Estimating determinants of transportation and warehousing establishment locations using U.S. administrative data

Carpenter, Craig; Dudensing, Rebekka; Van Sandt, Anders Article Estimating determinants of transportation and warehousing establishment locations using U.S. administrative data REGION Provided in Cooperation with: European Regional Science Association (ERSA) Suggested Citation: Carpenter, Craig; Dudensing, Rebekka; Van Sandt, Anders (2022) : Estimating determinants of transportation and warehousing establishment locations using U.S. administrative data, REGION, ISSN 2409-5370, European Regional Science Association (ERSA), Louvain-la-Neuve, Belgium, Vol. 9, Iss. 1, pp. 1-27, https://doi.org/10.18335/region.v9i1.366 This Version is available at: https://hdl.handle.net/10419/312624 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc/4.0/ Volume 9, Number 1, 2022, 1–27 journal homepage: region.ersa.org DOI: 10.18335/region.v9i1.366 ISSN: 2409-5370 Estimating Determinants of Transportation and Warehousing Establishment Locations Using U.S. Administrative Data Craig Carpenter1, Rebekka Dudensing2, Anders Van Sandt3 1Michigan State University, East Lansing, MI, USA 2Texas A&M University, College Station, TX, USA 3University of Wyoming, Laramie, WY, USA Received: 26 April 2021/Accepted: 13 December 2021 Abstract. Interactions between transportation and warehousing and other industry clusters are not widely explored and the determinants of logistics locational determinants is limited in the U.S. context. These gaps in the literature, along with the U.S. transportation and warehousing sector’s decentralization from urban areas and concentration in regions, highlight the importance of understanding the effects of place-based factors and interindustry clusters on the locations and employment of transportation and warehousing industries. The analysis uses restricted-access U.S. Census Bureau data aggregated to the county level, along with secondary data sources, to estimate the locational determinants of transportation and warehousing (TW) industries based on transportation infrastructure as well as sociodemographic and institutional variables. The analysis takes a crosssectional (non-causal) approach to focus on time-invariant location factors while testing and implementing zero-inflated count data distributions to model the data generation processes more accurately. Results indicate that subsectors are affected differently by infrastructure, sociodemographic, and institutional variables. Additionally, different factors are associated with industry presence versus size. Finally, we show that data using aggregated industries obscures locational factors’ importance for individual sub-sectors and, further, that industrial aggregation obscures TW sectors’ relationships to other clusters. 1 Introduction Transportation and warehousing (TW) industries provide or support the transport of passengers and storage of cargo via roads, rails, water, air, and pipelines. Given these modes of transportation, and in addition to local labor force composition, certain locations may offer comparative advantages to local industries based on their economic and placebased assets, leading to clusters of transportation and warehousing establishments (Kang 2020,Ng, Gujar 2009) 1 .Porter (2000,2001) popularized the idea that inter-related industries harness regional competitive advantage to strengthen innovation, productivity, and other economic outcomes. Transportation infrastructure is often an institutional factor used to predict or describe economic outcomes. 1 This article necessarily uses many abbreviations and acronyms. In addition to in-text first-use definitions, we provide a list of acronyms in appendix Table A.1 for reference. 1 2 Synergistic effects between transportation and warehousing and other industry clusters are not widely explored, and in general, the literature on logistics locational determinants is fairly limited in the U.S. context (Rivera et al. 2014,2016). These gaps in the literature, along with the U.S. transportation and warehousing sector’s decentralization from urban areas and concentration in some U.S. regions (Cidell 2010), highlights the importance of understanding the effects of place-based factors and inter-industry clusters on the locations and employment of transportation and warehousing industries. While transportation topics have been more explored in the European context, less attention has been given to the potentially different U.S. transportation context, likely due to prior data limitations. Fewer inter-regional differences in transportation policy exist between U.S. states than European countries (Rodrigue, Notteboom 2010), which may have provided richer research opportunities regarding both transportation and its relationship to economic competitiveness. Additionally, state ownership of some transportation networks (Clausen, Voll 2013) is more common in Europe, which may prompt published studies of both industries and infrastructure. Regardless, understanding the relationships between TW location decisions and local economic, demographic, and infrastructure measures is pertinent to both geographies as infrastructure spending has waned in recent years in both the US and Europe, but may be increasing in the near future. Using confidential U.S. administrative data aggregated to the county level, the focus of this article is to illustrate how more refined data and unique data generation processes may be utilized to explore the influence of local industry, socioeconomic, and place-based factors on the size of local transportation and logistics (including warehousing) industries. This illustration includes accounting for clustering by examining the association of various other industries on TW location. The analysis takes a cross-sectional (associative and non-causal) approach to focus on time-invariant location factors while testing and implementing zero-inflated count data distributions to most accurately model the data generation processes of TW industries. The following section reviews the literature on TW locational determinants and transportation’s effects on economic competitiveness. The novel data and methods underlying our approach are then presented, followed by study results for selected transportation sectors. The conclusion discusses general observations about TW cluster predictors and the propensity of TW sectors to support other clusters, opening the door to future U.S. TW locational choice research. 