Past Studies of Targeted Tax Incentives

Analysis of the role of tax policy on economic growth enjoys an extensive treatment by economists. A 1997 Federal Reserve Bank review of research findings cited over 90 studies that evaluated the role of fiscal policy in economic growth in the United States (see, for example, the research of Michael Wasylenko in the New England Economic Review).[91] If anything, the past few years have seen an acceleration of this analysis accompanied by the development and widespread application of more robust statistical techniques that enable analysts to evaluate impacts.

Scholarship on business tax incentive programs is mixed, but generally negative.

Many of these papers attempt to explain differences in growth, wages and industrial composition through analysis of interstate tax policy. An equally large number of studies also evaluate whether expenditures (as evidenced by infrastructure) influence growth (see, for instance, the research of William Fox and Sanela Porca in 2002).

A considerably smaller number of studies have attempted to evaluate the influence of individual targeted tax policies on economic growth. A number of these have been reviewed in a study in 2002 by Timothy Bartik, senior economist at the Upjohn Institute and co-editor of Economic Development Quarterly, a scholarly journal on economic revitalization.

Despite extensive analysis of fiscal incentives in general, the literature does not yet suggest a consensus on their impact on local economic conditions. Many studies find no impact on some important policy variables (e.g. income, employment) while those that do find impacts report rather modest taxation elasticities on growth, in the range of ‑0.1 to ‑0.4.[92] These figures mean that for every 1 percent decrease in taxes, we would see between a 0.1 percent and 0.4 percent increase in economic activity.

Scholarship on business tax incentive programs is mixed, but generally negative as to the impact of government economic development efforts to create jobs and additional wealth or other announced economic goals of these programs. For instance, Todd Gabe of the University of Maine and David Kraybill of The Ohio State University examined state economic development incentives on 366 Ohio manufacturing and nonmanufacturing establishments that began large expansions between 1993 and 1995. They found empirical evidence to suggest that the incentives offered these firms had little if any actual impact on expected employment growth. The small impact that was seen suggested a slightly negative effect on actual growth.[93]

In a February 2001 review of more than 300 scholarly papers on economic development programs, Terry Buss, then a professor of public management at Suffolk University in Boston, found that "studies of specific taxes are split over whether incentives are effective, although most report negative results."[94]

These findings of questionable and even negative economic impact would not surprise many scholars. In their 2004 paper "The Failures of Economic Development Incentives," University of Iowa economists Peter Fisher and Alan Peters explain the findings of their metareview of academic literature. (A metareview is simply a review and summation of many literature reviews; literature reviews are themselves summations of the research of fellow scholars on particular subjects.)

Fisher and Peters examined three questions surrounding government business development programs. First, do incentives improve growth and development where offered more than would occur on its own? Second, is this development directed to low-income populations? Third, they ask, "How costly to government is the provision of these incentives compared to alternative policies?"[95]

Their conclusion was also mixed, as are many literature reviews, but on balance Fisher and Peters surmise that these programs are either ineffective or carry costs that exceed the alleged benefits derived from them. As to their first question, they conclude that:

The upshot of all of this is that on this most basic question of all — whether incentives induce significant new investment or jobs — we simply do not know the answer. Since these programs probably cost state and local governments about $40-$50 billion a year, one would expect some clear and undisputed evidence of their success. This is not the case. In fact, there are very good reasons — theoretical, empirical and practical — to believe that economic development incentives have little or no impact on firm location and investment decisions.[96]

The two economists think there may still be a role for government to play in economic development, but it should focus more on the fundamentals, such as infrastructure and education, as well as worker training. That said, Fisher and Peters conclude: "(T)he most fundamental problem is that many public officials appear to believe that they can influence the course of their state economies through incentives and subsidies to a degree far beyond anything supported by even the most optimistic evidence."[97]

In addition to the presence of a range of findings in the literature, policy recommendations are further challenged by the absence of findings extrapolated to a benefit-cost framework. Even if a robust econometric finding of a positive impact of targeted fiscal incentives were to occur, it would not necessarily translate into a clear policy recommendation in favor of such incentives. For instance, if a study of a state or region concluded that there were a statistically significant link between targeted tax incentives and new jobs, the tax incentives might still be bad policy if each new entry-level job was purchased at the cost of a million dollar state tax investment.

Further, as mentioned earlier, evaluation of targeted incentives on local economic activity has been more sporadic than analysis of general fiscal policy. Also, the analytical methods employed by state economic development agencies are better suited to managing programs than to evaluating economic growth. In particular, the use of firm-specific reports of gross job flows may be a useful management tool, but it is particularly ill-suited to economic analysis. Thus, a review of findings regarding targeted tax incentives will leave an unbiased reader hungry for more substantive analysis.

