B1667
Title: Time-varying panel data models with latent group structures
Authors: Shahnaz Parsaeian - University of Kansas (United States) [presenting]
Ali Mehrabani - Southern Illinois University (United States)
Abstract: Joint estimation and identification of latent group structures in a time-varying panel data model is proposed that allows the coefficients to vary across both individuals and time. It is assumed that the coefficients change smoothly over time and form different groups across individual units, where the number of groups and the group membership are unknown a priori. When treated as smooth functions of time, the individual functional coefficients are heterogeneous across groups but homogeneous within a group. To identify the individuals' group identities and to estimate the group-specific functional coefficients, a penalized sieve estimation procedure is proposed. The proposed approach is implemented by an alternating direction method of multipliers algorithm. The proposed method is further illustrated by simulation studies, which demonstrate the finite sample performance of the method in both classification and estimation.