2 Literature Review Economic studies of transportation and warehousing (TW) tend to focus on the costs and barriers of moving either goods or people. Analyses of the benefits and costs (including external costs) of infrastructure often account for both people and freight in traffic volume (Behiri et al. 2018,Li, Madanu 2009,Mayeres et al. 1996). Researchers accept that the people and freight require different treatment, but common models are readily adapted to address either passenger or freight concerns based on quantities of goods to be transported between specific locations, distribution flows between those locations, modal splits, and assignment to transportation networks (de Jong et al. 2004). In essence, these models orient networks and transportation infrastructure to minimize transportation distances and costs. Despite its importance in moving goods as well as labor and consumers, transportation emerged as a topic within economic geography relatively recently (Hesse, Rodrigue 2004), and the literature on TW firm location decisions, in particular, remains relatively sparse (Holl, Mariotti 2018). A comprehensive review of the extensive firm location literature is beyond the scope of this article; however, this section discusses key advances in the literature since McFadden’s (1973) discrete choice work. Much of the literature pertaining to transportation in an economic development context is focused on transportation infrastructure and its role in supporting development (Chen et al. 2016,Maparu, Mazumder 2017,Melecky et al. 2019). Demand increasingly influences supply chains and TW logistics, particularly with the rising importance of e-commerce and its associated distribution (Bowen 2008,Hesse, Rodrigue 2004,Hughes, Jackson 2015, REGION : Volume 9, Number 1, 2022 3 Rodrigue 2020). Beyond road densities, transportation hubs, flows, and networks are becoming increasingly important (Crang 2002). Inland hubs and intermodal facilities are becoming more important, and warehouses are moving to places with strong multi-modal networks, especially as technology promotes larger warehouses and distribution centers, which are increasingly independent firms rather than divisions of manufacturing or retail firms (Bowen 2008). Cidell (2010) expands on this work, using Economic Census and County Business Patterns Data to measure the concentration of warehousing and distribution establishments across the U.S. by calculating Gini coefficients for each county, although they note that employment and payroll data are not reliably available from those sources for the warehousing sector, thus limiting her choice of dependent variable. Independent variables represent demand attributes, such as population and income, and physical infrastructure attributes such as interstate and railroad miles, enplanements, and distance. Infrastructure such as highways, rail, and inland waterways were not always significant predictors of freight establishments within a metropolitan area (Cidell 2010), although other researchers have continued to find these factors critical (Guerin et al. 2021,Holl, Mariotti 2018). Overall, warehouses and distribution centers decentralized and moved toward suburbs over time. Cities with strong warehousing sectors, as measured by the Gini coefficient, tended to have more concreated establishments near the city center. Concentration early in the study period was also associated with greater decentralization within the metropolitan area over time, although many cities added the largest number of freight establishments within the central county with the densest transportation infrastructure. Other researchers have found similar results related to suburbanization (Allen et al. 2012, Cidell 2010,Dablanc et al. 2014,Holl, Mariotti 2018) and inland hub development (Bowen 2008,Holl, Mariotti 2018,Monios, Wilmsmeier 2013,Rodrigue 2020). While cost minimization drives TW location decisions, lower transportation costs often arise from agglomeration economies (Cidell 2010,Hesse, Rodrigue 2004). Further, the prevalence of railroad and highway miles, interstate highway and airport presence, and related transportation costs are often predictors of regional economic competitiveness, both for the TW sector itself (Cidell 2010,Guerin et al. 2021,Holl, Mariotti 2018) and for other sectors dependent upon the logistics (Belleflamme et al. 2000,Dudensing 2008). The range of variables included in recent logistics firm location models, such as gross domestic product (GDP) (Guerin et al. 2021), population density and accessibility to residential population and manufacturing (Sakai et al. 2020) and urban structure (Holl, Mariotti 2018), suggests that researchers are beginning to think beyond infrastructure density to broader factors that affect regional TW competitiveness. In the U.S. context, rural areas can be more attractive to certain types of firms because wages, property taxes, and land costs are all lower than in most metro areas (Parajuli, Haynes 2017,Wilkerson 2001). Economic clustering – geographic concentrations of industries related through shared knowledge, skills, inputs, demand and/or other linkages – is a popular way of building economic competitiveness (Delgado et al. 2016). These geographic concentrations build upon regional competitive advantages (Porter 1990) and agglomeration factors (Marshall 1920). In Porter’s (1990) framework, competitive advantage is influenced by four elements: firm strategy, structure, and rivalry; demand conditions; factor conditions; and related and supporting industries. Firm strategy, structure, and rivalry describes the size, number, ownership, and goals of firms. Demand conditions reflect needs of domestic (local) buyers, which can include the number of potential buyers, their purchasing needs, and their ability to pay. Factor conditions include the availability of factors of production, including land, labor, capital, and infrastructure. Related and supporting industries are those who require the goods or services of a given target industry, provide inputs, or use similar labor or technology. Each element is part of a system supporting the development of clusters that foster economic competitiveness within industries, and through those industries, of national and regional economies. Clusters have been incorporated into a handful of TW location studies. Sakai et al. (2020) and Durmu¸s, Turk (2014) find the presence of industrial clusters has a positive effect on logistic establishment location decisions. Van den Heuvel et al. (2013) observes REGION : Volume 9, Number 1, 2022 4 within-sector clustering as Belgian TW establishments tend to develop in proximity to existing logistics establishments, and they call for additional research into drivers and benefits of clustering within the TW sector. 3 Conceptual Framework The focus on cost-distance minimization models in the transportation literature, gravity models in transportation and trade/regional economics literatures, and economic clusters within the regional economic literature seem to converge where the rubber meets the road. That is, for an entrepreneur or economic developer looking for opportunities to start or grow a transportation-based business, establishment and employment agglomeration (clustering) signal locations with existing advantages. Locations with similar demographic, transportation, and economic characteristics should be similarly suited to transportation business success. Durmu¸s, Turk (2014), Sakai et al. (2020), and Van den Heuvel et al. (2013) all incorporate clustering into their location choice models but fall short of placing their work within the broader clustering framework. They incorporate agglomerations of TW or other industry establishments with other elements of traditional cost-minimization location modeling. However, economic competitiveness signaled by clusters is a function of cost factors, including infrastructure, population density, and labor availability and quality. In fact, these factors are represented within the four elements of competitiveness in the Porter (1990) model. Thus, we embrace the broader framework of economic competitiveness and treat factor conditions (infrastructure, labor), demand conditions, industry structure, and related industries as part of the competitive locational choice decision. The competitiveness model also facilitates the exploration of how TW and other industry clusters interact. While interesting clusters themselves, the TW sectors in this study serve numerous other clusters and in fact may be considered parts of those clusters. In addition, subsectors within the broader TW sector often reinforce and strengthen each other. Figure A.1 represents our adaptation of the Porter framework, emphasizing the interactive relationship between firm structure and related industries, as well as government’s direct influence on factor condition through infrastructure investment. 