We find that providing incentives to firms in Michigan lifted both employment and sales at those firms by 7.1% and 9.9%, above their nonincentivized counterparts, respectively. The average incentive amount per job created, however, worked out to be $593,913 per year. This is calculated by taking the average incentive offered ($3.32 million) and dividing it by the average employment growth of firms that had received offers, then multiplying that by the effect size from our model (7.1%). In other words, the incentivized establishment created more than six jobs on average, but at a high cost per job.[*]
The per-job cost of incentives is a vital measurement. The state repeatedly sells its programs to the public as a method for increasing employment. The questions are, do they add more jobs than would have existed otherwise, and at what cost? If the state forgoes tax dollars of $100,000 to create a job that pays $50,000, it would be hard to make the case that the economy or the state treasury has enjoyed a net gain. This is particularly true when you consider the opportunity costs associated with such programs. Allowing people and companies to keep more of what they earn would create jobs too, as would spending more state tax dollars on different priorities, such as improving Michigan’s roads and infrastructure. In short, the incentive programs we examined are unlikely to pass a basic cost-benefit analysis.
The high cost of creating jobs through these incentive deals might come as little surprise to those who have read previous studies of Michigan’s economic development programs. The Mackinac Center has been studying these initiatives since the late 1980s and has seen massive subsidy offers by state and local government to private, for-profit corporations, including multinationals such as Foxconn. Some of the details surrounding deals offered by the Michigan Economic Growth Authority remain remarkably secretive.
Michigan ranked first in a 2013 national accounting of the largest incentive deals. That study, published by Good Jobs First, was titled, “Megadeals: The Largest Economic Development Subsidy Packages Ever Awarded by State and Local Governments in the United States.” It examined more than 240 deals, exceeding $75 million each. Michigan had 29 such deals — more than any other state. The average incentive per job in that study came out to $456,000.
To test the robustness of our estimates, we ran our model with seven different specifications and found similar outcomes. Across all the alternate specifications, the annual cost of incentives per job ranged from a low of $128,000 to a high of $461,000. (See Table 2 in the appendix for more details.) Further, we test the outcomes with employment levels at one year, two years and three years after a company received an incentive and find consistent results.
We then ran the first model with data from 1990 through 2015 for each of the nine types of incentive programs. This time, we did not include firms that received any specific incentive twice, or an incentive from any of the other programs included in this analysis. This found positive and statistically significant results for three incentive programs: the 21st Century Jobs Fund, Michigan Business Development Program and Michigan Economic Growth Authority. Results for five of the other six types of incentives programs revealed no statistically significant impact on employment. A sixth type of incentive, Seed Capital Funds, showed a statistically significant and negative impact.
We then limited our analysis to the period from 2010 through 2015 and find that employment effects are not statistically significant in the form of our model employing “fixed effects,” but they are in the form using “random” ones. The difference here is important.
Fixed effects models imply statistical control for what are known as “time invariant factors” that may change across firms but are often constant over time. These may include the size of the firm, technological change, profitability, pay, worker productivity and ownership. All of these are items that could be controlled for by using a fixed effects estimation. By contrast, in a random effects model, an estimation is made about the varying factors by relaxing such assumptions and allowing for estimation of individual firm effects.
The incentives offered in the model with random effects ranged from $371,900 to $500,500 per job per year.
When we add in the 412 firms that had received more than one incentive award (treatment) to the mix — whether from the same incentive category or not — we find that the annual incentive cost per job ranges from $109,300 to $547,400 over the larger pool of 2,302 firms. This analysis likewise was made across all programs and model specifications.
Three programs showed statistically significant positive results. This next section discusses the findings about them.
[*]This is based on both incentivized and nonincentivized firms having the average of 79 jobs before the incentives were received.