Summary
State and local governments regularly compete to offer large discretionary subsidies to attract firms, and the accompanying jobs, to their jurisdictions. Many argue that this is a classic race to the bottom. However, there is a potential allocative efficiency gain to allowing states to compete for firms with subsidies. This paper looks at the overall effects of subsidy competition, which has historically been an under-studied topic due both to the limited availability of data and a lack of clear counterfactuals over where firms might have located in the absence of subsidies.
This paper uses new, hand-collected data on state-level incentive spending and firm-level subsidy deals across the U.S. between 2002 and 2017, together with techniques from the empirical auction literature, to estimate a model of the subsidy competition “market” and quantify its welfare effects. In the model, the firm develops—and approaches—a shortlist of viable locations. The firm’s decision is treated as an English auction, in which bidders (the states, in this case) compete with increasing bids until only one remains.
Differences in location characteristics mean that the firm does not necessarily locate in the state that offers the largest subsidy. The auction model suggests the winning subsidy should just compensate the firm for locating in the winner, instead of the runner-up, location. This paper uses data on winning subsidies and the identity of the runner-up location to recover firms’ preferences over location characteristics. Findings suggest, for example, that if the winning state increases the corporate tax rate by one percentage point, the state needs to offer $11 million more in subsidies to attract the average-sized manufacturing firm, all else equal.
With these estimates of firms’ preferences over location characteristics, the states’ distributions of willingness to pay for firms can be recovered. The estimated distributions of state valuations and firm profits can then be used to evaluate a counterfactual policy in which state and local governments do not offer discretionary subsidies.
In the no-subsidy counterfactual, about half of the firms in this dataset choose different locations. This leads to a decrease of total welfare (firm profits plus state valuations) of about 4%, because the higher-valuation places are not able to express those values through bidding. The firms are the clear winners from subsidy competition. Across the dataset, states transfer a total of $40 billion to firms for $13 billion in gains. The benefits to firms also shrink substantially where the firm costs of engaging in subsidy competition (i.e., hiring site selection consultants) are taken into account, or if states are assumed to be over-optimistic in their valuation of firms.
These results suggest that the scope for discretionary subsidies to be an effective tool to reduce geographic inequality in the U.S. is extremely limited. Further research is needed to explore the effects of alternative subsidy policy regulation—including the substantially different system that prevails in the EU—especially in light of the increased enthusiasm for industrial policy, both in the U.S. and abroad.
Main article
State governments regularly compete to offer large discretionary subsidies to attract firms to their jurisdictions. This paper contributes to the historically understudied question of the welfare effects of this subsidy competition. Using a new U.S. dataset and treating the firm’s location decision as an English auction, this research finds that banning subsidies would decrease total welfare by about 4%. Firms emerge as the clear winner from subsidy competition, however; discretionary subsidies appear to have an extremely limited scope to be an effective tool to reduce U.S. geographic inequality.
State and local governments regularly offer large discretionary subsidies to attract firms, and the accompanying jobs, to their jurisdictions. In fact, this deployment of subsidies is estimated to be one of the largest economic development tools used in the United States (Bartik 2017). The incentives can be staggering—for example, in 2006 Honda received a package of tax breaks and subsidies worth $145 million to locate a new assembly plant in Indiana. The same year, Samsung received $233 million to build a new semiconductor plant in Texas.
The large subsidy numbers are a function, in part, of competition between states. Many places want to attract a new auto assembly plant, and they therefore offer subsidies to make their location more attractive in the eyes of the firm. Many argue that this is a classic race to the bottom. However, there is a potential allocative efficiency gain to allowing states to compete for firms with subsidies. The place that is struggling more to create jobs and may therefore have the greatest value in winning a firm—but is not a priori the most profitable location—can express this high value in the form of the subsidy. In that case, subsidy competition would lead firms to locate in places where they create larger positive externalities, and therefore such competition can possibly be an effective tool to address rising geographic economic inequality within the United States.
The deployment of subsidies is estimated to be one of the largest economic development tools used in the United States.
This paper tackles this question: what are the welfare effects of subsidy competition? Given the pervasiveness of subsidy-giving as a policy tool, one would think the answer to this question would already be clear. However, two main difficulties have prevented researchers from making much progress. The first is data. There is no comprehensive dataset on large discretionary subsidies, due to limited transparency from state and local governments on the subsidy-setting process. The second issue is that, though we can observe where firms locate and the size of the subsidies they receive, we do not know the counterfactual location firms would have chosen in the absence of subsidies. What we observe is an equilibrium outcome that is a function of the firm’s preferences over location characteristics and the competition between state and local governments.
