Corruption and stoCK marKet development: neW evidenCe from gCC Countries

. The theoretical relationship between corruption and stock market development has been debated quite extensively in the literature, yet the evidence on the impact of corruption on stock market development remains contradictory and ambiguous. This paper investigates the impact of corruption, as measured by Corruption Perception Index (CPI) published by Transparency International, on stock market development focusing exclusively on Gulf Cooperation Council (GCC) countries with its special characteristics of combining richness with relatively high level of corruption. Results from an estimation of alternative regression models on a panel of six GCC countries over the period 2003–2011, through which CPI is legitimately comparable, confirms a positive impact of corruption on stock market development, where the latter is measured by market capitalization. This is consistent with the view that corruption greases the wheels of economy by expediting transactions and allowing private firms to overcome governmentally imposed inefficiencies.


Introduction
Factors affecting stock market development in many coun tries have received a great deal of attention, especially Latin America, Middle East and Central Asia, which have shoul dered an increasing share of world growth. Recently, a num ber of papers have examined the impact of corruption on stock market development (see Ng 2006;Pinheiro 2010;Bolgorian 2012;Jain et al. 2012;Lau et al. 2013;Shahbaz et al. 2013;Tvaronavičienė et al. 2013Tvaronavičienė et al. , 2014Chêne 2014;Giriūnas, Mackevičius 2014;Peker et al. 2014;Šabasevičienė, Grybaitė 2014;Caurkubule, Rubanovskis 2014a, 2014bRaudeliūnienė et al. 2014;Kaminskienė et al. 2014;Tunčikienė, Korsakienė 2014;Kanapinskas et al. 2014;Tunčikienė, Drejeris 2015), but little interest was given to the impact of corruption on stock market development in Gulf Countries Council (GCC), the richest countries in Middle East.
Country Risk Guide, and Transparency International's Corruption Perception Index (CPI).
One of the most renowned indices is the Corruption Perception Index (CPI) published by Transparency International. This index is defined as "abuse of public power for private benefit" (see Freckleton et al. 2011); it is an aggregate indicator that classifies countries based on the degree to which corruption is perceived to exist among politicians and public authorities.
The aim of this paper is to examine the impact of cor ruption on stock market development in GCC. We explore the correlation between CPI and stock market development; employing market capitalization as a proxy for stock market development. The study covers a sample period from 2003 to 2011 of six members of GCC, using pooled regression, random and fixed.
The main contribution of this paper to the ongoing dis cussion in the literature lies in two points. First, it investi gates the impact of corruption on stock markets develop ment focusing exclusively on GCC countries. It is interesting to investigate this relationship in such group of homoge neous countries considering the unique characteristics of these countries combining richness with relatively high level of corruption. This is important since many studies neither distinguish between developing and developed countries in their tests nor consider the special characteristics of certain groups within each category. This is reflected in many challenging and conflicting results. Second, this pa per challenges the view that perceptionsbased measure of corruption is not good enough as a measure of corruption. Perceptionbased measures have been criticized on argu mentative and empirical grounds, causing some researchers to resort to other measures or proxies that fit better into their econometric models. However, this does not seem to be the case with GCC as perceived levels of corruption perform rather well in explaining the variation in financial market development.
The paper is structured as follows. Section two provides a brief review of theoretical and empirical Studies on the impact of corruption on stock market development. Section three presents descriptive statistics and explain research methodology. Section four reports regression results and section five concludes.