Timothy Bartik’s 2002 study[98] provides an admirable survey of methods for evaluating targeted incentive policies. The estimation provided in this study is a direct result of Bartik’s recommendations and conforms to the multiple methods of econometric estimation reviewed in his paper.

Theoretical Considerations in the Model

Before we review our model, there are some additional considerations that direct the research.

First, a major challenge in many of the fiscal incentive studies is the holistic treatment of state fiscal policy. Clearly, firms and individuals both respond to incentives in choosing their place of residence through both taxes and amenities. The latter of these two variables includes government expenditures on such things as parks, roads and police. An econometric comparison of regions that does not account for the fullness of tax policy differences runs the great risk of misestimating the role a particular incentive plays.

For example, in a nationwide study of targeted tax incentives, any analysis that does not estimate effective tax rates (distinct from the targeted incentive policy) will not properly specify the causative relationship between taxes and firm location decision. A similar argument regarding infrastructure may be offered. Thus international or interstate studies of fiscal policy impacts will necessarily require a comprehensive estimate of tax burdens — not simply expenditures or credits in a targeted incentive program.[99] Fortunately, the intrastate study we conduct here largely avoids this concern, since the bulk of fiscal differences will occur at the state, not local, level.

In addition to fiscal considerations, a number of studies have cast doubt upon the magnitude of the regional economic impact of large firms, which are the most likely to receive fiscal incentives. These studies include papers published in 2004 by Kelly Edmiston,[100] William Fox and Mathew Murray[101] and Michael Hicks.[102]

Edmiston finds that the impact of new large firms is almost always overstated, with multipliers often less than one. He further finds that expansion of existing firms generates substantial effects. Fox and Murray test the local impacts of large firm relocation, finding no significant net impacts in the communities in which the firms locate. Using a quasi-experimental approach, Hicks finds that large gambling and wholesale-retail facilities locating generate no net employment or income gains in the counties in which they locate. In these studies it is not only the effectiveness, but the very potential for effectiveness — the "efficacy" — of targeted business incentives that are cast into doubt.

In evaluating the impact of the MEGA program, this analysis is aided by the fact that only the state of Michigan will be investigated. While Michigan is one of the most geographically difficult U.S. states to model (due to the physical split between the upper and lower peninsulas), the commonality of the federal and state tax instruments suggests that a relatively simple model may be effectively employed to test the impact of the MEGA credits upon the state’s economic growth, incomes and employment.

The analysis presented below will evaluate a rather limited, but important question: Did the MEGA credits influence either aggregate or business-sector growth in Michigan’s counties through 2002? Since this study is confined to a single state, the overall fiscal condition of the state is not analyzed. This makes the analysis more limited, but more tractable in scope. It also offers the potential for results to change if overall policy were to be modified. For example, our findings are conditioned upon the policies in place before and during the MEGA credit period. A change in labor market or fiscal policy in the coming years may render our findings inappropriate as a forecast. We can only speak to what has happened.

Other considerations beyond the scope of econometric analysis matter. For example, any targeted incentive will inevitably treat firms differently. There is a considerable potential range of noneconomic impacts that can result from a policy that permits elected officials to distribute public funds to individual firms. In the upcoming analysis, we can only estimate whether the MEGA credits have changed Michigan’s economic landscape, not whether they are an appropriate policy, even if they do improve incomes and employment. Such concerns are discussed elsewhere in this report.

Modeling the Economic Impacts of MEGA

In order to assess the impact of the MEGA Credits, we have constructed an econometric model of the type recommended by Bartik (2002) as an advanced statistical measure of economic development credits. The full technical details and modeling considerations are contained in "Appendix A: The Model of MEGA’s Economic Impacts" and have been subject to peer review. Here, we briefly summarize the method.

Econometric analysis of each of Michigan’s counties from 1990 through 2002 provides a basis for assessing whether or not the MEGA credits actually influenced economic activity, either in aggregate, or within specific industries. The time period was selected to allow five years of modeling of Michigan’s economy prior to the implementation of MEGA. We extend our analysis to 2002, since it is the most recent year for which county-level data on income and employment have been published by the Bureau of Economic Analysis at the U.S. Department of Commerce and the Bureau of Labor Statistics at the U.S. Department of Labor. This allowed us to investigate the impact of the first seven years of the program.

We are most interested in changes to income, employment and the unemployment rate in counties where firms received specific MEGA credits. The econometric method we use in this effort is designed specifically to account for the impact of MEGA credits.

As we selected the model, we were aware of the problem of identifying the actual MEGA credit amount and date of impact. Also, we are aware that despite the selection criterion, some attempt to target more distressed areas may also influence the decision to offer a firm a MEGA credit. Further, we are aware that the impact of surrounding counties or regions and regional trends may influence the impact of the MEGA credits, or more generally economic growth as measured by incomes and employment. Each of these characteristics of the MEGA credit program were considered and empirically tested when we chose both the type of model and its specific form. Again, the details of the model are contained in Appendix A.

