As a robustness test, we evaluate several different combinations of model specifications. We run Model 1 with random effects specification (Table 2; Models 1 and 2) and find consistent effects. We further include matching group fixed effects — where a group consists of one treated establishment and five control establishments (Table 2; Models 3 and 4) and find our results are robust. Finally, we include the age of firms as an additional control to our original specification (Model 1) and find consistent results (Table 2; Models 5, 6 and 7). Across the specifications, we find the effect size of incentives ranging from $128K per job to $461K per job.
Table 2: Alternate Specifications and controlling for age of firm
It is plausible that the effects of incentives would be reflective on job growth in the future years. To test that possibility as a robustness test, we include three additional outcome variables — employment at 1 year after incentive, 2 years after incentive and 3 years after incentive. Table 3 shows estimates from our preferred fixed effects specification. We find that Michigan incentives increases employment by 6.92% to 7.54% in the next three years of receiving incentives.
Table 3: Incentive effects on future employment