Essays on several contemporary issues in econometrics and financial economics

The dissertation consists of two essays on several topics in econometrics and financial economics. Chapter 2 addresses the issue of statistical testing under nonstandard conditions in the case of Markov regime switching models, which have been wildely used in economic and financial modeling. It is shown that if we depart from the time domain maximum likelihood framework and transform the problem into a frequency domain testing problem, we only have to face the problem of unidentified nuisance parameters in a nonlinear context, with no singularity problem in the information matrix. Two test statistics, a Difference Test (DT) and a Lagrange multiplier (LM)-type test, are proposed and their asymptotic distributions are derived. Since the DT test requires estimating parameters both under the null and the alternative, the LM test has an advantage in terms of simulating and bootstrapping finite sample empirical distributions. Compared to Hansen's (1992) LR bound test, the LM test performs better in terms of finite sample power. We have the exact asymptotic distribution for the tests while Hansen (1992) only has the asymptotic bound for the LR statistic. Also, his test requires three-dimensional grid search while the tests here only need a one-dimensional grid search. The LM test is applied to test Engel and Hamilton's (1990) exchange rate model and Hamilton's (1989) GNP model. The null of random walk is not rejected in the former model, but the null is rejected for the GNP model. While Chapter 2 addresses an issue of statistical testing, Chapter 3 addresses an issue of econometric modeling in financial economics. We focus on the feedback relationship between stock market returns and economic fundamentals, which has been an important issue in both theoretic and empirical asset pricing literature. A vector-autoregression (VAR) model is used to investigate the variations of expected and unexpected returns in an emerging market such as Korea. It is shown that expected returns in the Korean stock market vary with a set of macroeconomic variables, and that the predictable component is substantial. However, a likelihood ratio test rejects the constraint to the variation of expected returns implied by the intertemporal consumption-based pricing model. We further find that economic fundamentals can explain only a small portion of the variation in unexpected stock returns. We can also conclude that the VAR model captures richer dynamics in the data than the usual single equation modeling employed in the literature.