On selecting a measure of labour activity: evidence from registered nurses, 1981 and 1989

The purpose of this paper is to demonstrate the effects of the selection of measures of labour activity on the conclusions that are derived from the analysis of individual labour supply. The results of the analysis using three commonly specified single measures (Annual Hours Worked, Hours Worked per Week, and Weeks Worked per Year) as the dependent variable in TOBIT equations are compared with those obtained using a PROBIT-2SLS in which the individual labour supply is measured by Hours Worked per Week and Weeks Worked per Year. The RHS variables are those that are used in earlier individual labour supply studies and are the same in all of the models analysed. Among the findings are (1) wage rate effects that are quite strong in single equation models and disappear when a multi-equation approach is used, (2) the effects of children on labour supply is more complex than in indicated in single equation results and (3) duration dependence is much stronger than indicated results using a single measure of individual labour supply. These results suggest that using a single measure of labour supply will lead to inappropriate conclusions as to the effects of such variables as wage rates, number and ages of children, spouse's employment state, and income from other sources on individual labour supply.