Predicting Worker Turnover: An Assessment of Intent on Actual Separations
通过分析横截面和纵向调查数据，发现离职意图与实际离职受不同因素影响，意图对预测实际离职作用不大，并构建了优化预测模型。
The predictive ability of most turnover models emanates from a key construct measuring a worker's intention to withdraw from a work organization. Empirical evidence to collaborate this cardinal assumption has, however, only moderate support. To test the predictive value of intention to leave, a logistic regression model of combined cross-sectional and longitudinal survey data on workers'potential and actual turnover was initiated. The results demonstrate that actual turnover and intent are influenced by a separate set of factors with intent a poor predictor of turnover behavior. Age, tenure, wage level, and perceived chances for improvement at the job were found to have a significant impact on actual separation, but not on the intent to leave. Work repetitiveness, importance of improvement, and perception of co-workers' intent influence the intent to move, but not turnover. Perception of co-workers' intentions to leave and importance of improvement played a dual role for both intent and turnover, but much more so for intentions. Further confirmation of these differences was sought by in/excluding intentions into the logistic regression of all the independent variables on actual separation. No improvement in the model occurred indicating the negligible impact of intent on explaining turnover. A parsimonious regression equation to optimize prediction of actual turnover was created, leading to a high and low risk leaver-stayer profile with probabilities ranging from 85.3% to 3.9%. The overall conclusion confirms the distinctiveness of intention and turnover as distinct constructs which are only superficially linked. As they differ in their basic explanatory antecedents, only variables linked to turnover predict work separations. Intentions do not.