Abstract

We aim at proposing alternative statistical tools for analyzing misreported data, a problem that frequently occurs when dealing with social science data. Despite misreporting being a well-recognized issue in a large variety of study fields, assessing its rate and pattern of occurrence still remains an open and critical task from a methodological point of view. The main reason lies in the fact that misreporting is typically treated as a nuisance factor to be removed from the analysis with ex-post correction methods, rather than to be considered as a structural component in the model specification. In this perspective, the incorporation of the
misreporting mechanism in the probabilistic modelling of the outcome of interest, as a further source of uncertainty to be accounted
for, represents a new and ambitious research direction in the statistical literature. Additionally, the development of effective
methods for appropriate handling of misreported data requires specialized approaches in relation to both the specific type of
responses (i.e. categorical, ordinal, or numerical) and to the data collection techniques (i.e. registry or survey data). The analysis of
survey micro data may help in defining a profile of the  phenomenon under investigation and in having evidence of the potential misreporting and
its determinants. On the other hand, registry data give a more extensive and continuous monitoring of the phenomenon and, if
opportunely treated, they may provide a global realistic picture of the phenomenon under investigation. An integration of these
different sources of information may lead to more accurate estimates.