Theoretical literature
While there is a large consensus in the literature on the negative impact of corruption on economic growth, some researchers continue to argue that corruption has a moti ve factor on growth; there are two competing approaches on how the corruption may affect the economic growth: corruption greases the wheel of economy and corruption sands the wheel of economy.
In theory; corruption may not necessarily be bad for stock market development, Early studies documents a posi tive impact of corruption on stock market development (see Leff 1964;Lui 1985). The pioneering theoretical work of Leff (1964) stated that corruption works like the engine of economic growth in the situation when strict/inefficient regulations forced by the government because corruption "Enables the privates agents to buy their way out of politi cally imposed inefficiencies". Ahlin and Pang (2008) claim that corruption raises the need for liquidity and thus makes financial improvements more potent.
Corruption may hold a positive impact on stock market development through its impact on FDI, acting as "grease" by expediting transactions and allowing private firms to overcome ineffective regulations and governmental institu tions (see Leff 1964;Huntington 1968;Bardhan 1997;Egger, Winner 2005;CuervoCazurra 2008;Aidt 2009). Lau et al. (2013) argue that bribery (a proxy of corrup tion) helps firms to overcome the inefficiencies in the eco nomic system and decreases uncertainty, which may lead to a positive performance. Furthermore, Chêne (2014) argues that corruption can overcome red tape and institutional weaknesses and "grease the wheels" of the economy in high ly regulated countries that do not have effective government institutions and governance systems.
In contrast, a number of studies suggest there is a nega tive impact of corruption on stock market development. Mashal (2011) argues that corruption loots economic growth by decreasing competition in domestic market which impairs efficiency of domestic and foreign firms. Moreover, corruption makes it more difficult and costly to conduct foreign operations through obtaining licenses and permits (see Habib, Zurawick 2002;Voyer, Beamish 2004;CuervoCazurra 2008). Ng (2006) argues that managers might participate in projects that they otherwise would not just so they can ac cept bribes creating waste and increasing transaction costs in the economy. Moreover, corruption may negatively affect stock market development through its influence on FDI. Wei (2000), Lambsdorff (2003), and Voyer and Beamish (2004) find a negative relationship between a host country's corruption and the FDI it receives. Other academics refer to corruption as "sand in the gear" (see Jain et al. 2012;Lau et al. 2013) viewing government officials as bribetakers who will try to set up hurdles for businesses so that they can extract more bribery.
Some studies suggest asymmetric impact of corruption on stock market development on the basis of the country's development, i.e., whether emerging or developed. De Rosa et al. (2010) find that corruption has a negative effect on firm productivity in Central and Eastern European countries. Hillman and Krausz (2004) show that corruption brings short term performance advantages. For example, Wang and You (2012) suggest that corruption is likely to increase firm growth when financial markets are underdeveloped whereas it deters firm growth when there are more devel oped financial markets. On the same vein, Pinheiro (2010) suggests that, in more developed countries, corruption is inversely related to stock market returns while, in develop ing economies, higher levels of corruption impact positively on stock markets returns In contrast, Olken (2007) finds that corruption can seriously hamper the distributive efforts of developing countries. That is, the growth impact of reducing corrup tion is higher when the financial system is less developed. Conversely, a rise in corruption is more prohibitive in a less developed financial system (see Ahlin, Pang 2008).
A contradictory view argues that in developed econo mies like the UK, corporate bribery is not that obvious as in the emerging markets (Lau et al. 2013), so a financially developed country is hurt less by a given increase in cor ruption, since funds can be borrowed more readily (see Ahlin, Pang 2008).
Another strand of the empirical literature suggests that corruption plunders economic development/growth by increasing the cost of business and also uncertainty in the decision making process. Since Asset prices are determined based on future discounted cash flows; therefore, they are good measures for evaluating the cost of corruption from the investors' points of view (see Ciocchini et al. 2003). Hence, corruption should increase the firm's borrowing cost, decrease stock valuation and stock market develop ment (see Ng 2006).
Corruption does not only affect stock market devel opment but also affects its volatility. Pastor and Veronesi (2012) argue that if investors consider bribery as a resource for firms, it decreases uncertainty about government poli cies and helps overcome the inefficiencies in the country. In that sense, bribery may lower stock market volatility es pecially in emerging markets.

