EVALUATING FIRM-SPECIFIC LOCATION INCENTIVES:
An Application to the Kansas PEAK Program
Bobonis, Gustavo J., and Howard J. Shatz. 2007. “Agglomeration, Adjustment,
and State Policies in the Location of Foreign Direct Investment in the United
States.” The Review of Economics and Statistics 89(1): 30–43.
Buettner, Thiess, and Martin Ruf. 2007. “Tax Incentives and the Location of FDI:
Evidence from a Panel of German Multinationals.”
International Tax and Public Finance 14(2): 151–64.
Easson, Alex. 2004. Tax Incentives for Foreign Direct Investment. The Hague:
Kluwer Law International. ICAincentives.com. 2013.
The Only Global Incentive Deal Database.
International City/County Management Association and National League of Cities.
2009. Economic Development 2009 Datasets.
Jensen, Nathan M., Edmund Malesky, and Matthew Walsh. 2014. Competing for global
capital or local voters?
The politics of business location incentives. Working Paper, Washington
University in St. Louis.
Klein, Patrick, and Enrico Moretti. 2013.
People, places and public policy: some simple welfare economics of local
economic development programs.
National Bureau of Economic Research Working Paper 19659.
Klemm, Alexander and Stefan Van Parys. 2012. Empirical evidence on the effects
of tax incentives. International Tax and Public Finance 19 (3): 393-423.
Legislative Division of Post Audit. 2013. Performance audit report.
Economic development: determining which economic development tools are most
important and effective in promoting job creation and economic growth in Kansas.
Neumark, David, and Jed Kolko. 2010.
“Do Enterprise Zones Create Jobs? Evidence from California’s Enterprise Zone
Journal of Urban Economics 68: 1–19.
Neumark, David, Brandon Wall, and Junfu Zhang. 2011. “Do Small Businesses Create
More Jobs? New Evidence from the National Establishment Time Series.” Review of
Economics and Statistics 93(1): 16–29.
The Pew Center on the States. 2012. Evidence Counts: Evaluating State Tax
Incentives for Job Growth.
The use of financial incentives to attract and retain companies has become one
of the most common economic development strategies of U.S. states and
municipalities. Despite the widespread debate on the effectiveness of these
programs, few systematic academic studies have examined how incentives affect
job creation and local economic development. The result is that policymakers
often lack objective data from which to draw conclusions about the benefits of
At a time in which state and municipal budgets are increasingly strained, new
tools that allow policymakers to evaluate and understand the costs and benefits
of incentive programs are needed. This paper attempts to provide policymakers
with such a tool by exploring the impact of the Promoting Employment Across
Kansas (PEAK) incentive program and other incentives. The paper is part of a
larger project funded by the Ewing Marion Kauffman Foundation that seeks to
examine the effects of incentives on job creation in the Kansas City region as
part of a two-year study of incentive competition.
The paper’s main finding is that, when comparing firms receiving PEAK incentives
to a similar set of “control” firms, PEAK incentives recipients are
statistically not more likely to generate new jobs than similar firms not
receiving incentives. A secondary set of findings shows that firms relocating to
Kansas, with or without incentives, do not experience job growth at higher rates
than existing firms.
More important than the specific analysis of the PEAK program, this paper
provides a model for the evaluation of incentive programs that could be applied
to both state and municipal incentive programs. In the conclusion, I offer some
suggestions for reforms of both the reporting of incentives and the analysis of
the economic impact of incentives, and alternative economic development
The use of financial incentives to attract and retain companies has become one
of the most common economic development strategies of U.S. cities and
municipalities. In a survey of U.S. municipalities, 95 percent of respondents
indicated they utilized some form of fiscal incentives to attract firms, while
every U.S. state has a menu of incentives1 to offer firms and many of these
states have shifted towards offering fewer, but much larger “megadeals.”2 While
many states have increased their scrutiny of their incentive programs, only four
states have integrated evaluation of incentives into the state policy process.3
While these incentives come in many different forms, ranging from tax holidays,
grants, and low-cost loans to infrastructure improvement, these government
policies targeted at individual firms have come under increased scrutiny from
academics, NGOs, and the media. The Kansas City metropolitan region, which
straddles the Missouri-Kansas border, has become a symbol of the problems with
incentive competition within the United States. The New York Times exposé on
incentives devoted a full installment of the series to Kansas City.4
Both critics and supporters of incentive policies can find examples of firms
receiving incentives that support their respective stories. Some incentives can
be credited with luring investment or facilitating an expansion that generates
direct jobs and tax revenues, which have much larger spillovers to the
community. More common, however, are criticisms of incentive programs that
illustrate the inefficiency and ineffectiveness of incentives as a job creation
This existing debate is an important starting point in documenting how and when
incentives work or don’t work, but fails to provide a more holistic picture of
the costs and benefits of incentives. This working paper is part of a larger
project funded by the Ewing Marion Kauffman Foundation that explores how
incentives affect job creation in the Kansas City region, as part of a two-year
study of incentive competition. This paper provides preliminary evidence about
one of the most important Kansas incentive programs, Promoting Employment Across
Kansas (PEAK). The main finding is that, when comparing firms receiving PEAK
incentives to a similar set of “control” firms, firms that receive PEAK
incentives are not statistically more likely to generate new jobs than similar
firms not receiving incentives. A secondary finding is that attracting new
investment, while clearly generating new jobs in the short run, has a limited
impact on job creation. Firms that relocated to Kansas are no more dynamic in
their job creation outcomes than already established firms, although they can
shift existing jobs from an existing location to Kansas.
