-
5,161
-
It is difficult to measure the level of unobservable corruption. This study develops a new empirical model that minimizes limitations of existing methods none of which is satisfactory for its original needs. A dynamic multiple-indicators-multiple-causes (DYMIMIC) model considers the level of corruption as an unobservable variable and applies a factor analysis which utilizes multiple explanatory variables of corruption and multiple indicators as proxies for corruption. In addition, several dummy variables and fictitious variables are added for robustness and stability of the parameter estimates derived by maximum likelihood estimation and by an expectation-maximization step algorithm.
Applying the model to 62 countries during the period of 1991 through 2002 results in the following points: (i) Globalization and democratization deter corruption; (ii) Government regulation rather than its size facilitates corruption; and (iii) Countries of Latin America and Africa show higher levels of corruption than others which suggests that culture, history and socioeconomic system of a country significantly influence its level of corruption. How to reduce the multicollinearity due to the lack of reliable data remains a task to be pursued in the succeeding study.
