Date of Award


Document Type


Degree Name

Doctor of Philosophy (PhD)


Agricultural Economics

First Advisor

Hector O. Zapata


Over the past two decades, developments in time series analysis have brought new approaches for combining structural characteristics of market models with stochastic processes that better represent available data. One line of research is the works of Zellner and Palm, which is known as structural econometric and time series analysis (SEMTSA). The other approach is the structural vector autoregressive model (SVAR), which is an economic-theory enhancement to the standard VAR approach. Empirical evaluations of how well these approaches may work in explaining the dynamics of commodity markets are lacking. The current study provides an empirical evaluation of these two approaches for the U.S. rough rice market. Transfer functions (TF), derived from a dynamic structural econometric model of the U.S. rice market, were estimated. The RMSE and MAPE evaluation revealed that the TF model greatly reduces forecasting errors relative to the existing structural and ARIMA models for the seven rice market variables (acreage planted, yields, production, domestic consumption, exports, ending stocks, and rough rice prices) in an out-of-sample period (1990--1999). A turning point evaluation indicated that forecasts generated by the TF model closely follow the actual movements of all variables except ending stocks. The research also addressed the empirical usefulness of combining structural-statistical properties of economic data in commodity modeling. A comparative analysis of the impulse response functions revealed that the estimated effects in the VAR model of specific behavioral shocks often do not appear economically intuitive. Having imposed structural relationships in a time series context, the study found that most impulse response functions in the SVAR model are in conformation with economic logic, with empirical results far superior to those generated from a VAR in levels. These empirical findings in favor of the TF and SVAR models stem from a common methodological approach, which combines economic theory with statistical properties of time series. The research findings suggest that a significant contribution to commodity modeling can be derived from this type of approach. This conclusion is supported by the empirical findings from economic model of the U.S. rough rice market.