Abstract
Research on local economic development (LED) policy remains limited in its access to detailed information about what policies are practiced where. This study helps to fill important gaps with a unique data set of 106 localities’ responses to Amazon.com's 2017 “HQ2” RFP. These data are analyzed qualitatively to assess patterns in policy practice and their relationship to local characteristics. Highlighting the balance between use of incentives to firms versus broader public investments, we find that most localities embrace a relatively balanced mix of the two in ways that suggest isomorphic decision making. Localities practicing extremely high numbers of development policies specialize in either firm-level incentives or public investments, a choice that appears to be associated with local incomes. All policy mix categories include diverse types of localities, indicating significant room to maneuver. These findings offer new perspectives on the range and sources of local approaches to place-based policy and community development.
Keywords
Introduction
Local economic development (LED) policy has a questionable reputation. While it is often seen as holding important potential to spur long-term prosperity, its practice is frequently criticized due to the widespread use of corporate subsidies. Based on these concerns, research on LED policy highlights the distinction between incentives to individual firms and broader investments in public goods (Nager, Lowe Reed, and Langford 2019; Neumark and Simpson 2015; Osgood, Opp, and Bernotsky 2012). Yet despite a longstanding interest in distinguishing between these two major types of LED policy, research on this topic suffers from important limitations. These include difficulties in gathering reliable, detailed information about the kinds of policies practiced and the degree of commitment across a wide range of cities and urban regions, as well as establishing consistent definitions of policy types for comparative purposes (Storper 2010; Wolman and Spitzley 1996).
Despite these limitations, prior research agrees that incentives to individual firms are a core part of the LED policy toolkit, despite a preponderance of evidence that they garner inferior results to public investments in local economic productivity and specialization (Bartik 2019; Peters and Fisher 2004; Warner and Zheng 2013). Together, these observations present puzzles about how local actors and organizations make decisions about LED policy—puzzles that could be better addressed with more complete and accurate information about what policies are practiced where.
This paper attempts to help clarify these issues by employing a novel data set of 106 (87 full, 19 partial) responses to Amazon’s 2017 Request for Proposals (RFP) for a planned second headquarters. These proposals represent the most complete and thorough documentations of economic development policy undertaken by the localities 1 producing them, addressing many of the limitations of more commonly used survey data. While the data set does not observe shifts since 2017, it offers a comprehensive snapshot of LED policy in the United States prior to the COVID-19 pandemic, cataloguing many policies whose implementation continued for years thereafter and providing a window into ongoing evolutionary processes in local urban policy and governance. The texts of these proposals, along with related local media coverage, are analyzed qualitatively to help clarify how localities balance their use of firm-level incentives versus broader investments in LED, including what is the range of policy “mixes” that localities practice, as well as how variations in policy practice relate to local attributes such as population size, geography, income, and institutions authored the policy documents.
Previous research using smaller samples of the proposals used here reached divergent conclusions regarding whether localities emphasize “incentive” versus “investment” policies (Jensen 2019; Nager, Lowe Reed, and Langford 2019). Our analysis suggests that there are a few “policy mixes” (Agrawal, Hoyt, and Wilson 2022; Bartik 2020; Feldman and Lowe 2018) which characterize the landscape of economic development policy practice—the vast majority occupying a “big middle” that employs a minimal, but balanced mix of incentives and investments, as well as two main types of much more intensive policy approaches. One intensive approach focuses on firm-level incentives, while the other emphasizes public investments. We find that while the choice between these two subtypes of extreme policy adoption is associated with local income levels, there remains significant variation within each mix in terms of population size, geographic region, economic strength, and institutional authorship. Other key findings include evidence of isomorphic (i.e., replicating what appears to be normatively dominant) policy selection, particularly within the “big middle”; a bias toward reporting policies with simple organizational structures; and few, if any significant relationships between institutional authorship and policies reported.
These findings offer suggestions for future research and policy practice regarding the factors driving local policy decision making, as well as the opportunities for greater local adaptation and flexibility. Even if the product of conformity, the “big middle” policy offers a more nuanced view of LED policy selection, suggesting that most localities attempt a “both and” approach to implementing firm-level incentives alongside public investments. And while a large majority of proposals fell into this seemingly isomorphic category—which is unexpected in light of past research finding that economic development policies are heavily influenced by local economic and political conditions (Jensen and Malesky 2018; Lobao, Adua, and Hooks 2014; Reese 2006)—the diverse range of policies and mixes expressed among the exceptional cases suggests multiple pathways into policy differentiation. Our findings also help bring further clarity to the relationship between local economic conditions and LED policy commitments by showing how the elevated levels of policy practice previously observed at the extremes of the local income distribution differ in approach.
We continue by examining what is at stake in local commitments to incentive versus investment policies, as well as some longstanding limitations in our ability to accurately gather information about the state of LED policy practice. We then describe how the data used in this paper can help fill some of these informational gaps as well as the methods employed in this study. These are followed by analyses of the types of policies practiced, their prevalence, varieties of LED policy “mix,” and their relations to local characteristics. Finally, we discuss the implications of our findings in light of contemporary trends in urban economic development.
LED Policy: Known Impacts and Puzzling Choices
The era of “entrepreneurial” inter-local competition for economic growth and development has now persisted for roughly forty years (Eisinger 1988; Weaver 2016). Since this era began in the 1980s, LED policy has seen multiple “waves” of popular themes, from attracting outside businesses to investing in local community assets, quality of life, small firms, and/or workforce development (Bradshaw and Blakely 1999; Qian and Acs 2023). In a context of manufacturing decline and increasing inter-regional economic inequality (Kemeny and Storper 2024; Moretti 2012), current LED policy discussions focus on how these policies can address both inter- and intra-regional inequalities of economic opportunity (Austin, Glaeser, and Summers 2018; Rodríguez-Pose 2018), while also foregrounding the centrality of LED policy in processes of adaptation to both technological and climate change (Feldman and Kogler 2010; Hanson, Rodrik, and Sandhu 2025; Hughes and Hoffmann 2020).
This study aims to enhance our understanding of LED policy practices through a qualitative analysis of comprehensive economic development policy documents from over 100 US and Canadian cities. We address the following questions:
How do localities balance between policies that specifically benefit individual firms (“incentives”) and policies more broadly aimed at enhancing local economic productivity and specialization (“investments”)? What are the most common LED policies, both individually and in terms of composite policy “mixes”? Are variations in commitment to policies and policy mixes associated with local characteristics such as population size, geography, or household incomes? Do different types of institutional authors impact the LED policies described?
