Technical Change and Model Specification: U.S. Agricultural Production

U.S. agricultural crop and livestock relationships are examined in the context of both duality and time-series econometrics. Based on time-series test results, cointegrated models are estimated. Traditional models generally overestimated the precision of statistical relationships and gave a considerable number of spurious results, particularly with regard to technical change. When time-series properties of the data were addressed, there was much less evidence of bias in technical change. Model specification test results were sensitive both to the time-series specification of the maintained model and its functional form. Preferred functional form depended on the choice criterion. Copyright 1997, Oxford University Press.