Three Essays in Macroeconomics and Financial Economics

Abstract

This dissertation presents empirical analysis of linear and nonlinear models in macroeconomics and financial economics. It conveys the message about the substantial benefit in the analysis stems from a little departure from the standard models. By relaxing some assumptions, especially the linearity, it demonstrates some significant improvements of analysis performed in terms of accuracy and theoretical consistency. Empirical Analysis of A Core Inflation Measure in An Estimated DSGE Model The first chapter presents the analysis of inflation by allowing an ad-hoc time-varying inflation target given by one of the best core inflation measures, namely the PCE trimmed mean core inflation. At the same time, we are evaluating the core inflation measure by directly incorporating them into a dynamic general equilibrium model. The analysis of the inflation dynamics, especially in correspondence to its broken-down components is interesting and worth exploring further. It is argued that the Fed has been actually targeting a time-varying inflation target consistent with the underlying inflation dynamics. Analysis of New Keynesian Phillips Curve Relationship in An Estimated Nonlinear DSGE Model This paper estimates a nonlinear DSGE model based on Amisano & Tristani (2010) with US data. The model is approximated up to the second order. Conditional particle filter is used to calculate the likelihood and Bayesian method is used to simulate the posterior distribution of the parameters. The analysis of the nonlinear NKPC better accommodates short-term sharp-turns of the dynamics in the economy. It shows different impulse responses that are conditional on high or low inflation rate in the initial period. As a result, the relationships between inflation, its inertia, the expected inflation and output are more consistent with the theory suggested by the model. Value at Risk (VaR) Based on GARCH-Type Estimated Volatility of 5 Stock Markets The Great Recession has stirred up debate about risk management practices. Value-at-Risk (VaR) is often blamed for imprudent excessive risk taking leading to the crisis. VaR-based potential loss calculation is based on the assumption of normality of the shocks. In reality, shocks distributions are often highly kurtotic. VaR will be more accurately representing the real risks if such distributions are used in the calculation. EGARCH(1,1) with Student t-distribution is shown to be more reliable than the simple standard RiskMetrics and the standard GARCH(1,1) approaches.