Bidding for firms An asymmetric auction model of interjurisdictional competition

Bidding for Firms: An Asymmetric Auction Modelof Interjurisdictional CompetitionLaurent Martin*Department of EconomicsUniversity of WashingtonSeattle, WA 98195martinl@This version: September 8, 2000AbstractThis paper analyzes firm-specific interjurisdictional competition as an auction. Beforechoosing a site, a firm solicits bids (i.e., subsidies, tax breaks) from several jurisdictions.Its costs differ across locations so, unlike in standard high-bid auction models, the seller(i.e., the firm) has preferences over who among the bidders (i.e., the regions) wins. Usingmechanism design, it is shown that the firm’s optimal location decision rule solves atrade-off between subsidy extraction and cost minimization in a way that is inconsistentwith total surplus maximization, so the firm’s location can be inefficient. Positiveimplications of the model are shown to be consistent with some recent stylized facts.Extensions of the basic model discuss the political economy of bidding and the impactof negative interjurisdictional spillovers from location decisions. Should bidding wars bebanned? The model frames the current policy debate by suggesting that the policymakertypically faces an efficiency/equity trade-off. Whether or not location contests arepositive-sum games from society’s standpoint hinges primarily on whether the localbenefits from business locations vary sufficiently across locations to outweigh the welfarecosts of interjurisdictional bidding.Keywords: Interjurisdictional competition, auctions, asymmetric auctions, auctions with seller preferences.JEL classification: H73 (Interjurisdictional Differentials and Their Effects), D44 (Auctions)* I am most grateful to Peter Cramton, Robert Schwab, Lawrence Ausubel, and Wallace Oates for their advice and comments. I thank participants at the Stanford Institute for Theoretical Economics (SITE) Workshop on Interregional Competition in Public Economics where this paper was first presented, especially Gregory Besharov, John Conley, William Hoyt, Scott Page, and David Wildasin. The paper also benefitted from remarks and suggestions made by Gary Biglaiser, Philippe Jéhiel, Benny Moldovanu, Tomas Philipson, Jacques Robert, Jesse Schwartz, by an anonymous referee, and from discussions during seminars at University of Maryland, Concordia University, Université de Montréal, Dalhousie University, Simon Fraser University, University of Washington, and Université du Québec àMontréal. I thank the Social Sciences and Humanities Research Council (SSHRC) of Canada for its financial support.1. IntroductionBefore choosing a site, firms often initiate bidding wars in which competing communities offer escalating recruitment incentives. Location contests resemble auctions yet most models of interjurisdictional competition for business are analogous to the competitive market model.1 In contrast, this paper models firm-specific interjurisdictional competition as an auction designed and conducted by a firm with all the bargaining power. Operating costs typically differ across potential sites so, unlike in a standard high-bid auction, the firm (the “seller”) has preferences over who among the jurisdictions (the “buyers”) wins. The optimal location decision rules in this asymmetric auction are derived and compared to auction rules yielding efficient locations2In the United States, competition for business has been compared to an “economic war among the states.” Is it a zero-sum or a negative-sum game that should be banned, as some suggest?3 The model distinguishes circumstances in which firm-specific interjurisdictional competition is efficiency enhancing from those making it more likely to be harmful.Incentives-based competition for investments is a global phenomenon (Oman 2000). In recent decades, bidding wars for firms intensified in the number of participants, in the frequency of occurrence, and in the size of the amounts involved (Donahue 1997a,b, UNCTAD 1996, Hanson 1993). Mazda built a large plant in Flat Rock, Michigan – where the city and the state offered a $120 million package – after it had actively solicited bids from several jurisdictions (Milner and Moore 1989). The location decision for the Saturn plant, in Spring Hill, Tennessee, with financial incentives worth $70 million, followed a contest involving 38 states. Micron1 In Oates and Schwab (1988, 1991) and Wildasin (1988), for example, many small communities compete for small shares of a homogeneous stock of capital owned by a large number of individuals. For recent surveys of the literature on interjurisdictional competition in public economics, see Wilson (1999) and Oates (1996).2 In the literature, auctions are asymmetric when bidders’ valuations differ in ways that are commonly known(e.g., Maskin and Riley 1998). In an asymmetric auction, a revenue-maximizing seller should give preferential treatment to some bidders (Myerson 1981, McAfee and McMillan 1987, 1989, Ayres and Cramton 1996, Corns and Schotter 1999), even though she has no preferences per se. This paper extends these studies by introducing seller preferences as a source of asymmetry. Among the few authors who have, explicitly or not, analyzed the role of seller preferences in auction design are Branco (1994) and Naegelen and Mougeot (1998) who analyze procurement auctions when a government has preferences for domestic firms, Cornelli and Li (1997) who study privatization auctions of state enterprises with foreign bidders, and Laffont and Tirole (1991) who