Investment and corruption: a look at causality.

ABSTRACT

Using the Granger test on pane/data of corruption and national accounts for 58 countries, this paper investigates causality between investment and corruption. While controlling for country and time fixed-effects and ensuring against the existence of a unit root of each time series, we find strong evidence of investment Granger-causing corruption. This finding is robust up to four lags of the investment variable. The F-test of the fixed-effect variables suggests that time and place have no effect on causation. We find no evidence of reverse causality of corruption on investment. Using a similar mode/for economic growth and corruption, we find no evidence of causality, nor reverse causality.

1. INTRODUCTION

Levels of corruption, globally, are arguably on the rise (Transparency International, 2001, Tanzi and Davoodi, 1999). Its effect on economic variables is widely accepted as negative. However, the determination of causal relationships between national investment and corruption, or between corruption and economic growth is pending and paramount. To economists, the misspecification of the corruption variable in an econometric model can lead to false conclusions. To policy makers, knowing the direction of causation, if it exists, will help formulate appropriate policies and economize on the use of scarce resources, particularly in the case of developing or emerging economies. Abed and Davoodi (2000) find that macroeconomic measures tend to dominate the level of corruption in transition economies.

Contemporary economic thought regards corruption as investment and economic growth-inhibitive. This finding is a recent contribution from the proliferation and usage of various proxies of corruption. Using a bureaucratic efficiency index as a proxy for actual levels of corruption and the ethnolinguistic fractionalization statistic as an instrument variable in a two-stage least squares model, Mauro (1995, pg. 705) concludes that corruption "causes" reduced levels of investment. Most other studies have not gone as far as to suggest a "causal" relationship. However, many researchers (Leite and Weidmann (1999), Tanzi, and Davoodi (1997) and Kaufmann and Wei (1999), and Mo (2001)) have used similar models to show that higher levels of corruption are associated with lower levels of investment and lower economic growth.

To date, there is one direct test of causality relating corruption to investment or corruption to economic growth. Alberto and Calderon (2000) find that, for developing countries, the quality of institutions not only causes increased economic growth, but economic growth also causes institutional quality (improvements in corruption). The fact that there is little work in this area by economists adds credence to the notion that econometricians tiptoe around this delicate issue. Zellner (1979, pp. 9), in his assessment of the subject of causality, makes this point when he criticizes econometricians for "... excluding terms like causality and cause from their textbooks."

This paper will further explore the relationship between corruption and investment by testing for a causal relation using the Granger-Wiener test on panel data. An alternate means of determining causality between two variables would be to use a simultaneous equations system. This model, however, lacks the time ordering element critical in determining causation. In so doing, this model primarily determines instantaneous causation. Moreover, simultaneous models are not concerned with the non-stochastic nature of the relevant variables. In his seminal paper on causality, Granger (1969, p. 430) argues that the flow of time is a natural argument and serves as a rational foundation and "... plays a central role" in defining causality, since the future cannot predict the past. This model differs from the simultaneous equation model that determines endogeneity from the inclusion of minimally a third variable. These models are very sensitive to the instrument variable used and can often lead to false conclusions. An example of this misidentification can be found in Mauro (1995) and Leite and Weimann (1999). Each author uses a different instrument variable to arrive at a different conclusion on endogeneity.

This paper will proceed with a brief discussion on causality, followed by a discussion on the data and the Granger causality test. The next section proceeds with a stationarity test on each identified series. This will followed by a discussion of the empirical results and a brief conclusion.

2. A DISCUSSION ON CAUSALITY

According to Feigl (1953) and Jefferys (1967), causation cannot occur without a theoretical underpinning or a set of laws. The authors referenced above maintain that employing a statistical methodology to determine causation could simply be a theoretical exercise, void of any credibility. This implies that using Granger's (1969) definition of causality, without a just theoretical basis, is quite dangerous. An illustrative example can be found in Gujarati (1995, pg 622). In this example, Gujarati performs a Granger causality test of auto sales and the Treasury bill rate and finds that using four lags of both variables, auto sales Granger causes the rates on Treasury Bills. Albeit a valid statistical outcome, this result is void of economic theory as it is obviously more plausible to conceive the opposite; that Treasury bill rates Granger-cause auto sales.

