The Role of Stochastic and Nonstochastic Specification in Transformed Models: A Case Study of Quarterly Beef Demand

摘要

The role of auto correlated residuals, a heteroskedastic error variance, and various transformation of variable schemes is simultaneously examined via a quarterly beef demand model. Principal results indicate that autocorrelation must be corrected in Box-Cox models, heteroskedasticity can be induced into the error variance of transformed equations (and, therefore, the analytic form of the hetroskedasticity must be simultaneously estimated with the nonstochastic part of the model), the method of applying a different Box-Cox transformation parameter to each model variable is preferred over other transformation schemes, and elasticities are not robust as to both functional and/or error term specification.