The Cover Page

Title of the paper:

A study of Causal Relationship among Economic Freedom, Economic Growth, and Corruption: Empirical Evidence from Panel Data

1. Tsangyao Chang, Professor, Department of Finance, FengChiaUniversity,Taichung, Taiwan. TEL: 886-4-2451-7250 ext. 4150. FAX: 886-4-2451-3796. E-mail: .

2. Corresponding author: Wen-Chi Liu, Lecturer,Department of Business Administration, Da Yeh University, Changhua, Taiwan. Ph.D. Candidate,College of Business, FengChiaUniversity,Taichung, Taiwan. TEL: 886-4-8511888ext. 3372. Cell phone: 886-9-36300868 FAX: 886-4-8511382 E-mail: .

A study of Causal Relationship among Economic Freedom, Economic Growth, and Corruption: Empirical Evidence from Panel Data

Tsangyao Chang[1] Wen-Chi Liu[2]

Abstract

This study uses panel data to investigate the causal relationship among corruption, economic freedom and economic growth for 83 countries over the 1998-2006. Empirical evidence reveals that feedback exists between economic freedom and corruption, and one way causality runs from economic growth to economic freedom and corruption, respectively. Based on the results of panel Granger causality test, the economic freedom and corruption are respectively employed as the dependent variables to perform the panel OLS estimation. Conclusively, on the whole, the higher economic growth and lower corruption in the previous year will grease the wheels of economic freedom next year but the higher economic growth and freedom in the previous year will sand the wheels of corruption next year.

Keywords: Economic Freedom, Corruption,Economic Growth, Granger Causality, Panel OLS Estimation

I. Introduction

Economic freedom as defined by the Fraser Institute, a think tank that publishes EconomicFreedom of the World since 1996, is composed of personal choice, voluntary exchange,freedom to compete and protection of person and property.

A positive relation between economic freedom andgrowth hadbeen gotten by some empirical studies (e.g., Barro, 1991; Vanssay andSpindler, 1994; Torstensson 1994). Nonetheless, itis less clear whether freedom causes growth, growth causes freedom, or the twoare jointly bilateral. Farr et al. (1998)was one of the earliest studies on causality toexamine the relationship between economic freedom and the level of GDP. The empirical result wasfeedback existed between economic freedom and the level of GDP. Then, Heckelman, J. C. (2000)employed the annual freedom indicators developed by the Heritage Foundation to performthe causalrelationship with economic growth. The tests suggested the average level of economic freedomprecedes economicgrowth. Haan and Sturm (2000) also pointed out that economic freedom broughtcountries to their steady state level of economic growth more quickly,but did not increase the rate of steady state growth. Vega-Gordillo et al. (2003) employedKiviet’s method andyielded interesting results that politicaland economic freedoms appeared to enhance economic growth. Another causal relationship proves to be significant, is thatprior higher growth rates foster political freedom (Lipset’s hypothesis), but the results showed no statistically significant causality workingfrom growth to economic freedom. These results always were interpreted with cautionas Haan and Sturm (2000: 231) noted that “One possible objectiontowards our analysis so far could be that the choice of our sample ofcountries, although only based on data availability, may have influencedour results.”

In recent years, development economists as well as international financial institutions and policymakers have paid moreattention to the harmful effects of bureaucratic corruption. However, corruption is not a new question. The current literature on corruption exhibited its harmful effects on growth (e.g., Shleifer and Vishny 1993, Mauro 1995, Cheung 1996, and Bardhan 1997). However, Lui (1985)showed that corruption couldefficiently lessen the time spent in queues. The reason was that bribes couldgive bureaucrats an incentive to speed up the process, in an otherwise sluggishadministration (see also Leys, 1964). Furthermore, Huntington (1968)argued that corruption could help surmount tedious bureaucratic regulationsand foster growth. According to him, such a phenomenon had been observedin the 1870’s and 1880’s in the United States, where corruption by railroad,utility and industrial corporations resulted in faster growth. For another relative study, Felix Fofana et al. (2005)linked between corruption, poverty andgrowth was analyzed in a panel of 18 African countries for the 1996-2001 time periods. The empirical resultssuggested that: 1) it was poverty that caused growth but not the other way around. This impliedthat past information of the state of human development help improvedprediction on growth;2) it was the state of growth that caused corruption and inequality; 3) It was corruption that causedinequality; 4) corruption and poverty together caused growth; 5) poverty and growth togethercaused corruption; 6) and lastly, inequality together with growth caused corruption.