4 Transportation and Warehousing Data We use restricted-access 2014 data from the Longitudinal Business Database (LBD) and Integrated Longitudinal Business Database (ILBD) for the continental U.S. along with 2014 secondary data sources to econometrically estimate the location decisions of five TW industries. In Europe, Eurostat offers a program similar to the U.S. Federal Statistical Research Data Center program (Eurostat 2020). As noted above, researchers often measure industry size in a location with public establishment counts, which is a nonnegative integer count and includes numerous zeros that increase with more geographically refined units of observation. This measure is limited both because of disclosure issues and because an establishment count is a poor measure of regional industry size. Employment may serve as a better measure of industry size, while still using count data methods. We thus examine both non-employer establishments (i.e., establishments without any paid employees), employer establishments, and total employment as measures of industry size. The Census Bureau’s County Business Patterns (CBP) is among the most utilized county-level public data sources for establishment and employment counts for industries within the North American Industry Classification System (NAICS), including TW industries. Other federal data programs publish information about the economic activity in TW, including the Quarterly Census of Employment and Wages (QCEW), the Quarterly Workforce Indicators (QWI), and Non-employer Statistics (NS). Although public versions of these data are available, the exact counts of a particular NAICS code are also often suppressed. These limitations are particularly prevalent in some TW industries due to the small number of employers within some counties, especially rural counties. Given the extensiveness of the disclosure limitations, exact counts yield more precise estimates; specifically, Carpenter et al. (2021) show that the measurement error in public U.S. regional economic data implies substantially biased estimates. A Federal Statistical Research Data REGION : Volume 9, Number 1, 2022 5 Center (FSRDC) is thus a natural place to make improvements to existing research using the LBD and ILBD, which we aggregate to the county-level to develop unsuppressed non-employer establishment counts, employer establishment counts, and total county employment (as in Carpenter et al. 2021,Van Sandt et al. 2021), all within specific TW industries. The zero-inflation methods described below capitalize on this refined data by being able to accurately discern the zero-generating regime for unsuppressed counties with no TW industries. The LBD is an annual series produced by the U.S. Census Bureau based on establishment records from the Business Register (Jarmin, Miranda 2002). The Business Register (BR) acts as the source of information for both the public CBP as well as the restricted LBD; however, the LBD undergoes more edits for longitudinal consistency and does not contain suppressed or noise infused values. Consequently, the LBD is a fundamental dataset for studying the determinants of firm strategy, structure, and rivalry, including entry, growth, and exit at the establishment, firm, industry, and economy-wide level (Carpenter, Loveridge 2018,2019,2020,Davis et al. 2006,Foster et al. 2006,Haltiwanger et al. 2013). Comprehensive longitudinal establishment-level data are similarly available in other countries, with economists often using them to examine firm establishment location decisions, though TW locational analysis remains understudied (Arauzo-Carod, Viladecans-Marsal 2009,Chen, Moore 2010,Devereux et al. 2007,Figueiredo et al. 2002, Holl 2004). Location quotients (LQs) of groups of non–TW industries explore the effects of related industries and the potential for inter-industry clustering. Despite the LBD and the ILBD being at the establishment level, the Census Bureau requires us to conduct our analysis at an aggregated county level due to the sensitivity of the data. However, this unit of analysis is useful for the inclusion of important secondary data sources that capture the potential influence of factor conditions, including local internet access, travel infrastructure, water coverage, urban influence, and human capital, as well as demand conditions including other demographic and locational variables. Further, county-level aggregation facilitates the implementation of the previously discussed count data methods. Table 1provides descriptive statistics for the publicly available county-level data 2 . Variables are grouped by Porter Element with demographic variables representing either demand or factor conditions; infrastructure and institution variables reflect factor conditions. A handful of variables may cross elements, but each variable is listed only once. The TW establishment and employment data, of course, reflect firm strategy. The detailed nature of the data allows us to model the location and employment attributes for a range of TW industries, which fall within Transportation and Warehousing (NAICS 48-49) and Professional, Scientific, and Technical Services (NAICS 54). Specifically, we examine General Warehousing & Storage (NAICS 493110); Process, Physical Distribution, & Logistics Consulting (NAICS 541614); Pipeline Transportation (NAICS 486); Support Activities for Road Transportation (NAICS 488490); General Freight Trucking (NAICS 4841). Other sectors and industries are feasible and interesting, but we limit the presentation to example sectors here to cover a breadth of TW sectors, for which locational determinants have been shown to vary in international contexts (Holl, Mariotti 2018,Kang 2020,Rivera et al. 2014), while maintaining relative brevity 3 . Figure 1shows a map of the distribution of Pipeline Transportation (NAICS 486) employment and data suppression issues. Maps of the other industries examined herein are available in the appendix (Figures A.2-A.5). Motivation for the included determinants of transportation location draw on both past literature and virtual focus groups. Following a method established by Loveridge et al. (2013) the authors conducted virtual focus groups with U.S. economic development practitioners, small business support professionals, and entrepreneurs. Commonly found location determinants in published studies include local land and labor costs, taxes, infrastructure, market size, and agglomeration economies (Blair, Premus 1987). Proxy variables often include per capita income or wage rate, high school and bachelor degree 2Table A.1 in the appendix provides additional descriptive statistics. 