Forecasting Performance of Alternative Error Correction Models

It is well established that regression analysis on non-stationary time series data may yield spurious results. An earlier response to this problem was to run regression with first difference of variables. But this transformation destroys any long-run information embodied in the levels of variables. According to ‘Granger Representation Theorem’ (Engle and Granger, 1987) if variables are co-integrated, there exist an error correction mechanism which incorporates long run information in modeling changes in variables. This mechanism employs an additional lag value of the disequilibrium error as an additional variable in modeling changes in variables. It has been argued that ECM performs better for long run forecast than a simple first difference or level regression. This process contributes to the literature in two important ways. Firstly empirical evidence does not exist on the relative merits of ECM arrived at using alternative co-integration techniques. The three popular co-integration procedures considered are the Engle-Granger (1987) two step procedure, the Johansen (1988) multivariate system based technique and the recently developed Auto regressive Distributed Lag based technique of Pesaran et al. (1996, 2001). Secondly, earlier studies on the forecasting performance of the ECM employed macroeconomic data on developed economies i.e. the US and the UK. By employing data form the Asian countries and using absolute version of the purchasing power parity and money demand function this paper compares forecast accuracy of the three alternative error correction models in forecasting the nominal exchange rate and monetary aggregate (M2).