Considering differential importance and non-linearities in indices of productivity

Researchers have argued that any measure of productivity should consider both the differential importance of its various indices and the nonlinearities present between a given index and overall productivity. The purpose of this research was to empirically evaluate the contributions of these two features in a measure of productivity called ProMES. Several simpler measures of productivity were derived from ProMES which do not account for differential importance and/or non-linearities. Bivariate correlations were computed between ProMES and the simpler measures of productivity to estimate the variance lost by dropping one or more features. It was tentatively concluded that there was a small but meaningful loss in variance when the non-linearities feature was dropped; however, dropping additional features did not result in a meaningful loss in variance. Limitations and implications of this study are discussed.