Functional forms and the estimation of a state's sales tax revenue: Tennessee as a case study

Abstract

Problems of specifying a proper functional form for estimating sales tax revenues of a state are discussed. A general functional form is introduced, in which linear and log-linear forms are special cases. The maximum likelihood approach and data for Tennessee are used to estimate parameters in the general form. Empirical results disclose that the null hypotheses indicating the functional form is linear or log-linear are both rejected. The best functional form obtained from the general form provides not only better R2 and Durbin-Watson statistic, but also generates estimates of sales tax collections with smaller errors than the linear and log-linear forms.