There are a myriad of theoretical causes of corruption conjectured in the economic and political science literature. Rose-Ackerman (1999) implies, from the list of preventive measures prescribed, that the level of corruption is endogenously determined from the organizational structure of a political regime. Shleifer and Vishny (1995) also use this (organizational) structural model of corruption as the central basis for the level of rent-seeking activity found in an economy, vis-a-vis market and subsistence forms of production in developing economies. Perrson (2002), in turn Myint (2000), argue that political institutions shape fiscal policy, which has a net effect on the incidence of corruption. However, the conjecture widely touted as a plausible explanation of the cause of corruption (Tanzi (1998) and others) argues that the level of corruption is highly, inversely related to low wages of public sector employees. That is, owing to the lower opportunity cost of being detected, caught and punished, lower-waged public sector have a higher probability of engaging in corrupt behavior, i.e. receiving bribes. If this theory holds true, one corrective course of action is to simply increase public sector wages, financed with an increase in private investment.

3. BASIC GRANGER MODEL AND DATA

In its simplest form, the Granger test states that given two stationary series {[y.sub.t]} and {[x.sub.t]}, and the equations:

(1) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],

(2) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII];

where [[epsilon].sub.t] and [[eta].sub.t] are two un-correlated white-noise series, [x.sub.t] is said to Granger-cause [Y.sub.t] ([x.sub.t] [??] [y.sub.t]) if any [b.sub.j] [not equal to] 0. Likewise, [y.sub.t] is said to Granger-cause [x.sub.t] ([y.sub.t] [??] [x.sub.t]) if any [d.sub.j] [not equal to] 0. However, if the former and latter are true, then there is feedback (bilateral causality) between the two series, and causality is most likely caused by a third variable.

The author utilizes the basic Granger model above on balanced panel data taken from the Penn World Table and the ICRG for 58 countries, over years 1982 to 1992. The ICRG corruption index takes on a value from 0 to 6, with 0 identifying the highest level of corruption, and 6 the least. The investment and economic growth are taken from Penn World Table (PWT 5.6). The ICRG table is selected since it is the only series available to the author that span the same time frame as the national accounts data. To control for spurious results that can occur from a missing variable bias, the author will allow for country-specific and time-specific differences by including two vectors of dummy variables: one for time and the other for country-specific effects.

The equations to be estimated are:

(3) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

(4) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

(5) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

(6) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

where [corrupt.sub.i,t], [growth.sub.i,t] and [Inv.sub.i,t] are the level of corruption, economic growth and investment in each country in each year, respectively. [D.sub.i,t] and [Y.sub.i,t] are dummy variables for country and year, respectively, [z.sub.l,t], [y.sub.l,t],, [x.sub.l,t] and [w.sub.l,t] are white noise error terms.

The principal assumption of the granger test is stationarity of each series used. In the presence of a unit-root, the regression coefficients will produce spurious results, leading to false conclusions. To test for stationarity, the author will use a unit-root test for heterogeneous panel data sets developed by lm et al (1995).

4. STATIONARITY TESTS

The test for the existence of a unit-root in a corruption series {[C.sub.t]} is a straightforward process. Suppose that [C.sub.t] is defined in the following way:

(7) [C.sub.t] = constant + [alpha] [C.sub.t-1] + error.

If the null hypothesis that [alpha]=1 cannot be rejected, then the series contains a unit-root. Similarly, subtracting [C.sub.t-1] from both sides of equation (7) produces

(8) [DELTA] [C.sub.t] = constant + [beta][C.sub.t-1] + error,

the most common form of the unit-root test. In this form, note that [beta]=[[alpha]-1] and hence, the null hypothesis is that [beta]=0 with a Dickey-Fuller distribution. The null hypothesis is rejected if the t-statistics exceeds the critical value determined by the Dickey-Fuller table.

The past decade has witnessed a burgeoning of unit-root tests for panel data sets, with Quah (1994) being the most common in the literature. However, their simplifying assumptions of no serial correlation across time and cross-sectionally make them quite limiting in their application. In the case of the three series that the author plans to use, there are plausible grounds to suspect serial correlation and heteroskedasticity cross-sectionally. To control for these possible occurrences, the author will employ a unit-root panel test for heterogeneous data sets developed by Im et al (1995).

