Essays in term structure models and GMM estimation with incomplete knowledge

This dissertation consists of two separate projects. The first project (in Chapter 1) expands the standard generalized method of moment (GMM) framework into the situation with incomplete knowledge on the correct model specification and correct moment conditions and derives consistent model and moment selection criteria in this context. The second project (in Chapters 2–4) develops a state-space approach for estimating a large class of dynamic term structure models and explores the use of these models for bond portfolio hedging, mean-variance portfolio management, and pricing of fixed-income derivative securities.; Chapter 1 develops consistent model and moment selection criteria for GMM estimation with incomplete knowledge. The criteria select the correct model specification and all correct moment conditions asymptotically. The criteria are then applied to dynamic panel models with unobserved individual effects, where they are used to select the lag length for lagged dependent variables, to detect the number and locations of structural breaks, to determine the exogeneity of regressors, and/or to determine the existence of correlation between some regressors and the individual effect. Finally, Monte Carlo experiments are conducted to evaluate the finite sample performance of the criteria.; Chapter 2 studies discrete-time affine term structure models. Most of these models have unobserved factors and time-varying market prices of risks that imply nonlinear state-space forms. To handle these problems, the chapter applies the extended Kalman filter to estimate model parameters and extract information on the factors. The chapter also conducts specification analysis in these non-nested models to test cross-section restrictions on the yield curve and to determines relative importance of factors, such as the stochastic long-run mean, stochastic volatility, and square-root specification.; Chapter 3 uses the method in Chapter 2 to extract information on bond yields and returns from dynamic term structure models. This information is then used for hedging, portfolio management, and derivative pricing. In particular, a new dynamic hedging method utilizes information on conditional volatilities of bond returns; the mean-variance portfolio management utilizes information on both conditional means and volatilities of bond returns; and a simulation method for derivative pricing utilizes information on the whole conditional densities of bond yields.; Chapter 4 proposes a general state-space method that is applicable to both affine and nonaffine term structure models in continuous- and discrete-time frameworks. The method uses empirical distributions to approximate the true conditional densities for prediction and filtering and the approximation error can be made arbitrarily small. Furthermore, for continuous-time models, the formation of the likelihood function does not require knowledge of the transitional densities of diffusion processes.