Government jobs for talents, cronies or sale?
In this paper, I provide a new approach to estimate government worker skills that helps to quantify the quality of government workers and their selection. The approach consists of estimating skills from wages in comparable jobs in the private sector, relating these skills to a large number of skill-related observables using Machine Learning estimators and then predicting government worker skills out-of-sample. This method is particularly useful in settings where government output is unobserved, wages of government workers are uninformative about differences in skills and the researcher observes a variety of individual-specific skill-related observables such as educational background and test scores. I apply this method to Indonesian household-level panel data from 1988 to 2014. I study how the quality of the government workforce has evolved over time and assess whether the government has become better at selecting civil servants over the course of democratization, decentralization and civil service reform processes in the early 2000s. I find both small improvements in the skills of the government workforce as well as small improvements in how the government selects workers that are both indistinguishable from zero. I can rule out larger positive effects and thus provide evidence that democratization and decentralization reforms have not led to large improvements in government hiring in Indonesia up until 2014.