The advantages of the autoregressive distributed lag (ADL) model for estimating the effects of generic advertising on market demand are evaluated by applying the model and attendant methods to data used in a recent study of Norway’s export promotion program for whitefish.
The dynamic specification differed greatly depending on model selection criteria (Akaike Information, Hannan-Quin, Schwarz, and Adjusted R2) (Table 1). Despite this there was little to choose between the specifications in terms of the estimated long-run demand elasticities. The estimated short-run elasticities differed among the specifications, with the model selected by the Hannan-Quin criterion indicating a more elastic response to income than the model selected by the Schwarz criterion. The bounds test for cointegration, a special feature of the ADL approach, proved useful in distinguishing between the appropriateness of quantity- and price-dependent specifications of the demand equation. Tests for weak exogeneity of the regressors indicated adjustments in quantity are 5.5 times more important than adjustments in price in resolving dynamic disequilibria caused by random (monthly) shocks to long-run demand.