Some empirical evidence on the non-normality of cost variances.

Abstract Cost variance data was collected from 14 production departments of a medium size manufacturing plant belonging to a Fortune 500 company. The data consists of weekly standard direct labor costs and direct labor efficiency variances in 1985 and 1486. For each of the 14 departments, we tested whether the following two types of variances are normally distributed: the direct labor efficiency dollar variances and the direct labor efficiency percentage variances. Various tests of normality were considered, and the two most powerful ones (the Bowman-Shenton Ks² and Shapiro-Wilk W tests) were used. For the direct labor efficiency dollar variances, only seven out of the 14 departments can be considered to have normally distributed data. For the direct labor efficiency percentage variances, only one out of the 14 departments can be considered to have normally distributed data. Our results indicate that one should not simply assume normally distributed variances indiscriminately.