An Investigation into Equal Pay in Germany: How Suitable are the Process Generated Data from the National Insurance?

In Germany, process-generated data from the German national insurance have recently become available for scientific analysis. With these data for the first time it is possible to access comprehensive employment records for the major part of the population. There are numerous advantages to these administrative data in comparison with German survey data such as the socio-economic panel, the microcensus or the income and consumption sample. Due to the purpose of their collection they are of high quality and go back as far as the beginning of the 20th century. Panel mortality may occur for administrative reasons. Employment records are mostly complete from the 14th year of one's life until death. At first glance these data seem to be predestined for long-term analysis of income inequality. A drawback of the data is their restriction to incomes liable to social security payments. Furthermore, incomes are truncated at a contribution ceiling. In total, 10 to 14 percent of higher incomes are not fully represented. To appropriately analyse income inequality, this problem must be dealt with. The author presents an imputation method for incomes above the contribution ceiling based on the assumption of pareto-distributed incomes. The method is applied to the samples of retirement accounts from the research data centre of the German Retirement Insurance from the years 2005 and 2006. The income distributions generated by this method are compared with alternative statistical methods to examine their plausibility. Finally, using men from Western Germany aged 20 to 59 years, the long-term development of wage and salary inequality is analysed.