<![CDATA[Objective: This study examines the causal effect of Indonesia’s Kartu Prakerja program on the Open Unemployment Rate (OUR). Existing evaluations rely largely on short-horizon data that are vulnerable to reverse causality, leaving the program’s effectiveness unresolved. Design/Methods/Approach: The analysis employs a balanced panel of 38 districts in East Java from 2016 to 2024. Two-Way Fixed Effects (TWFE) models with continuous treatment intensity and lagged treatment specifications are used to identify delayed program impacts. Robustness is assessed through four sensitivity tests: Local Projections, wild bootstrap inference, and leave-one-out analysis. Findings: Kartu Prakerja has no significant short-run effect but reduces unemployment after three years. The estimated coefficient at the three-year lag is −0.023 (p = 0.006), consistent with the lock-in mechanism documented in the training evaluation literature. The result remains robust across all sensitivity checks, is independently confirmed by Local Projections, and is supported by a 95% wild bootstrap confidence interval of [−0.040, −0.005]. The strongest effects are observed in districts with high poverty rates and a large agricultural sector. Originality/Value: This study provides one of the first long-horizon causal evaluations of Kartu Prakerja using treatment intensity and lagged exposure measures, offering new evidence on active labor market policies in developing countries. Practical/Policy implication: Short-term evaluations may underestimate the effectiveness of training programs. Policymakers should adopt longer evaluation horizons and prioritize implementation in economically vulnerable regions where the program generates the largest employment gains.]]>
