Technical Considerations in the Modeling Used by MEGA

MEGA engages economists at the University of Michigan to help the authority determine a proposed MEGA deal’s fiscal impact. That the authority consults economists at the University of Michigan is a credit to the MEGA program. To date, the state has spent about $986,000[108] on the University of Michigan’s analyses of MEGA deals.

The university’s economists employ a computer-based economic model to help them determine the economic impact of a proposed MEGA agreement. The model employed to make such forecasts is known as REMI, short for Regional Economic Models Inc. It is a respected, popular model, and it used by both public and private sectors.

Nevertheless, a model is inherently a simplification of reality that involves assumptions over which reasonable people can disagree. While we would emphasize again that the REMI model users at the University of Michigan appear to use the program skillfully,[109] the jobs claims they produce are based on assumptions that, though made in good faith, may overstate the impact of the MEGA program for a variety of reasons, which are detailed below.

Direct Cost of State Tax Relief

MEGA approved 76 deals between April 1995 and December 2000. Of these, 59 analyses performed by University of Michigan economists came with important qualifiers, such as, "These calculations do not include any revenue losses due to the property tax abatement. If the cost of the abatement were included, the net revenue gain to state government would be slightly less."[110] Similar statements were made in cases in which the REMI modelers ignored the cost of Capital Acquisition Deductions and Investment Tax Credits.

To the modelers’ credit, this is an intellectually honest qualification. The REMI model would have in fact lowered its government revenue forecast if the cost of abatements were included. For instance, more than $154.9 million in state-level property tax relief has been offered to 147 firms that have received MEGA packages — relief that could be deducted in part from state government revenues.

The effect of the SBT incentives on government revenue estimates is harder to assess. State officials — particularly MEGA officials — have argued that in reality, the SBT credits involve no net cost to the state at all. According to this argument, no taxes have been forgone, since MEGA assumes (and asks company officials to guarantee) that the private economic activity would not have occurred without the MEGA tax breaks in the first place. Further, the officials argue, the effect on tax revenues is actually positive, since the MEGA credit generates spin-off economic activity that is not subject to the credit and therefore creates new state tax revenue that would not have been produced otherwise.

But despite the best intentions of state and business officials, it’s not clear that the full value of a given MEGA SBT tax credit was needed to influence a firm’s location or expansion decision. It is possible that a lesser amount might have sufficed. This point is probed in greater detail in "The Problem of Ensuring MEGA Credits Are Necessary."

For the moment, we would simply say that if a firm would have located or expanded in Michigan without some or all of the benefits of a MEGA package, the University of Michigan impact analyses would inevitably overstate the benefits of that MEGA deal for the state and taxpayers.[111]

The Local Incentives in MEGA Deals Are Not Directly Modeled

As mentioned earlier, between April 1995 and December 2004, more than $957 million[112] had been offered to MEGA companies by local units of government or local economic development agencies as part of their overall MEGA package. These incentives are not directly included in the REMI model. One reason these are not directly incorporated is that the REMI model employed by University of Michigan economists is only a statewide, or "one-area," model.[113] (A more complex model that would allow local-area impacts to be measured would probably be more expensive.) It cannot directly accommodate the costs of local incentives locally.

It is possible that the modelers attempt to factor in the cost of local incentives in the one-area model through indirect methods. Unfortunately, determining whether this has been done is difficult from the available public documents of the model’s "input assumptions" — i.e., the record of what the computer was told to assume in the model, such as the direct number of employees and expected payroll resulting from the deal. (For an example of MEGA input assumptions, see Graphic 7.)

Graphic 7: REMI Assumption Worksheet

In any event, if the REMI model does include an indirect estimate of the local costs, the influence of the incentives would not be well-defined, meaning, for instance, that it would not be possible to determine where in the state the job and revenue impact would be realized.

The value of these incentives is derived primarily from abatements on personal or real property over six- or 12-year periods. Other local incentives include donated land, waived permitting fees, free transportation to MEGA companies in the form of busing, and free or subsidized golf or other recreation for employees of MEGA companies.

Including the net cost of these incentives directly would require a different model, but it would also likely lower the revenue impact analyses and perhaps the number of jobs the program could claim, too. We say "likely lower" in part because some of the local incentives mentioned here may not have an economic cost. If a local municipal golf course, for instance, has excess capacity, little cost may be incurred in allowing additional users to access the fairways. On the other hand, some of the local abatements almost certainly do have a cost, and some of them have been claimed by firms that failed to generate the direct jobs assumed as part of the MEGA package, such as R.J. Tower and Kmart.

The Cost of New State and Local Government Services

Not only are the costs of local incentives not modeled; the impact estimates are not made with regard to increased costs placed on state and local government to maintain public services given the increased demands a MEGA project can create. As noted in a review of economic development policy in Michigan by Upjohn Institute senior economist Timothy Bartik, Wayne State University political scientist Peter Eisinger, and Upjohn Institute senior regional analyst George Erickek:

Despite its sophistication, the MEGA analysis (using REMI) omits a full fiscal analysis that would consider impacts of MEGA projects on local revenues. Neither does it estimate required spending on state and local public services to keep service quality constant as population increases in response to the project.[114]

Doing so would likely mute forecasts of MEGA job creation. The REMI model would recognize that a larger population brings costs as well as benefits. Such costs would include the increase in spending on public services, such as police and fire, to accommodate the larger population.

Differences Between High- and Low-Unemployment Areas

As Bartik has also noted, the REMI analyses do not differentiate between the jobs impacts in high- and low-employment areas.[115] Incentives may have a greater impact in areas with higher unemployment than those with lower unemployment. This has important implications, because REMI may be overstating the jobs impact for deals that have occurred in lower unemployment areas.

The academic literature on this dynamic is mixed, but it would be nonetheless worth investigating. On two occasions, the Mackinac Center for Public Policy has found that 60 percent of MEGA deals go to firms in counties with unemployment rates that are lower than the state average overall.[116] Indeed, no county in Michigan has been home to more MEGA deals than Oakland (46 altogether). Among counties with a population of 1 million or more, Oakland County ranks[117] as the fourth-richest county in America, and between January 1995 and December 2004, its monthly unemployment rate was continuously below the state average.

Failed Employment and Wage Assumptions

Documents from the state and the University of Michigan show that by 2005, 127 MEGA deals should have produced facilities achieving the employment counts expected at the time the MEGA deals were signed. Of these, only 10 can be shown to have directly created the number of jobs promised within the expected time frame. Three of these 10 have since performed poorly. Kmart, one of the three, has even declared bankruptcy and downsized its workforce to levels significantly lower than it had at the time it signed the MEGA agreements. (For the methodology used to make these calculations, see "Appendix C: Determining MEGA’s Job Counts.")

This means that the "REMI inputs" involving employment and wages would be overstated for the vast majority of these 127 MEGA deals. Thus, based on what we know now about the underperforming MEGA packages, the "spin-off" benefits attributed to the projects are almost certainly too high.

This observation has implications for more than the 127 MEGA deals theoretically operative by 2005. If the input assumptions for these past cases have typically been too high, it is likely that the assumptions are also too optimistic in the REMI projections for MEGA projects that have yet to be completed and staffed. It is perhaps worth noting that the initial direct job estimates used in the REMI assumptions appear to be provided by the companies themselves, rather than independent sources or state officials.