban Saha 1 School of Accountng Fnance and Economcs, Edth Cowan Unversty, 270 Joondalup Drve, Joondalup, WA-6027, Australa (Emal: s.saha@ecu.edu.au) and Jen-Je Su Department of Accountng, Fnance and Economcs, Grffth Unversty, Nathan QLD 4111, Australa (Emal: j.su@grffth.edu.au) Abstract: Ths paper explores the nteracton effects of economc freedom and democracy n controllng corrupton for 100 countres by usng quantle regresson technque. The man contrbuton s to explore the nteracton effects throughout condtonal dstrbuton of corrupton across natons. Our results renforce some fndngs n the lterature, but also provde new conclusons. The fndngs suggest a stronger and sgnfcant nteracton effect n reducng corrupton, especally n the most-corrupt countres. However, democratc and economc freedoms alone may not cure corrupton effectvely n the most-corrupt natons, a sound democratc reform can elmnate corrupton substantally only after achevng a threshold level of economc freedom. INTRODUCTION Most cross-country emprcal research on the consequences of corrupton affrms ts detrmental effect on growth and development. Gven ths negatve mpact, much stands to be ganed from understandng the causes of and ways of reducng corrupton. Economsts and polcy makers have emphaszed the role of nsttutonal factors such as economc and poltcal freedom n combatng corrupton (Tresman 2007, Graeff and Mehlkop 2003, Blss and D Tella 1997, 1 Correspondng Author: s.saha@ecu.edu.au. 389

3 Investgatng the Interacton Effect of Democracy and Economc Freedom on Corrupton: A Cross-Country Quantle Regresson Analyss Rose-Ackerman 1978, and Krueger 1974). Some consensus s slowly emergng n the lterature, although several aspects reman unresolved. For nstance, even though most studes confrm that greater economc freedom reduces corrupton, the role of democracy remans unclear (Rock 2007, Montnola and Jackman 2004, Mohtad and Roe 2003, Tresman 2000, and Ades and D Tella 1999). Furthermore, the current lterature s lackng behnd to answer, whether more economc freedom and greater democracy consstently reduce corrupton among the most and the least corrupt natons. We address ths ssue as corrupton control s a crucal concern from a polcy perspectve especally for the developng countres. Ths paper analyzes the jont effects of democracy and economc freedom n combatng corrupton by utlzng quantle regresson technque. Quantle regresson technque has ts advantages over ordnary least squares (OLS), nstrumental varables, and panel estmaton as t estmates the nteracton effect on corrupton throughout the condtonal dstrbuton of corrupton. 2 Saha, Gounder and Su (2009) (hereafter, SGS) assessed the nteractve effect of democracy and economc freedom n controllng corrupton, and the fndngs reflect that democracy and economc freedom were complementng n reducng corrupton. Specfcally, whle economc freedom reduces corrupton n any poltcal envronment, t s more effectve wth a hgher level of democracy. However, democracy can ncrease corrupton f economc freedom s low and wll reduce corrupton only above a certan threshold of economc freedom. Bllger and Goel (2009) n contrast, found that among the most-corrupt natons, greater economc freedom does not appear to reduce corrupton, but greater democracy does. However, unlke SGS (2009), Bllger and Goel (2009) dd not consder the nteracton effect. Ths paper nvestgates SGS (2009) by utlzng quantle regresson. In partcular ths paper focuses on the dstrbuton of the corrupton percepton ndex to dentfy whether the nteracton effects of democracy and economc freedom vary over dfferent quantles.e., does the nteracton effect dffer n hghly-corrupt natons compared to the least-corrupt ones? There may be extensve nsttutonal dfferences between clean and corrupt countres that may nfluence the effcacy of combatng corrupton even when the same level of democracy and economc freedom exsts n the most-corrupt and clean countres. 3 In that case the fndngs have sgnfcant polcy mplcaton for the most corrupt and the least corrupt countres.e., the polcy recommendatons to cure corrupton could be vared n nature dependng on the partcular characterstcs of the most corrupt and the least corrupt countres. The nteracton effect results support some fndngs n the lterature, but also contrbuted to the lterature by addng the results of the senstvty of the dstrbuton of the dependent varable corrupton. Quantle regresson results manly dffer n terms of the most and least corrupt natons suggestng that some of the polcy mplcatons regardng ant-corrupton reform based on ordnary least squares analyss requre to be reconsdered. The next secton dscusses the model, data and estmaton methodology followed by results and concluson. 2 OLS, nstrumental varables and panel regressons estmate the parameters of nterest at the mean evaluaton by a condtonal dstrbuton of the dependent varable (Bllger and Goel 2009 and Okada and Samreth 2012). 3 Goel and Nelson (2007) show the contagon effects of corrupton for the Unted States. 390

