Corruption and the size of government: causality tests for OECD and Latin American countries

Abstract

The purpose of this article is to examine the causality between government size and corruption, and to verify if there is a different pattern of causality between developed Organization for Economic Co-operation and Development (OECD) countries (excluding Mexico) and developing countries (Latin American countries) during the period 1996 to 2003. Applying Granger and Huang's (1997) methodology we find evidence that size of government Granger causes corruption in both samples. Since a larger government involvement in private markets today will be followed in future by a higher level of corruption a policy advice would be to enhance governance. The promotion of good governance helps to combat corruption given that it complements efforts to reduce corruption more directly, and it is strongly recommended by the International Monetary Fund, other multilateral institutions, and all worried with the negative impacts of corruption on economic activity.

Notes

¹An extensive review of the empirical literature on the causes and the consequences of corruption can be found in the work of Lambsdorff (1999).

²Tanzi and Davoodi (1997) and Johnson et al. (1999) use as measure of government size the ratio government revenue/GDP. We do not test causality considering this variable inasmuch as it would reduce even more the size of our sample.

³TI's choice to increase the sample of countries over the years, of course, brings some comparability problems since the change in the rank of countries in the sample may be simply the result of the inclusion of a new country and not to an increase or decrease in corruption. TI explicitly warns against year-to-year comparisons of its index on the ground that its composition may change from one year to another. However, other index like the International Country Risk Guide's (ICRG) corruption index presents other problems (Lambsdorff, 2003).

⁴The results are not shown here but are available from the authors upon request.

⁵The post-sample forecast, on the other hand, is done by estimating Equation 2(2) excluding T periods from the end of the sample of all the cross-section units. Equation 3(3) is then estimated based on the forecast errors of the T observations excluded from the sample.

⁶A great number of lags in a first difference panel imply an excessive loss of degrees of freedom. Given our small sample size we choose to work with at most two lags.

⁷For other evidence on the effects of corruption in Latin America, see Gani (2007) who discusses the relationship between foreign direct investment and corruption control.

⁸Corruption may in fact negatively influence governance quality, and indirectly economic growth (Everhart et al., 2007).