This article attempts to rehabilitate Bradford Smith, who in 1926 published a short article in the Journal of the American Statistical Association in which he considered the implications of detrending time series by either deviations-from-trend or first-differencing prior to regression. His discussion covers such topics as the permanent-transitory innovation distinction inherent in difference and trend stationary processes, common factor restrictions and general-to-specific modelling. These have all become mainstays of modern time-series econometrics, yet there does not appear to be even one reference to Smith's paper in all the years since its publication.
No takes yet. Share an insight, caveat, or question.
Terence C. Mills (2010) studied this question.
Synapse has enriched 5 related papers. Consider them for comparative context: