A0151
Title: The output convergence debate revisited: Lessons from recent developments in the analysis of panel data models
Authors: Hashem Pesaran - Trinity College, Cambridge (United Kingdom) [presenting]
Ron Smith - Birkbeck University of London (United Kingdom)
Abstract: There has been a resurgence of interest in the debate on whether output per capita has been converging across countries, and the possible determinants of the cross-country differences in output per capita. The analysis of dynamic heterogeneous panels with interactive effects provides a fruitful framework within which the assumptions and econometric techniques employed in this literature can be examined. Much of the convergence debate assumes that country-specific outputs follow a partial adjustment process towards an unobserved steady state, partly driven by a latent factor taken as a proxy for global technology. Empirical studies within this and similar literatures tend to use panel or cross-section estimators that impose strong homogeneity restrictions on adjustment speeds and time effects, the so-called parallel trends assumption. Heterogeneity across units that is correlated with the regressors causes the estimates to be inconsistent, and such correlated heterogeneity is inherent in dynamic models. A systematic, step-by-step, theoretical analysis is conducted of the implications of parameter heterogeneity in intercepts, slopes, and factor loadings, showing how neglected heterogeneity can result in inconsistent estimates and distorted inference. It also examines the treatment of both time-varying and time-invariant covariates that could explain the differences in cross-country outputs. The theoretical results are illustrated using Penn World Table data.
