Three Essays on Nonstationary Panels with Unobserved Heterogeneity

This dissertation develops several econometric techniques to address the unobserved heterogeneity in nonstationary panels, namely identifying latent group structures in cointegrated panels, studying nonstationary panels with both cross-sectional dependence and latent group structure, and estimating panel error-correction model with unobserved dynamic common factors. Chapter 1 considers a panel cointegration model with latent group structures that allows for heterogeneous long-run relations across groups. We extend Su, Shi, and Phillips' (2016) classifier-Lasso (C-Lasso) method to the nonstationary panels and allow for the presence of endogeneity in both the stationary and nonstationary regressors. In an empirical application, we study the potential heterogeneous behavior in testing the validity of long-run PPP hypothesis. The results confirm that at least some countries favor the long-run PPP hypothesis in the post-Bretton Woods period. Chapter 2 proposes a novel approach, based on Lasso, to handle unobserved parameter heterogeneity and cross-sectional dependence in nonstationary panel models. We establish a mixed normal asymptotic distribution for our estimators, which permit inference using standard test statistics. Finally, we apply our approach to study the international R&D spillovers model with unobserved group patterns. The results shed new light on growth convergence puzzle though the channel of technology diffusions. Chapter 3 proposes a novel econometric model that accounts for both long-run and short-run co-movements in panel error correction models. We establish asymptotic properties for two Lasso-type estimators. In an empirical application, we estimate long-run cointegration relationships between bid and ask quotes in stock market. We introduce a new measure for efficient price, which is weighted average of bid and ask quotes.