D-Separation, Forecasting, and Economic Science: A Conjecture

The paper considers the conjecture that forecasts from preferred economic models or theories d-separate forecasts from less preferred models or theories from the actual realization of the variable for which a scientific explanation is sought. Dseparation provides a succinct notion to represent forecast dominance of one set of forecasts over another; it provides, as well, a criterion for model preference as a fundamental device for progress in economic science. We demonstrate these ideas with examples from three areas of economic modeling. Word Count: 8348 JEL: B41; C45; C53 _______________ The authors are at Texas A&M University, College Station, Texas, USA. An early draft of this paper was presented at the 2009 SAS forecasting conference in Cary, North Carolina, USA. We thank Tom Fomby for the invitation to address the SAS group. Further, we thank two anonymous referees for substantive comment on an earlier draft of the paper.