Making public job training work : Evidence from decentralized targeting using firm input

Job training programs are used by governments as a solution to underemployment and to improve the match between skill supply and skill demand. Targeting such programs with input from local employers has long been hypothesized as a way to improve their effectiveness. We test this hypothesis using a unique situation in Brazil in which two national skill training programs ran in parallel, with one taking local employer input in choosing course offerings while the other retained a traditional, government-led structure. The employer-informed program nearly doubled the short-term effect on trainees’ employment and earnings relative to the traditional program. The differential effectiveness of the employer-informed program is not attributable to differences in course or student composition across programs. Our findings provide evidence that limited, structured input from the private sector can improve alignment between skills trained and skill demand and increase the employment and earnings of the underemployed.