This dissertation is composed of three chapters centering on nonparametric econometrics with applications to consumer demand system analysis, value-at-risk analysis of commodity future prices, and credit risk analysis of home mortgage portfolios.
The first chapter, based on my joint research with Abdoul Sam considers a semiparametric estimation model for a censored consumer demand system with micro data. A common attribute of disaggregated household data is the censoring of commodities. Maximum likelihood and existing two-step estimators of censored demand systems yield biased and inconsistent estimates when the assumed joint distribution of the disturbances is incorrect. This essay proposes a semiparametric estimator that retains the computational advantage of the two-step methods while circumventing their potential distributional misspecification. The key difference between the proposed estimator and existing two-step counterparts is that the parameters of the binary censoring equations are estimated using a distribution-free single-index model. We implement the proposed estimator using household-level data obtained from the Hainan province in China. Horrowitz and Härdle (1994)'s specification test lends support to our approach.
The second chapter is an empirical application of a nonparametric estimator of Value-at-Risk on the cattle feeding margin. Value-at-Risk, known as VaR is a common measure of downside market risk associated with an asset or a portfolio of assets. It has been used as a standard tool of predicting potential portfolio losses for twenty years in the financial industry. Recently VaR has gained popularity in agricultural economics literature since the market price risks associated with agricultural commodities are under evaluation. As initial empirical findings suggest that the performance of any VaR estimation technique is sensitive to the types of data set (portfolio composition) used in developing and evaluating the estimates, agricultural data provides a unique laboratory to further explore VaR and its estimation approaches. This essay as a first attempt applies a distribution-free nonparametric kernel estimator of VaR in an agricultural context, the cattle feeding margin using futures data. The empirical results suggest that the nonparametric VaR estimates enjoy a significant efficiency gain without losing much accuracy compared to the parametric estimates.
The third chapter measures credit risks associated with residential mortgage loans. Credit risk is the primary source of risk for real estate lenders. Recent advancements in the measurement and management of credit risk give lenders with sophisticated internal risk models a significant comparative advantage over other lenders in terms of capital optimization and risk controlling. This manuscript helps understand the determinants of credit risk and acquire perspectives on how it is distributed in the current or future loan portfolios. This essay contributes to the existing volume of literature as it incorporates the nonparametric estimation technique into default risk analysis. The CreditRisk model is modified and estimated using the consumer side of information. The model identifies the factors determining household default risks and generates a full loan loss distribution at the portfolio level using consumer finance survey data. In the end, portfolio management strategies are discussed.