3 The authors acknowledge that the choice of sectors could be criticized as ad hoc, rather than covering a breadth of industries. Nonetheless, these sectors are common to transportation and warehousing research, and as the results indicate, this breadth of sectors allows for interesting cross-sector comparisons of locational determinants and future researchers are encouraged to examine additional sectors. REGION : Volume 9, Number 1, 2022 6 Table 1: Descriptive statistics Variable Obs. Mean Std. Dev. Source Demographics – Demand Conditions Population 3,107 101,931.30 327,468.00 ACS Population Density 3,107 0.27 1.80 ACS Median Age 3,107 40.85 5.18 ACS Per Capita Income (thousands $ ) 3,107 39.59 11.64 BEA Percent of Residents in Poverty 3,107 16.84 6.55 ACS Demographics – Factor Conditions Unemployment Rate (5yr avg.) 3,107 7.89 2.68 BLS Social Capital Index 3,107 0.01 1.26 NERCRD Opiates Prescribed / 100 People 2,944 85.72 49.37 CDC Percent Work in Another County13,107 30.10 17.69 ACS Percent Black 3,107 0.09 0.15 ACS Percent Hispanic 3,107 0.09 0.14 ACS Percent – Bachelors or Above 3,107 13.24 5.48 ACS Infrastructure and Institutions – Factor Conditions Median Home Value ( $ 1,000’s) 3,107 135.77 79.20 ACS Avg. Combined Sales Tax Rate23,107 7.01 1.68 Tax Foundation Avg. Effective Property Tax Rate 3,107 1.06 0.51 SmartAsset.com Internet Service Providers (ISPs) 3,106 5.19 1.12 FCC Metro - Urban Influence Code 3,107 0.37 0.48 ERS, USDA Micropolitan Metro Adjacent - UIC 3,107 0.33 0.47 ERS, USDA Micropolitan Non-metro Adjacent - UIC 3,107 0.30 0.46 ERS, USDA Interstate Density33,107 1.79 3.07 Census Shapefiles Highway Density33,107 0.37 0.24 Census Shapefiles Percent Covered by Water 3,107 4.50 11.15 ERS, USDA Community Colleges 3,107 0.33 0.84 NCES Universities or Colleges 3,107 0.72 2.40 NCES Military Bases 3,107 0.04 0.22 US Census Bureau Census Region Fixed Effects 3,107 N/A N/A US Census Bureau TW Industry Establishments and Employment – Firm Strategy, Structure, and Rivalry General Warehousing & Storage (493110) Estab. 3,106 3.32 13.22 WholeData General Warehousing & Storage (493110) Emp 3,106 209.12 881.42 WholeData Warehousing and Storage (4631) Non-emp 1,681 5.21 24.31 NS Management Consulting Services (54161) Estab. 3,106 41.78 198.01 WholeData Management Consulting Services (54161) Emp. 3,106 293.25 1,705.54 WholeData Management, Scientific, and Technical Consulting Services (5416) Non-emp. 3,072 228.31 924.11 NS Pipeline Transportation (486) Estab. 3,106 1.33 3.93 WholeData Pipeline Transportation (486) Emp. 3,106 16.89 187.20 WholeData Pipeline Transportation (486) Non-emp. 511 1.23 5.91 NS Support for Road Transportation (488490) Estab. 3,106 0.81 3.31 WholeData Support for Road Transportation (488490) Emp. 3,106 10.89 59.73 WholeData Support for Transportation (488) Non-emp. 2,924 44.28 229.92 NS General Freight Trucking (4841) Estab. 3,106 22.18 85.38 WholeData General Freight Trucking (4841) Emp. 3,106 293.68 1,014.61 WholeData General Freight Trucking (4841) Non-emp. 3,100 160.87 781.12 NS Notes: Due to disclosure prevention limitations, these descriptive statistics are based on public data sources, while the main regression results are based on the limit-access Longitudinal Business Database and Integrated Longitudinal Business Database. All data are based on 2014; 2014 is chosen for practical reasons related to the availability of many of the secondary data sources. The internal and unsuppressed data used in regressions differ slightly in addition to the inclusion of unsuppressed cells. The more refined 541614 NAICS was not publicly available. Industry size measure abbreviations: non-employer establishments (Non-Emp.); employer establishments (Est.); total employment (Emp.). 1Percent of all residents age 16 and older. 2State-level variable. 3Miles of road per hundred square miles. Abbreviations: ACS: American Community Survey, BLS: Bureau of Labor Statistics, BEA: Bureau of Economic Analysis, CDC: Center for Disease Control and Prevention, FCC: Federal Communications Commission, ERS: Economic Research Service, NS: Non-employer Statistics, NCES: National Center for Education Statistics, NERCRD: Northeast Regional Center for Rural Development, USDA: US Department of Agriculture REGION : Volume 9, Number 1, 2022 7 Notes: Map shows employment ranges in every county in the continental U.S. “Suppressed” indicates that the exact value of the data is suppressed to prevent improper disclosure of identifiable information and placed into bins (e.g., “20-99 employees”). For the sake of these maps, we replace these ranges with their midpoint. Figure 1: Employment Distribution of Pipeline Transportation (NAICS 486) shares, unemployment rate, highway and interstate coverage, property taxes, rurality spectrum codes, population density, climate, racial composition, and location quotients (Coughlin, Segev 2000,Guimar˜aes et al. 2000,2004). The virtual focus groups supported many of these previously found factors but added insights into the potential importance in the rural U.S. of broadband access and issues related to workforce turnover and opioid misuse, particularly in rural areas and particularly related to TW industries (Joudrey et al. 2019,Rigg et al. 2018). This article uses the opioid prescription rate as a proxy for opioid misuse; though imperfect, given regional variation in prescription rates and negative outcomes associated with opioid misuse (Quast 2018), we continue to include this variable given the relevance of the opioid epidemic in the U.S. noted in our qualitative focus groups and in past research (Rigg et al. 2018). Finally, researchers define the transportation and logistics clusters based on research interests and purposes, and definitions may affect study outcomes. For example, the Kumar et al. (2017) transportation and logistics cluster includes both freight and passenger transportation, suggesting that passenger transportation might affect the prevalence of the cluster in metropolitan areas. For clarity in the presentation of results, this article relies on Location Quotients (LQs) of groups of non–TW industries to examining the potential for inter-industry clustering in this article4. 5 Methods Given the nature of geography-based establishment and employment data, and the ability to allow for two zero-generating processes through an added logit link function, count data models are useful for empirical estimation. Resultantly, count data estimators have become common in locational and threshold models (Carpenter et al. 2021), though researchers often fail to appropriately apply and compare the various models, as we will demonstrate. For non–TW establishments, there are numerous applications of Poisson (Arauzo-Carod, Viladecans-Marsal 2009,Papke 1991), Negative Binomial (NB) (Conroy et al. 2016,Holl 2004,Smith, Florida 1994), Hurdle Poisson (HP) (Chakraborty 2012, 4 This article uses the entire U.S. as the reference for the location quotient because, as Figure 1highlights, that TW industries tend to cluster in areas above alternatives, such as at the states. Additionally, many states are quite small and have few TW industries, confusing the interpretation of the LQs. As noted later, regressions include regional fixed effects. REGION : Volume 9, Number 1, 2022 8 Henderson et al. 2000), Zero-Inflated Poisson (ZIP) (Chakraborty 2012,List 2001,Reum, Harris 2006). We also consider the Zero-Inflated Negative Binomial (ZINB) here within TW when using a large geographic scope without data suppression. The Poisson model is useful as a starting point for comparison to other estimation procedures. However, the Poisson model’s assumption that dependent variable’s conditional variance is equal to the conditional mean is often violated in practice due to multiple sources of overdispersion, in which case the NB is more efficient. Overdispersion is likely to exist in TW industries due to large infrastructure endowments in some geographical units motivating the need for many TW establishments and levels of employment 5 . Zero-inflated and hurdle versions of the Poisson model are often used to account for excess zeros in the data when it is believed that the zeros arise from two separate regimes. In the current context, the two zero-generating regimes in zero-inflated models may be interpreted as structural and non-structural zeros, i.e., counties that have zero establishments due to a lack of a requisite resource like water and counties that have zero establishments due to chance or extant economic conditions. On the other hand, HP only assumes one type of zero and truncates the Poisson distribution after the logit zero-generating, which is less flexible6. Finally, researchers can fail to account for overdispersion that remains after the zero-inflation. This is particularly important when examining TW sectors because remaining overdispersion (from unobserved heterogeneity or excessive concentration of firms) increases in economies of scale and agglomeration, which are common in TW sectors. If authors are interested in modeling both an inflation stage and accounting for these multiple sources of overdispersion, ZINB is underutilized; indeed, we find it to be preferred for many TW industries. We compare results and diagnostics from the Poisson, NB, and their zero-inflated complements to provide results across five TW subsectors. This flexible approach takes advantage of the count data, while using the potential for overdispersion caused by (1) numerous zeros, which depend on industrial sector specificity, spatial monopsony, and economies of scale, (2) long-tailed distributions, resulting from spatial concentration, which varies depending on spatial resource dependence and economies of agglomeration, and (3) unobserved heterogeneity (Carpenter et al. 2021). To choose among the various count data models, many researchers incorrectly use Vuong’s Statistic to compare these various count data estimators (Wilson 2015). To provide a consistent comparison, we use graphical distribution and information criteria comparisons as suggested in Greene (1994). All models make use of White’s robust standard errors, and variance inflation factors did not indicate any potential concern over loss of efficiency from multicollinearity. 6 Results While TW includes dozens of industries spanning many sectors of the economy and NAICS sector codes, we focus on investigating a subset of these industries. To summarize, the purpose of this section is not to present an exhaustive analysis of all TW industries’ locations, but instead to provide examples of different data generation processes and the advantages of different industry size measures. Through these example models, we illustrate how more refined data and unique data generation processes may be utilized to explore how local industry, socioeconomic, and place-based factors influence the size of 5 The NB approach allows unobserved heterogeneity between subjects and implies overdispersion, but where the amount of overdispersion increases with E ( yi|xi ) (Wooldridge 2010). Although this may hold when modeling TW employment (e.g., due to large employers), this relationship is unlikely to hold when measuring establishments due to excess zeros driving much of the overdispersion while decreasing E ( yi|xi ). Additionally, we note that, though less efficient, Poisson would still produce consistent estimates with fewer assumptions than NB. 6 The log-likelihood of the Zero-Inflated Poisson model may be written, ln L = Pi∈Sln {F(γ′Zi) + [1 −F(γ′Zi)] exp(−λi)} + Pi/∈S{ln[1 −F(γ′Zi)] −λi+yiβ′Xi−ln(yi! )} , where S is the set of observations taking on a zero value ( yi = 0), F is the logit link function that determines the odds of a zero belonging to regime one or two, Z is a vector of covariates that describe the participation decision, X is a vector of covariates that describe the amount decision, and γ and β are the participation and amount parameters of interest to be estimated, respectively. REGION : Volume 9, Number 1, 2022 15 spatial econometric techniques to measure effects of neighboring counties. The increasing availability of disclosed data through multiple U.S. agencies also presents opportunities to further explore differences in locational determinants of TW in the U.S. and European contexts. Finally, future research could expand the study of inter-cluster dependence, which could reduce hazards associated with cluster targeting in economic development (Barkley, Henry 2009) and improve overall competitiveness opportunities (Porter 2001). Such research could support efforts to capture inter-industry linkages and define industries within local clusters (Delgado et al. 2016). References Allen J, Browne M, Cherrett T (2012) Investigating relationships between road freight transport, facility location, logistics management and urban form. Journal of Transport Geography 24: 45–57. CrossRef Arauzo-Carod JM, Viladecans-Marsal E (2009) Industrial location at the intrametropolitan level: The role of agglomeration economies. Regional Studies 43: 545–558. CrossRef Barkley DL, Henry MS (2009) Targeting industry clusters for reginal economic development: The REDRL approach. In: Goetz S, Deller S, Harris T (eds), Targeting Regional Economic Development. Routledge, London, 183–197 Bartik TJ, Biddle SCY, Hershbein BJ, Sotherland ND (2018) Wholedata: Unsuppressed county business patterns data: Version 1.0 [dataset]. Kalamazoo. Retrieved from https://upjohn.org Behiri W, Belmokhtar-Berraf S, Chu C (2018) Urban freight transport using passenger rail network: Scientific issues and quantitative analysis. Transportation Research Part E: Logistics and Transportation Review 115: 227–245. CrossRef Belleflamme P, Picard P, Thisse JF (2000) An economic theory of regional clusters. Journal of Urban Economics 48: 158–184. CrossRef Blair JP, Premus R (1987) Major factors in industrial location: A review. Economic Development Quarterly 1: 72–85. CrossRef Bowen JT (2008) Moving places: the geography of warehousing in the US. Journal of Transport Geography 16: 379–387. CrossRef Calafati L, Froud J, Haslam C, Johal S, Williams K (2021) Diversity in leading and laggard regions: Living standards, residual income and regional policy. Cambridge Journal of Regions, Economy and Society 14: 117–139. CrossRef Carpenter CW, Loveridge S (2018) Differences between latino-owned businesses and white-, black-, or asian-owned businesses: Evidence from census microdata. Economic Development Quarterly 32: 225–241. CrossRef Carpenter CW, Loveridge S (2019) Factors associated with latino-owned business survival in the United States. The Review of Regional Studies 49: 73–97. CrossRef Carpenter CW, Loveridge S (2020) Business, owner, and regional characteristics in latinoowned business growth: An empirical analysis using confidential census microdata. International Regional Science Review 43: 254–285. CrossRef Carpenter CW, Van Sandt AT, Dudensing R, Loveridge S (2021) Profit pools and determinants of potential county-level manufacturing growth. International Regional Science Review.CrossRef Carpenter CW, Van Sandt AT, Loveridge S (2021) Empirical methods in business location research. Regional Studies, Regional Science 8: 344–361 REGION : Volume 9, Number 1, 2022 16 Chakraborty K (2012) Estimation of minimum market threshold for retail commercial sectors. International Advances in Economic Research 18: 271–286. CrossRef Chen MX, Moore MO (2010) Location decision of heterogeneous multinational firms. Journal of International Economics 80: 188–199. CrossRef Chen Y, Salike N, Luan F, He M (2016) Heterogeneous effects of interand intracity transportation infrastructure on economic growth: Evidence from Chinese cities. Cambridge Journal of Regions, Economy and Society 9: 571–587. CrossRef Cidell J (2010) Concentration and decentralization: The new geography of freight distribution in US metropolitan areas. Journal of Transport Geography 18: 363–371. CrossRef Clausen U, Voll R (2013) A comparison of North American and European railway systems. European Transport Research Review 5: 129–133. CrossRef Conroy T, Deller S, Tsvetkova A (2016) Regional business climate and interstate manufacturing relocation decisions. Regional Science and Urban Economics 60: 155–168. CrossRef Coughlin CC, Segev E (2000) Location determinants of new foreign-owned manufacturing plants. Journal of Regional Science 40: 323–351. CrossRef Crang M (2002) Between places: Producing hubs, flows, and networks. Environment and Planning A: Economy and Space 34: 569–574. CrossRef Dablanc L, Ogilvie S, Goodchild A (2014) Logistics sprawl. Transportation Research Record 2410: 105–112 Davis SJ, Faberman RJ, Haltiwanger J (2006) The flow approach to labor markets: New data sources and micro-macro links. Journal of Economic Perspectives 20: 3–26. CrossRef de Jong G, Gunn H, Walker W (2004) National and international freight transport models: An overview and ideas for future development. Transport Reviews 24: 103–124. CrossRef Delgado M, Porter ME, Stern S (2016) Defining clusters of related industries. Journal of Economic Geography 16: 1–38. CrossRef Devereux MP, Griffith R, Simpson H (2007) Firm location decisions, regional grants and agglomeration externalities. Journal of Public Economics 91: 413–435. CrossRef Dudensing R (2008) Biomedical devices clusters in South Carolina and the United States. Clemson, SC Durmu¸s A, Turk SS (2014) Factors influencing location selection of warehouses at the intra-urban level: Istanbul case. European Planning Studies 22: 268–292. CrossRef Eurostat (2020) Access to microdata. Retrieved november 1, 2020, from https://ec.europa.eu/eurostat/web/microdata/overview Figueiredo O, Guimar˜aes P, Woodward D (2002) Home-field advantage: Location decisions of Portuguese entrepreneurs. Journal of Urban Economics 52: 341–361. CrossRef Foster L, Haltiwanger J, Krizan CJ (2006) Market selection, reallocation, and restructuring in the U.S. retail trade sector in the 1990s. Review of Economics and Statistics 88: 748–758 Greene WH (1994) Accounting for excess zeros and sample selection in poisson and negative binomial regression models. NYU Working Paper No. EC-94-10. Retrieved from https://papers.ssrn.com/sol3/papers.cfm?.abstract id=1293115 REGION : Volume 9, Number 1, 2022 17 Guerin L, Vieira JGV, de Oliveira RLM, de Oliveira LK, de Miranda Vieira HE, Dablanc L (2021) The geography of warehouses in the S˜ao Paulo metropolitan region and contributing factors to this spatial distribution. Journal of Transport Geography 91: 102976 Guimar˜aes P, Figueiredo O, Woodward D (2000) Agglomeration and the location of foreign direct investment in Portugal. Journal of Urban Economics 47: 115–135. CrossRef Guimar˜aes P, Figueiredo O, Woodward D (2004) Industrial location modeling: Extending the random utility framework. Journal of Regional Science 44: 1–20. CrossRef Haltiwanger J, Jarmin RS, Miranda J (2013) Who creates jobs? Small versus large versus young. Review of Economics and Statistics 90: 347–361 Henderson JW, Kelly TM, Taylor BA (2000) The impact of agglomeration economies on estimated demand thresholds: An extension of Wensley and Stabler. Journal of Regional Science 40: 719–733. CrossRef Hesse M, Rodrigue JP (2004) The transport geography of logistics and freight distribution. Journal of Transport Geography 12: 171–184. CrossRef Holl A (2004) Transport infrastructure, agglomeration economies, and firm birth: Empirical evidence from Portugal. Journal of Regional Science 44: 693–712. CrossRef Holl A, Mariotti I (2018) The geography of logistics firm location: The role of accessibility. Networks and Spatial Economics 18: 337–361 Hughes C, Jackson C (2015) Death of the high street: Identification, prevention, reinvention. Regional Studies, Regional Science 2: 237–256. CrossRef Jarmin RS, Miranda J (2002) The longitudinal business database (No. 02-17). Retrieved from https://www.census.gov/ces/pdf/CES-WP-02-17.pdf Joudrey PJ, Edelman EJ, Wang EA (2019) Drive times to opioid treatment programs in urban and rural counties in 5 US states. JAMA - Journal of the American Medical Association, American Medical Association 322: 1310–1312 Kang S (2020) Warehouse location choice: A case study in Los Angeles, CA. Journal of Transport Geography 88: 102297 Kumar I, Zhalnin A, Kim A, Beaulieu LJ (2017) Transportation and logistics cluster competitive advantages in the U.S. regions: A cross-sectional and spatio-temporal analysis. Research in Transportation Economics 61: 25–36. CrossRef Li Z, Madanu S (2009) Highway project level life-cycle benefit/cost analysis under certainty, risk, and uncertainty: Methodology with case study. Journal of Transportation Engineering 135: 516–526. CrossRef List JA (2001) US county-level determinants of inbound FDI: Evidence from a twostep modified count data model. International Journal of Industrial Organization 19: 953–973. CrossRef Loveridge S, Nawyn S, Szmecko L (2013) Conducting virtual facilitated discussions. Community development practice. Retrieved from http://www.commdev.org/images/pdf/Conducting-virtual-facilitated-dicussions template-new 1.pdf Maparu TS, Mazumder TN (2017) Transport infrastructure, economic development and urbanization in India (1990-2011): Is there any causal relationship? Transportation Research Part A: Policy and Practice 100: 319–336. CrossRef Marshall A (1920) Principles of Economics. Palgrave MacMillan, London. CrossRef Mayeres I, Ochelen S, Proost S (1996) The marginal external costs of urban transport. Transportation Research Part D: Transport and Environment 1: 111–130. CrossRef REGION : Volume 9, Number 1, 2022 18 McFadden D (1973) Conditional logit analysis of qualitative choice behaviour. In: Zarembka P (ed), Frontiers in Econometrics. Academic Press, New York, NY, 105–142 Melecky M, Roberts M, Sharma S (2019) The wider economic benefits of transport corridors: A policy framework and illustrative application to the China-Pakistan economic corridor. Cambridge Journal of Regions, Economy and Society 12: 17–44. CrossRef Monios J, Wilmsmeier