The Im et al (1995) test produces a t-bar statistic from the augmented Dickey-Fuller tests produced for each panel section. This t-bar statistic has a standard normal distribution under the null hypothesis of a unit root. The augmented Dickey-fuller tests are constructed for each section using the regression equation:

(9) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

where [y.sub.i,t] = [x.sub.i,t] - [bar]xt, [bar]xt = 1/N [SIGMA] [x.sub.it], [DELTA][y.sub.it] = [DELTA][x.sub.it] - [bar][DELTA]xt, [bar][DELTA]xt = 1/N [SIGMA][DELTA] [x.sub.it] and x refers to each series in the panel data, i.e. corruption, investment or growth. The results of the de-meaned ADF(1) equations are listed in Table I. Note that based on the 58 equations, only a limited number of countries' corruption, growth and investment series, respectively, would have passed the unit root test at the 5% level of significance. Using the standard unit root test on these countries would severely reduce the degrees of freedom and increase the standard errors of our estimated coefficients. This reduction would severely reduce the power of the Granger tests.

Table II lists the results of the lm et al. (1995) panel unit root test. Note that the null hypothesis of a unit root is rejected at the 1% level of significance for each of the series in question. Therefore, we can conclude that the panel data of the three series: corruption, investment and economic growth do not contain a unit root. This important conclusion allows us to fulfill the stationarity assumption of the Granger causality test, and thus reach a necessary condition for its usage.

5. REGRESSION RESULTS

Equations (3) through (6) are estimated and the results listed in Tables III and IV, For each equation, each lag is allowed to enter the equation sequentially, so as to observe its marginal effect. To economize on space and since the autoregressive lags of the left hand-side variable are trivial, only the causal variables whose coefficients are tested for causal relationships are listed. Table III produces strong evidence that investment Granger causes corruption. The F-tests reject the null hypotheses that the investment coefficients are jointly zero as their values are 5.75, 2.25 and 4.10, respectively, for specifications (1), (2) and (3). These values are all significant at the 1% level. The F-statistic of the joint significance of the lags of investment coefficients in specification (4) is significant at the 10% level. However, note that the sign of the coefficient alternates from negative to positive, making a net impact more difficult to discern. To shed better light on this issue, an F-test on the sum of the coefficients is also provided. This F statistic has a value of 7.59 and is statistically significant at the 1% level and the sign of the coefficient of the sum of investment lags variable is positive. All four specifications produce findings to suggest that an increase in the investment level Granger causes a reduction in the level of corruption. The F-statistics of specifications (5) through (8) of Table I have values of .23, 1.01, 1.18 and .47, respectively. The null hypothesis that the coefficients of the lags of the corruption variable are jointly statistically different from zero cannot be rejected at the standard levels of significance. Therefore, we can reject the possibility of feedback between corruption and investment for the full sample of countries.

The F tests of the coefficients of the fixed effects suggest that causal effect of investment on corruption in the USA is different that from that of other countries for the first two lags. This distinction dissipates after lags three and four enter specifications (3) and (4).

Table IV lists the results for equations (5) and (6). The F statistics in specifications (1), (2), and (3) are .03, .49 and 2.00, respectively. Accordingly, the null hypothesis that the causal coefficients (lags of growth) are not jointly statistically different cannot be rejected at the standard level of significance.

The F statistics in specification (4) is 4.29, statistically significant at the 1% level. This rejects the null hypothesis that the coefficients of the lag terms are jointly equal to zero. Again, the reversal of sign of these coefficients and the magnitude of these coefficients suggest a net zero effect on corruption. As a test, the F statistics that the sum of the growth coefficients is zero is .09 and statistically insignificant at the standard level. Therefore, the sum of the lagged growth terms has a net zero effect on corruption. The F statistics of a reverse causal effect on economic growth are .23, 1.01, 1.18 and .47 for specifications (1), (2), (3) and (4), respectively. These statistics reject the theory that economic growth Granger causes corruption.

6. CONCLUSION

The effect of corruption on investment and economic growth is widely assumed to be negative. The past decade has witnessed a burgeoning of empirical works highlighting this negative relationship between corruption and investment. However, the causal relationship of corruption and the two variables have received little attention. This paper attempts to blaze this trail by performing a Granger Causality Test using panel data. Using data for 58 countries over 12 years, we find that investment Granger-causes corruption. An increase in investment causes reduced levels of corruption, and not vice versa. This finding is statistically highly significant and robust over four lagged terms. We find no such relationship between corruption and economic growth, or any sign of reverse causality of economic growth on corruption.