4 Shraban Saha and Jen-Je Su II. THE MODEL, DATA AND ESTIMATION METHODOLOGY Followng SGS (2009), the model s specfed as: CPI = β + β UNEM β1demo + β 2EF + β3demo * EF + β 4 log( RGDP) + β ALR 7 + ε + β GINI 5 (1) where CPI s corrupton percepton ndex, the last four terms are soco-economc control varables.e., real gross domestc product per capta (RGDP), Gn ndex for ncome nequalty (GINI), unemployment rate (UNEM) and adult lteracy rate (ALR) and ε s the error term and denotes country. In estmatng Equaton (1), we are partcularly concerned wth the coeffcent β3, whch captures the nteracton effect. The coeffcents of RGDP, GINI, UNEM and ALR n Equaton (1) are of less nterest although they help to establsh the plausblty of the overall results. We expect that perceved corrupton wll fall, the hgher the per capta ncome and lteracy rate and the lower the unemployment and ncome nequalty. Corrupton defned as msuse of publc offces s a percepton based subjectve ndex. Transparency Internatonal s annual corrupton percepton ndex (CPI) s used as the measure of corrupton (the dependent varable). 4 The democracy ndex (DEMO) s constructed by combnng Freedom House poltcal rghts, cvl lbertes and press freedom ndces. 5 The economc freedom ndex (EF) s a composte ndex obtaned from the Hertage Foundaton. 6 We re-scale CPI, DEMO and EF from 0 to 10, wth a low value of CPI representng a low level of corrupton and low DEMO and EF ndcatng lower freedoms. The data for ths study ncludes cross-sectonal observatons for 100 countres wth all varables averaged over the perod Equaton (1) s estmated by usng OLS and the regresson quantle estmaton. Unlke the standard OLS estmaton whch yelds a result at the condtonal mean of the dependent varable (corrupton), the quantle regresson estmates results throughout the condtonal dstrbuton of corrupton. Ths allows comparson of how dfferent percentles of the corrupton level.e., the least-and-most-corrupt (the low and hgh-quantle, respectvely) countres are affected by democracy, economc freedom and ther nteracton. Quantle regresson, developed n Koenker and Bassett (1978), estmates the regresson parameters at dfferent ponts n the condtonal dstrbuton of the dependent varable. Specfcally, the θ th regresson quantle solves the followng mnmzaton problem: ' ' mn[ θ y X β (1 θ ) y X β ] K t t t t β R t { t: yt xt β } t { t: yt < xt β } + (2) where 0 < θ < 1. In Equaton (2), y t s the corrupton ndex (CPI) and the vector X t contans 4 For detals see 5 See for detals. 6 Economc freedom ndex s an arthmetc mean of eght ndvdual freedoms (.e., busness freedom, trade freedom, monetary freedom, freedom from government, fscal freedom, property rght, nvestment freedom and fnancal freedom). See for detals. 7 The data sources are same as those n SGS (2009). 391

5 Investgatng the Interacton Effect of Democracy and Economc Freedom on Corrupton: A Cross-Country Quantle Regresson Analyss the ndependent varables shown n equaton (1). The condtonal quantle of y t gven X t s Q y (θ x t ) = x t β θ where β θ s the slope at the θ th quantle. We hypothesse that democracy and economc freedom should have a complementary effect n brngng down corrupton and consequently the nteracton effect (ß 3 ) s negatve. The partal effects of democracy and economc freedom on corrupton are computed to dentfy the ndvdual mpact of democracy (economc freedom) for a gven level of economc freedom (democracy) for dfferent quantles. Specfcally, the partal effects of democracy and economc freedom on corrupton at quantle (θ) are computed as follows: Δ CPI * EF (3a) / ΔDEMO ( gvenefand θ ) = β1 θ + β3 θ Δ CPI * DEMO (3b) / ΔEF ( gvendemoand θ ) = β2 θ + β3 θ III. EMPIRICAL RESULTS The results of the base model (1) for OLS and regresson quantles (gven q=0.1, 0.3, 0.5, 0.7, 0.9) are presented n Table 1. OLS estmates provde a baselne of mean effects, and we compare these to estmate for separate quantles n the condtonal dstrbuton of corrupton. 8 The followng observatons are noted. Frst, the OLS result (at the condtonal mean) and the quantle result at q=0.5 (at the condtonal medan) are smlar across dfferent parameters. Second, among the fve quantles, there s consderable varablty for each parameter (for example, at q=0.1, the coeffcents for DEMO and EF are 0.16 and -0.66, respectvely; whle at q=0.9, they are 1.10 and -0.19). The results suggest that democracy s more effectve n the most corrupt countres, whle the effectveness of economc freedom ncreases for the least corrupt countres. Ths result s consstent wth Goel and Bllger (2009). Thrd, the nteracton effect of DEMO and EF s (sgnfcant at the 1% level) for OLS and quantle (q=0.5), mplyng a complementary effect of DEMO and EF at (condtonal) mean and medan. In other words, the result reflects that the jont effect of democracy and economc freedom combat corrupton sgnfcantly. The quantle regresson results confrm our hypothess that the nteracton effect of democracy and economc freedom s consstent throughout the condtonal dstrbuton of corrupton. Interestngly, the quantle results also show that the nteracton effect appears to be stronger among corrupt countres than clean ones (β 3 =-0.05 at q=0.1 and at q=0.9). Fourth, the socoeconomc control varables, except lteracy rate, retan the expected sgn although not beng sgnfcant n every quantle (except for RGDP). That s hgher values of ncome nequalty and unemployment ncrease corrupton, whereas a hgh ncome level decreases corrupton. The next step examnes the partal effect of DEMO (EF) at dfferent levels of EF (DEMO) throughout the condtonal dstrbuton of corrupton across countres. We depct the relatonshp gven by equatons 3a and 3b n Fgure 1. Fgure 1(A) ndcates that EF sgnfcantly reduces corrupton at any poltcal envronment as well as at any corrupton level (as the partal 8 To nterpret the sgns of the coeffcents, one should note that smaller values of the dependent varable denote less corrupton. 392