These questions are explored using 2017 data, providing a comprehensive pre-pandemic snapshot of urban economic development strategies. While predating many current phenomena, many of the observed policies are multi-year in nature, and research suggests that LED approaches show stability through macroeconomic crises (Hinkley and Weber 2021; Lobao, Adua, and Hooks 2014). As a result, the analyses in this study offer a window into an ongoing evolution of urban governance, inter-city competition, and public–private coordination that can be usefully integrated and compared with post-pandemic analyses.
Policy Impacts: What “Works” in LED
The use of direct cash incentives and subsidies for individual firms to secure local employment has long been a cause for concern. Estimates suggest that approximately $50–100 billion are spent annually by state and local governments in the United States on business incentives (Bartik 2020; Mitchell et al. 2019), a sum that was recently given a major boost through federal place-based investments (Hanson, Rodrik, and Sandhu 2025; Lincicome, Joffe, and Chanwong 2024). While research shows that firm-level subsidies can generate significant long-term increases in employment (Freedman 2017; Greenstone, Hornbeck, and Moretti 2010), the downsides and opportunity costs of these firm-level subsidies are considerable. Scholars consistently find that incentives’ impacts are markedly inferior to broader investments in local economic capacity and job quality (Bartik 2020, 2022; Benner and Pastor 2015; Rogerson 1999), with incentives costing multiple times more per job created than investments in workforce training and customized business services. When localities choose to give tax dollars to firms, they are implicitly choosing to forego potential public investments, and empirical evidence indicates that there can be a direct negative impact from firm subsidies on public investments in areas such as education (Dye and Merriman 2000; Malizia et al. 2021; Weber 2003). Research also finds that many incentives subsidize decisions that businesses would have made regardless (Bartik 2019; Mast 2020), and that their use can trigger “race to the bottom” dynamics that further ratchet up the dollars spent per job created (Mast 2020; Zheng and Warner 2010).
The persistence of firm-level incentives in spite of their inefficiency is often attributed to the dynamics of electoral politics and/or locally specific path dependency (Jensen and Malesky 2018; Lobao, Adua, and Hooks 2014; Reese 2006). However, these explanations remain partial, given that place-based investments also offer potential political benefits and mayors express a great deal of ambivalence toward incentives (Levine Einstein et al. 2018).
Based on these previous findings, we expect incentives to be a common, even a default practice among the localities analyzed. But we aim to build on previous research by leveraging this study's data for a more complete picture of the practice of incentives as well as alternative LED policies, both within and across localities.
Filling in Gaps: Surveying Local Policy Practice
Previous research establishes incentives’ popularity, especially during macroeconomic downturns (Zheng and Warner 2010; Warner and Zheng 2013). However, our understanding of how incentives are deployed in balance with alternative policies is limited due to inconsistent definitions of types of policies practiced and reliance on survey data with known shortcomings.
One can imagine different forces driving convergence or variation in policy practice, whether universal competitive pressure for growth (Logan and Molotch 1987; Peterson 1981), or distinct local profiles of industries, economic growth trends, and regional competitors. Many studies suggest that deteriorating local economic conditions motivate firm-level subsidies to quickly attract employment growth (Betz et al. 2012; Osgood, Opp, and Bernotsky 2012; Rubin and Rubin 1987; Slattery and Zidar 2020). Conversely, a locality's size, resources, and affluence have also been found to be positively related to LED policy use (Basolo 2000; Lobao, Adua, and Hooks 2014; Pagano and Bowman 1995). These findings imply a “U”-shaped curve of LED policy adoption relative to local economic strength. However, this body of previous research varies in what is counted as an “economic development policy,” sometimes distinguishing between firm-specific incentives and public investments, sometimes not, while also including varying types of public investments in their analyses.
To better address these inconsistencies, we organize our analysis around a distinction between “incentives” that directly subsidize individual firms and broader “investments” in local economic capacity and specialization. This distinction has been recognized as important elsewhere, albeit sometimes under different labels (Nager, Lowe Reed, and Langford 2019; Neumark and Simpson 2015; Osgood, Opp, and Bernotsky 2012). We also focus on policy “mixes,” an important level of analysis which can help to better capture how localities pursue economic prosperity through a variety of simultaneous policy commitments (Agrawal, Hoyt, and Wilson 2022; Bartik 2020; Feldman and Lowe 2018; Markusen and Schrock 2006).
Previous studies using smaller samples of the documents analyzed here reached contradictory conclusions regarding whether the center of gravity in LED policy practice lies with incentives versus investments (Jensen 2019; Nager, Lowe Reed, and Langford 2019). Our qualitative analysis of a larger sample aims to clarify whether either approach dominates, as well as to identify whether any distinct policy mixes can be discerned. We expect both lower-income and higher-income localities to both deploy more LED policies than their peers, while exploring potential differences in approaches at the extremes. We also seek to better understand the patterns and range of LED practices at the middle of the distribution of incomes and policy intensity, an analytic “gray zone” that has tended to be overshadowed in quantitative analyses of predictors of more intensive policy use (Goertz 2008).
We engage in these questions using comprehensive economic development profiles produced by teams of local public- and private-sector economic development professionals. These profiles offer in-depth information on a wide range of LED policies, including levels of commitment and implementation progress. We complement these LED profile documents with concurrent local media reports on the documents, their authors, and the local discussions surrounding them. These data are used to assess patterns of local policy practice in medium and large US localities. 2 We focus on the popularity of individual policies, policy mixes, degrees of local policy commitment, and variations in policy use based on local characteristics and the documents’ institutional authorship. The data used here help to address another important limitation in previous research, which frequently relies on surveys with low response rates, dichotomous policy measures, and low reliability in question response within and across respondents (Farley 2020; Lamothe, Lamothe, and Bell 2018; Wolman and Spitzley 1996).
The findings from this pre-pandemic dataset provide a useful baseline for understanding how localities’ economic development strategies continue to evolve in response to recent challenges and opportunities, including the shifting geography of work and residency locations, technological change, and changes in federal policies related to regional investment and international trade. The analysis of this data as a recent baseline is intended to support the efforts of urban policymakers and researchers seeking to craft effective, equitable economic development strategies in a rapidly shifting urban landscape.