This finding lends support of the hypothesis that corruption is an endogenous variable in an economic model. Moreover, the goal of reducing corruption can be obtained simply by increasing investment. This is important information for policy makers seeking to maximize the welfare of their constituents. Utilizing scarce resources to develop anti-corruption measures may not be very efficient. It would be a more efficient use of scarce public resources to stimulate investment; doing so will increase standard of living and reduce corruption.

Table I. AUGMENTED DICKEY-FULLER STATISTICS

                      [[tau].sub.[mu]]

Country               Corrupt             Inv       Growth

Algeria                -2.37              -2.01     -0.99
Australia              -4.16              -2.68     -2.15
Austria *             -14.93              -2.20     -3.96
Bangladesh              0.31              -4.93     -0.52
Belgium *              -0.83              -1.36     -0.89
Bolivia                -1.00              -0.75     -2.25
Brazil                 -2.00              -1.48     -2.86
Cameroon               -1.26              -0.92     -1.43
Canada *               -1.59              -2.23     -1.84
Chile                  -5.16              -4.05     -0.76
Colombia               -1.59              -2.66     -2.86
Costa Rica             -1.59              -7.61     -2.08
Denmark *              -1.59              -1.10     -1.68
Dominican Republic     -1.59              -3.31     -1.31
Ecuador                -4.16              -3.39     -1.87
Egypt                  -1.64              -2.64     -0.81
EL Salvador            -1.59              -2.62     -1.19
Finland *              -1.59              -3.55     -7.95
France *               -1.92              -1.35     -4.27
Gabon                  -0.34              -2.79     -1.12
Germany *              -5.72              -2.32     -0.56
Ghana                  -1.65              -2.48     -2.53
Guatemala              -1.59              -1.19     -1.49
Honduras               -1.59              -3.71     -1.03
Hong Kong              -4.16              -3.37     -2.40
Iceland                -1.59              -1.02     -2.36
India                  -0.72              -1.58     -2.27
Indonesia              -1.65              -2.26     -3.32
Iran                   -3.25              -1.99     -2.24
Ireland *              -4.16              -1.84     -1.20
Israel                 -1.59              -3.14     -2.06
Italy *                -8.38              -1.08     -6.24
Kenya                  -1.97              -1.94     -0.54
Malawi                  0.41              -2.00     -3.83
Malaysia               -1.88              -2.25     -2.69
Morocco                -1.62              -3.68     -2.29
Netherlands *          -1.59              -3.64     -2.25
New Zealand *          -1.59              -1.61     -5.25
Nigeria                -9.39              -1.44     -3.22
Norway *               -1.59              -2.34      0.57
Pakistan               -2.43              -2.01     -1.60
Panama                 -4.16              -1.63     -1.88
Paraguay               -1.71              -3.65     -0.51
Peru                   -9.39              -2.64     -2.56
Philippines            -1.83              -2.91     -4.14
Singapore              -0.81              -2.97     -1.91
Spain *                -2.40              -1.43     -1.56
Sri Lanka              -1.59              -2.71     -2.54
Switzerland *          -1.59              -0.05     -1.26
Thailand               -4.16              -1.67     -0.48
Togo                   -1.59              -1.52     -0.48
Tunisia                -1.66              -1.28     -1.78
Turkey *               -2.57              -3.10     -0.09
U.K. *                 -0.65              -0.94     -2.05
U.S.A. *               -5.16              -2.93     -2.53
Uganda                 -1.09              -2.80     -3.52
Uruguay                -1.59              -2.68     -2.56
Venezuela              -9.39              -2.05     -2.26

Notes: The numbers are estimated augmented Dickey-Fuller statistics
calculated independently from an autoregression for the corruption
index (corrupt), economic growth (growth) and Inv (inv/GDP);
[[tau].sub.[mu]] indicate the inclusion of a constant term in the
autoregression. * denotes a member of the OECD as of 1989.

Table II. Panel Unit Root Tests

           t-bar

Corrupt    -9.24 ***
Inv/GDP    -6.49 ***
Growth     -4.71 ***

Note: T-bar is a Z score of the average t-bar statistic developed by
Im et al. (1995). The t-bar statistic is defined as z=
[N.sup.1/2]([t.sup.*]-E[[t.sub.T](1,0)])/S.D.([t.sub.T](1,0)),
[t.sup.*]=average ADF(1) for all panels and [t.sub.T](1,0) is the
t-statistic of the estimated [beta] coefficient in the stochastic
simulation of equation (9). *** denote the 1% level of significance.

