Error correction in DHSY

In this note, we consider the contradiction between the fact that the best fit for the UK consumption data in Davidson et al. (1978) is obtained using an equation with an intercept but without an error correction term, whereas the equation with error correction and without the intercept has better post-sample forecasting properties than the former equation. This contradiction is explained and the two equations reconciled in a nonlinear framework by applying a smooth transition regression model to the data.