consider potential collusion between the seller and a buyer arising from favoritism. The model also complements the work of Jéhiel, Moldovanu, and Stacchetti (1996, 1999) who analyze auctions with externalities in which the bidders have preferences over the identity of the winner. See Klemperer (1999) for a recent survey of auction theory.3 For a broad presentation of this debate, see “The Economic War Among the States,” The Region 10, June 1996, and consult /sylloge/econwar, respectively published and maintained by the Federal Reserve Bank of Minneapolis. See also the series on “Corporate Welfare” in Time Magazine, November 1998, a OECD study (Oman 2000), and the UNCTAD (1996) report on Incentives for foreign direct investment.Technologies received over 300 offers from most states and from more than 70 countries in 1994 when it announced a plan to build a microchips plant – it accepted a $200 million deal and moved to Lehi, Utah. Before it built a new plant in Rio Rancho, New Mexico, where it received a $114 million incentive package, Intel had circulated an “ideal incentive matrix” to a number of states. Alabama gave $250 million to Mercedes in 1993. BMW received a package of incentives worth about $150 million to build a plant in South Carolina in 1992. Indiana signed a $290 million deal with United Airlines in 1991 after a 22 month bidding war (Breagy 1996). Many American cities spend large amounts of public funds to attract or retain sports teams (Noll and Zimbalist 1997). Outside the U.S., LG Group from Korea chose Wales instead of Scotland to build a $2.6 billion plant after Wales offered a subsidy worth about $46,000 per job (The Times, London, July 11, 1996), and Brazil states compete to attract foreign car makers (Oman 2000).Recruitment subsidies are often huge, but losing bids are sometimes even higher. Saturn chose Tennessee, but had received larger offers from New York ($1.5 billion over 30 years), Michigan ($660 million) and Illinois (at least $170 million). Local advantages, such as access to market, transportation, quality of life, cost of electricity, training and infrastructure assistance, were important in the firm’s decision. BMW rejected a $330 million package from North Carolina, and accepted a lower offer from South Carolina. The location chosen facilitated distribution to its potential major markets and provided lower long-term operating costs.4 These examples are analogous to auctions: some communities bid to attract a firm with characteristics, differentiated from the characteristics of other potential projects, that they value. These are auctions with seller preferences since, as in the last two cases, location-specific costs and recruitment subsidies both matter to the firm. If revenues do not depend on plant location, relative cost advantages measure the intensity of the firm’s preferences over the jurisdictions.Case studies (e.g., Buchholz 1998) confirm that an ascending auction setting provides the best analytical framework for interjurisdictional location contests, as opposed to a sealed-bid auction setting. Firms, who often hire site selection consultants, typically solicit fiscal incentives from many jurisdictions and use the offers received to negotiate counteroffers.5 Nunn, Klacik, and Schoedel (1996) study the location of new maintenance operating centers (MOC) for two airline4 For more details about these examples, see Breagy (1996). For a case study of the Saturn location decision, including a comparison of potential operating costs in several areas, see Bartik et al. (1987).5 Indeed, Anderson and Wassmer (1995) find “evidence of the validity of the oft-told anecdotal story that a firm solicits an incentive offer from one community with negative location characteristics, and then uses this incentive offer to get a similar incentive from the community where it would have located without any incentive at all.”companies, United (UAL) and American, and summarize the bidding process as follows: “(1) Airline companies framed the competition by releasing information about the MOC and its location criteria; (2) cities ... made increasingly higher bids; and (3) airline companies used the bidding tactically to land a MOC site.”The UAL case is instructive. In December 1989, the company asked 20 cities to submit proposals by April 1990. After reviewing the initial offers, with values up to $100 million, UAL shortened the list to 9 cities in December. The bidding activity continued until October 17, 1991, when UAL and representatives of the four remaining cities met in Chicago for final negotiations. The standing offers were approximately $340 million (Louisville), $300 million (Denver), $200 million (Oklahoma City), and $155 million (Indianapolis). The talks went on for a few days, during which Louisville dropped out. UAL put a final pressure on Indianapolis to come up, in 24 hours, with an extra $85 million to close a gap with a competitor’s bid. Even though its site had the best location characteristics for UAL, Indianapolis acquiesced, raising its winning offer to an estimated $290 million (Nunn, Klacik, and Schoedel 1996, Hudnut 1995).How would the firm want to conduct such an ascending auction? Following the modeling strategy adopted in the main body of work on mechanism design, the firm is assumed to have commitment power, allowing it to choose, and credibly commit to, its preferred (optimal) auction format among all possible ones, enabling the bidders to choose their strategies in the absence of uncertainty about the firm’s ex post behavior.6 If the firm