The debate on the impact of corruption on economic performance goes beyonda “moralistic view” that unequivocally condemns corruption. The moraljudgment on corruption may bias the understanding of its economic consequences. One strand of the literature argues that corruption may take placein parallel with a low quality of governance and can, therefore, reduce theinconvenience of such low quality. This is the “grease the wheels” hypothesis.Another strand stresses that although bribery may have benefits if the qualityof governance is low, it may as well impose additional costs in the samecircumstances. The existence of such costs provides a rationale for the “sandthe wheels” hypothesis. The core of the debate on the “grease” vs. the “sand the wheels” hypotheseslies in the combination of corruption with a low quality of governance. While there are many aspects of governance that corruption may grease orsand, the literature has mainly focused on two. One concerns the ill functioningof bureaucracy (i.e., its failure to accomplish assigned goals; see Leff,1964) while the other refers to policy options by public authority.

After these literature reviews, only a small number of recent studies, few general conclusions can be derived as to the relationship among economic freedom, economicgrowth and corruption with panel causality test. This research tries to link them and investigatewhether they exists anycausalrelationship by using more 83 countries in order to increase the robustness. Furthermore, let we judge by the coefficients of panel OLS to conclude whether we support the “grease” or the “sand the wheels” hypotheses.

The specific objectives are:

 To determine whether corruption causes economic growth or vice-versa;

 To determine whether economic growth causeseconomic freedom or vice-versa;

 To determine whether economic freedom causes corruption or vice-versa...

This paper is organized as follows. Section II presents the data used and summary statistics. Section III first describes the methodology employed, then discusses the empirical findings. Section IV concludes. .

II. Data

This study uses panel data to investigate the causal relationship among corruption, economic freedom and economic growth for 83 panel countries over the 1998-2006. Details about these 83countries(i.e., 14 low, 24 lower middle, 16 upper middle, and 29 high income countries on the average) please seethe table 1. The data for 83 countries begin in 1998 since the CPI in annual frequency is only availablefrom that time.

CPI (Corruption Perceptions Index) is the proxy for political corruption and bureaucratic corruptionpublished by Transparency International(website: since 1995. Transparency International has devised a CPI based on opinion surveys of business people, professional risk analysts and the public. Original scores are ranged from 0 (completely corrupt) to 10 (clean). There are only 40 countriespossessing complete data since 1995 and few low income countries are included. Consequently, for robustness, we decide to capture the data including 83 countries since 1998 to include more low income countries.

EF(Index of economic freedom) defined over the range 0 to 100, where the number of 100 means the most free. The index developed by the Heritage (website: Economic freedom is defined as the absence of government coercion or constraint on the production, distribution, or consumption of goods and services beyond the extent necessary for citizens to protect and maintain liberty itself. In other words, people are free to work, produce, consume, and invest in the ways they feel are most productive.

At last, EG (Economic growth) was captured through the annual growth of per capita GDPat 1998 constant pricesbetween and. The data come from International Monetary Fund (website: prices

Tables 2-4 provide summary statistics for the Corruption Perception Index (CPI), Index of economic freedom (EF), and Economic growth (EG). For details, we divide 83 countries into four groups-- low, lower middle, upper middle, and high income countries by the GNI per capita, Atlas method (current US$) of the World Bank over the 1998-2006. According tothe average EF in Table 2,we could find that less economic freedom exists in low income countries but it hasimproved for the last three years, and more economic freedom exists in high income countries. Bylooking at the average EG in table 3, we find that less economic growth exists in both low and high income countries and higher economic growth exists in upper middle income countries. Based on the average CPI in table 4,we could capture that less corruptionexists in high income countries and more corruption exists in low income countries, but it has alsoimproved for the last three years.