Data and Methods
This study examines LED policy practices in the United States, centering its inquiry on policy prevalence, local policy “mixes,” and variation among localities. We utilize data that are more comprehensive and reliable than more commonly used survey data by analyzing every available proposal submitted for Amazon.com, Inc.'s 2017–2018 “HQ2” headquarters site selection competition.
These documents provide unprecedented breadth and depth of information on LED policies. The nearly eight-page request for proposals required detailed narratives on business environment, incentives, workforce skills, education institutions and resources, infrastructure, diversity, and quality of life (Amazon 2017). Many observers noted the proposals’ exceptional comprehensiveness (Bond 2018; Nager, Lowe Reed, and Langford 2019, 4; Starner and Jones-Kelley 2017), with some calling them “the largest database of political and civic intelligence at the local level in the world (Parilla 2018).” Amazon's stated criteria were in keeping with a growing recognition that Fortune 500 firms have become increasingly preoccupied with headquarters locations with combinations of cost of living, housing availability, and services and amenities such as strong public education systems and accessible green space that will allow them to competitively attract scarce high-skilled talent (Booker and Kargbo 2017; Sisson 2024).
Local officials described the proposal authorship processes to local media as intensive, team-based efforts, 3 often comprising the first time a wide range of institutionally scattered and/or previously unwritten information was compiled together in one document. Local officials also frequently stated that these documents would become their de facto economic development profiles to be shared with future potential investors.
While these features suggest high data quality, there are still potential sources of bias. Localities could have emphasized policies that they expected Amazon to care more about, such as business incentives. However, the company's broad criteria explicitly encouraged a comprehensive approach, which is corroborated by the company's ultimate selection of localities that focused on the full array of criteria (i.e., Northern Virginia, New York) over neighboring ones that focused on maximizing their incentive offers (e.g., Baltimore, Newark). Furthermore, the data's inclusion of a diverse geographic and demographic range of localities, as well as an analytic focus on both the extremes and the middle of the distribution of policy commitments, mitigate against these and other potential limitations.
Amazon originally claimed that the winner of its HQ2 competition could result in “as many as” 50,000 jobs and $5 billion of investment over a ten-to-fifteen-year period. Although some were skeptical of the plausibility of these claims (DePillis 2018; Elliott 2018; Leroy 2017), their unprecedented scale could also have had a distorting effect on the information localities included in their proposals. However, as described above, localities were explicitly given specific and lengthy requirements for information of unprecedented breadth and depth about economic development policies. If these criteria elicited anything different from cities in terms of what information they produced, it would be in the direction of increased completeness regarding economic development policies and investments.
Furthermore, this bidding contest, while noteworthy for its scale and the public discussion it elicited, had precedent through a long line of incentive “megadeals”; according to the corporate subsidy watchdog group Good Jobs First, from 2000 to 2024 there have been 34 corporate incentive packages in the United States with a value at the time of agreement of at least $1 billion, with the largest recorded being $8.7 billion provided to Boeing in 2013 in the state of Washington. Northern Virginia's ultimately successful bid in the Amazon.com “HQ2” competition ranked 45th in the most recent (2024) list-based dollar value of incentives provided to an employer, and at $20,000 per job was well below this category's average of $658,000 per job, as well as an overall average across all local incentive packages of $196,000 per job (Bartik 2020, 17, 21; goodjobsfirst.org 2025; Leroy 2017).
Two main reasons explain why smaller localities would bother to submit detailed proposals to a competition whose selection clearly indicated large population centers. First, many of these smaller locales emphasized their close proximity to such population centers. As the table in Appendix shows, several dozen of the proposals collected for this project were smaller communities located within or immediately adjacent to metropolitan areas with a population size at the scale Amazon was seeking (1 million or more). Smaller neighboring communities often presented themselves as part of a larger metropolitan area with distinct advantages such as more open developable space, greater regulatory flexibility, and/or enhanced quality of life.
Second, many localities who recognized that they were unlikely to be selected saw their proposals as functioning within longer timelines than the HQ2 competition—possibly for future Amazon investments down the road in logistics and/or special project offices—as well as for a broader audience, such as for other large corporations, especially technology firms, including suppliers or vendors to Amazon, seeking new sites in the future. Amazon explicitly encouraged this approach to proposal submission as a bid for future attention beyond the HQ2 competition in statements such as their announcement of the HQ2 competition's 20 finalists, in which the company stated: “Through this process we learned about many new communities across North America that we will consider as locations for future infrastructure investment and job creation (Forbes 2018).”
We collected proposals in response to the RFP (“proposals”) from public archives (e.g., MuckRock.com, Github.io) as well as through individual searches on Google.com for all US metropolitan statistical areas. 4 The individual searches were also used to collect relevant local news reports, which often provided supporting information about a locality's proposal, as well as information about submissions from nearby localities. Proposal documents were being released as late as 2020, and searches continued afterwards for relevant news coverage.
Of 238 submissions reported by Amazon (Shoemaker 2018), 5 we confirmed the existence of 181 proposals and retrieved 106 with at least partial text, including 87 full-text proposals. 6 The sample includes 35 of the top 50 most populous US cities 7 and 44 of 53 MSAs with over one million residents. 8 While localities in the Northeast region of the United States are somewhat overrepresented (38 percent of proposals), the sample contains relatively even observations from other regions of the United States at a wide range of population sizes. 9 Figure 1 places the locations of all partial and full proposals on a map to illustrate their geographic distribution, and Table 1 shows the count values of proposals by region.
Map of Locations of all Proposals Collected.
Proposals (Partial and Full Original Text) by Region (%).
Proposals Gathered.
To analyze the proposals, we employed a version of Large-N Qualitative Analysis (LNQA) (Goertz 2017; Goertz and Haggard 2023). This involves examining a large percentage of relevant cases to test one or more research propositions. Our analysis began with a descriptive “what” question (Gerring 2012)—“What economic development policies do localities practice?” and progressed to identifying key patterns and categories. We also leverage the in-depth contextual information in the proposals to consider the degree of local commitment to policies. Local media reports help to supplement the proposals with context about the proposal authors, timelines, and local discussions about the proposal documents. Finally, we evaluate potential patterns of association between kinds and intensities of policy practice on the one hand and local characteristics, such as population size, region, economic strength, and the institutional entities authoring the proposals, on the other. Where relevant, we employ descriptive statistics to help answer questions about possible relationships between a locality's characteristics and its LED policy practices.
