Practical Considerations when Estimating in the Presence of Autocorrelation


  • Sanjiv Jaggia
  • Alison Kelly-Hawke


When correcting for autocorrelation, most econometrics texts suggest using a quasi-differencing procedure. A number of issues arise. First, it is found that the results from popular two-step procedures may differ dramatically from those obtained from iterative processes. Second, while it is true that most regression packages implement an iterative procedure, the methodology itself is not conveyed in a straightforward manner to students of econometrics. Third, given the various iterative methods in the literature, it is not always clear which method is superior. Fourth, for autocorrelated errors, the importance of the correction factor in simple forecasting is often overlooked. Finally, regression packages report an R2 that is not comparable to that from the Ordinary Least Squares (OLS) estimation. This paper succinctly outlines the procedure for performing iterative procedures, explicitly accounts for autocorrelation among errors when generating forecasts, and identifies the necessary transformations for making proper comparisons relating to R2.