III.Methodology and Empirical Result

A. Panel Unit Root Test

It is well known that, in small samples, traditional unit root tests have low power against nearstationary alternatives. Panel data circumvent the low power problem of standard unit roottests by increasing the number of observations.

The conventional ADF test for single-equation is based onthe following regression equation:

, [1]

where i = 1,2,…,n countries (cross-section),  is the first difference operator, is the variable under study, is a white-noise disturbance term with a variance of , and t = 1, 2,….,T time period (time series). The unit root null hypothesis of is tested againstthe one-side alternative hypothesis of <0, which correspondstobeing stationary. The test is based on the test statistic(where is the OLS estimate ofin Equation [1] and is its standard error) since the single-equation ADF test may havelow power when the data are generated by a near-unit-root but stationaryprocess. Levin et al., (2002, hereafter, Levin-Lin-Chu) found that thepanel approach substantially increased power in finite samples whencompared with the single-equation ADF test. They proposed apanel-based version of Equation [1] that restrictedby keeping itidentical across cross-sectional regions as follows:

[2]

where i =1,2,…N indexes across cross-sectional regions. Levin-Lin-Chu test the null hypothesis of(i.e.,all the entities have a unit root)againstthe alternative of(i.e.,all the entities do not have a unit root), with the test based on the teststatistic(where is the OLS estimate ofin Equation [2], and is its standard error).

Table 5reported the results of the panel unit root test for threevariables of EF, EG and CPI, respectively. Based on the results, we found that the null hypotheses of non-stationaritywere strongly rejected at the 1percent significant level. These indicate that all three variables EF, EG, and CPI for 83 countries areall stationary.

B. Panel VAR Causality Test and Panel Least Squares Estimation

In this context, we employed the Hurlin and Vent (2003) panel data Grangercausality procedure (HV, hereafter). The introduction of a paneldata dimension permits the use of both cross-sectional and time-seriesinformation to test any causality relationships between two variables. Inparticular, by increasing the number of observations, this procedure raisesthe degrees of freedom. Thus, it noticeably improves the efficiency ofGranger causality tests.

Consider a time-stationary VAR representation, adapted to a paneldata context. For each individual i we have, t [1, T]:

[3]

,where are i.i.d.(0,).

The HV-procedure is based on the following homogenous non-causalityhypothesis:

The null hypothesis states non-existence of causal relationships acrossN. If this null is rejected, there is evidence of Granger causality. In thegeneral case, the test statistic can be computed by the following Wald testproposed by HV:

[4]

where SN denotes the total number of observations, denotes therestricted sum of squared residuals obtained under the null hypothesis,and is the unrestricted sum of squared residuals computed fromequation 3.

This new procedure also follows a standard Granger causalitywhere the variables entered into the system need being time-stationary. Thus, the three variables are subjected to unit root testing.

Since we found all the three variables EF,EG, and CPI are stationary, in the part A, we can further proceed to use HV procedure to test whether there exists any causal relationship among these three variables EF, EG, andCPI using our stationary panel data.

Since we found that the casual directions among these three variablesfor both lag one and two give us the similar results, we only reportthe results for lag order two.Table 6 reposts the results of the panel VAR causality test. The empirical evidence reveals that economic growth causesboth economic freedom and corruption respectively, whereasneither economic freedomnor corruptioncauseseconomic growth. Moreover,we found feedback exists between economic freedomand corruption. Based on the above result, the EF and CPI were respectively chosen as the dependent variables when we performed the panel OLS estimation.

After performing the panel VAR causality test, we employed the panel least squares estimation to investigate the signs of coefficients amongeconomic freedom, corruption, and economic growth. The second column of table 7presents the results of the panel OLS estimationby choosing the EF as the dependent variable. The coefficients of EG and EF withlag order oneare closed to be significant at 1% level. These results indicate thatthe higher economic growth and higher CPI (i.e., lower corruption) in the previous year would have significant positive effects on economic freedom this year for all 83 countries.

The third column of table 7 presents the results of the panel OLS estimationby choosing CPI as the dependent variable. The coefficients of EG and EF withlag order oneare closed to be significant at 10% level. On the whole, these resultsindicate thatthe higher economic growth and higher economic freedom in the previous year would have significant positive effects on CPI (i.e., lower corruption)this year for all 83 countries.

