Educational attainment and selection into the labour market: The determinants of employment and earnings in Indonesia

This paper uses household survey (Sakernas) data from 2004 to estimate the determinants of earnings in Indonesia, a country where non-salaried work is widespread and where earnings data are available for salaried employees only. We deal with the selection bias by estimating a full-information maximum likelihood system of equations, where earnings are observed for salaried employees, and selection into the labour market is modelled in a multinomial setting. We also deal with reverse causality between educational attainment and earnings by instrumenting years of schooling in both the multinomial selection and the earnings equations. Our identification strategy, following Duflo (2001), uses information on exposure to a large-scale school construction programme implemented in the 1970s. Duflo recognizes that schooling may affect an individual's probability of working as a salaried employee, which creates a simultaneity bias, but does not directly deal with this issue. We find that the parameters of the earnings equation estimated under multinomial selection differ from standard OLS estimates, which ignore the selection bias, and from a binomial selection procedure "à la Heckman" (1979). In particular, the estimated parameters that vary the most are those related to the variables with the strongest impact on individual selection into the different labour-market statuses. We also find that workers with higher educational attainment are most likely to find a job as salaried employees, and that non-salaried work is as an alternative to inactivity.