A system for forecasting corporate-tax revenue based on fuzzy logic and fuzzy set theory

Abstract

A system based on fuzzy set theory for forecasting the Florida corporate income tax revenue has been developed in response to several years of unsuccessful forecasts generated by conventional econometric methods. The system consists of two subsystems. The first utilizes time series of revenue and of the real per capita GNP. Both time series are transformed into trend vectors through the moving-average technique. Both vectors are divided into strings of growth patterns such as “accelerating growth,” “decelerating growth,” and “negative growth.” The system checks for several indications of systematic relationship between such strings in the GNP vector and the tax revenue vector. Based on that relationship, the forecast of the corporate tax revenues is generated in fuzzy terms, such as “very rapid growth,” “slightly negative growth,” etc. The fuzzy forecast from the first subsystem of the system constitutes input into the second subsystem, which in turn generates the range of forecasted revenue in millions of dollars. A control mechanism, which is built into the system, continuously checks forecasted rates of change in tax revenues against the actual throughout the history of the time series to make sure that the cumulative forecasting error will not reach an unacceptable magnitude.