Empirical studies
Similar to the theoretical literature on corruption and stock market development, the findings from the empirical stu dies on the impact of corruption on stock market develo pment are mixed and conflicting.
A good starting point for this empirical literature review is with the paper of Ayaydın and Baltaci (2013) who focus on the role of corruption and banking sector development on stock market development using a panel data of 42 emerg ing economies for the period 1996 to 2011. They found that corruption level is significantly and negatively associated with stock market development. Bolgorian (2011) analyzed a data set of corruption and stock market development measures such as market capi talization and total value of share trading for 46 countries around the world for the period 2007-2009, using a quan titative approach for investigating the dependence of the Corruption Perception Index (CPI) on stock market devel opment. He found that countries with higher relative stock market development are less corrupt, and the powerlaw relation between level of corruption and stock market de velopment is significant at the 5% level.
Abdul Qadir and Yaroson (2013) highlighted the role of selected macroeconomic variables and corruption in ex plaining stock market development in Nigeria for the pe riod 1998-2011. Employing the Augmented Dickey Fuller (ADF) unit root test to analyze the stationary properties of the variables adopted in the study, they found that corrup tion has significant impact on the development of the stock market; the turnover ratio as a share of GDP which is used to test the market liquidity, has a negative impact on the stock market; and the real interest rate, foreign direct investment and value of shares traded to be significant in determining stock market development in Nigeria. Cherif and Gazadar (2010) examine the relationship between institutional indicators and stock market devel opment using International Country Risk Guide (ICRG) to proxy for corruption. Using data from 14 MENA coun tries for the period from 1990 to 2007 and applying panel data and instrumental variable methods, they find there is a negative relationship between corruption and stock market development. Yartey (2010) examines the impact of corruption as a part of institutional determinants of stock market devel opment using a panel data of 42 emerging economies for the period 1990 to 2004. He finds find there is a negative relationship between corruption and stock market develop ment. He also proves that macroeconomic factors such as income level, gross domestic investment, banking sector de velopment, private capital flows and stock market liquidity are important determinants of stock market development in emerging market countries. Shahbaz et al. (2013) examine the long run relationship between financial development, corruption and economic growth in Pakistan using ARDL bounds testing approach and applying cointegration tests and VECM granger causal ity method to examine the direction of causality between the variables for the period of 1987-2009. They find that a rise in corruption has a positive impact on financial development.

Sample and data
The study made use of data which were sourced from the Transparency international organization, World Development Indicators (WDI), and British Petroleum (BP) Statistical review of world energy 2014.
In order to test the relation between corruption and stock market development, we use the Corruption Perception Index (CPI), Market Capitalization (MC), Oil Price, Foreign Direct Investment (FDI), Domestic credit to private sector (% of GDP), and GDP growth (G) for Gulf Cooperation Council (GCC) countries (Bahrain, Kuwait, Oman, Qatar, Saudi Arabia, and the United Arab Emirates) for the period 2003-2011 with 53 observations. The study period is limited to 2011 because the methodology of calculating CPI has been updated in 2012 by Transparency International. This means that CPI scores before 2012 are not compatible with those of 2012 and beyond. 1 The following are the variables specified in the study so as to achieve the primary objectives of the study.

Corruption (CPI):
This paper employs the Corruption Perceptions Index (CPI), published by Transparency International, as pro xy to measure the level of corruption. CPI is compiled by Transparency International and it draws on surveys from different businessmen, country experts, international institutions and the local populace to report the extent of perceived corruption amongst government officials and politicians. The index is scaled from 0 to 10 where 0 means that a country is perceived as highly corrupt and 10 means it is perceived as the least corrupt (very clean).

Market capitalization (LNMC):
This paper employs the natural logarithm of capi talization ratio to proxy stock market development. Capitalization ratio is calculated as the value of domestic equities traded on the stock exchange relative to GDP. This index is obtained from the World Development Indicators (WDI) issued by World Bank.

Oil price (OILPRC):
The variable that we use to proxy for oil prices is crude oil price compiled by British Petroleum (BP) Statistical re view of world energy 2014. Given that GCC economies are oildependent, it is plausible that oil price volatility is an important determinant of stock market development. In real terms, the oil price drastically decreased in the period 2003-2011 (with the exception of a price peak in 2009).

Foreign direct investment (FDI):
Foreign direct investment is defined as investment made by a company or entity based in one country, into a com pany or entity based in another country; Open economies with skilled workforces and good growth prospects tend to attract larger amounts of foreign direct investment than closed, highly regulated economies. We use foreign direct investment as a proxy of business environment. This varia ble is extracted from WDI of World Bank.

