Semester of Graduation

Summer 2026

Degree

Master of Science (MS)

Department

Agricultural Economics and Agribusiness

Document Type

Thesis

Abstract

This thesis investigates the best-fit empirical distribution for daily U.S. corn and soybean futures price changes and examines how distributional choices affect Value-at-Risk (VaR) and Expected Shortfall (ES) estimates. The dataset comprises daily nearby futures prices for corn (11/16/1994–04/23/2025) and soybeans (10/16/1979–02/15/2023), yielding 7,626 and 10,853 observations, respectively. An ARCH-LM test confirms the presence of Autoregressive Conditional Heteroskedasticity (ARCH). Generalized Autoregressive Conditional Heteroskedasticity (GARCH) models were estimated, and to account for potential asymmetric effects, an EGARCH model was used. EGARCH (1,1) estimation reveals an inverse leverage effect, where positive price shocks generate higher volatility. Six heavy-tailed distributions were evaluated, and the Anderson–Darling goodness-of-fit criterion indicates that the mixed Burr Type XII provides the best empirical fit for the standardized residuals over the observed period. In all cases, AIC selection criterion was used to identify the best model, and the Ljung-Box statistics on standardized residuals and squared standardized residuals to further ensure white noise and no further ARCH effects. To validate the risk forecasts, a 1,000-day rolling backtest was performed. At the 95% confidence level, the mixed Burr XII and parametric normal distribution, provide statistically valid coverage under the Kupiec Proportion of Failures (POF) test. The mixed Burr XII distribution emerges as the most consistent distribution across all confidence levels. Sizeable differences in the estimated VaR and ES were discovered. These results suggest that future work may consider closer examination of the statistical properties of alternative flexible distributions for modeling non-normality in futures price changes. One fruitful extension of this work is the estimation and validation of risk in price changes using bootstrap methods which do not require a specific probability distribution.

Date

5-29-2026

Committee Chair

Hector Zapata

LSU Acknowledgement

1

LSU Accessibility Acknowledgment

1

Download

Included in

Agribusiness Commons

Share

COinS