Bounds for timely estimates of average household income

This paper proposes a novel set identification approach to produce timely estimates of average household income that are robust to the sample selection bias driven by item non-response, which may affect surveys. Our method covers a large number of practical situations and considers several, increasingly restrictive assumptions. Extensions to other functionals beyond average income are possible. As a practical example, we use data from Banca d'Italia's Household Outlook Survey. Starting from wide nonparametric bounds based solely on brackets, we progressively narrow the identified set by exploiting additional information: unfolding brackets, exact income responses, and monotonicity assumptions leveraging the panel and cross-sectional dimensions of the data. Our method identifies bounds that contain the official EU-SILC estimate of average Italian household income and are available almost one year earlier than the EU-SILC release. The economically plausible assumptions we impose narrow the initial bounds by 25%. The method is simple, transparent, and broadly applicable to other contexts where timely, unbiased income measures are needed.