Tackling misclassification in surveys about undeclared work via the EM algorithm

For the development of adequate policy measures which deal with undeclared work, it is important to knowledge its extent and structure. Direct surveys of individuals aim to fill this gap. However, in view of the sensitivity of the subject, collected information may be affected by misclassification errors. In this work, we aim to estimate the individual propensity to work off-the-book, tackling misclassification errors. To this end we used the approach proposed by Ward et al. (2009, Biometrics) for modeling presence-only ecological data via logistic regression, based on the expectation-maximization (EM) algorithm