A new U.S. dataset on state-level incentive spending and firm-level subsidy deals
To make progress, I gather new, hand-collected data on state-level incentive spending and firm-level subsidy deals across the U.S. between 2002 and 2017. These subsidies are composed of tax breaks, grants, loans, and in-kind transfers, from both the state and local government. The contribution from the state is usually the majority of the total, and therefore I use the state as the main government agent of interest. The data includes about 400 subsidy deals, with details on the terms of each deal (i.e., what the state offered the firm and what the firm, in turn, promised the state) and details on the underlying competition for the firm (i.e., the number of states vying for the plant, and the identity of the top contenders). I then develop a tractable model of the subsidy competition “market,” which encompasses many real-world features of subsidy competition in the United States. I leverage the new data, together with techniques from the empirical auction literature, to estimate the model and quantify the welfare effects of subsidy competition.
The average subsidized firm in the data promises to create 1,400 jobs and receives a subsidy worth $150 million over 10 years.
The average subsidized firm in the data promises to create 1,400 jobs and receives a subsidy worth $150 million over 10 years. Although job creation is the focus of much of the discussion around these deals, the size of the subsidy is not easily explained by the number of jobs promised by the firm. This could partly be due to location characteristics—places that may be less attractive locations for firms may need to offer larger subsidies, all else equal. In the raw data, local characteristics that are favorable to firms, like right-to-work laws and the prevalence of research universities, are correlated with smaller subsidy sizes. Meanwhile, places with higher unemployment rates also offer larger subsidy deals, perhaps reflecting the high value that governments place on job creation.
Modelling firm location choice as an English auction
A model is necessary to disentangle differences in firm profits across locations from the location’s value of winning the firm. The model mimics how subsidy competition works in practice. Once a firm has determined it would like to build a new plant, for example, it hires a team of consultants to do research on potential sites to find a list of places that would be suitable for the new investment. Given the research, the firm develops a shortlist of viable locations and then contacts the governments in those places. At this stage, the firm tells the government it is considering investing in their state, but would like to know what the government can do for them. This is where the subsidy comes in.
Because the firm does this for multiple states at the same time, a bidding war can break out, where the firm goes back and forth between states to make them aware of competing offers. You can think of firms “shopping” across locations for the most attractive place, not unlike the process a consumer might go through for a car, or a loan (Allen, Clark and Houde, 2019; Cuesta and Sepulveda, 2019). Finally, the firm will locate in whichever state is the most attractive, i.e., where the firm will be most profitable. This is a function of the subsidy offer, as well as the location characteristics that are relevant to the firm, such as taxes, the labor market, and infrastructure.
High-tech manufacturing firms value both college-educated population and agglomeration, as measured by the industry establishment share in the location.
In auction parlance, this is an English auction—bidders compete with increasing bids, until only one remains. The classic example is an auction house, like Sothebys, auctioning off fine art, with bidders sitting with paddles to express their willingness to increase their bids. Of course, the firm does not gather all the state governments in a room, but the logic holds. The benefit of modeling the competition as an auction, besides the similarity to the real-word institutional details, is that there are techniques developed in the empirical auction literature to estimate bidders’ willingness to pay using observed bids.
Uncovering firm preferences around location
A central difficulty remains that the firm does not necessarily locate in the state that offers the largest subsidy. In the case of Amazon HQ2, multiple cities offered larger incentive packages than the winning locations. Therefore, the empirical approach needs to adjust for differences in location characteristics that make some places more or less attractive. The auction model suggests the winning subsidy should just compensate the firm for locating in the winner, instead of the runner-up, location. In other words, the winning subsidy is equal to the difference in the firm profits in the runner-up and winning location plus the runner-up’s subsidy offer. I use data on winning subsidies and the identity of the runner-up location to recover firms’ preferences over location characteristics.
To see the intuition for this approach, imagine two subsidy deals for automobile manufacturing facilities of identical size. One plant locates in Alabama with a subsidy of $100 million and the other locates in Ohio with a subsidy of $140 million. In both cases, the runner-up in the subsidy competition is South Carolina, so the runner-up valuation and profit are held constant. Now suppose Alabama and Ohio have almost all of the same location characteristics: the same tax rate, the same wages, the same skilled workforce. The only difference between the two states is that Alabama is a right to work state and Ohio is not. Then, the $40 million difference in the two observed subsidy deals would be attributed to how much automobile manufacturers prefer to locate in a right to work state.