The first stage of analysis involved coding proposals for the presence of policies in key thematic areas established in the literature (e.g. Bartik 2019; Betz et al. 2012; Warner and Zheng 2013): cash/tax incentives, workforce development and education, business support, and infrastructure. The definition and scope of each category is discussed in the sections below. All proposals were coded qualitatively in a separate document. These documents included notes regarding policies named with page references to the source document, along with key contextual information regarding policy design, resources, actors, timelines, and extent of implementation thus far.
Consistent with much of research on LED policy, the presence and type of policies are the main outcome of interest (Agrawal, Hoyt, and Wilson 2022; Betz et al. 2012, 369; Wolman and Spitzley 1996). This focus on the number of concrete policy actions is for several reasons: first, it facilitates comparison across diverse localities by focusing on analogous policy commitment decisions that might take place at different scales or with different modalities based on the local context. Second, it follows directly from the main research question regarding local policy priorities by focusing the analysis on what localities choose to include versus omit in their policy commitments. Third, it reduces potential biases from inconsistent reporting or overclaiming. Only policies with clear evidence of committed resources and ongoing implementation processes were included. This approach, as opposed to using dollar amounts or predicted policy impacts, minimizes distortions from reporting inconsistencies and lack of consensus on metrics, while highlighting the comparative landscape of local policy decisions.
Maintaining clear standards of evidence to only include concrete policies also helps to reduce potential biases from localities overclaiming their policies practiced. Only policies described with clear evidence of resources committed and a process underway were counted as such—“aspirational” policy ideas without such evidence were noted but not included in the counts and analyses of actual policies discussed below. We chose discrete policies as opposed to other possible outcome measures (e.g., dollar amounts) to both reduce distortions from reporting inconsistencies as well as to focus on the challenges of policy design and implementation that are common across many types of localities.
Key comparative categories such as types of LED policy mix were developed using what is often referred to as “abduction” or “double fitting (Ragin and Amoroso 2019; Timmermans and Tavory 2012),” wherein each case's analytic value is maximized to reveal, modify, or disconfirm patterns identified thus far. We used spreadsheets populated with content from all of the proposals’ coded documents to compare the relevant features of localities’ policies, policy mixes, evidence of commitment, and local structural characteristics. Particular attention was paid to patterns both at extreme values and within “gray zones,” the latter of which, due to their not falling near the extremes of a distribution, can have distinct patterns that may be “washed out” of statistical analysis whose calculations may be driven more by values at extreme points of the distribution (Goertz 2008, 107). In other words, we looked for patterns and relationships at all parts of the distribution of values, to make sure we were not letting conclusions about the whole distribution to be driven disproportionately by one or another part and give due attention to distinct policy “mixes” at different parts of the distribution of policy adoption.
Ultimately, only extreme values at the high end of policy adoption were found to merit a separate policy mix category, as those at the low end tended to resemble the policy mix at the middle of the distribution (the “big middle”). Extreme values at the high end were identified based on clear cutoff points in the histograms for number of policies within the main policy categories. This meant that a given locality could qualify as extreme in more than one policy area. These histograms are included in policy category sections below.
LED Policies: Patterns and Variation
The first analysis responds to the initial question of the balance and types of practice within and between the basic “incentive” versus “investment” LED policy categories. The key categories are cash and tax incentives, workforce development and education, business support, and infrastructure/built environment. Issues that arise include different kinds of policy “packages” within a policy category, an apparent narrative bias among proposals toward organizationally “simple” policies, and the room for variation in who adopts certain types and sets of policies.
Cash and Tax Incentives
Five Most Frequent Cash/Tax Incentive Policies.
Distribution of Cash/Tax Incentive Policies Per Proposal.
Fourteen percent (12/87) of proposals contained no incentive policies of any kind. Although some of the places with no incentives were smaller and likely had less-than-average administrative capacity, some, such as Boston, MA and the San Francisco Bay Area, CA, were quite large and had among the most robust arrays of workforce and infrastructure policies. These observations agree with established literature that incentives are highly prevalent, while showing that different kinds of localities may deliberately avoid this type of policy.
Two patterns stand out in the cash/tax incentive policy category. First, cities and states located near places known to be of particular interest to Amazon tended to post some of the biggest incentives in terms of dollar amounts. For example, the states of New Jersey and Maryland, likely aware of Amazon's interest in New York and Washington, DC (Florida 2018; Parilla 2018), both offered among the very largest tax incentive packages ($5 billion and $6.5 billion, respectively). Localities within these states also often sought ways to build on top of these state-level offers.
The other noteworthy pattern was the intensity and breadth of tax breaks offered by Midwestern US cities. Along with localities near New York and Washington, DC, those from the Midwest comprise nearly all of the multi-billion-dollar incentive packages that were offered. Chicago's partially released proposal illustrates the often extreme scale of the incentives more prevalent in the Midwest (30-year TIF and an estimated $1.32 billion in payroll tax credits), while Cleveland's illustrates the scope (TIF, sales tax increment, payroll tax reimbursement, local wage tax credit, free land on the Cleveland State University campus, and a cash grant from the state workforce board).
This does not mean, however, that all proposals from the Midwest had abnormally large cash and tax incentive policies. 12 Nor does it mean that proposals from this region neglected investment-oriented policies. 13 This significant room for local variation aligns with other findings in this study regarding considerable room, even when geographic and economic conditions are similar, for locally unique policies and policy mixes.
Workforce Development and Education Policies
Five Most Frequent Workforce Development/Education Policies.
Distribution of Workforce/Education Policies Per Proposal.
Localities with the greatest commitment to workforce development and education policies described a wide range of programs, usually at all education levels, with specific curricular efforts for formal education and degrees as well as complementary support programs (e.g., mentorship, coaching, teacher training, Summer youth programs). Some (e.g., Northern Virginia; Detroit, MI; Hamilton, Ontario (Canada); Washington, DC) attempted to work backwards from target numbers for workers or graduates in specific fields or occupations that their policies were intended to produce. Others, such as Boston, MA, New York, NY, and Providence, RI, described policies roughly as in-depth and comprehensive, but without specific workforce targets. One commonality across nearly all of these was a high degree of coordination between the state and local levels, with state governments committing to significant expenditures and technical support (e.g., standard-setting along with support to meet state standards).
