A Logit Analysis of the Effect of Relocation on Job - Quit Probability

  • Published 2016

Abstract

Binary logit analysis is adopted to estimate job-quit probability in terms of explanatory variables representing the effect of housing re¬ location and other characteristics of workers. We utilize a cross-sectional sample of households rehoused by the City Council of Glasgow from condemned properties due to re-development schemes. The logistic model is adopted for its computational convenience, al¬ though a behavioural justification is attempted through the random utility choice theory. The simultaneous logistic model is proposed to allow for the 'within-family' dependence and its properties explored with comparison to a simultaneous linear probability model. Deriving a simple test we support the complementarity hypothesis of husband/wife labour force parti¬ cipation responses. While exploiting the likelihood estimation techniques we introduce a goodness-of-fit measure which is approximately F-distributed. The most interesting variables turn out to be: the pre-move weekly work hours, change in travel time, change in housing costs, age, skill, and availability of employers in the new area. Policy implications of these findings are outlined. Using the curve we have established the U-shaped property of the married females' quit probabilities implying concentration near zero and one probability. We argue that the S curve is a useful tool for turnover D analysis compared to the survival curve of the demographic approach. On the basis of the curve we develop and test an arc-elasticity formula for average (quit) probability which is easy to compute. Aggregation bias due to neglect of individuals' heterogeneity in cross-sectional data is analyzed for the logistic model. Criteria are developed for the variance-elasticity measure of average response proba¬ bility, and the appropriate formula derived.

Cite this paper

@inproceedings{2016ALA, title={A Logit Analysis of the Effect of Relocation on Job - Quit Probability}, author={}, year={2016} }