In the wake of shrinking public coffers, policy makers are demanding greater accountability from their economic development initiatives. In a discipline known for ‘claiming anything that falls,’ attempts to objectively evaluate economic development programs have been stymied by ill-suited data sources and methods. Survey research is expensive and responding firms have an incentive to lie about the effectiveness of subsidies. Publicly available data on employment, wages, and other outcomes are highly aggregated and lack the power to capture impacts from anything other than the most dramatic, large-scale initiatives. Confidential employee- and establishment-level (micro) data holds considerable promise for more rigorous and objective approaches to program evaluation, but have rarely been used for this purpose. This paper uses data on employee and employer characteristics to evaluate the short-term impacts of a training-subsidy initiative in the State of Maine. I match employee records in the state’s Unemployment Insurance database to establishment records in the Quarterly Census of Employment and Wages to create a longitudinal database containing both firm and worker characteristics. This database is matched to program information on participating businesses and trainees and analyzed using a quasi-experimental approach with propensity-score matching. I found some evidence of a small positive impact on participant earnings and employment levels one year after training, but no significant immediate impacts.