G (2013) The role of intermodal transport in port regionalisation. Transport Policy 30: 161–172. CrossRef Ng KYA, Gujar GC (2009) The spatial characteristics of inland transport hubs: Evidences from Southern India. Journal of Transport Geography 17: 346–356. CrossRef Papke LE (1991) Interstate business tax differentials and new firm location: Evidence from panel data. Journal of Public Economics 45: 47–68. CrossRef Parajuli J, Haynes KE (2017) Panel data models of new firm formation in New England. REGION 4: 65–76. CrossRef Porter ME (1990) Competitive advantage of nations. Harvard Business Review 68: 15–34. CrossRef Porter ME (2000) Location, competition, and economic development: Local clusters in a global economy. Economic Development Quarterly 14: 15–34. CrossRef Porter ME (2001) Clusters of Innovation: Regional Foundations of U.S. Competitiveness. Council of Competitiveness, Washington, D.C Quast T (2018) State-level variation in the relationship between child removals and opioid prescriptions. Child Abuse and Neglect 86: 306–313. CrossRef Reum AD, Harris TR (2006) Exploring firm location beyond simple growth models: A double hurdle application. The Journal of Regional Analysis & Policy 36: 45–67 Rigg KK, Monnat SM, Chavez MN (2018) Opioid-related mortality in rural America: Geographic heterogeneity and intervention strategies. International Journal of Drug Policy 57: 119–129. CrossRef Rivera L, Gligor D, Sheffi Y (2016) The benefits of logistics clustering. International Journal of Physical Distribution and Logistics Management 46: 242–268. CrossRef Rivera L, Sheffi Y, Welsch R (2014) Logistics agglomeration in the US. Transportation Research Part A: Policy and Practice 59: 222–238. CrossRef Rodrigue JP (2020) The distribution network of Amazon and the footprint of freight digitalization. Journal of Transport Geography 88: 102825 Rodrigue JP, Notteboom T (2010) Comparative North American and European gateway logistics: The regionalism of freight distribution. Journal of Transport Geography 18: 497–507. CrossRef Sakai T, Beziat A, Heitz A (2020) Location factors for logistics facilities: Location choice modeling considering activity categories. Journal of Transport Geography 85: 102710. CrossRef Smith DF, Florida R (1994) Agglomeration and industrial location: An econometric analysis of Japanese-affiliated manufacturing establishments in automotive-related industries. Journal of Urban Economics 36: 23–41. CrossRef Van den Heuvel FP, de Langen PW, van Donselaar KH, Fransoo JC (2013) Spatial concentration and location dynamics in logistics: The case of a Dutch province. Journal of Transport Geography 28: 39–48. CrossRef REGION : Volume 9, Number 1, 2022 19 Van Sandt AT, Carpenter CW, Dudensing RM, Loveridge S (2021) Estimating determinants of health care establishment locations with restricted federal administrative data. Health Economics 30: 1328–1346. CrossRef Wilkerson C (2001) Trends in rural manufacturing. Federal Reserve Bank of Kansas City. Retrieved from https://www.kansascityfed.org/documents/363/mse-Trends in Rural Manufacturing .pdf Wilson P (2015) The misuse of the Vuong test for non-nested models to test for zeroinflation. Economics Letters 127: 51–53. CrossRef Wooldridge JM (2010) Econometric Analysis of Cross Section and Panel Data (2nd ed.). MIT Press, Cambridge, MA © 2022 by the authors. Licensee: REGION – The Journal of ERSA, European Regional Science Association, Louvain-la-Neuve, Belgium. This article is distributed under the terms and conditions of the Creative Commons Attribution, Non-Commercial (CC BY NC) license (http://creativecommons.org/licenses/by-nc/4.0/). REGION : Volume 9, Number 1, 2022 20 A Appendix: Notes: This figure adapts the Porter Diamond model to reflect the specific interaction of TW sector firm structure and related industries. This interaction recognizes that TW is often considered a support sector to other clusters and that various subsectors of TW reinforce the broader TW sector. The adapted model also shows the substantial direct effect of government on factor conditions through infrastructure investment. The bold arrow indicates the substantial effect that government investment (typically) has on factor (input) conditions. Figure A.1: Porter diamond model adapted for TW industry clustering Notes: Map shows employment ranges in every county in the continental U.S. “Suppressed” indicates that the exact value of the data is suppressed to prevent improper disclosure of identifiable information and placed into bins (e.g., “20–99 employees”). For the sake of these maps, we replace these ranges with their midpoint. Figure A.2: Employment Distribution of General Freight Trucking (NAICS 4841) REGION : Volume 9, Number 1, 2022 21 Figure A.3: Employment Distribution of Support Activities for Road Transportation (NAICS 488490) Figure A.4: Employment Distribution of General Warehousing & Storage (NAICS 493110) Figure A.5: Employment Distribution of Process, Distribution, & Logistics Consulting (NAICS 541614) REGION : Volume 9, Number 1, 2022 22 Table A.1: List of Abbreviations and Acronyms ACS American Community Survey BEA Bureau of Economic Analysis BLS Bureau of Labor Statistics BR Business Register CBP County Business Patterns CDC Center for Disease Control and Prevention Emp. Total employment ERS Economic Research Service Est. Employer establishments FE Fixed Effects FCC Federal Communications Commission FSRDC Federal Statistical Research Data Center GDP Gross Domestic Product HP Hurdle Poisson ILBD Integrated Longitudinal Business Database ISP Internet Service Provider LBD Longitudinal Business Database LQ Location Quotients NAICS North American Industry Classification System NB Negative Binomial NCES National Center for Education Statistics Non-Emp. Non-employer establishments NERCRD Northeast Regional Center for Rural Development NS Nonemployer Statistics QCEW Quarterly Census of Employment and Wages QWI Quarterly Workforce Indicators TW transportation and warehousing USDA United States Department of Agriculture ZINB Zero-Inflated Negative Binomial ZIP Zero-Inflated Poisson Table A.2: Additional TW Industry Summary Statistics Industry and size measure Min. Median Mean Max. Source General Warehousing & Storage (493110) Estab. 0 2 8 411 CBP General Warehousing & Storage (493110) Emp 0 0 209 17,576 WholeData Warehousing and Storage (4931) Non-emp 0 0 3 572 NS Management Consulting Services (54161) Estab. 0 5 49 5,380 CBP Management Consulting Services (54161) Emp. 0 6 293 37,724 WholeData Management, Scientific, and Technical Consulting Services (5416) Non-emp. 0 25 226 23,562 NS Pipeline Transportation (486) Estab. 0 2 3 155 CBP Pipeline Transportation (486) Emp. 0 0 17 10,067 WholeData Pipeline Transportation (486) Non-emp. 0 0 0.2 110 NS Support for Road Transportation (488490) Estab. 0 1 3 110 CBP Support for Road Transportation (488490) Emp. 0 0 11 1,282 WholeData Support for Transportation (488) Non-emp. 0 9 42 6,786 NS General Freight Trucking (4841) Estab. 