In this article the authors investigate the impact of the choice of time series method, the length of the data stream used to estimate the model, and the frequency of the data on forecasting accuracy for own source, non‐tax general fund revenue for six Connecticut municipalities. The authors find that exponential smoothing models are generally the most accurate. They also conclude that local government officials should rely on bimonthly rather than monthly or quarterly data and retain, in a readily usable format, more than three years of data.

With the availability of powerful, inexpensive computer hardware and software, computer-intensive statistical methods are becoming more commonly applied in the social sciences. These methods frequently offer a number of advantages over traditional parametric procedures. This article briefly reviews four computer-intensive methods (permutation/randomization tests, bootstrapping, the jack knife, and cross-validation), discusses some of the strengths and weaknesses of these approaches, provides example applications, and discusses five commercially available software packages that may be used to implement these methods.