Analysis of labour participation behaviour of Korean women with dynamic probit and conditional logit

We analyse the dynamic labour participation behaviour of Korean women. State dependence under unobserved heterogeneity is considered, where the heterogeneity may be unrelated, pseudo-related, or arbitrarily related to regressors. Three minor methodological contributions are made: interaction terms with lagged response are allowed in dynamic conditional logit; a three-stage algorithm for dynamic probit is proposed; and treating the initial response as fixed is shown to be ill-advised. The state dependence is about 0.6 x SD(error), higher for the married or junior college-educated, and lower for women in their twenties and thirties. While education increases participation, college education has negative effects for women in their forties or above. Marriage has a high negative short-term effect but a positive long-term effect.