Abstract Empirical modelling is analysed and illustrated for a small system of aggregate demand relationships in the UK, using estimators specifically developed to tackle the problems of autocorrelation and simultaneity. The tests discriminate between cases where (vector) residual autocorrelation arose from autoregressive errors, or from mis‐specification of the dynamics of the observed data series. Tests also check if the dynamics of the pre‐specified model were adequate, but these should not be used constructively: that a null hypothesis is false does not imply that the postulated alternative must be true. Thus, generalizing initial simple models after test rejections does not provide a valid modelling strategy, partly because a later rejection invalidates all earlier inferences, and to correctly interpret any of the empirical evidence, one must account for all results. The final estimates seek to do that for a range of earlier specifications.
Stored only when both papers are in the corpus — 56 citations exist upstream.
Imported from OpenAlex (CC0) — nothing here is generated.