In the conclusion, I offer some suggestions for reforms of both the reporting
of incentives and the analysis of the economic impact of incentives, and
alternative economic development strategies.
INCENTIVES: ARGUMENTS FOR AND AGAINST
There are two broad rationales for using firm-specific incentives.5 First,
proponents argue that the attraction of even a single major firm can serve as a
catalyst for local economic development, having positive impacts on wages,
property values, and ultimately tax revenue.6 If a small incentive can swing the
decision of a large firm, the benefits of these incentives far outweigh the
A second rationale is the classic “market failure” concept that many of the
benefits of a firm aren’t simply captured by profits in the firm, but also have
spillover effects in the community. For example, imagine two firms. One firm
will create ten jobs but have few other spillover effects in the Kansas City
area. A second firm will create ten direct jobs, and by sourcing from suppliers
and using local distributors, that company will create an additional ten jobs.
While both firms have the same payroll, sales, and ultimately profitability, the
second firms is much more valuable to the region.
These firm-specific incentive programs are not without their critics. Markusen
and Neese (2007) argue that incentive competition is a net loss. Easson (2004,
63), puts it bluntly:
According to the conventional wisdom, tax incentives for
investment—in particular for foreign direct investment (FDI)—are not
That is the view held almost universally by theorists and by the international
bodies that advise on tax matters.
Tax incentives are bad in theory and bad in practice.
Without dwelling on the many details that critics have
highlighted, we can briefly note the main criticisms. First, much of the
literature on incentives, and tax policy in general, finds that incentives are
rarely the main factor for shaping investment location decisions, or in the
decision for expansion.7 Incentives often are what firms look for to sweeten the
deal once they have made a decision. Thus, they are not especially effective in
luring new firms to a region.
Second, they are often excessively costly relative to the number of jobs they
create. Even if we take the amount of jobs created on the face value, in many
cases the dollar amount per job doesn’t make sense. More problematic is that
most studies of incentives find that a large percentage (often in excess of 75
percent) of the jobs “created” by incentives were going to be created anyway.
Thus, the “redundancy” rates of incentives make them three, four, or five times
more costly than a simple dollars per job calculation.
Third, and perhaps most obvious to observers of the ‘border war’ in Kansas City,
is that incentives can distort economic decisions. The irony is that the few
times when incentives are effective, they could be for simply maximizing
government subsidies and not for efficiency or market-seeking motivations. The
story of Applebee’s jumping back and forth across the Kansas and Missouri border
is a striking example. These incentives affected the location decision of the
company, but it is difficult to make that case that this was an economically
efficient use of taxpayer money.
Taking stock of this literature, there are a number of theoretical positives and
negatives of incentives, yet most of the literature on incentives mainly just
examines the firms that received incentives. In this paper, I propose an
EVALUATION OF THE EFFECTIVENESS OF INCENTIVE PROGRAMS
Evaluations of incentive programs are notoriously difficult. The first problem
is data limitations. Many countries, states, and cities provide very few details
about their incentive programs, and even less detail on the companies that
received incentives. While this lack of transparency has been well documented in
other studies, this relates to the second issue: A proper evaluation of an
incentive program requires the generation of a counterfactual. What would the
company have done without an incentive? Would jobs, sales, and profitability be
less if the company didn’t receive the incentive? Would the company have moved
to another location, or possibly gone out of business?
These are difficult questions to answer, and most of the information required is
in the hands of companies seeking incentives. Thus, this information asymmetry
(only the company knows if the incentive would be necessary) can lead
governments to provide excessive incentives to firms that would have undertaken
the same activity with or without government support. To properly evaluate an
incentive program, then, we cannot just look at a firm that received an
incentive tied to a new investment or expansion. Of course we will see a
correlation between new capital investment and more jobs with this incentive
program. But what inferences can we draw on how much of this outcome we should
attribute to the incentive program?8
Even a study of a company over time can lead to erroneous attribution of
positive outcomes to incentives. For example, imagine a company that is in
business for twenty years and engaged in three expansions of employment. In Year
1, the company started with ten jobs; in Year 5, an expansion of an additional
five jobs (fifteen jobs in total); and in Year 10, another expansion of five
jobs (twenty jobs in total). If the company received an incentive in either or
both years, most statistical models would find a positive relationship between
incentives and job creation.