The distinct level and scope of commitment to workforce education and skills in these proposals emerged quite clearly in the data set. For example, while Northern Virginia's proposal was widely discussed in the media as focused on a commitment to build a new Virginia Tech satellite campus, the Virginia Tech component was actually a facet of a much more elaborate STEM workforce plan. The plan specified STEM workforce targets, with state commitments to fund new higher education degree programs in more universities, new job internships and work-based learning, instructor hiring and training, youth mentoring and Summer camps, and educational partnerships with government agencies. Overall, these proposals with highly robust workforce and education commitments indicate an important and potentially critical role for multi-level coordination in the production of investment-oriented LED policies, something that merits further attention in terms of the political and organizational resources required to support it.
Another noteworthy pattern across nearly all proposals, especially in light of the aforementioned, was that policies tended to be described in organizationally simple terms—i.e., almost always described as having one key theme and being the initiative of one or two institutions. Programs that were more organizationally complex were rare and may have been under-reported. For example, we were surprised to see less than a handful of “workforce intermediaries” (WI) mentioned, given that these number somewhere between the hundreds to a thousand nationwide (Giloth 2022, 8; Marschall 2021, 12–13), and that they frequently train workers in the kinds of advanced IT skills and occupations for which Amazon had job postings listed in Northern Virginia (Amazon.jobs 2023; Conway and Giloth 2014; National Fund for Workforce Solutions 2017).
By cross-referencing to recent reports tracking workforce intermediaries nationwide, nine instances could be identified of localities with full proposals in the data set which also had active WIs. 14 Of these, just two (Washington, DC and New York, NY) explicitly mention and describe a WI program. These omissions raise questions about why localities would fail to mention these programs. Although there could be different reasons, it seems striking that these proposals which failed to mention existing WIs—which operate at the nexus of often complex local organizational networks related to training, employment, and social services—contained a great deal of information about organizationally simpler policies such as traditional courses of education in K-12 and higher education. One possibility this raises is that there may be issues regarding the “legibility” of more institutionally complex policies, whether as a matter of what registers as legible to local actors, or what they anticipate will register most with outsiders. At a minimum, the pattern of omission appears to merit further inquiry, as it raises questions regarding how local institutions document and represent their governance activities and capacities (Foster and Barnes 2012; Hamilton, Miller, and Paytas 2004; Hughes and Hoffmann 2020).
Business Support
Five Most Frequent Business Support Policies.
Business support policies showed less variation than other categories, making it difficult to single out an obvious set of extreme cases in this category (See Figure 4 for the distribution of the number of proposals (y-axis) that showed different quantities of business support policies per proposal (x-axis)). As with workforce intermediaries, this is somewhat surprising in light of research which identifies custom business services as a high-yield intervention (see e.g. Bartik 2019, 55, 104‒5). This may also relate to previously mentioned observations regarding potential reporting biases in favor of more organizationally “simple” policies. In the business support category, localities frequently focused on commitments of local public employees as a “concierge” service, or to provide “expedited” processes to reduce permitting timelines, rather than discuss the kinds of technical and organizational capacity-building among multiple partners with complementary forms of specialized skills and knowledge that researchers describe as at the core of these more “complex” policies’ contributions to economic development (Bartik 2019).
Distribution of Business Support Policies Per Proposal.
Infrastructure and Built Environment
Five Most Frequent Infrastructure/Built Environment Policies.
Distribution of Infrastructure/Built Environment Policies Per Proposal.
Although the extreme cases in this policy area tended to be larger population centers, 15 the overall correlation between the number of infrastructure policies and the local population was relatively weak (R2 = .17). Extreme cases stood apart in being more likely to describe a broad transportation or transit funding package involving multiple interventions, 16 and moreover, in describing significant commitments to local climate/energy-related policies, along with usually at least one other policy type—often a publicly owned utility and/or an intervention in housing supply or affordability. Proposals at the middle of the distribution of the number of infrastructure policies described interventions in fewer modes of transportation and were less likely to name climate-related policy interventions.
Extreme infrastructure policy cases were not necessarily the wealthiest or most prosperous cities. Within all parts of the distribution, in fact, one observes cities with varying levels of household income in a wide range of geographic regions. These add further evidence that localities have a fair amount of room to maneuver in breadth and intensity of the policies they endorse—not only the largest and richest places go heaviest in their investment policies. Given the capital intensity of infrastructure projects, it could be that aspects of political coalitions in a given locality, and perhaps their connections to state-level actors and organizations, are more consequential in this policy area. Future research might pursue this avenue with more in-depth observation of networks of local officials and their relations to higher levels of government.
Policy Mixes: The “Big Middle” and The Extremes
Our analysis suggests a first, fundamental decision in LED policy mix: to specialize in a highly distinctive set of policies or not. The specialists are the fifteen extreme cases; the remainder, over 80 percent of total proposals (72/87), show a fair degree of coherence as a group, and are referred to as the “big middle.” Big middle proposals originate from cities and metropolitan areas in all regions and of nearly all sizes. While they vary in the specific policies they select, they tend to adopt a balanced mix of incentive and investment policies, generally employing the following criteria:
Two to four tax breaks and cash incentives, usually with at least one property tax break, but also possibly job creation tax credits or special sales tax incentives, plus possibly free and/or subsidized land; One or two main workforce development and education interventions, with an emphasis on custom training for firms via community colleges and extra K–12 STEM and/or computer science curricular offerings; One or two business support policies, most frequently some effort to assure expedited permitting and/or some dedication of public personnel to attend to a firm's needs; Roughly three to five infrastructure-related policies, most commonly new investments in mass transit, sometimes also investments in pedestrian and/or bicycle infrastructure, and possibly some investment in commercial infrastructure (e.g., airport, seaport, railroad, or intermodal).
In light of previous literature, this common decision to place a fairly equal weight on what are usually treated as rival approaches to policy is in itself somewhat surprising. Given the broad conformity to this approach regardless of diverse local conditions, this “big middle” policy mix is also suggestive of isomorphic decision making. That is to say, the policy selection in most places appears to be more a function of legitimacy in the eyes of outsiders as opposed to rational selection to maximize local performance (Andrews, Pritchett, and Woolcock 2013; DiMaggio and Powell 1983). In the neoinstitutional literature, some of the primary factors driving isomorphic decision making include pressures to signal legitimacy, as well as an environment of uncertainty. These factors are prominent in the field of LED policy, whether through the imperative to present a good “business climate” to outside investors, or the high degree of uncertainty in local policy implementation and impact measurement (e.g. Aberbach and Christensen 2014; Green Leigh and Blakely 2017; Kingdon 1995; Molotch 1976). The possibility that LED policymaking has powerful isomorphic drivers can be inferred from research more broadly on policy diffusion (e.g. Krenjova and Raudla 2018; Shipan and Volden 2008). Other studies find evidence of more localized diffusion among proximal cities (Brueckner and Saavedra 2001; Mast 2020). But the factors that would drive stability and change in the tendencies observed here, which appear operate at a national level—such as how policy mixes come to be normalized, and the conditions under which innovation takes place in spite of strong pressures to conform—have been less-discussed in research on LED policy and local governance.
Extreme cases broke away from the big middle both quantitatively, in terms of the sheer number of policies they endorsed, and qualitatively, in terms of how they tended to show strong commitments to specific policy categories with a raft of distinct but related policy efforts. Table 7 illustrates some of the major tendencies across the entire data set versus the “big middle,” proposals with extreme values, and key subsets among the proposals with extreme values in terms of policies practiced by policy type, number of large population centers, and average median household income. The fifteen extreme cases are shown with their policy counts and the categories in which they qualified as extreme in Table 8.
Policy Counts (Mean/Min/Max) and Demographic Features of Different Policy Mixes.a
Population and income data exclude Canadian localities and Expo Park, TX, USA.
Atlanta, GA registered extreme values on both incentive and investment policies.
Proposals With Extreme Values.
Note. Categories with extreme values shaded.
How do the extreme cases differ from the big middle? Although the extreme proposals are more likely to be large population centers than the overall sample, they are not exclusively so—of the fifteen extreme cases, five had a population under one million. 17 Also striking is a clear division among extreme cases—virtually none are extreme in both cash/tax incentive policies and any “investment”-oriented policy categories (workforce/education or infrastructure). 18 Associated with this division among extreme cases in incentive versus investment policies is the local median household income—localities with an extreme number of incentive policies have median household incomes almost $22,000 less on average than those with an extreme number of investment policies, and over $14,000 less than the overall sample. This can help to clarify some of the previous observations in the literature regarding higher policy adoption at the extreme ends of local incomes. One possibility raised here is that the strongest relationship between incentives and local economic hardship is largely at the far low end of the income distribution among localities, and that the two ends of the income spectrum, while both deploying policies more intensively than the middle, diverge in the kinds of policy they are practicing with heightened intensity.
One way to interpret the apparent “either/or” decision between extreme levels of incentive versus investment policies is through the lens of specialization. As with individuals and firms, specialization is costly, and it is plausible that cash/tax incentives versus long-term investments in public goods require their own distinct commitments in time, learning, and coordination. The possibility the different capacities required align roughly according to the incentive/investment policy distinction is supported by the fact that there were several extreme cases in both workforce/education and infrastructure/built environment policies. Local processes of “placemaking” and achieving “distinctiveness” have been observed elsewhere in the literature (Markusen and Schrock 2006; McCann 2002), and the possibility that incentives and investments represent distinct axes of distinctiveness merits further investigation. Meanwhile, these specialization options are situated in contrast to a much more common default, in which localities hedge between incentive and investment policies by using a “big middle” policy mix that demonstrates a solicitousness to business without overextending local capacity.
Proposal Authorship
We find little evidence of a strong relationship between institutional authorship and a proposal's policy content. Considering the box and whisker plots included in Figure 6, we see that while proposals with just private-sector authors might have somewhat lower policy averages, and proposals with state and local-level authors might have higher ones, the quartile ranges have significant overlaps across nearly all of the categories of policy and authorship. Given that state governments are likely to seek partnerships with the most capable, resource-endowed metropolitan areas to begin with, it is also difficult to isolate the effect of local-state authorship. Moreover, the extreme policy cases were evenly divided across six different categories of institutional authorship.
Policy Distributions (Quartiles, Medians, Outliers) by Policy Type and Institutional Authorship Row). 19
The relatively weak relationship between institutional authorship and policies endorsed suggests that we look elsewhere for the sources of variation. In the discussion below we will integrate this with other observations from the study to consider some more specific suggestions for future research.
Toward a Better Understanding of Local Policy Mixes and Their Sources
Utilizing a novel sample of in-depth LED policy documents, we have examined the patterns of policy practice across medium and large cities and metropolitan regions primarily in the United States with an emphasis on balances between the use of incentives to individual firms versus broader public investments. We have also considered possible relations between the policies practiced and a locality's population size, region, economic performance, and policy document authorship. Key findings include three primary types of policy mix—a predominant “big middle” as well as an “extreme” category that bifurcates into two sub-categories, “incentive specialists” and “investment specialists”—whose content and incidence offer several insights for policymakers and researchers of urban economic development and governance.
The big middle describes over 80% of proposals, which exhibit a largely balanced approach between investment and incentive policies. This challenges assumptions that these two policy modes are competing substitutes and that local elected officials are “locked in” to incentive policies (Dye and Merriman 2000; Jensen 2019; Jensen and Malesky 2018; Malizia et al. 2021; Weber 2003). This is not to say that adopting firm-level incentives does not imply foregoing some possible public investments, but rather, that the most common practices evince a basic logic of “both and” as opposed what has commonly been interpreted as “either/or” thinking regarding these main policy categories.
Moreover, the similarity of “big middle” policy mixes across a wide range of local conditions indicates elements of isomorphic decision making (Andrews, Pritchett, and Woolcock 2013; DiMaggio and Powell 1983)—that is, many localities may adopt what appears to be a normatively “middle of the road,” balanced mix of LED policies in order to signal legitimacy to outside investors and minimize the risks of policy uncertainty and failure. This would suggest that path dependence in LED policy may be less a matter of local characteristics and more a matter of larger-scale organizational fields, such as the metropolitan regions of a country (Lobao, Adua, and Hooks 2014; Reese 2006). It also poses questions for future research regarding the roles of legitimacy and uncertainty in establishing the normatively dominant policy mix, as well as in local actors’ decisions to either conform or swim against the current.
These issues also relate to this study's observations about the predominance of organizationally “simple” (one or two responsible agencies, one main policy function) policies in the proposals, along with the evidence that more organizationally complex policies, such as workforce intermediaries, are being under-reported. Future research might consider whether such under-reporting is because local actors themselves are not aware of more organizationally complex local policies, or instead because they anticipate what they view as the preferences of their intended audience.