Domestic credit to private sector (% of GDP) (DCP)
In the last few decades, private sector has emerged as major participants in emerging stock markets. The domestic credit to private sector (DCP) indicates financial resources provided to the private sector by financial corporations. This variable is a natural candidate to proxy banking system activity and is computed as percentage of domestic credit to private sector as compared to GDP.

GDP growth (G)
This variable is obtained from the World Development Indicators (WDI) issued by World Bank to proxy economic growth. Table 1 provides a summary statistics for the above mentioned variables for the sample period.  Table 1 clearly show a considerable level of variation in almost all variables across all countries included in the sample. Figure 1 depicts the behavior of each variable, by country, over the study period. CPI shows that although GCC countries have almost similar levels of corruption, two countries (United Arab Emirates and Qatar) enjoy lower levels of corruption compared to the remaining four countries.
Economic variables show unstable performance across time in Figure 1. The generally declining economic growth rate shows some improvements beginning from 2009. Each country seems to develop its unique story in terms of its ability to attract foreign direct investment. On the other hand, financial variables seem to follow a more homogenous pattern across countries. As far as market capitalization is concerned, the general trend is downward sloping, and the last few years of the series indicate a continuing decline towards belowaverage levels.

Methodology
The literature suggests a wide spectrum of variables to proxy those elements of development that promotes the expansion of financial market activities. In practice, de velopment research suffers numerous data issues, and the availability of data dictates the number of alternatives at the disposal of the researcher to a considerable extent. 2 On the other hand, financial market development is measured in a number of ways. The common measures are the value of shares traded, market capitalization, and the turnover ratio value of shares traded, all being taken as a percentage of GDP. 3 It is not easy to isolate explanatory variables specific to each of these measures. Macroeconomic performance, for example, is expected to affect them all positively, but no distinction can be made on how this effect may differ for each. Therefore, the practice of regressing a variable that measures financial market development on a number of 3 These are available from the World Development Indicators dataset. suggested explanatory variables remains an empirical issue on both sides of the equation. The fact that corruption is also measured in different ways by different institutions adds another complication.
The usual starting point, the simplistic case, is to try the OLS regression, where the model can be written as: This form assumes neither time effect nor heterogeneity or individual effects (i.e. countryspecific effects, u i ). The above model treats all countries in the dataset as identical individuals. That is, it assumes that there is nothing unique about any country, compared to others, that might have an effect on its financial market development. If individual country or time effects do not exist, then OLS is expected to produce efficient and consistent parameters estimates. Obviously this is not necessarily the case, as it implies that withinindividual errors are uncorrelated. For example, if some unobserved factor such as business orientation of a given country, which is not directly captured by the in cluded explanatory variables, is higher than GCC average in a given year, then it should be the case that it is higher than average in other years too. A populationaveraged estimation of the same model with clusterrobust standard errors relaxes this restrictive assumption and leads to more efficient estimators.
The most widely used panel regressions are the fixed effect and the randomeffect models. While a fixedmodel treat this heterogeneity though allowing each individual to have its own intercept, a random model deals with hetero geneity through allowing each group of countries to have its own specific error variance. The fixed model in our case can be written as follows: Under this model, constant slope and constant variances are assumed across all countries included in the sample. Practically, an individual effect is a part of the constant, and it is allowed to be correlated with other variables included in the model.
The fixedeffects model (Eq. (2)) assumed that all coun try differences were captured by differences in the intercept parameters to be estimated using least squares estimator. Unlike the fixed effect model, a random effects model does not allow the intercept to vary across individual countries, but rather assumes a common (average) intercept for all countries and recognizes that these countries are a sample of a wider population, so that individual differences are treated as random rather than fixed. In other words, the intercept is considered to be composed of two parts, "fixed" population average (a) and random differences around the average (u i ). The model can then be written: where these random effects are assumed to have a zero mean, a constant variance and zero correlation across in dividuals. Hence, each individual country to express its own heterogeneous though having its own error variance, i.e. its own component of the composite error term that combines the usual regression disturbance term and the random individual effect. 4 Although the dataset is way far from rich, it is still wor thy to try a randomcoefficient model where randomness in coefficient is applied to cpi only.

