„Ökonometrische Analyse von aggregierten Tendenzdaten aus Panelerhebungen"

Abstract

It is shown that the explanation of aggregated tendency data by other quantitative or qualitative variables can be reasonably performed by use of the logistic model. Since empirical results based on aggregated IFO tendency data concerning price expectations and production plans (Ronning 1980) showed only small explanatory power, the simulation study presented in this paper was designed to answer the question whether the unsatisfactory results were due to the (wrong) specification of explanatory variables or to the stochastic model assumed as the data generating process. The results of the simulation study reveal that the assumption of "product multinominal sampling" which was used in empirical research, causes the main trouble, since the assumption does not recognize the time series character of the aggregated tendency data. Two alternative estimation methods (Maximum Likelihood and the approach proposed by Berkson and Theil) as well as two measures of goodness of fit (Pearson Chi-Square and Likelihood Ratio) were considered. The simulation study indicates that for aggregated tendency data based on a moderate sample size, the Berkson/Theil approach and the Pearson measure should be chosen.