Assessing forecast performance in a cointegrated system

摘要: This paper examines the forecast performance of a cointegrated system relative to comparable VAR that fails recognize is characterized by cointegration. The we examine composed three vectors, money demand representation, Fisher equation, and risk premium captured an interest rate differential. forecasts produced vector error correction model (VECM) associated with this are compared those obtained from corresponding differenced autoregression, (DVAR) as well autoregression based upon levels data (LVAR). Forecast evaluation conducted using both ‘full-system’ criterion proposed Clements Hendry (1993) comparing for specific variables. Overall our findings suggest selective improvement (especially at long horizons) may be observed incorporating knowledge cointegration rank. Our general conclusion when advantage appears, it generally longer horizons. consistent predictions Engle Yoo (1987). But also find, (1995) gain in clearly depends chosen transformation.