Table III. GRANGER CAUSALITY TESTS
INVESTMENT AND CORRUPTION

INV [??] Corruption

Dependent Variable: Corruption

                            (1)             (2)

IN[V.sub.t-1]               .009            .004
                            (2.40) ***      (0.66)
IN[V.sub.t-2]                               .009

IN[V.sub.t-3]                               (1.35)

IN[V.sub.t-4]

# of countries              58              58
Adj. R Square               .97             .96
F test of fixed effect      [49,538]        [48,478]
dummies                     1.91 ***        1.56 ***
F test                      [1,143]         [2,478]
F statistic(direction)      5.75 ***(+)     2.25 ***(+)
F statistic (of
summation of
investment lags)

                            (3)             (4)

IN[V.sub.t-1]               .004            -.01
                            (0.52)          (-0.55)
IN[V.sub.t-2]               .005            .22
                            (0.50)          (.75)
IN[V.sub.t-3]               .008            -.03
                            (1.10)          (-1.05)
IN[V.sub.t-4]                               .01
                                            (0.50)
# of countries              58              58
Adj. R Square               .96             .96
F test of fixed effect      [47,418]        [46,358]
dummies                     1.26            1.20
F test                      [3,418]         [4,358]
F statistic(direction)      4.10 ***(+)     2.13 *
F statistic (of                             [1,362]
summation of                                7.59 ***(+)
investment lags)

Corrupt [??] INV

Dependent Variable: INV/GDP

                            (5)             (6)

[CORRUPT.sub.t-1]           0.11            -.15
                            (0.92)          (-0.47)
[CORRUPT.sub.t-2]                           0.21
                                            (0.72)
[CORRUPT.sub.t-3]

[CORRUPT.sub.t-4]

# of countries              58              58
Adj. R Square               .94             .94
F test of fixed effect      [49,538]        [48,478]
dummies                     1.53 ***        1.14
F test                      [1,538]         [2,478]
F statistic(sign)           .84             .35

                            (7)             (8)

[CORRUPT.sub.t-1]           -.34            -0.30
                            (-1.05)         (-.83)
[CORRUPT.sub.t-2]           .40             .44
                            (0.87)          (0.92)
[CORRUPT.sub.t-3]           .002            -0.42
                            (0.006)         -0.86
[CORRUPT.sub.t-4]                           0.39
                                            (1.13)
# of countries              58              58
Adj. R Square               .94             .94
F test of fixed effect      [47,418]        [46,358]
dummies                     1.02            1.05
F test                      [3,418]         [4,358]
F statistic(sign)           .60             .64

Notes: The CORRUPT variable is taken from ICRG for 58 countries for
years 1982-1989. Zero and six bound this series, where zero represents
the most corrupt country and six is the least corrupt. The INV data
is taken from the Penn-World table for the same years as the corrupt
series. Coefficients significant at the 1%, 5% and 10% level of
significance are indicated with and ***, **, and *, respectively.

TABLE IV. GRANGER CAUSALITY TEST ECONOMIC GROWTH AND CORRUPTION

Growth [??] Corruption

Dependent Variable: Corruption

[Growth.sub.t-1]     .05         -0.21       -0.23          -0.28
                     (0.17)      (-0.28)     (-.67)         (-.70)
[Growth.sub.t-2]                 0.30        .26            .30
                                 (0.98)      (-.79)         (.83)
[Growth.sub.t-3]                             -.79           -1.09
                                             (-2.39) ***    (-3.27) ***
[Growth.sub.t-4]                                            .89
                                                            (2.72) ***
# of countries       58          58          58             58
Adj. R Square        .96         .96         .96            .96
F test of fixed      [49,538]    [48,478]    [47,418]       [46,358]
effect dummies       1.81 ***    1.38 **     1.06           1.20
F test               [1,538]     [2,478]     [3,418]        [4,358]
F statistic(sign)    .03         .49         2.00           4.29 ***
F statistic (of                                             [1,362]
summation of                                                .09
investment lags)