Toverify the robustness of our panel VAR causality test results, we also performed the panel OLS estimationby using the EG as the dependent variable. We found that the coefficients of EF and CPI with lag order oneor two are not significant at 10% level, and the results for the lag order two are reportedin the fourth column of table 7. The result reveals that economic growthis not well explained by the two independent variables of economic freedom and corruption. Apparently, economic growthis influenced by other factors in these 83 countries. These resultsare consistent with those found in ourpanel VAR causality testthatone way causality running from EG to EF and CPI, respectively.

IV. Conclusion

This study uses panel data to investigate the causal relationship among corruption, economic freedom and economic growth for 83 panel countries over the 1998-2006.

According to the results of panel Granger causality test,the empirical evidence reveals that one way causality runs from economic growth to economic freedom and corruption, respectively, but feedback exists betweeneconomic freedom and corruption. Then, we choose economic growth, corruptionand economic growthas the dependent variables, respectively,to perform the panel OLS estimation. Conclusively, we find thatthe higher economic growth and lower corruption in the previous year will grease the wheels of economic freedom next year, butthe higher economic growth and economic freedom in the previous year will sand the wheels of corruption next year.

Reference

Bardhan, P. (1997), “Corruption and Development: A Review of the Issues,” Journal of Economic Literature 35: 1320-46.

Barro, R.J. (1991),“Economic Growth in a Cross Section of Countries,” Quartely Journalof Economics 106: 407 – 443.

Bonferroni, C. E. (1936),“Teoria statistica delle classi e calcolo delle probabilità,” Pubblicazioni del R Istituto Superiore di Scienze Economiche e Commerciali di Firenze, 8:3-62.

Cheung, S. N. (1996), “Simplistic General Equilibrium Theory of Corruption,” Contemporary Economic Policy 14 (3): 1-5.

Dawson, J. W. (2003)“Causality in the Freedom-Growth Relationship,” European Journal of Political Economy, Vol. 19, 479-495

Farr, K.W., R.A. Lord,and J.L. Wolfenbarger(1998),“Economic Freedom, Political Freedom and Economic Well-Being: A Causality Analysis,” Cato Journal, Vol. 18, no. 2, 247-262.

Fofana, F. andC. J. Francis (2005), “The Causality Between Corruption, Poverty and Growth:A Panel Data Analysis,”Unites States Agency InternationalDevelopment (USAID) and International Development Research Centre (IDRC), Ottawa, Canada.

Haan, J. deand C. L. J. Siermann (1998), “Further Evidence on the Relationship between EconomicFreedom and Economic Growth,” Public Choice 95 (3/4): 363–380.

Haan, J. deand J. Sturm (2000), “On the Relationship between EconomicFreedom and Economic Growth,” European Journal of Political Economy16: 215–241.

Hadri, K. (2000),“Testing for Stationarity in Heterogeneous Panel Data,” TheEconometrics Journal 3, 148-161.

Heckelman, J. C. (2000),“Economic Freedom and Economic Growth: A Short-Run Causal Investigation,” Journal of Applied Economics, Vol. III, no.1, 71-91.

Huntington, S.P. (1968)“Political Order in Changing Societies,”New Haven: YaleUniversityPress.

Hurlin, C. and B. Venet (2003), “Granger Causality Tests in Panel Data Models with Fixed Coefficients,” LEO and University Paris IX Dauphine.

Im, K. S., M. H.Pesaran, and Y.Shin (2003), “Testing for Unit Roots in Heterogeneous Panels,”Journal of Econometrics, 115, 53–74.

Kiviet, J. F. (1995), “On Bias, Inconsistency, and Efficiency of Various Estimatorsin Dynamic Panel Data Models,” Journal of Econometrics 68: 53–78.

Leff, N.H. (1964), “Economic Development through Bureaucratic Corruption,”American BehavioralScientist 8: 8–14.

Levin, A., C. F.Lin, andC. Chu (2002), “Unit Root Tests in Panel Data: Asymptotic and Finite-Sample Properties,”Journal of Econometrics, 108, 1–24.