Nonlinear Models for the Prediction of Early Quits and Dismissals

The inverse relationship between length of service and labour turnover is a consistent finding in the literature. This paper demonstrates by a longitudinal cohort analysis of 40,816 British Rail workers recruited during 1979-1980, the increased accuracy and sensitivity obtained by concentrating separately on quits and dismissals.Comparisons between quit rates and dismissal rates during the first ten months of employment pointed to the following major results: (1) quit rates and dismissal rates are better explained by nonlinear functions rather than by a linear one; (2) quits are best described by a parabolic function whereas dismissals are best described by a nearly inverse 'U'-shaped function; and (3) nonlinear patterns of quit rates and dismissal rates remained consistent for 1979 and 1980, although the unemployment levels, quit rates and dismissal rates in 1979 differed substantially from those in 1980.