ON IMPROVING ECONOMETRIC ANALYSES OF GENERIC ADVERTISING IMPACTS

It is possible to obtain robust estimates of structural parameters using observational data, but it is difficult to do so. Necessary, but not sufficient, conditions are to adopt a modeling philosophy and to undertake a comprehensive evaluation of the results. Using a general-to-specific modeling philosophy, we obtained stable estimates of the long-run advertising elasticity for fluid milk. This result contrasts with an earlier, published model which did not provide stable estimates as new data points became available. It is difficult, however, to apply the general-to-specific modeling approach because it requires the researcher to specify an initial general model. But analysts are unlikely to agree on this initial model, and if this is true, then the "generality" of the model is in question. Moreover, it is a fact that the quality of the available data is sometimes insufficient to obtain the desired stable estimates.