0 9 23 3,929 CBP General Freight Trucking (4841) Emp. 0 45 294 20,782 WholeData General Freight Trucking (4841) Non-emp. 0 50 161 23,074 NS Notes: Due to disclosure prevention limitations, these descriptive statistics are based on public data sources, while the main regression results are based on the limit-access Longitudinal Business Database and Integrated Longitudinal Business Database. All data are based on 2014; 2014 is chosen for practical reasons related to the availability of many of the secondary data sources. The internal and unsuppressed data used in regressions differ slightly in addition to the inclusion of unsuppressed cells. The more refined 541614 NAICS was not publicly available. For details on WholeData, see Bartik et al. (2018). Abbreviations: Estab: employer establishments; Non-Emp: non-employer establishments; Emp: total employment; CBP: County Business Patterns; NS: Non-employer Statistics. REGION : Volume 9, Number 1, 2022 23 Table A.3: Poisson Marginal Effects for Locational Determinants of T&W Sectors Process, Physical Industry (NAICS) General Warehousing Distribution, & Logistics Pipeline Transportation Support Activities for Road General Freight Trucking (4841) & Storage (493110) Consulting (541614) (486) Transportation (488490) Size Measure Est. Emp. Est. Emp. Non-Emp. Est. Emp. Non-Emp. Est. Emp. Non-Emp. Est. Emp. Interstate Density 0.076** 16.08*** -0.006 0.125 0.059*** 2.324*** 0.061*** 0.262*** -1.396*** 0.008 -2.273*** 0.887*** 0.070 Interstate Density2 0.002 -0.208*** 0.0009 -0.039*** -0.014*** -0.145*** -0.003 0.004 0.074*** -0.000004 0.342*** 0.839*** 0.022*** Highway Density 3.332*** 291.8*** 1.320*** 5.340*** -0.224* 8.134*** -0.075 -1.537 19.29*** 0.094 46.41*** 280.4*** 7.885*** Highway Density2 -1.578*** -154.5*** -0.418*** 2.447*** 0.305*** -17.58*** -0.226 -0.952* -11.25*** -0.100 -22.60*** -131.6*** -3.697*** Water coverage -0.027*** -4.546*** -0.007*** 0.054*** -0.001 -0.111*** -0.004* -0.067*** -0.091*** -0.002 -0.590*** -2.298*** -0.037*** Sales tax -0.058** 0.205 -0.038 -0.949*** -0.024 -0.708*** 0.034* -0.324*** 0.036 0.0005 2.937*** -9.601*** -0.192*** Property tax rate 0.014 -19.64*** -0.141 -2.433*** 0.004 -0.522** -0.064 -2.973*** 1.391*** -0.003 9.717*** -27.31*** 2.530*** Military bases -0.113* -8.909*** 0.044 2.803*** 0.083*** 0.519*** -0.156** -2.770*** 0.412*** -0.147*** -18.22*** -21.12*** -2.162*** Community colleges 0.122*** 20.69*** -0.061*** -1.023*** 0.016** -1.835*** 0.056*** -0.467*** -0.734*** -0.009 -0.117 10.27*** -0.200*** Universities -0.066*** -8.727*** 0.017** 0.110*** -0.003 1.136*** 0.013 0.390*** 0.266*** 0.005 2.317*** -1.468*** 0.508*** ISP count 0.049 2.117*** -0.153*** -0.485*** -0.005 -3.371*** -0.116*** -0.499*** 0.143* 0.020 -0.009 1.454*** 0.691*** Social capital 0.316*** 23.97*** -0.115 0.928*** 0.036* 2.302*** 0.169*** -2.677*** 0.585*** 0.037 2.218*** 21.96*** 1.004*** Micro Metro-Adj. 0.191 4.955*** 0.268 -3.723*** -0.089 -2.853*** -0.115 -1.896*** -2.750*** -0.151 -24.87*** -19.72*** -4.514*** Metropolitan 0.509** -13.37*** 0.631*** 8.756*** -0.128** -3.609*** -0.103 -8.930*** 3.582*** 0.063 -61.42*** 36.12*** -6.539*** Out-commute % -0.008*** 0.263*** -0.006** -0.074*** 0.002* -0.161*** -0.006*** 0.223*** -0.064*** -0.003** 0.816*** -1.424*** -0.034*** Opioid RX rate 0.013*** 0.269*** 0.001 0.161*** 0.001** -0.017*** 0.001 0.004 0.026*** 0.009 -0.485*** 0.793*** -0.022*** Poverty rate -0.001*** -0.107*** -0.0002* -0.009*** -0.00004 -0.003*** -0.0002*** 0.003*** -0.003*** -0.0001 -0.007*** -0.042*** -0.005*** Median age -0.100*** -13.27*** 0.053*** -0.181*** -0.011** -0.602*** -0.046*** 0.585*** 0.396*** -0.004 0.319*** -9.345*** -0.278*** Unemployment rate 0.021 -3.166*** -0.005 0.569*** -0.003 -0.244*** -0.036** -0.533*** -0.592*** -0.017 0.527*** -22.54*** -0.088 Bachelor’s degree % 0.086*** -4.197*** 0.146*** 0.978*** -0.00001 -1.101*** -0.048*** -0.662*** 0.450*** 0.004 -4.168*** 1.563*** -0.183*** ln(Population) 3.644*** 259.6*** 2.688*** 34.50*** 0.317*** 16.60*** 0.902*** 45.85*** 11.42*** 0.810*** 165.7*** 294.2*** 20.01*** Population density -0.0002 -0.106 -0.018** 0.274*** -0.034*** -0.429*** -0.037* -0.346*** -0.150*** -0.026*** -0.179* -5.858*** -0.018 Per capita income 0.003 0.136*** 0.008*** -0.013 0.004*** 0.322*** 0.016*** 0.051*** 0.022*** -0.002 0.649*** -0.884*** -0.018 Hispanic % 0.048*** 0.325*** 0.019*** 0.111*** 0.004*** -0.080*** 0.001 0.487*** 0.229*** 0.010*** 1.981*** 3.276*** 0.205*** Black % 0.020*** -0.203*** 0.005 0.015 -0.0009 0.084*** 0.017*** 0.073*** 0.075*** 0.003 0.151*** 1.576*** -0.019 Home Value -0.005*** -0.554*** -0.001** -0.030*** -0.000003 -0.036*** -0.004*** 0.032*** -0.011*** 0.0004 -0.339*** -0.642*** -0.024*** Significance levels: *** p <1%, ** p <5%, * p <10% Note: Industry size measure abbreviations: non-employer establishments (Non-Emp.); employer establishments (Est.); total employment (Emp.). Regressors in Table A.3 are included in same respective regressions as Table A.4. REGION : Volume 9, Number 1, 2022 24 Table A.4: Poisson Marginal Effects for Locational Determinants of T&W Sectors Process, Physical Industry (NAICS) General Warehousing Distribution, & Logistics Pipeline Transportation Support Activities for Road General Freight Trucking (4841) & Storage (493110) Consulting (541614) (486) Transportation (488490) Size Measure Est. Emp. Est. Emp. Non-Emp. Est. Emp. Non-Emp. Est. Emp. Non-Emp. Est. Emp. Location Quotients TW 1.102*** 66.41*** -2.497*** -72.76*** 0.049*** 1.195*** 0.035 4.571*** 5.500*** 0.247*** 15.78*** 82.31*** 3.767*** Finance -0.305*** 3.089*** -0.172*** -4.355*** -0.263*** -1.430*** 0.001 -0.991*** -0.966*** 0.033 4.882*** 14.45*** -0.394** Real Estate -0.283*** -48.51*** 0.009 -1.288*** 0.051*** 2.287*** 0.223*** 2.329*** -0.282* 0.047 -4.488*** 5.991*** 0.117 Professional and tech -0.552*** -68.38*** 1.850*** 55.11*** -0.073** 2.016*** -0.056 -1.453*** -3.574*** -0.037 -4.804*** -80.04*** -3.565*** Education -0.181*** -4.207*** -0.077* -2.064*** -0.055** -2.163*** -0.117*** -1.191*** 0.219*** -0.012 -8.694*** -9.805*** -1.545*** Health services -0.776*** -33.54*** -0.825*** -13.27*** 0.126*** 1.005*** 0.111** -6.506*** -3.118*** -0.031 -19.85*** -73.12*** -2.183*** Art and recreation -0.087 9.509*** 0.032 -3.366*** 0.017 -0.987*** -0.052** 0.213 -0.663*** 0.010 -3.518*** -12.62*** -0.548*** Accommodation & Food -0.237** -24.41*** -0.190** -2.526*** -0.146*** -3.186*** -0.098* -1.369*** 2.705*** 0.017 -20.45*** -118.9*** -5.207*** Mining and gas -0.082*** -6.507*** 0.013** 0.167*** 0.003** 0.230*** 0.021*** 0.087*** 0.103*** 0.010*** -0.542*** -3.890*** -0.034** Construction -0.625*** -45.65*** -0.268*** -3.843*** 0.022 2.388*** 0.173*** -1.003*** -2.726*** -0.045 -0.880** -11.95*** 0.235 Retail -0.683*** 10.30*** 0.061 -0.839 0.056 -5.767*** -0.075 1.937*** -1.395*** 0.074 12.20*** -45.02*** 1.097*** Regional FE YES YES YES YES YES YES YES YES YES YES YES YES YES n3,063 3,063 3,063 3,063 3,063 3,063 3,063 3,063 3,063 3,063 3,063 3,063 3,063 Significance levels: *** p <1%, ** p <5%, * p <10% Note: The exact observation count is suppressed by U.S. Census Bureau disclosure review process. We use all counties in the continental U.S. in 2014. Industry size measure abbreviations: non-employer establishments (Non-Emp.); employer establishments (Est.); total employment (Emp.). TW LQs are calculated on all of NAICS 48, 49, and 541614, except for the respective NAICS under consideration. TW LQs are calculated on all of NAICS 48, 49, and 541614, minus the respective NAICS under consideration. Regressors in Table A.4 are included in same respective regressions as Table A.3. Regional fixed effects (FE) based on Census regions. REGION : Volume 9, Number 1, 2022