Table 1 shows the estimates obtained from establishment fixed effects specification. Model 1 shows the baseline specification with establishment and year fixed effects. Model 2, our preferred specification, includes matching group specific time trends to the baseline model. We find that Michigan incentives increases establishment-level employment by 7.1% and sales by 9.9%. We also perform a test of overidentifying restrictions and find that fixed effects model is our preferred model [Sargan-Hansen statistic 2.2 × 1010; p-value <0.0001].
In order to find the effect size of job growth, we first find average incentive amount of the treated group, which is $3.32 million per incentive. We then find the average employment of the treated group, which is about 79 jobs per establishment. The effect sizes expressed as dollar incentives per job can then be estimated as $593,913 per job (i.e., $3.323 million divided by 0.0709 ×78.9 jobs).
We prefer Model 2 because it accounts for the least restrictive set of assumptions, controlling for year, establishment and control group fixed effects. Importantly, the selection was based on controlling these assumptions, not on model results.
Table 1: Fixed effects estimates of incentives on employment and sales