6 Shraban Saha and Jen-Je Su Table 1: Interacton Effects of Democracy and Economc Freedom on Corrupton: OLS Versus Quantle Regresson OLS Q0.1 Q0.3 Q0.5 Q0.7 Q0.9 DEMO 0.61** (1.98) 0.16 (0.26) 0.63 (1.57) 0.52 (1.42) 0.43 (0.89) 1.10** (2.05) EF -0.48*** (2.90) (1.53) -0.47* (1.89) -0.46* (1.90) (1.00) (0.79) DEMO*EF -0.12*** (2.73) (0.55) -0.13** (2.23) -0.12** (2.16) -0.12* (1.67) -0.20*** (2.61) Log(RGDP) -1.09*** (4.97) -1.20*** (3.10) -1.06*** (3.15) -0.83** (2.31) (0.92) -0.85** (1.98) GINI 0.03** (2.07) 0.04 (1.53) 0.05** (2.44) 0.03 (1.16) 0.02 (0.78) (0.14) UNEM 0.02 (0.99) 0.03 (0.73) 0.01 (0.64) 0.02 (1.31) 0.03 (1.24) 0.01 (0.40) ALR 0.04*** (3.54) 0.05** (2.03) 0.04*** (2.77) 0.03** (2.43) 0.02 (1.44) 0.02* (1.63) Constant 14.05*** (7.36) 13.17*** (3.71) 12.27*** (4.52) 12.49*** (4.26) 10.10*** (2.64) 14.50*** (3.87) Observatons F-Statstc (p-value) a (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Notes: Dependent varable s corrupton percepton ndex. Lower quantles (e.g., Q0.1) sgnfy less corrupt natons. Absolute t-statstcs are n the parentheses. ***, **, and * denote sgnfcance levels at the 1%, 5% and 10%, respectvely. a F-statstc and assocated p-values are reported for the test of all slope parameters jontly equal to zero. 393

7 Investgatng the Interacton Effect of Democracy and Economc Freedom on Corrupton: A Cross-Country Quantle Regresson Analyss Fgure 1: Partal Effects of Democracy and Economc Freedom (A) Partal effect (Economc Freedom) at dfferent levels of democracy and corrupton DEMO=1-0.4 Partal Effect (EF) DEMO=3 DEMO=5 DEMO=7 DEMO= Quantle (B) Partal effect (Democracy) at dfferent levels of economc freedom and corrupton Partal Effect (DEMO) EF=1 EF=3 EF=5 EF=7 EF= Quantle 394