But the problem is that if companies only apply for and receive incentives in
the years they already were considering expanding, it is highly like we are
really erroneously concluding that incentives help generate jobs. This is akin
to claiming that hospitals kill people, because many more people die in
hospitals. Companies that receive incentives for job creation create some jobs,
but this does not mean that the incentives were effective in creating these
While there are few clear fixes to this problem of causal inference, in this
working paper I outline a relatively comprehensive database of firm
establishments that gives us some leverage on this problem. While we know that,
as mentioned above, companies that already are considering expanding or
relocating are more likely to apply for and receive incentives, we can use this
rich dataset to perform “matching methods” to attempt to compare firms that
received incentives with other very similar firms. Thus, we can explore if the
firms that received incentives perform better than their peer groups after
receiving incentives. In the next section I give an overview of the Kansas PEAK
Program and the establishment level dataset that will be used for matching PEAK
firms with similar firms in Kansas.
THE KANSAS PEAK PROGRAM AND THE NATIONAL ESTABLISHMENT TIME-SERIES DATA
The Promoting Employment Across Kansas (PEAK) program is an incentive program
enacted in 2009 that has many similarities to other state programs. PEAK
provides an incentive (retaining up to 95 percent of the payroll withholding
taxes of eligible employees) to encourage firms to relocate, expand, or stay in
Kansas. This program, administered by the Kansas Department of Commerce and the
Kansas Department of Revenue, was one of two programs evaluated as part of the
Kansas Post Audit Committee.9 The first part of their audit provides a detailed
overview of this program.
While there are clear eligibility conditions, such as paying above the county
median wage for the establishment and provisions for enforcement, this program
has not been without criticism. Part 1 of the Kansas Legislative Audit
identifies problems with the administration of the program and company
self-reporting that was never verified. In short, the existing data on Kansas
incentive programs hampers both the functioning and evaluation of this program.
The Legislative Division of Post Audit (2013, 11–12) highlights the lack of
actual, as oppose to estimated results, is a major constraint on the evaluation
of this program.
While there is no silver bullet to overcoming the lack of information collected
on companies receiving incentives, existing data on the employment and sales of
PEAK companies relative to other establishments in Kansas is available. To
assess the impact of PEAK incentives on individual firms requires fine-grained
establishment-level data. This data, created by Walls & Associates using Dun &
Bradstreet data provides one of the most comprehensive databases of
establishment-level data. This National Establishment Time Series (NETS) data,
in contrast to many other sources of firm data, disaggregates each establishment
of a firm. This is critical since most incentive programs provide funding for a
single establishment, for example, the single location of a company that has
multiple Kansas locations.
This data has been used by other researchers and compared to existing databases.
The most comprehensive analysis can be found in Neumark, Wall, and Zhang (2011).
In their study, they examined the correlations in employment numbers between the
NETS data and U.S. Current Population Survey and Current Employment Statistics,
which yielded an overall correlation of 0.99 and 0.95 respectively, although the
NETS data generally had higher estimates of employment and lower levels of
The most comprehensive data starts in 1992, when Dun & Bradstreet were allowed
to purchase Yellow Pages data to directly call individual firms. This massive
data collection effort has resulted in a database of millions of firms. This
data includes detailed information on 500,000 firms located in Kansas.
Using public records requests, documentation from the Kansas Legislative Audit,
and news media sources, we linked seventy-two PEAK incentive recipients to the
NETS data. As outlined in the Kansas Legislative Audit, between 2009 and 2013,
117 companies had signed PEAK agreements, although only ninety-four companies
were provided incentives and were active during the review. Thus, this working
paper captures the majority of the PEAK incentive recipients.
Comparing PEAK firms to all 500,000 establishments in the NETS data would be an
unfair comparison. PEAK firms tend to be much larger in both employment and
sales, and may be concentrated in different sectors. Thus, central to evaluating
this program is finding the correct comparison set of firms.
Luckily, there is a large amount of literature on the use of “matching methods”
to analytically compare treatment firms (firms getting PEAK incentives) with a
control group (firms that are similar to PEAK firms but did not receive
incentives). I utilize the most well known of these methodologies: propensity
score matching, using the five
“nearest neighbors.” These are not necessarily geographic neighbors. Rather,
they are firms that looked very similar to the firms receiving PEAK incentives.