It was not possible to directly test the disaggregated impact of the MEGA credits to either high technology firms or offices, since there is no data series that treat either offices or high technology activities differently from other firms within their respective industries. Thus, we tested aggregate incomes and employment impacts in three major sectors potentially affected by MEGA: manufacturing, wholesale and construction.[103]

We were able to test the aggregate impact of these MEGA credits. The major strength of our model is that it evaluates what actually occurred, not what was hypothesized to occur, subsequent to the awarding of a MEGA credit.

Results of the Estimation

County-Level Results

Our first tests were on the impact of county-level employment, incomes and the unemployment rate during the period 1990-2002. Our model worked very well, proving sufficiently flexible to accommodate the data that we have available and performing similarly to a number of other regional growth models.

In the case of county-level changes to per-capita income, employment and the unemployment rate, the impact of MEGA credits is unambiguously nonpositive. County-level changes to these economic measures range from zero (the most common result) to modestly negative. The results clearly and strongly suggest that, as a charitable interpretation, in aggregate, the MEGA credits have been unsuccessful in improving per-capita income, employment and the unemployment rate.

Two objections could be raised in response to these results. One is that the data end in 2002, while the local impacts may require a longer period to materialize.

It is true that the lag in local impact could be large. Still, the data would reflect the impact of MEGA credits implemented in 1995, and these have not produced a net positive impact on county-level employment or incomes. If the failure to detect an economic impact is due to a lag, the lag is at least seven years.

A second issue is the period being studied, which included periods of recession and a weak economy. Some might wonder if MEGA might have prevented local economic conditions from being worse, even if it didn’t produce the intended economic growth.

This does not appear to have happened. The many Michigan counties whose companies did not receive MEGA credits fared no worse than the counties whose businesses did receive MEGA credits. The evidence clearly suggests no benefit to a Michigan county from a private facility in that county receiving MEGA incentives.[104]

State-Level Results

The failure of MEGA credits at the county level has an important corollary: The state of Michigan as a whole has not received an economic benefit in per-capita income, employment or the unemployment rate from the MEGA program.[105]

Business-Sector Results: Manufacturing, Warehousing and Construction

It is possible that MEGA credits fail to improve economic growth, but nevertheless shift economic activity between business sectors. We therefore investigated, as mentioned earlier, the impact of MEGA credits on wages and employment in three different business sectors: the industrial sectors of manufacturing and warehousing activities, and the construction industry.

With manufacturing and warehousing activities, there were no impacts from MEGA credits on employment that came close to being economically or statistically significant. There was similarly no statistically significant impact on warehousing-related wages. In contrast, there was a statistically significant reduction in manufacturing wages, but it was so small as to be economically insignificant.

Only when we assessed the impact of the MEGA credit on construction employment in a county did we find that under certain circumstances MEGA credits had an impact. This impact was positive: One new construction job was created for each $123,000 in MEGA credits approved in a county.

Unsurprisingly, however, these jobs were temporary, and we found that 75 percent of the net MEGA credit impact on construction employment disappeared in the first year the project started, and the full net increase in construction employment was gone by the end of the second year. The new jobs also carried lower wages than those already in existence — although, as with the decline in manufacturing wages, the resulting decline in average construction wages was so small as to be virtually economically meaningless. (Specifically, for each $1,000,000 MEGA credit, the average construction worker sees his total annual wages drop by less than 25 cents.). Finally, we would add that even the interpretation of construction job growth should be made with caution, as the model’s estimates did not generate typical levels of statistical significance.

The transience of the construction jobs is consistent with most findings of construction employment dynamics. The findings are also consistent with the economic challenges in Michigan for significant portions of the period being analyzed. Recessions reduce wages, and the lower wages were likely the product of a reduction in hours worked (a common business cycle result).

Other findings from the model involving variables other than the MEGA credits are discussed briefly in Appendix A. The results relevant to MEGA, however, appear above, and they clearly indicate that the MEGA credit program has failed to increase either employment or incomes, although it has shifted business activity towards the creation of some temporary construction jobs.[106]

Summary of Findings

Our empirical analysis of MEGA credits from the beginning of the program through 2002 suggests that the largest impact of MEGA credits was a transient increase in construction employment that lasted about two years. This increase represented a shift in economic activity toward construction, with the cost per new construction job being approximately $123,000 in MEGA credits (plus MEGA program costs and other opportunity costs).[107]

The analysis suggests no net economic benefit to the counties that hosted firms receiving MEGA grants. MEGA credits had no effect on a county’s per-capita income, employment and unemployment rate. Similarly, the MEGA credits had no measurable impact on the state’s per-capita income, employment and unemployment rate.