Results, discussion and limitations
Regression outputs for a number of alternative loglinear forms / specifications are summarized in Table 2. Columns (1), (2), and (3) report results for three pooled OLS regres sions. While Column (1) drops g from the list of explana tory variables, Column (2) drops out dcp, with Column (3) dropping both, so that only significant coefficients are included. These two variables are significantly correlated, and, therefore, do not appear together in any of the reported models. Next, each successive block of three columns fol lows the same specification on a different panel model. That is, Columns (4) through (6) report panel regression results for the same three specifications on a populationaveraged model (robust error). Next, Columns (7) through (9) report results from the application of fixedeffect model, Columns (10) through (12) provides a similar report from the esti mation of a randomeffect model, and, finally, the last three columns do the same job for a randomcoefficient model, where coefficient randomness is assumed to apply only to the coefficient of our concern, the level of corruption (cpi). To preserve space, results from major statistical tests and other important statistics are reported in the lower part of the same table at the columns corresponding to the model to which the test was applied. The final row indicates that the sample size is equal across all the regressions reported.
Obviously, the results from the pooled OLS regression are not very impressive, with an R 2 ranging between 0.17 and 0.18. However, the Ftest strongly rejects the joint hy pothesis that all coefficients are not significantly different from zero. The coefficient of oil price is quite modest in magnitude and significant at the 5% level, but its sign needs some attention. Naturally, one would expect that higher oil prices result in more generous public investment spending, as public revenues increases. Indeed, GCC public budgets are obviously tied to oil revenues, but extra revenues do not necessarily seem to translate into expansion in public invest ments subcontracted to firms listed on stock exchange. In fact, higher oil prices do make richer public budgets, but better economic performance in this region in general has been dependent on political stability. Oil shocks are rather harmful to investors' confidence, as it causes entrepreneurs to reconsider their plans, and confuses financial market players. Therefore, a negative relation between oil prices and financial market development is more reasonable than one would expect. However, the magnitude of this coefficient is rather small and almost insensitive to the inclusion of other control variables such as the rate of economic growth.
The coefficient of foreign direct investment, fdi, which reflects the development of business environment, has the expected sign. The magnitude of this coefficient decreases slightly with the dropping of control variables (g and dcp). The coefficient of our focus, corruption index, is signifi cantly different from zero at the 1% level of significance, and its magnitude is reasonable and robust against the inclusion of other control variables. 5 The above model treats all countries in the dataset as identical individuals. That is, it assumes that there is noth ing unique about any country, compared to others, that might have an effect on its financial market development. If individual country or time effects do not exist, then OLS is expected to produce efficient and consistent parameters estimates. Obviously this is not necessarily the case, as it implies that withinindividual errors are uncorrelated. For example, if some unobserved factor such as business ori entation of a given country, which is not directly captured by the included explanatory variables, is higher than GCC average in a given year, then it should be the case that it is higher than average in other years too.
A populationaveraged estimation of the same model with clusterrobust standard errors relaxes this restrictive assumption and leads more efficient estimators, as can be seen from Column (4) through (6). Here, country specific effects are assumed to be random and are averaged out. It is clear that the coefficients differ considerably from those of the pooled OLS. All the coefficients in Model (6) are sig nificantly different from zero at the 1% level of significance. Their magnitudes as well as their efficiency are almost totally insensitive to the introduction of either g or dcp. As far as our focus variable is concerned, this indicates that, on average, an extra point on the ladder of corruption is associated with almost 0.18 point increase in logged market capitalization.
Considering the value range of these two variables, the effect of corruption on market capitalization cannot, indeed, be described as a negligible effect. However, while a clusterro bust populationaveraged estimator outperforms the basic pooled one, it is known that its standard error is not fully efficient. Panel regressions are usually used in accounting for heterogeneity across groups (i.e. countries) while not violating OLS assumptions. Columns (7) through (9) report regression output for the fixed effect model (Eq. (2)). The Waldtest strongly re jects the OLS model in favor of the fixed effect model. The magnitudes of the coefficients are considerably different from those of OLS in general, and all coefficients, except g and dcp are significantly different from zero. The magnitude of cpi is fairly close to its populationaveraged counterpart.
Columns (10) through (12) report the estimation results for Eq. (3). The coefficients are quite close to those of the fixedeffect model. The value of the statistic rho indicates that about 89 percent of the total error variance is attrib uted to country heterogeneity, also indicating a good level of goodness of fit. 6 The test statistic of Breusch and Pagan LM test for random effects is estimated at Chibar 2 (01) = 121.9 which is significant at the 0.001 level of significance, indicating a strong rejection of the OLS model estimates in favor of the randomeffect model. Therefore, both fixed and random effects are superior to the pooled model. A common test used in model selection between these two models is the Hausman test. The null hypothesis of this test states that the difference in coefficients between the two models is not systematic. Applying the Hausman test gives a test statistic the follows a Chi distribution with four degrees of freedom, Chi 2 = 0.10, which is too small to reject the null hypothesis of the test. Therefore, the test says that it is better to fit a randomeffect rather than a fixed effect model. However, although statistical testing tends to favor the randomeffect model, there is a concern that our coefficient of interest might change significantly across groups (of countries), so a Swamy's test can be used to test constancy of the param eters. The test statistic for this test is Chi 2 = 200.8 which is significant at the .001 level of significance, indicating a strong rejection of parameters' constancy.
This result leads us to investigate the application of a randomcoefficient model. Columns (13) through (15) from a randomcoefficient regression, where randomness in coefficient is applied to cpi only. 7 The values of the coef ficients are generally quite close to the results obtained from the previous two models. All     Note: standard errors in parentheses (* p < 0.10, ** p < 0.05, *** p < 0.01).