handicaps the bid of a high cost region, relative to the bid of a low cost region, by the size of the cost differential between them, the auction yields efficient location outcomes. In contrast, the firm’s optimal auction is biased. As in the UAL example, the firm optimally handicaps a high cost region’s bid by less than its cost disadvantage to encourage ex ante preferred regions to bid more aggressively. This underhandicapping rule makes the auction more competitive, raising the expected recruitment subsidy, but it may produce a locational inefficiency.7 A central government can correct this distortion by taxing the recruitment subsidies of lower-level governments.86 Perhaps the firm’s horizon is infinite and it initiates a location contest every time it builds a new plant. As a result, it has an incentive to build a reputation for “sticking to its announced auction rules.” For a discussion of the commitment issue in the context of auction theory, see McAfee and McMillan (1987).7 For similar reasons, but with a different interpretation, the optimal mechanism is also biased in a standard asymmetric auction (McAfee and McMillan 1987, 1989). For related results in procurement auctions, when the government has preferences for domestic firms, see Branco (1994) and Naegelen and Mougeot (1998).8 U.S. Representative David Minge (Minnesota) introduced a resolution in Congress, in 1999, that contained such a provision. See section 6.3.The model predicts that a location contest is more profitable when a firm’s operating costs differ less across jurisdictions. Indeed, significant reductions in transportation and communication costs in recent decades, making jurisdictions closer substitutes to one another from the point of view of mobile firms, have coincided with an intensification of interregional competition for firms and with an upward trend in the amounts involved (Donahue 1997a,b). There is evidence that some jurisdictions “overbid” for new firms. Winning bids sometimes exceed the actual value of the investment to the community. The paper offers two possible explanations. First, in a median voter model of bid-making, when the benefits from the firm’s location are not uniformly distributed geographically or across residents, the local official is willing to overbid when the decisive voter can shift part of the fiscal cost to others who would benefit less.9 Second, when losing the location contest is costly (relocation of economic activity or political cost), the regions’industrial recruitment efforts can be too aggressive.What are the social benefits and costs of firm-specific interjurisdictional competition? What would be the welfare impact of banning it? Consider two regions eager to attract a firm. If the high cost region would greatly benefit from the firm’s arrival, it could overcome its cost disadvantage by submitting a sufficiently high bid. With a ban, the low cost location is chosen, preventing the realization of potential efficiency gains. The firm may also receive a transfer from the low cost jurisdiction, where it would have located anyway in the absence of incentives. Such public funds paid to firm owners, with no effect on location outcomes, shift distortionary taxes to local taxpayers and raise equity concerns. Location contests can thus be negative-sum games, as most participants to the public debate argue (e.g., Burstein and Rolnick 1995).This suggests an efficiency/equity trade-off in the design of restrictive policies. Whether or not location contests are positive-sum games from society’s standpoint hinges primarily on whether the local benefits from business locations vary sufficiently across locations to outweigh the welfare costs of interjurisdictional bidding.10 It also depends on the extent to which regions do, in fact, “overbid” for new firms, which might occur when the local bid-making process is politically distorted and when firm locations generate negative interjurisdictional spillovers.The paper suggests a way to frame the controversy over bidding wars and contributes to a long-standing debate in public economics about the normative implications of interjurisdictional9 Underbidding may also arise in the median voter model. As a matter of fact, the evidence on whether communities underbid or overbid for business is mixed. See section 8 for results and references.10 Bartik (1991) provides empirical evidence suggesting that the benefits from economic development vary sufficiently across regions to make state and local development policies beneficial for the nation.competition. A first tradition, going back to Tiebout (1956), views competition in the public sector as a force encouraging efficient government performance. Another tradition (e.g., Break 1967, Zodrow and Mieszkowski 1986) emphasizes the detrimental effects of a competition viewed as a “race to the bottom” yielding suboptimal public good levels and more regressive tax structures. Both views are relevant in assessing firm-specific interregional competition. It is sometimes the “(not-so) invisible” hand that channels society’s resources where they are valued the most, and sometimes an instrument of “industrial blackmail.”11Few studies use an auction framework to analyze interjurisdictional competition. Black and Hoyt (1989) suggest that, if new residents reduce the per capita cost of providing public goods and services, jurisdictions will be willing to bid to attract a large firm. Interjurisdictional bidding is efficiency enhancing when information is perfect, but location outcomes can be inefficient when the