Simulations have shown that if two independent time series, each being highly autocorrelated, are put into a standard regression framework, then the usual measures of goodness of fit, such as t and R-squared statistics, will be badly biased and the series will appear to be ‘related’. This possibility of a ‘spurious relationship’ between variables in economics, particularly in macroeconomics and finance, restrains the form of model that can be used. An error-correction model will provide a solution in some cases.
KeywordsAutocorrelation Durbin–Watson statistic Econometrics Ordinary least squares (OLS) Regression Serial correlation Spurious regression Weiner process
- Baltagi, B. 2001. A companion to theoretical econometrics. Oxford: Blackwell.Google Scholar
- Dickey, D., and W. Fuller. 1979. Distribution of the estimates for autoregressive time series with a unit root. Journal of the American Statistical Association 74: 427–431.Google Scholar
- Granger, C. 2001. Spurious regression in econometrics. In A companion to theoretical econometrics, ed. B. Baltagi. Oxford: Blackwell.Google Scholar
- Kendall, M. 1954. Exercises in theoretical statistics. London: Griffin.Google Scholar
- Pilatowska, M. 2004. Realization of the congruence postulate as a method of avoiding the effects of a spurious relationship. In Dynamic economic models, ed. Z. Zielinski, vol. 6. Torun: Nicolaus Copernicus University.Google Scholar