Cpi
against linear regression clearly rejects the liner regression in favor of the random coefficient model. The random ef fect parameter for cpi is significantly different from zero and both variance components are statistically significant. However, as far as our focus coefficient, cpi, is concerned, the coefficient is obviously significant and robust across all competing specification, a matter that enhances the view that corruption does grease the financial wheel in GCC countries.

Conclusions
This paper investigates the impact of corruption on stock market development focusing exclusively on GCC coun tries. It provides empirical evidence, from the Middle East, on the effect of corruption on financial market, where the former is measured by CPI of Transparency International and the latter by market capitalization relative to gross do mestic product.
Econometric analysis of this relation shows that corrup tion in GCC countries is positively associated with financial market development, and therefore confirms the view of many previous studies (see Leff 1964;Huntington 1968;Bardhan 1997;Egger, Winner 2005;CuervoCazurra 2008;Aidt 2009) that corruption greases the wheel of financial development.
As far as GCC countries are concerned, the frequently criticized measure of corruption based on public percep tions, CPI, can actually explain considerable variation in financial market capitalization during the study period. This result challenges the argumentative view that recommends avoiding the use of perceptionsbased measure of corrup tion in empirical analysis.
Given the mixed empirical findings obtained from other studies applied to different economies, it does seem that the relationship between corruption and financial markets does not necessarily take a similar form across different econo mies at a particular time. However, it should be emphasized that despite data limitations, the relation between corrup tion and GCC financial market development is statistically significant and robust against changes in model forms and specifications.
Finally, it should be noted that the main element that complicates the empirical study of the relation between cor ruption and financial markets development is the fact that both sides of this relation are being described in the litera ture by different competing measures. Therefore, future re search should pay more attention to the nature of differences between different types of measures, particularly those of corruption. Until a specific measure of corruption proves superior, the choice of corruption measure will remain an empirical question. Future research should also try to ex pand the sample to other countries in the region covering a longer time period, a matter that requires better access to economic and financial data for countries of the region.

Disclosure statement
Authors declare that they don't have any competing finan cial, professional, or personal interests from other parties.