Masters Theses

Date of Award

8-1987

Degree Type

Thesis

Degree Name

Master of Science

Major

Agricultural Economics

Major Professor

Dan L. McLemore

Committee Members

Charles B. Sappington, Emmit L. Rawls

Abstract

The purpose of this study was to develop forecasting models capable of predicting basis for Tennessee feeder cattle six months in advance. Both econometric and time series models were developed, estimated and tested out-of-sample. The econometric model was estimated using five continuous inde-pendent variables. These were: transfer cost, futures price level, feeder cattle supply, stage of the cattle cycle and local grazing conditions. Dummy variables were used to represent the futures contract change to a cash settlement system and to represent the seasonality of feeder cattle production and marketing. The first type of time series model used was a univariate ARIMA model. Two ARIMA model specifications were used in the study, both of which accounted for seasonal components in the autoregressive scheme. Transfer function models were the second type of time series models used. Two forms of transfer function models were used with each including the futures contract change dummy variable in addition to seasonal autoregressive components. The second transfer function also used a transportation cost index as an exogenous variable. The five models were used to predict values for 12 10-day market-ing periods out-of-sample. These forecasts were compared to actual values using Theil's U2 coefficient, root mean square error and graphics. Four models were able to predict better than a naive no-price-change model as indicated by coefficients less than one. Values for the U2 coefficients ranged from 0.28 to 1.12 and the cor-responding RMSE ranged from $0.87 to $3.49. The transfer function model with the dummy variable to represent the futures contract change to cash settlement and the transportation cost index was the superior model based on the above criteria. The econometric model was second best and the transfer function model with only the cash settlement dummy variable ranked third. The pure time series models ranked fourth and fifth, consistently overestimating the Tennessee feeder cattle basis.

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

Share

COinS