firm knows privately its location-specific costs. In King, McAfee, and Welling (1993), two regions bid to attract a firm, that can potentially relocate in a second period. The firm’s arrival creates a surplus, unknown ex ante, and the firm uses a sequential auction to maximize its share of the surplus. As in King, McAfee and Welling (1992), they also consider an extension where the regions can, before bidding, invest to enlarge the expected surplus. This paper is more specific about the nature of the surplus, and complements these earlier studies by analyzing the case where the jurisdictions have private information about their willingness to bid. In two models applied to interstate bidding wars, Biglaiser and Mezzetti first analyze an elected official’s willingness to pay for a new project when she has reelection concerns (Biglaiser and Mezzetti 1997) and then study an incentive auction in which multiple principals bid for the exclusive services of an agent (Biglaiser and Mezzetti 2000). In the previous papers, the bidding process is modeled ad hoc as either an English, a second-price, or a first-price auction, whereas this paper formally studies how a firm with commitment power would design it.There are also a few papers in the international economics literature that examine the competition for specific investments12 Haaparanta (1996) suggests a model of menu auctions to study the competition for foreign direct investments by multinational corporations. Doyle and Wijnbergen (1994) provide a model in which a firm obtains an early tax holiday from a host government that fades over time once the sunk cost of moving in the country has been incurred.11 This expression was used by the mayor of Flat Rock, Michigan, where Mazda installed a plant after a bidding war involving several cities and states (Milner and Moore 1989).12 For cases of international contests, see Haaparanta (1996), The Economist (02/01/1997), and Oman (2000).13 Throughout the paper, the terms “jurisdiction,” “region,” and “community” are used interchangeably. 14 There are various possible explanations for these benefits. The firm’s arrival generates positive spillovers,like agglomeration economies (Taylor 1992, Head, Ries and Swenson 1995, 1999), it has a long-lasting impact on local real wages (Bartik 1991, De Bartolome and Spiegel 1995) or property values (Bartik 1991), it creates new jobs for residents of an area with high unemployment (Bartik 1991, Courant 1994), or it creates a local consumer surplus if the firm supplies a good or service consumed locally as in the case of sports teams (Noll and Zimbalist 1997). Jurisdictions will also be willing and able to bid for business if there is an excess capacity of public infrastructure (Oates 1996), if the preexisting taxes exceed benefit taxes (Courant 1994), or if the local public goods are provided with a technology that exhibits increasing returns to population scale (Black and Hoyt 1989). 15 The case of heterogeneous communities is treated in section 8.Bond and Samuelson (1986) suggest a signaling motive. When its productivity level is private information, a country can use tax holidays to signal the high quality of its business climate.The rest of the paper is as follows. The model is presented in section 2. Definitions ofsocial welfare and locational efficiency are provided in section 3. The firm’s optimal auction is derived in section 4, it is implemented with an ascending auction in section 5, and its rules are interpreted in section 6. Positive implications of the model are derived in section 7. The political economy of bidding and the impact of negative spillovers are studied in section 8, and a welfare analysis of a ban on targeted incentives is performed in section 9. The conclusion is in section 10.2. ModelBefore choosing a site for a new indivisible investment, a firm conducts an auctionmechanism in which n jurisdictions participate.13 The auction rules specify the location decision and the monetary transfers to the firm given the jurisdictions’ strategies. These transfers represent the monetary equivalent of the incentive packages offered by the regions. Such packages typically include a mix of recruitment subsidies, public infrastructure investments, tax breaks, loanguarantees, land preparation fees, and worker training programs. The firm’s arrival generates net benefits worth v i to jurisdiction i N = {1,...,n} if it is chosen.14 Local officials, who havereelection concerns, choose bidding strategies consistent with the median voter preferences. With homogeneous communities,15 in which the residents’ benefits from the firm location and the tax prices associated with the incentive package are identical, v i measures the jurisdiction’swillingness to pay to attract the project. Valuations are private information to the regions. From the firm’s and the other jurisdictions’ perspective, they are the realization of independent randomvariables with distributions F i (·), continuous positive densities f i (·), and supports I i = [v i ,v*i ], where 0 v i < v *i < for all i N. Jurisdictions are in general asymmetric , unless F i (·) = F j (·), for all i,j N, in which case they are symmetric . Uncertainty about the valuations captures the firm’s16 Many distributions are regular, e.g., the uniform, the exponential, and the gamma. In the context of auctions,regularity implies that the winner’s expected payment is increasing in his valuation (McAfee and McMillan 1987). 