Moment-Based Estimation of Macroscopic Dynamic Models in Macroeconomics and Finance
The thesis is divided into two parts. Part I (macroeconomic models) is devoted to empirically examine the role of backward-looking behavior in a standard New-Keynesian model and its behavioral variant. This part includes three chapters. In chapter 2, the structural parameters of the New-Keynesian model (NKM) are estimated from a historical data set of the US economy. The moment-matching method is used to discuss the importance of backward-looking behavior in the New-Keynesian Phillips Curve, and its empirical results are contrasted to the Bayesian estimation. Then the model with purely forward-looking expectations and its hybrid variant are compared using a formal test. Chapter 3 discusses the persistent dynamics of inflation and output in the NKM, and analyzes statistical properties of the estimation methods and the choice of moment conditions. The empirical performance of model selection methods is examined using information criteria along the lines of the maximum likelihood estimation. Chapter 4 demonstrates that market euphoria and depression in the economy can be explained by switching between heterogeneous investors along the lines of the discrete choice theory. This chapter investigates a bounded rationality model with historical Euro Area data and discuss the effects of the investors' over(or under)reaction on market structure. Part II (financial market models) contains two chapters concerning the relevance of behavioral heterogeneity in financial markets. Chapter 5 discusses the effect of heterogeneous trading rules on the return volatility in the asset pricing model. The adaptive belief system is estimated using the simulated method of moments estimator. Especially, two types of the noise term in the model dynamics are investigated by means of simulations (i.e.\ additive and multiplicative). In a structural stochastic volatility model, the two trading mechanisms (i.e.\ wealth and herding) are compared according to a formal test. Finally, chapter 6 examines the social interaction effects of a market microsimulation model on various historical FX data. In particular, simulation based inference is used to examine the validity of the group behavior when the analytical expression for moment conditions is fairly complicated.