In the no-subsidy counterfactual, about half of the firms choose different locations.
The results on firm preferences can be interpreted as the changes in subsidy needed given changes in location characteristics. For example, I find that if the winning state increases the corporate tax rate by one percentage point, the state needs to offer $11 million more in subsidies to attract the average-sized manufacturing firm, all else equal. High-tech manufacturing firms value both college-educated population and agglomeration, as measured by the industry establishment share in the location. A location that experiences a one standard deviation increase in the proportion of the population that has a college degree (equivalent to 6 percentage points) will be able to offer a $36 million smaller subsidy to attract the high-tech manufacturing firm, all else equal.
Estimating the welfare effects of subsidy competition
Given the estimates of firms’ preferences over location characteristics, I can proceed with recovering the states’ distributions of willingness to pay for firms. I then use the estimated distributions of state valuations and firm profits to evaluate a counterfactual policy in which state and local governments do not offer discretionary subsidies. This shift away from discretionary subsidies has been proposed by some state legislatures in the past, and is effectively the policy in the EU, with some exceptions.
The firms are the clear winners from subsidy competition.
In the no-subsidy counterfactual, about half of the firms choose different locations from those observed in the data. These locations are places where firms are more profitable (i.e., they have better location fundamentals), but the location has slightly lower valuations for winning the firm. This leads to a decrease of total welfare (firm profits plus state valuations) of about 4%, because the higher-valuation places are not able to express those values through bidding.
The firms are the clear winners from subsidy competition. Most of the aggregate welfare gain is transferred to the firms; the total subsidy spending over the sample amounts to over $40 billion, while state valuations only increase by about $13 billion under competition. In other words, states transfer a total of $40 billion to firms for $13 billion in gains. Although, in the aggregate, states are better off under the ban, my results suggest that any type of subsidy “truce” would likely be hard to sustain—the results are highly unequal across geographies. In the counterfactual with no competition, many of the states in the Midwest and South lose the majority of the firms that they had attracted with subsidies, while states like New York, California, Texas and Virginia retain all their firms or are net gainers of firms under a subsidy ban.
A few additional considerations temper the estimated welfare gain from subsidy competition:
- Firms have costs to engaging in subsidy competition that are not explicit in the model. Namely, firms hire site selection consultants to research locations and negotiate with governments. When I incorporate a conservative estimate of consulting costs into the analysis, the welfare gain from competition shrinks from 4.3% to 1.2%.
- States may overestimate the benefit of winning any particular firm. In a back-of-the-envelope estimate I find that if states are slightly over-optimistic about the valuation of a firm, the welfare gain from subsidy competition quickly dissipates.
- States with governors facing re-election are willing to pay more for a manufacturing firm, all else equal (Slattery 2024). This raises the issue that the “welfare” of the state decision maker may not necessarily align with that of a social planner.
The scope for discretionary subsidies to be an effective tool to reduce geographic inequality in the U.S. is extremely limited.
In short, I find that the scope for discretionary subsidies to be an effective tool to reduce geographic inequality in the U.S. is extremely limited. In fact, the places that are struggling the most are rarely even considered by firms as viable locations and therefore are not observed participating in subsidy competition. The policy in the EU is substantially different, allowing only these most distressed places to offer discretionary subsidies. Further research is needed to compare these two systems and explore alternative subsidy policy regulation, especially in light of the increased enthusiasm for industrial policy, both in the U.S. and abroad.
This article summarizes ‘Bidding for Firms: Subsidy Competition in the United States’ by Cailin Slattery, published in the Journal of Political Economy in August 2025.
Cailin Slattery is at the University of California, Berkeley.
References:
Allen, Jason, Robert Clark, and Jean-Francois Houde. 2019. “Search Frictions and Market Power in Negotiated-Price Markets.” Journal of Political Economy, 127(4): 1550–1598.
Bartik, Timothy J. 2017. “A New Panel Database on Business Incentives for Economic Development Offered by State and Local Governments in the United States.” W.E. Upjohn Institute.
Cuesta, Jose Ignacio, and Alberto Sepulveda. 2019. “Price Regulation in Credit Markets: A Tradeoff between Consumer Protection and Credit Access.” Working Paper.
Slattery, Cailin. 2024. “The Political Economy of Subsidy-Giving.” Working Paper.
