Household Asset Surveys contain quantitative questions asking the household, for example, to state the value of its holdings in a given financial or real-estate product. Such questions often invite respondents to indicate a range rather than a specific amount. This strategy reduces the non-response rate. The downside is that the amounts declared are not precise values that can be used to directly compute inequality indicators. Our article describes a general procedure for a one-time estimation of inequality indices and for obtaining confidence intervals. The method is suitable for sampling-based collection and for data expressed in interval form. It relies on a two- or three-“stage” modelling of inequality indicators. The stages are nested to form a “hierarchical” model. The first stage describes the sampling, the other two the process for generating data on total assets. In addition to the observations of socio-demographic variables available in the survey, the process modelling uses various other information sets. They include the value ranges declared by households for their asset holdings and additional information such as wealth-tax liability. The procedure yields confidence intervals that allow for sampling variation and the uncertainty over monetary amounts, which are observed imprecisely. We discuss the modelling of the “total gross assets” variable more specifically and examine two models: the first directly describes the “total gross assets” variable collected in the survey; the second is a simultaneous-equation model that concurrently describes several aggregated components of total gross assets. Using the fullest information set, we estimate average gross assets at about EUR205,000 in early 2004, with a Gini inequality index of approximately 0.65—a value comparable to the inequality measured from past Household Asset Surveys. However, by adding information on wealth-tax liability, we significantly reduce the width of the confidence intervals.