An empirical investigation of teacher attrition

Using data from a national longitudinal survey, current duration analysis techniques are used to examine the first spell in teaching for a sample of certified elementary and secondary school teachers. A rich set of observable demographic and school characteristics is included in a proportional hazard model which also accomodates unobservable differences (heterogeneity)in both individuals and schools. The estimates of the model indicate that the length of a teacher's first spell in teaching is more responsive to wages than improved working conditions such as smaller student-teacher ratios. The paper also finds that marriage and fertility variables, which have not been included in previous models of teacher attrition, are important determinants of the length of time that a teacher remains in the field. (C) 1998 Elsevier Science Ltd. All rights reserved.