Modeling Demand For Unionization With Nontraditional Data Analysis Methods

Upon reviewing the extant literature on determinants of unionism, it becomes clear that many areas that have had a plethora of research attention do not converge upon singularly directional findings. This study explores a potential cause of such an apparent anomaly: nonlinearity of data. An exploratory examination of correlation coefficients among typical union determinant variables seems to show different patterns of relationships at different levels of union demand. Thus, a break from traditional linear data analysis techniques is explored in the interest of explaining more variance with typical, theoretically derived variables by using neural network analysis. Results of analyses on industry level data reveal that using neural network analysis to model union demand explained over four times as much variance as multiple regression analysis.