Susanne Raessler
Institute of Statistics and Econometrics
Faculty of Business Administration, Economics and Social Sciences
Friedrich-Alexander-University Erlangen-Nuremberg
Lange Gasse 20
D-90403 Nuremberg
This paper illustrates the effects of missing data in panel surveys on the results of multivariate statistical analysis. Large data sets of the German IAB Establishment Panel are used which typically contain more than 10000 cases in each wave. Due to item-nonresponse continuous as well as categorical variables of interest are affected by missing values. Using only the available cases for the estimation task reduces the data set considerably. Thus, we multiply impute the missing data applying a Bayesian data augmentation algorithm. The imputer's model is based on a multivariate normal distribution for the data and some noninformative prior distributions for the parameters. The handling of so-called semi-continuous variables is explained to impute the incomplete mixed data suitably.
The analyst's model is based on a translog production function with labour and capital as input factors. Also the influence of industries and the use of modern technologies are considered. Furthermore, we include interaction variables that indicate deviations in the parameters concerning the two parts of Germany. Besides differences between industries, we find a higher productivity, if modern technologies are installed. The results for Eastern and Western Germany only differ for some industries and the constant. Using only the available cases valuable information seems to be discarded. Calculating the regression coefficients using the imputed data sets and combining the results according to the multiple imputation principle reduce these differences up to 11%-points. Additionally, the differences of the industrial branches between Eastern and Western Germany decrease when inference is based on multiple imputation.