To match these firms, I use a set of observational variables including the
firm’s previous employment, whether or not the firm is a subsidiary of a parent
company, and the sector of company (three-digit SIC code).
This comparison allows us to simply compare the total employment of the firm
in 2012, the most recent year of complete NETS data, between PEAK firms and
similar firms. In table 1, I present a comparison of these firms using the raw
2012 employment data.
Table 1. Comparing firms receiving PEAK incentives to other firms in Kansas
using propensity score matching
FIRM RELOCATIONS AND JOB CREATION
Thus far, my analysis has focused on job creation within a firm. The results
indicate that the companies that relocate to Kansas, with or without incentives,
are no more or less likely to generate jobs than similar companies in the area.
Yet this doesn’t mean that these companies won’t create new jobs when they
relocate to Kansas.
For example, according to data presented in the Kansas Audit, fifty-four
incentives were provided to new establishments and firms expanding existing
establishments. While these firms are not more likely generate more jobs in the
long run than existing Kansas companies, companies relocating to Kansas can
provide one-time job creation as jobs are shifted to Kansas from another state.
In our data, thirty-four PEAK incentives were provided to companies relocating
Unfortunately, it is difficult to evaluate how many of these PEAK-supported
relocations generated new jobs in Kansas since these relocations may have
happened even without a PEAK incentive, and some were moves across the state
line from Missouri. Can we associate PEAK incentives with job creation?
This is a difficult question and would require detailed data of individual
employees. What the data does tell us is that the vast majority of PEAK
incentives that went to relocations were for firms previously located in
Missouri (twenty-seven out of thirty-four relocations).
This bias toward attracting Missouri firms contrasts with the NETS data. Of the
over 45,000 firms in the dataset that relocated, almost 35,000 relocated from
another location in Kansas. While 79.4 percent of PEAK incentives provided to
relocating firms were targeted at Missouri firms, Missouri firms only represent
30.4 percent of the out of state relocating firms in the NETS data. Thus, while
many firms relocate to Kansas from large states like Texas (831 establishments)
and California (696 establishments), firms from these states very rarely receive
a PEAK incentive. This simple descriptive data suggests that a large number of
PEAK incentives firms may simply be shifting jobs across the Missouri-Kansas
Unfortunately there isn’t an obvious statistical fix for this problem unless we
can track individual workers within a company. Thus, the only recommendation is
to take care in interpreting job creation for companies that relocate within a
geographic distance that is easily commutable for existing workers. The NETS
database provides latitude and longitude information that would allow for
further exploration of these moves.
DISCUSSION AND CONCLUSIONS
This paper outlines a standard research methodology that can be applied to the
evaluation of the effectiveness of economic development programs. Central to
this evaluation is finding a comparison group of firms to be used as a “control
The accurate assessment of any evaluable program is enhanced with more detailed
data on the incentive programs, the recipients, and other firms not receiving
incentives. One simple policy recommendation is to aid the evaluation of
incentive programs through better management and sharing of data about incentive
programs. Basic information about the companies receiving incentives should
include identifying information beyond the company name. For example, if there
are multiple establishments, the establishment receiving the grant (and the
address of the establishment) should be provided.
A second, and perhaps less obvious, point is that the ideal comparison for any
incentive programs would be to have information not only on incentives granted,
but also on incentives not granted. Some states with discretionary funds for
relocation, such as the Texas Enterprise Fund, have a large number of applicants
that were not given incentives. These rejected applicant company names can (and
were by the author) accessed through a public records request. In contrast, the
State of Arkansas did not maintain data on companies that were rejected by their
discretionary incentives programs.
One simple solution is for discretionary, and even nondiscretionary, incentives
is to make the applications for PEAK grants available through public records
requests. This will provide much of the background information necessary for the
tracking of firms receiving incentives, and any information on firms that were
Also, the most important component of this evaluation is the use of other
non-incentive firms in an area as a control group. Thus, for a proper evaluation
of incentive programs we must not only focus on the firms receiving incentives.
We need to collect broader information about similar firms in the area to make a
proper comparison. This can be done after the fact for program evaluation, or
prior to implementation by using existing firms as a benchmark to evaluate an
Finally, we need to improve the overall method to collect data of firms
receiving state and local incentives. Those firms are supposed to report some
information to respective agencies, but collecting such information
retroactively is notoriously difficult even for agencies, as companies naturally
have the tendency not to disclose their internal information. There has to be an
explicit agreement at the beginning of receiving incentives about which company
information has to be reported to agencies.
My preliminary findings on the Kansas PEAK program, and Kansas incentive
programs more generally, is that there is no concrete evidence that they are
effective in generating jobs in Kansas. Yet a more comprehensive evaluation