부패 수준의 측정모형 - MIMIC 과 DYMIMIC

Models of Measuring Corruption: MIMIC and DYMIMIC

It is essential to figure out the degree of corruption in order to design appropriate anti-corruption mechanism. We review various existing methods, including the famous corruption perceptions index (CPI) published annually by the Transparency International (TI), of measuring public corruption levels, identify their merits and limitations, and conclude that none is satisfactory for its original needs. As alternatives, we suggest a multiple-indicators-multiple-causes (MIMIC) model and a dynamic MIMIC (DYMIMIC) model. The two models consider the degree of corruption as an unobservable variable and apply a factor analysis which utilize multiple explanatory variables of corruption and multiple indicators as proxy variables for corruption. Parameters of the two models are derived by maximum likelihood estimation and by an expectation-maximization step algorithm. MIMIC and DYMIMIC models make up for the weakness in the existing estimation methods resulting from the heavy dependence on the subjective surveys and the unreasonable consolidation of perceptions, causes and consequences of corruption.