Consequences, Detection And Forecasting With Autocorrelated Errors

Problem of autocorrelation arises if the assumption of the Classical Linear Regression Model that the errors terms are not autocorrelated is violated. As a consequence, the usual t, F, and ¿2 tests cannot be legitimately applied. This text uses various econometric approaches to critically observe the associated problems. Graphical method; Durbin-Watson method; Breush-Godfrey method; and The Runs Test were used to detect existence of autocorrelation among residuals of econometric data. In correcting autocorrelation, the method of first-difference, based on Durbin-Watson d-statistic and the dynamic forecasting techniques were used. The result gave a significantly reduced estimated autocorrelation coefficient. This improves the efficiency of the forecast and the use of various statistics in making inference.