Abstract
Basic econometrics essentially deals with violations to the classical assumptions. These occur regularly when analysing economic data, and the first violation we consider is that of dependence between errors in a time series regression. This is known as autocorrelation and leads to inefficient, and in some cases biased, estimates of the regression coefficients. It is thus very important to be able to test for the presence of autocorrelation, for which the standard statistic is that of Durbin and Watson. Estimation with autocorrelated errors is discussed using a detailed example concerning the UK consumption function, and further extensions for when a lagged dependent variable is included as a regressor are considered. The possibility of autocorrelation being a consequence of a misspecified model is also investigated.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
Notes
For a brief description of the scope and aims of the subject, see Terence C. Mills, ‘Econometrics’, in Adam Kuper and Jessica Kuper (editors), The Social Science Encyclopedia, Volume 1, 3rd edition (Routledge, 2004), 258–260.
G.S. Maddala, Introduction to Econometrics, 3rd edition (Wiley, 2001) remains an excellent introductory text.
The spurious regression phenomenon, then referred to as nonsense-correlations, was first pointed out by the famous British statistician G. Udny Yule in the mid-1920s: ‘Why do we sometimes get nonsense-correlations between time series? A study in sampling and the nature of time series’, Journal of the Royal Statistical Society 89 (1926), 1–63.
See Terence C. Mills, The Foundations of Modern Time Series Analysis (Palgrave Macmillan, 2011), chapter 5, and A Very British Affair: Six Britons and the Development of Time Series Analysis during the 20th Century (Palgrave Macmillan, 2013), chapter 2, for extensive discussion.
James Durbin and George S. Watson, ‘Testing for serial correlation in least squares regression I and II’, Biometrika 37 (1950), 409–428; 38 (1951), 159–177
Donald Cochrane and Guy H. Orcutt, ‘Application of least squares regressions to relationships containing autocorrelated error terms’, Journal of the American Statistical Association 44 (1949), 32–61.
A proof of this assertion is extremely complicated and will not be provided here. It may be found in G.S. Maddala and A.S. Rao, ‘Tests for serial correlation in regression models with lagged dependent variables and serially correlated errors’, Econometrica 41 (1973), 761–774.
James Durbin, ‘Testing for serial correlation in least squares regression when some of the regressors are lagged dependent variables’, Econometrica 38 (1970), 410–421.
A technical justification for this position is given by Grayham E. Mizon, ‘A simple message for autocorrelation correctors: don’t’, Journal of Econometrics 69 (1995), 267–288.
Trevor S. Breusch, ‘Testing for autocorrelation in dynamic linear models’, Australian Economic Papers 17 (1978), 334–355;
Leslie G. Godfrey, ‘Testing for higher order serial correlation in regression equations when the regressors include lagged dependent variables’, Econometrica 46 (1978), 1303–1310.
Author information
Authors and Affiliations
Copyright information
© 2014 Terence C. Mills
About this chapter
Cite this chapter
Mills, T.C. (2014). Autocorrelation. In: Analysing Economic Data. Palgrave Texts in Econometrics. Palgrave Macmillan, London. https://doi.org/10.1057/9781137401908_14
Download citation
DOI: https://doi.org/10.1057/9781137401908_14
Publisher Name: Palgrave Macmillan, London
Print ISBN: 978-1-349-48656-4
Online ISBN: 978-1-137-40190-8
eBook Packages: Palgrave Economics & Finance CollectionEconomics and Finance (R0)