Corruption [??] Growth

Dependent Variable: Growth

[CORRUPT.sub.t-1]    0.001       .009        .005           .002
                     (0.23)      (1.42)      (0.77)         (.26)
[CORRUPT.sub.t-2]                -.008       .40            -.005
                                 (-1.23)     (0.87)         (-0.06)
[CORRUPT.sub.t-3]                            .002           -0.006
                                             (0.006)        (-0.65)
[CORRUPT.sub.t-4]                                           0.002
                                                            (.24)
# of countries       58          58          58             58
Adj. R Square        .14         .14         .18            .24
F test of fixed      [49,538]    [48,478]    [47,418]       [46,358]
effect dummies       1.44 **     1.80 ***    1.67 ***       2.25 ***
F test               [1,538]     [2,478]     [3,418]        [4,358]
F statistic(sign)    .23         1.01        1.18           .47

Notes: The CORRUPT variable is taken from ICRG for 58 countries for
years 1982-1989. Zero and six bound this series, where zero represents
the most corrupt country and six is the least corrupt. The Growth data
is taken from the Penn-World table for the same years as the corrupt
series. Only the coefficients of the presumed causal variables are
listed above.

Coefficients significant at the 1%, 5% and 10% level of significance
are indicated with and ***, **, and *, respectively.

REFERENCES

Abed, D. T. and Davoodi, H. R., "Corruption, Structural Reforms, and Economic Performance in the Transition Economies", IMF Working Papers, 2000, WP/00/132.

Feigl, H., "Notes on Causality", Readings in the Philosophy of Science, Eds. H. Feigl and M. Brodbeck, Appleton-Century-Crofts, New York, 1953.

Granger, C. W. J., "Investigating Causal Relations by Econometric Models and Cross-Spectral Methods", Econometrica, Vol 37, Issue 3, 1969, 424-438.

Gujarati, D.N., Basic Econometrics, 3rd Edn, McGraw-Hill, New York, 1995.

International Country Risk Guide, Political Risk Services. New York: Political Risk Services, 1997.

Im, K. S. Perason, Hashem M. and Shin, Y. "Testing for Unit roots in Heterogeneous Panels", Department of Applied Economics Working Paper, No. 9526, University of Cambridge, 1995.

Jeffreys, H., Scientific Inference, 2nd Edn., University Press, Cambridge, 1967.

Kaufmann, D. and Wei, S., "Does "Grease Money" Speed Up the Wheels of Commerce", National Bureau of Economic Research, Working Paper 7093, 1999.

Leite, C. and Weidmann, J., "Does Mother Nature Corrupt? Natural Resources, Corruption, and Economic Growth", IMF Working Papers, 1999, WP/96/85.

MacDonald, R., "Panel Unit Root Tests and Real Exchange Rates", Economic Letters, vol 50, 1996, 7-11.

Mauro, R., "Corruption and Growth", The Quarterly Journal of Economics, vol 110, no. 3, 1995, 681-712.

Mo, P. H., "Corruption and Economic Growth", Journal of Comparative Economics, vol 29, 2001,66-79.

Myint, U., "Corruption: Causes, Consequences and Cures", Asia-Pacific Development Journal, vol 7, n2, 2000, 33-58.

Perrson, T., "Do Political Institutions Shape Economic Policy?", Econometrica, vol 70, no. 3, 2002, 883-905.

Quah, D., Exploiting Cross-Section Variations for Unit Root Inference in Dynamic Data, Economics Letters, 44, 1994,9-19.

Rose-Ackerman, S., Corruption and Government, Cambridge University Press, New York, 1999.

Schleifer, A. and Vishny, R., "Corruption", The Quarterly Journal of Economics, vol. 108, no. 3, 1993pp 599-617.

Summers, R., and Heston, A, "The Penn World Table (Mark 5 version): An Expanded Set of International Comparisons, 1950-1988", Quarterly Journal of Economics, vol 70, 1991.

Tanzi, V., "Corruption Around the World: Causes, Consequences, Scope and Cures," IMF Working Papers, 1998, WP/96/63.

Tanzi, Vito and Davoodi, H., "Corruption, Public Investment, and Growth", IMF Workinq Papers, 1997, WP/97/139.

Transparency International, Global Corruption Report, Available at http://www.qlobalcorruptionreport.org/, 2001.

Zellner, A., "Causality and Econometrics", in Three Aspects of Policy and Policymaking: Knowledge, Data and Institutions, Series in Carnegie-Rochester Conference Series on Public Policy, Eds. K. Brunnerand A. H. Meltzer, North Holland Press, 1979.

Dr. Ernst Coupet, Jr. earned his Ph.D. at the University of Illinois at Chicago in 2001. Currently he is an assistant professor of finance at Chicago State University.