8 Shraban Saha and Jen-Je Su effects are all below zero). The partal effect of EF becomes stronger n a more democratc envronment, gven the same corrupton quantle. Fgure 1(A) also shows that the mpact of EF on corrupton s not homogenous at dfferent corrupton levels for DEMO=1 (less democratc), EF s more effectve among clean countres but for DEMO=9 (democratc), t s more effectve among corrupt countres. In contrast, Fgure 1(B) shows that democracy at dfferent levels of economc freedom produces mxed results. At the early stages of economc lberalzaton, greater democracy ncreases corrupton opportuntes and the stuaton worsens n the most corrupt countres. Democracy only helps n brngng down corrupton (at all quantles) f the level of economc freedom s above a certan threshold (EF=7). Above ths threshold, the beneft of democracy n combatng corrupton becomes more notceable for more corrupt countres. Overall, our result confrms the fndngs of Saha et al. (2009) and Bllger and Goel (2009) that economc freedom does not help n reducng corrupton n the most corrupt countres. But t also added that wth the help of democratc freedom, economc lberalzaton can reduce corrupton substantally n a most corrupt country. In other words, t s evdent that a hgh level of economc freedom and democracy are more effectve and effcent n reducng corrupton n the most corrupt countres. The result s consstent wth some country specfc examples. For nstance, Inda has been a functonng and vbrant democracy for more than half a century but a hgher democracy level has been unable to reduce corrupton n Inda may be due to the exstence of below average level of economc freedom. On the other hand, South Korea llustrates a good example of the democratzaton process and ts deterrng effect on corrupton where there exsts a far greater than average level of economc freedom. Moreover, a hgher degree of economc freedom reduces corrupton sgnfcantly n closed to authortaran states (e.g., Hong Kong and Sngapore) but the degree of nfluence s much less n most corrupt countres (Chna and Russa) where economc freedom s farly low. The results manfest that the effect of economc freedom n curng corrupton s much stronger n most corrupt countres at a hgher level of democracy (above average). In other words, democracy can accelerate the ant-corrupton reforms n partcular for the most corrupt natons f there exsts a hgh level of economc lberalzaton. IV. CONCLUSIONS We have analysed how ncreased economc freedom and democracy deter corrupton n both tals of the condtonal dstrbuton of corrupton across countres. The n-depth partal effect quantle analyss reveals that the effects of democracy and economc freedom are not dentcal throughout the dstrbuton. Economc freedom sgnfcantly reduces corrupton at any poltcal envronment but the effectveness ncreases at hgher levels of freedom. We also fnd that the mpact of economc freedom on corrupton s not entrely homogenous at dfferent levels of corrupton. In contrast, democracy may not be effectve n combatng corrupton at a low level of economc freedom (for countres at any degree of corrupton), but democracy becomes effectve n curng corrupton for the most-corrupt countres when the level of economc freedom s hgh. The polcy mplcaton s that to reduce corrupton, t s mportant for democratzaton to be done n an envronment of hgh economc freedom. 395

9 Investgatng the Interacton Effect of Democracy and Economc Freedom on Corrupton: A Cross-Country Quantle Regresson Analyss ACKNOWLEGEMENTS We lke to thank the referees for ther valuable comments. The usual dsclamer apples. REFERENCES Ades, A. and R. D Tella (1999). Rents, Competton, and Corrupton, The Amercan Economc Revew. 89: Bllger, S.M. and R.K. Goel (2009). Do Exstng Corrupton Levels Matter n Controllng Corrupton? Cross-country Quantle Regresson Estmates, Journal of Development Economcs. 90: Blss, C. and R. D Tella (1997). Does Competton Kll Corrupton?, Journal of Poltcal Economy. 105: Goel, R.K. and M.A. Nelson (2007). Are Corrupt Acts Contagous?, Evdence from the Unted States, Journal of Polcy Modellng. 29: Graeff, P. and G. Mehlkop (2003). The Impact of Economc Freedom on Corrupton: Dfferent Patterns for Rch and Poor Countres, European Journal of Poltcal Economy. 19: Koenker, R. and G.J. Bassett (1978). Regresson Quantles, Econometrca. 46: Krueger, A.O. (1974). The Poltcal Economy of the Rent-Seekng Socety, Amercan Economc Revew. 64: Mohtad, H. and T.L. Roe (2003). Democracy, Rent Seekng, Publc Spendng and Growth, Journal of Publc Economc. 87: Montnola, G.R. and R.W. Jackman (2004). Sources of Corrupton: A Cross-Country Study, Brtsh Journal of Poltcal Scence. 32: Okada, K. and S. Samreth (2012). The Effect of Foregn Ad on Corrupton: A Quantle Regresson Approach, Economcs Letters. 115: Rock, M. (2007). Corrupton and Democracy. UN/DESA Workng paper, 55: Rose-Ackerman, S. (1978). Corrupton: A Study n Poltcal Economy. New York: Academc Press. Saha, S., R. Gounder, and J.J. Su (2009). The Interacton Effect of Economc Freedom and Democracy on Corrupton: A Panel Cross-Country Analyss, Economcs Letters. 105: Tresman, D. (2000). The Causes of Corrupton: A Cross-Natonal Study, Journal of Publc Economcs. 76: Tresman, D. (2007). What Have We Learned About The Causes Of Corrupton From Ten Years of Cross-Natonal Emprcal Research?, Annual Revews Poltcal Scence. 10: