This paper is concerned with the problem of modelling the tail of the wealth distribution with survey data when the data does not adequately cover the tail of the distribution. In order to deal with the problem post data collection, it is standard practice to either fit a Pareto tail to the data or to combine wealth survey data with observations from rich lists before fitting such a Pareto tail. This paper proposes a novel approach (‘rank correction’) to address such cases which does not require additional data-sources. Applying the rank correction approach to wealth survey data (HFCS, SCF, WAS) yields estimates of top wealth shares, which are closely in line with estimates from the World Inequality Database and therefore represent a significant improvement over the raw survey data. While the paper focuses on the distribution of wealth as a case in point the rank correction approach might generally prove useful in contexts, where the tail of a Pareto-distributed variable is not adequately covered by the available data.
Read full abstract