Expectation formulations and optimal decisions in cattle feedlot problems

Abstract

Economic agents are continually required or assumed to make decisions based upon variables which are unknown or perhaps known with some degree of uncertainty. In order to aid the decision maker in making intelligent decisions, the traditional approach has been to devise an expectations mechanism which will predict or estimate the future values of these relevant variables. As with any estimator, it is natural to attempt to attribute desirable statistical or analytical properties to such expectations. While this may produce optimal estimates in terms of selected predictive criteria they are not necessarily optimal in the context of individual decision models. That is, when viewed in the context of decision models which include controllable and uncontrollable exogenous variables, a structure relating endogenous to exogenous variables, and an objective function it is not clear that expectations which have good predictive properties are superior to alternative estimators. With respect to the role of expectations in decision models, very little has been done in agricultural economic research. This study consists of an investigation of the role of expectations in decision problems of the type just sketched. More specifically, the investigation is carried out in the context of a cattle feedlot operation. Several structural models describing the feedlot operation are developed. A number of objective functions are considered in connection with these structural models in which the expectation mechanism is regarded as part of the underlying structure. As it turns out, this type of situation poses some interesting problems with respect to optimal parameter estimates for the part of the structure describing the expectations generating mechanism. More precisely, it is shown that the assumed objective function is an important factor in determining the optimal estimator. Since the data available may include observations on endogenous and exogenous variables as well as prior probabilities on the parameters, the estimators examined include classical as well as Bayesian approaches. Analytically, optimal estimates of parameters for expectations in decision models were shown to depend upon the constants of the decision-maker's objective function. Experimentally, various feedlot models were developed using different expectation mechanisms and strategies. In every case, the appropriate strategy expectation mechanism combination resulted in higher returns than the constant policy strategy where the feedlot manager did not attempt to anticipate future values of cattle prices. In addition, the experimental results frequently pointed out the inadequacy of cattle-on-feed data. Alternative methods of providing information to decision makers were suggested based on the analytical results. More specifically, outlook information might be categorized so as to make it more useful to the decision makers for which it is intended.