Estimating Causal Effects from Panel Data with Dynamic Multivariate Panel Models

Abstract

Panel data are ubiquitous in scientific domains such as sociology and econometrics. Various modeling approaches have been presented for the analysis of such data including dynamic panel models, cross-lagged panel models, and their extensions. Existing panel data modeling approaches typically impose some restrictive assumptions on the data-generating process, such as Gaussian errors, effects that are constant in time, or univariate responses. With the goal of surmounting these restrictions, we present the dynamic multivariate panel model (DMPM) that supports both time-varying and time-invariant effects, multiple simultaneous responses across a wide variety of distributions, and arbitrary dependency structures of lagged responses of any order. We employ a Bayesian approach to the estimation of the model parameters and leverage state-of-the-art Markov chain Monte Carlo methods for simulation of the posterior distribution. We demonstrate the use of DMPM by applying the model to both real and synthetic data.