Title: Linear Models with Interval-Censored Variables
Speaker: David Pacini, University of Bristol
Time: 2025/03/20 17:00-18:30
Venue: ZOOM Meeting (Please scan the QR code below to join)
Abstract
This paper studies the problem of inference in a set identified problem defined by linear moment restrictions generated by multiple interval-censored variables. It applies to the case with interval-censoring of variables on both sides of a linear model as in our empirical application of interest concerning the relationship between risky assets and household wealth. We use the characterization of elements of the identified set developed by Bontemps, Magnac and Maurin (2011, Econometrica) as a condition about a dual convex auxiliary set. We construct a test procedure of this condition by minimizing the support function of this auxiliary set. We show how to regularize the test procedure when the minimization problem has a continuum of solutions. Furthermore, we show how this procedure extends to sub-vector inference in a natural way through duality again using a different regularization procedure. Although dependent on tuning parameters, asymptotically normal properties of the test statistic hold, and critical values need not be simulated. We derive weak conditions under which these properties are uniformly valid. Monte Carlo experiments evaluate the numerical performance of the novel inference procedure and compare it with existing ones.