17 One could assume non-zero payoffs if there are interjurisdictional spillovers. When Virginia bids successfully to attract a Motorola plant in Northern Virginia, Maryland can get a negative payoff (e.g, migration of workers) or a benefit (e.g., new contracts for suppliers located in Maryland). The search for the optimal auction becomes a multidimensional mechanism design problem in the context of an auction with externalities (Jéhiel, Moldovanu and Stacchetti 1999). Interjurisdictional spillovers are introduced in section 8, where a simpler approach is taken. 18 Thus, the firm’s tax liabilities under the jurisdiction’s tax code are computed as a cost and the firm-specific tax break is part of the transfer.19 Relaxing this assumption raises a technical issue: When the firm preferences are private information, would it be in its interest to announce an auction mechanism that differs from its preferred auction format in the case where these preferences are common knowledge? Perhaps would it want to announce a mechanism that would not convey any information about its preferences over the regions? This issue is discussed briefly in the conclusion.%i % x i c i .(1)motivation to conduct an auction in the first place. It also captures, among other things, the fact that key elements of a jurisdiction’s economic development plan are not common knowledge. The distributions are assumed to be regular, that is J i (v i ) = v i [1 F i (v i )]/f i (v i ) is increasing in v i .16Finally, the jurisdictions are risk neutral, so region i’s surplus is either v i x i , if it wins and pays x i ,or zero 17 if it loses and pays nothing.The firm is risk neutral and has preferences over the various jurisdictions. If jurisdiction i wins and pays x i , the firm gets a location-specific profit, when only the winner pays, ofThe constant % is a component of the profits that does not depend on location, perhaps because the firm sells its output nationwide. Each c i 0 is a location-specific component of the costs incurred by the firm when operating in region i. Cost differentials are defined as ij = c j c i , for all i,j N, so ij is interpreted as “jurisdiction i’s advantage over jurisdiction j.” They measure the intensity of the ex ante preference ranking of the firm over jurisdictions i and j. Firm costs differ across locations due to differences in regulation, wages, transportation costs, quality and quantity of public infrastructure, access to airports and interstate highways, proximity to research centers and universities, amenities valued by the employees, and differences in tax structures.18 The c i ’s are common knowledge, capturing the fact that the regions are fully aware of the relative quality of the local business climate they offer.19 The following assumptions are made:(A1): 0 = c 1 < c 2 < ··· < c n < ; then ij = c j c i > 0 when i < j, ~ i,j N.(A2): % 0.(A3): Max i N {J i (v i )+% c i } 0.20 Varying weights could also reflect a central government’s regional development objectives. The European Union (EU), for instance, has powers to control aid to industry. “A major objective of the EU in regulatingincentives is to ensure that the member countries have enough resources to devote to education, infrastructure and the needs of potential businesses, communities and regions. The EU’s incentive policy is also used to pursue its overall policy of closing the gap that exists between more prosperous nations and regions and poorer, undeveloped or underdeveloped nations and regions” (Breagy 1996). The EU blocks aid from developed nations, like Germany’s $12.5 million offer to a steel maker in 1995, but contributes funds to supplement offers from poorer regions, like in the case of the joint $510 million package offered in 1995 for AutoEuropa’s plant in Septubal, Portugal.SW i ( ) [v i x i ] #[% x i c i ](2)The first one is made for convenience. The second assumption stipulates that the firm’s project is profitable whether or not it receives fiscal incentives. The last assumption ensures, in proposition 1, that the firm never finds it optimal to choose no location and to dismantle the project.3. Social Welfare and Locational Efficiency3.1. Allocative and Distributive Implications of Location ContestsThe firm location generates a surplus, the sum of the local benefits and the firm’s profits,and location is efficient when this surplus is the largest. The social welfare created is a weighted sum of the payoffs. The weight on the regions’ benefits is normalized to one, and the weight on the firm’s profits is 0 < . If the firm locates in jurisdiction i, the social welfare created isThe social planner (say, the central government) could weigh differently the benefits of different jurisdictions. Attracting the firm in a low income, high unemployment area might be more desirable than a location in a wealthier area. Varying weights would reflect diminishing social marginal utility of income.20 One could also include a parameter representing the welfare cost of public funds, since the transfer to the firm either increases public expenditures (subsidy) or reduces revenues (tax break). This parameter would play a role opposite to the one played by .A value of less than 1 thus reflects a society’s concerns about the distributive impacts of location contests as well as the allocative distortions created by the use of public funds.3.2. Locational EfficiencyA special form of the social welfare function is worth mentioning. When = 1, socialwelfare maximization is synonym with total surplus maximization. This objective is achieved when the location decision of the firm, after a location contest, satisfies the following definition:。