Among localities adopting extremely high numbers of policies, we identified a clear distinction between “incentive specialists” and “investment specialists.” Relative to the overall sample of full proposals, those specializing in incentives had significantly lower-than-average median household incomes. Those specializing in investments were higher than the overall average, albeit by a smaller margin. This corroborates previous findings that firm-level incentives are more likely to be utilized in localities experiencing worse-than-average or deteriorating economic conditions (Betz et al. 2012; Warner and Zheng 2013). It also helps to clarify what seems to be a divergence in types or “styles” of policy intensity at the extremes of the distribution of local economic strength. Although population size (and by extension, organizational capacity) does seem to favor movement into a policy extreme, it is not a necessary condition.
Perhaps the policy mix that most defies simple explanation is the “investment specialist” category. While it is relatively easy to understand why most localities would default to the “big middle,” as well as why struggling ones would seek to increase their tax base with business incentives, there isn’t as obvious a common element among the localities that highly emphasized public investments in workforce/education and/or infrastructure/built environment. On first glance, it appears to include a mixture of affluent coastal metropolitan areas (San Francisco, CA; Boston, MA) and fast-growing sunbelt areas (Nashville, TN; Tulsa, OK). However, it also includes several places that fall outside of these categories. It furthermore raises questions about why, for example, Nashville and Tulsa seemed to distinguish themselves more than their sunbelt peers in workforce and education policies, or why Vancouver and Hamilton, Canada, seemed to push farther in infrastructure investment policies (and Hamilton in workforce and education) than their Canadian peers. These observations suggest that there are multiple pathways to extreme levels of public investment, a topic that merits further investigation.
Contrary to expectations, we found no strong patterned relationship between institutional authorship and policy document content. While the pairing of state- and local-level actors does seem to be associated with an increase in the policies endorsed, this is likely in large part due to self-selection (state governments seek the strongest localities for their partners). Furthermore, many of the strongest proposals in policy adoption and commitment had single institutional authors of various kinds. This suggests that while there may be advantages to including more types of actors in the LED policy discussion, the premise that a specific array of organizational actors in formal authorship roles is needed to enhance information-gathering and -processing may be overstated (Armstrong 2021; Benner and Pastor 2015; Muro, Parilla, and Ross 2023).
Overall, the observations in this study suggest high conformity to balanced policy mixes with a considerable range of types of localities embracing much more intensive and systematic approaches to LED. Over the course of this study, we encountered anecdotal evidence suggesting that key actors within LED organizations may merit further examination as influential agents shaping local policy conformity vs. deviance. Future research might consider LED managers’ prior training and career experience, as well as their connections to other local organizations. For example, the widely discussed case of Northern Virginia's strong program for talent investment was the initiative of a new regional economic development director with PhD-level training in economics who contracted the services of labor economist Enrico Moretti to help design the program (Mullins 2019). Research on related areas, such as prioritizing equity in planning for local resilience, further corroborates this channel of contingency in the experience and educational background of actors who hold key local institutional positions (Cowell and Cousins 2022). Further investigating how key actors’ backgrounds inform processes of local information aggregation and decision making might help to improve our understanding of the sources of local policy variation.
While this study helps to address several limitations of previous research, it also contains some limitations of its own. For example, the data, while comprehensive, cover a specific moment in time. Furthermore, the policy documents analyzed might have contained some reporting biases given common perceptions of Amazon.com and the nature of its contest. We have discussed factors that mitigate the reporting bias and illustrate the comprehensive nature of these documents, as well as why the data can be expected to be more reliable than most surveys. Nevertheless, potential limitations cannot be entirely dismissed.
Future research can usefully connect the data in this study to ongoing urban governance and policy processes. For example, how do cities’ prior policy mixes equip them to respond to ongoing shifts in work and commuting patterns, especially with respect to downtown business districts? Likewise for suburban cities seeking to reap the benefits of the centrifugal forces now at play in urban economic activity (Gupta, Mittal, and Van Nieuwerburgh 2023; Monte, Porcher, and Rossi-Hansberg 2023; Ramani and Bloom 2021). Furthermore, given the recent large-scale regional investments in priority industries in the United States (e.g., semiconductors, manufacturing, renewable energy) (Hanson, Rodrik, and Sandhu 2025; Muro et al. 2022), in what ways do the local impacts of these investments vary based on existing LED policy commitments? Similar questions could be asked about local responses to a rapidly evolving current landscape of international trade and federal support for state and local government.
The policy mixes present in 2017 are an influential antecedent to learning processes currently underway. Where will localities that stood out then for the breadth and depth of their policy commitments stand in 2028 or beyond? To what extent are current pressures affecting the most commonly accepted policy practices, and what local factors support increased policy innovation and distinction? Will the impacts of current changes such as shocks to central downtown economies, large-scale regional industrial investments, and/or shifting trade protections diverge on the basis of the existing set of LED policy commitments? Exploring these questions will be crucial to understanding a new era of change in urban economic development and governance.
Footnotes
Declaration of Conflicting Interests
The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author received no financial support for the research, authorship, and/or publication of this article.
Notes
Appendix. All Amazon HQ2 Proposals Collected by the Proposing Locality(-ies).
