Doctoral Dissertations

Date of Award


Degree Type


Degree Name

Doctor of Philosophy


Agricultural Economics

Major Professor

S. Darrell Mundy

Committee Members

John Booker, Luther Keller, Larry VanTassell, Alan Schlottman


In this study the effects of use of aggregate risk measures and risk models on the optimal farm organization of hurley tobacco farms in Tennessee were analyzed. One hypothesis was that measures of risk based on aggregate data and existing risk models yield valid solutions to farm planning problems for such farms. A second hypothesis was that farm management practitioners should not reject any of the more common models or procedures for use in application to any farm. Practitioners should select a model with data requirements and solution characteristics that most closely conform to the requirements of the problem at hand. Three objectives were established to test the two hypotheses. The first objective centered on estimation of the variance-covariance matrix of net returns necessary in the formulation of risk programming models. A second involved the identification and evaluation of alternative farm planning procedures that are characterized by estimates of expected returns and their variation for alternative farm organizations that vary in exposure to risk. The final objective involved validation of the risk planning procedures identified in farm applications. In fulfilling the first objective, several econometric procedures were applied to the time-series of net returns for various farm enterprises. The resulting variance-covariance matrices were compared both in quadratic programming applications to representative hurley tobacco farms and for conformity to various desirable characteristics of risk measures. Risk models compared in the second objective included quadratic programming (QP), marginal risk constrained linear programming (MRCLP), minimization of total absolute deviations (MOTAD) and the single-index portfolio model. Model results were compared across various representative hurley tobacco farms at differing levels of risk-aversion for conformity of enterprise selection and levels. Models were also evaluated at several points on the frontier of risk efficient choices for each farm. In the final objective, FINPACK, a farm financial planning package, was applied to the representative farms to analyze differences between reported farm organization and alternative organizations suggested by the risk programming techniques. An evaluation of the sensitivity analysis portion of the FINPACK package was made. While no statistical tests were applicable to the type of analysis performed, results supported the hypotheses of the study. Valid risk measures in terms of enterprise selection and levels were identified. Evaluated risk measures included those constructed from the nominal time-series of net returns, from the deflated time-series and from residuals of a simple moving average forecast of net returns, a weighted moving average forecast of net returns, a second-order trends removed forecast and from an integrated autoregressive and moving average forecast (ARIMA). Quadratic programming results from the generation of the mean-variance frontier of risk efficient choices for all representative farms showed substantial similarity across all these measures of risk. On the basis of conformity to desirable characteristics of risk measures, the measures formed from the residuals of weighted moving average forecasts of the time-series of net returns was superior to the other evaluated risk measures. The set of risk measures examined did not include all possible measures. Given current computational capabilities, no significant advantage was discovered regarding ease of calculation for any measure. Types of risk programming problems were identified so that choice among the evaluated alternative risk models could be made based on the characteristics of the available data and desired solution information. The range of covariances among possible enterprises was shown to be the most important factor in determining the type of model best suited for any application. Only when covariances were small or constant between alternatives were the computational advantages of the models that did not consider covariances shown to be significant. The single-index model was shown to be the least desirable model except in cases where aggregate risk data are incomplete. Possible uses of a computerized farm financial planning package was demonstrated by example to be an excellent source of the type of data necessary for risk planning procedures. This package also provided a format for the evaluation of alternative farm organizations suggested by the risk modelling procedures. In summary, this research demonstrates that appropriate aggregate measures of price and production risk exist that may be applied in several risk models. Solutions resulting from the application of these procedures in farm planning problems yield valid and useful farm management information .

Files over 3MB may be slow to open. For best results, right-click and select "save as..."