It is a well‐known criticism that if the distribution of wealth is highly concentrated, survey data are hardly reliable when it comes to analyzing the richest parts of society. This paper addresses this criticism by providing a general rationale of the underlying methodological problem as well as by proposing a specific methodological approach tailored to correcting the arising bias. We illustrate the latter approach by using Austrian data from the Household Finance and Consumption Survey. Specifically, we identify suitable parameter combinations by using a series of maximum‐likelihood estimates and appropriate goodness‐of‐fit tests to avoid arbitrariness with respect to the fitting of the Pareto distribution. Our results suggest that the alleged non‐observation bias is considerable, accounting for about one quarter of total net wealth in the case of Austria. The method developed in this paper can easily be applied to other countries where survey data on wealth are available.