Abstract This paper presents an alternative methodology for developing explanatory factors that are associated with Tax Court decisions. The WORDS computer program was used for measuring the content of narrative data in tax cases. The procedure is illustrated by developing factors that distinguish between capital gain and ordinary income in real estate transactions. The results are compared with another study on the same issue which used a traditional tax modeling methodology. Fifteen factors were identified by WORDS, of which seven were significant. A logistic; regression model was able to classify correctly 80 percent of the total observations in a jackknife holdout sample, indicating that the WORDS themes were consistent across the sample and are reliable measures of important factors that influence Tax Court decisions.
Taylor et al. (Thu,) studied this question.