PRICE FORECASTING WITH TIME-SERIES METHODS AND NONSTATIONARY DATA: AN APPLICATION TO MONTHLY U.S. CATTLE PRICES

Abstract

The forecasting performance of various multivariate as well as univariate ARIMA models is evaluated in the presence of nonstationarity. The results indicate the importance of identifying the characteristics of the time series by testing for types of nonstationarity. Procedures that permit model specifications consistent with the system's dynamics provide the most accurate forecasts.