| Full-text proposals | |||||
|---|---|---|---|---|---|
| Proposal number | Metro area | Secondary locality proposing within metro | State | Region (West, Midwest, Northeast, South) | Population (2018) |
| 1 | Albany | NY | NE | 883,169 | |
| 2 | Atlanta | GA | S | 5,949,951 | |
| 3 | Stonecrest | GA | S | 53,772 | |
| 4 | Atlantic City | NJ | NE | 265,429 | |
| 5 | Bakersfield | CA | W | 896,764 | |
| 6 | Baltimore | MD | S | 2,802,789 | |
| 7 | Columbia / Howard County | MD | S | 322,621 | |
| 8 | Boston | MA | NE | 4,875,390 | |
| 9 | Andover-Haverhill-Lawrence-Methuen (Lower Merrimack) | MA | NE | 788,183 | |
| 10 | Everett | MA | NE | 45,856 | |
| 11 | Lowell-Billerica-Tewskbury | MA | NE | 186,421 | |
| 12 | Lynn | MA | NE | 93,617 | |
| 13 | Marlborough-Westborough-Northborough- Southborough-Hudson | MA | NE | 84,540 | |
| 14 | Peabody | MA | NE | 52,865 | |
| 15 | Somerville | MA | NE | 80,434 | |
| 16 | Bridgeport | Danbury | CT | NE | 84,479 |
| 17 | Stamford (with Hartford) | CT* | NE | 3,572,000 | |
| 18 | Buffalo | (with Rochester) | NY | NE | 2,203,411 |
| 19 | Chicago | Gary | IN | MW | 76,677 |
| 20 | Kankakee County | IL | MW | 109,953 | |
| 21 | Cleveland | OH | MW | 2,057,009 | |
| 22 | Dallas | Expo Park | TX | S | N/D |
| 23 | Fort Worth | TX | S | 855,786 | |
| 24 | Denver | CO | W | 2,932,415 | |
| 25 | Durham-Chapel Hill-Raleigh | NC | S | 575,412 | |
| 26 | Fresno | CA | W | 994,400 | |
| 27 | Green Bay | WI | MW | 321,591 | |
| 28 | Las Vegas | NV | W | 2,231,647 | |
| 29 | Los Angeles | Garden Grove-Santa Ana-Orange | CA | W | 647,292 |
| 30 | Huntington Beach-Long Beach | CA | W | 669,489 | |
| 31 | Irvine | CA | W | 265,502 | |
| 32 | Louisville | KY, IN | S | 1,297,301 | |
| 33 | Madison | WI | MW | 660,422 | |
| 34 | Manchester | NH* | NE | 415,247 | |
| 35 | Miami | FL | S | 6,198,782 | |
| 36 | Minneapolis-St. Paul | MN, WI | MW | 3,629,190 | |
| 37 | Missouri (state-level) | MO | MW | N/A | |
| 38 | Nashville | TN | S | 1,930,961 | |
| 39 | New Haven | Waterbury | CT | NE | 108,672 |
| 40 | New York | NY, NJ,PA | NE | 19,979,477 | |
| 41 | Bayonne | NJ | NE | 65,300 | |
| 42 | Jersey City | NJ | NE | 261,746 | |
| 43 | Mt. Vernon | NY | NE | 68,035 | |
| 44 | New Rochelle | NY | NE | 79,205 | |
| 45 | Newark | NJ | NE | 280,463 | |
| 46 | Old Bridge Township | NJ | NE | 65,898 | |
| 47 | North Port-Sarasota | Manatee County | FL | S | 394,387 |
| 48 | Oxnard-Thousand Oaks | CA | W | 850,967 | |
| 49 | Philadelphia | PA, NJ,DE | NE | 6,096,372 | |
| 50 | Bristol Township | PA | NE | 53,625 | |
| 51 | Camden | NJ | NE | 74,608 | |
| 52 | Salem County | NJ | NE | 62,746 | |
| 53 | Wilmington | DE* | S | 965,749 | |
| 54 | Pittsburgh | PA | NE | 2,324,743 | |
| 55 | Portland | ME | NE | 535,420 | |
| 56 | Brunswick | ME | NE | 15,244 | |
| 57 | Portland | OR,WA | W | 2,478,810 | |
| 58 | Providence | RI,MA | NE | 1,621,337 | |
| 59 | Fall River | MA | NE | 89,339 | |
| 60 | New Bedford | MA | NE | 95,117 | |
| 61 | Riverside-San Bernadino | Riverside | CA | W | 323,935 |
| 62 | St. Louis | MO,IL | MW | 2,805,465 | |
| 63 | Edwardsville | IL | MW | 24,868 | |
| 64 | San Diego | CA | W | 3,343,364 | |
| 65 | Chula Vista | CA | W | 266,468 | |
| 66 | San Francisco-Oakland | CA | W | 4,729,484 | |
| 67 | San Jose- Sunnyvale | CA | W | 1,999,107 | |
| 68 | Sault Ste. Marie # | MI | MW | 80,025 | |
| 69 | Seattle | King County-Puget Sound | WA | W | 2,228,000 |
| 70 | Syracuse | (with Utica) | NY | NE | 650,502 |
| 71 | Tallahassee | FL | S | 385,145 | |
| 72 | Toledo | OH | MW | 602,871 | |
| 73 | Trenton | Princeton-West Windsor | NJ | NE | 58,773 |
| 74 | Tulsa | OK | S | 993,797 | |
| 75 | Virginia Beach – Norfolk – Newport | VA,NC | S | 1,728,733 | |
| 76 | Washington, DC | DC,VA, MD,WV | S | 6,249,950 | |
| 77 | Charles County | MD | S | 161,476 | |
| 78 | Northern Virginia (Arlington, Alexandria, Loudoun and Fairfax Counties) | VA | S | 1,942,052 | |
| 79 | Worcester | MA,CT | NE | 957,866 | |
| 80 | Gardner | MA | NE | 20,555 | |
| 81 | Leominster | MA | NE | 41,579 | |
| 82 | Calgary | Alberta | Canada | 1,477,000 | |
| 83 | Halifax | Nova Scotia | Canada | 409,124 | |
| 84 | Hamilton | Ontario | Canada | 759,000 | |
| 85 | Toronto | Ontario | Canada | 6,082,000 | |
| 86 | Vancouver | British Columbia | Canada | 2,531,000 | |
| 87 | Winnipeg | Manitoba | Canada | 800,000 | |
| Partial proposals | |||||
| Proposal # | Metro area | Secondary locality proposing within metro | State | Region | Population |
| 88 | Boston | Weymouth-Abington-Rockland | MA | NE | 91,630 |
| 89 | Charlotte | NC,SC | S | 2,569,213 | |
| 90 | Chicago | IL, IN, WI | MW | 9,498,716 | |
| 91 | Cincinnati | OH, KY. IN | MW | 2,190,209 | |
| 92 | Columbus | OH | MW | 2,106,541 | |
| 93 | Dallas | TX | S | 7,539,711 | |
| 94 | Detroit | MI | MW | 2,932,415 | |
| 95 | Grand Rapids | MI | MW | 1,069,405 | |
| 96 | Harrisburg | PA | NE | 574,659 | |
| 97 | Houston | TX | S | 6,997,384 | |
| 98 | Indianapolis | IN | MW | 2,048,703 | |
| 99 | Las Cruces | NM | W | 217,522 | |
| 100 | Los Angeles | Pomona | CA | W | 152,494 |
| 101 | Memphis | TN | S | 1,350,620 | |
| 102 | New Orleans | LA | S | 1,270,399 | |
| 103 | New York | Suffern-Rockland County | NY | NE | 325,522 |
| 104 | Orlando | FL | S | 2,572,962 | |
| 105 | Phoenix | AZ | W | 4,857,962 | |
| 106 | Sacramento | CA | W | 2,345,210 | |
Italicized entries indicate localities that fall within the most recently-listed metropolitan statistical areas.