Business support policies designed to raise productivity and employment are common worldwide, but rigorous micro-econometric evaluation of their causal effects is rare. We exploit multiple changes in the area-specific eligibility criteria for a major program to support manufacturing jobs ("Regional Selective Assistance"). Area eligibility is governed by pan-European state aid rules which change every seven years and we use these rule changes to construct instrumental variables for program participation. We match two decades of UK panel data on the population of firms to all program participants. IV estimates find positive program treatment effect on employment, investment and net entry but not on TFP. OLS underestimates program effects because the policy targets underperforming plants and areas. The treatment effect is confined to smaller firms with no effect for larger firms (e.g. over 150 employees). We also find the policy raises area level manufacturing employment mainly through significantly reducing unemployment. The positive program effect is not due to substitution between plants in the same area or between eligible and ineligible areas nearby. We estimate that "cost per job" of the program was only $6,300 suggesting that in some respects investment subsidies can be cost effective.