On Business Cycle and Asset Pricing Implications of Quantitative DSGE Models

Abstract

This thesis develops more-data-consistent dynamic stochastic general equilibrium (DSGE)models by incorporating the respective characteristics of two key elements – cyclical capital utilization and composite-good habit formation – that are routinely missing but empirically important. Based on improved DSGE models, this thesis explores channels and mechanisms of these elements that help explain observed salient domestic and international business cycles and asset prices. Chapter 2 incorporates a vector-error-correction-model process for the utilizationadjusted total factor productivity (TFP) into a two-country, one-good model with variable capital utilization, motivated by the empirical evidence supporting cointegration between adjusted TFP processes across countries. The simulation results show that by introducing endogenous capital utilization, the benchmark model succeeds in producing (i) observed positive crosscountry correlations of investment if the cross-country correlation of productivity shocks is moderate and (ii) historical cross-country hours worked, output, and consumption correlations if the cross-country productivity differential is very persistent. Chapter 3 studies the role of consumption-leisure composite habit formation in explaining both the observed procyclical labor supply and the observed high level of equity premium. The composite habit is featured by a free habit-persistence parameter and a free habit-intensity parameter that is empirically and theoretically evident but has not been considered in asset-pricing models. The analytical results suggest that as habit intensity increases, the wealth effect on hours worked dampens; also, the equity premium rises. Habit persistence has the reverse effect. The numerical analysis confirms these analytical steady-state relations when introducing the composite habit into a centralized production-based asset pricing model. However, the effect of habit persistence is marginal for a medium value. Chapter 4 extends the model presented in Chapter 3 to a decentralized setting and estimates this model using Bayesian likelihood techniques to fit four U.S. quarterly macroeconomic quantity data. The estimation results suggest a level of habit intensity near one and a medium level of habit persistence, but the data are not informative well on the latter. By using posterior modes to parameterize composite habit, the estimated model matches the data relatively well, and the non-linearly approximated baseline model can reproduce a realistic mean risk-free rate and equity premium.