Reinterpreting the DHSY (1978) consumption function with hindsight

Abstract

The justification for Davidson, Hendry, Srba and Yeo's (DHSY's) selection of their favoured equation is shown to be based upon contradicting assumptions about inflation. The empirically supported model is argued not to have a theoretically justified error-correction interpretation. The exclusion of the intercept biases the error-correction term such that it becomes significant and negative but undermines the theoretical plausibility of the long-run target. More generally, the statistical need to include an intercept in any regression to avoid parameter bias is demonstrated. The rise in inflation in the 1970s is not considered to be a credible explanation of the fall in the average propensity to consume (APC) from the 1950s and is shown to have questionable empirical support when a below unit income-elasticity is allowed for. The recent articles which have found support for DHSY's favoured formulation, augmented by time-varying parameters, using the most contemporary UK data should be viewed with caution. New interpretations for the DHSY model's persistent success are called for if it is not to be considered that dubious regressions have been resurrected.