Abstract
We analyze the conditions driving the organization of the territory near airports by studying the distribution of economic activities. We consider how commercial firms, service operators, and consumers compete for land. The theoretical setting identifies an aerotropolis (airport city) as a land equilibrium outcome characterized by the following spatial sequence: services area, commercial area, residential area. Using data on the distribution of establishments in the United States, we analyze the existence and determinants of aeropolitan configurations. Estimations performed with parametric methods detect some interesting dynamic patterns affecting the density and distribution of activities around selected U.S. airports. (JEL R12, R15)
I. INTRODUCTION
Logistics are becoming an increasingly important issue because firms are in search of flexibility. Speed and agility are already as important as price and quality in the strategy of firms that adopt the just-in-time good-supplying system. Firms choose their location to enhance their accessibility to markets. No longer are logistics seen as costs to be minimized, but as value-added activities in firms’ supply chain that need to be optimized.
Consequently, fast delivery is a key element (see Leinbach and Bowen 2004 for empirical evidence). In this context, airports are seen (especially by e-tailers) as a new kind of central business district (CBD) with enough capacity to leverage air commerce into high profits. In that spirit, Kasarda (2000) introduces for the first time the idea of an aerotropolis (airport city), namely, a large industrial area characterized by a high concentration of logistic facilities and commercial activities near specific airports.1 Arend, Bruns, and McCurry (2004) suggest that aerotropolises may extend up to 32 kilometers (20 miles), including a number of activities and infrastructures such as retail and distribution centers, light industrial parks, office and research parks, districts zoned for specific purposes, foreign trade zones, entertainment and conference facilities, and even residential development that contributes substantially to the competitiveness of firms in the area.2 Furthermore, empirical evidence also emphasizes that aerotropolises are also an important source of employment for the local territory; as we describe extensively in the subsection “Selected Airports: Some Descriptive Features,” they are able to account for about 10% to 30% of the employment in a region.
Our paper analyzes the conditions driving the organization of the territory near airports by studying the distribution of economic activities. More specifically, we study the existence and determinants of aeropolitan configurations.
The importance of airports as economic fuel for their surrounding areas has been underlined in the urban literature. McDonald and McMillen (2000) discuss the centripetal force of Chicago’s O’Hare International Airport for industrial and commercial activities, and Cohen and Morrison Paul (2003) evaluate the spillovers entailed by own-state airport infrastructure as a device for lowering manufacturing costs. By contrast, rather than focusing on the economic impact of airports, our analysis deals with the spatial organization of activities near airports and its evolution over time.
Starting from the setting of the location of divisible activities developed by von Thünen (1826), various models have tried to explain the configuration of the space in which house-holds commute to the CBD and form urban agglomerations around it.3 As pointed out by Fujita and Thisse (2013), the novelty of von Thünen’s work is that he introduced the notion of a bid-rent function: land is not homo-geneous and is assigned to the highest bidder.4 A piece of land at a particular location can be associated with a commodity whose price is not fixed by market supply and demand. Extending the von Thünian agricultural model to an urban context, Alonso (1964) suggests the rent each agent can bid at each location is explained by the savings in transportation costs with respect to a more distant site. Hence, land gives rise to a spatial heterogeneity, and agents stop bidding for the most distant land since no further savings can be enjoyed.5
In that spirit, we consider how commercial firms, service operators, and consumers compete for land. We define service operators as all firms developing activities associated with the use of an airport. Service operators provide a number of complementary services to commercial firms (e.g., freighter docks, bonded warehouse, mechanical handling, refrigerated storage, fresh meat inspection, mortuary, animal quarantine, livestock handling, health officials, security for valuables, decompression chamber, express/courier center, and equipment for dangerous and radioactive goods and large or heavy cargo).6
Our theoretical setting is simple. A group of service operators supplies a range of services near an airport, and commercial firms need to settle close enough to enjoy them. The spatial concentration of these services in the proximity of an airport allows firms to benefit from an easy access to many facilities. The two types of firms compete with consumers, who also aim at settling close to the airport. The theoretical setting models land competition across agents and analyzes the formation of aerotropolises. An aerotropolis requires a nondegenerated land equilibrium in which both service operators and commercial firms assign a higher value than consumers to land plots located in the proximity of the airport, and it is characterized by the following spatial sequence: services area, commercial area, residential area (as introduced by Kasarda 2000). Among other things, the land demand elasticity of service operators, commercial firms, and consumers determines the level of their respective bid rent functions (i.e., their willingness to pay).
The second part of the study proposes an empirical application to identify the presence of aeropolitan areas in the United States. Kasarda and Lindsay (2011) propose a qualitative identification by detailing the distinguishing features of four cases in which aerotropolises could be observed. Our contribution consists of suggesting a quantitative (econometric) strategy to detect the existence of these spatial structures. In particular, we are interested in assessing the conditions that make an airport a true center of attraction (i.e., a new CBD), which gives rise to the formation of aerotropolises’ structures. Information on firms’ distribution in the United States is collected to bring our model to the data: we collect part of the county business patterns data released by the U.S. Census Bureau, and we focus on the years 2000 and 2010. First, we extract the data concerning firms’ spatial distribution (by ZIP code), limiting our interest to a finite number of North American Industry Classification System (NAICS) codes to identify the groups of commercial and service activities. Then we select some representative U.S. airports (Memphis, Louisville, Los Angeles, and Newark) to study the land organization in their surroundings. Our approach owes to McMillen (2004a) the idea to think of airports as a source of formation of subcenters that could compete with the city center for the local land organization for productive activities. Estimations performed with parametric methods detect a few interesting evolutionary patterns of the density distribution of the two groups of activities for our sample of airports. The distance from/to the airport mostly drives the land organization, but the creation of an aeropolitan configuration is associated with the existence of some specific conditions. The degree of attractiveness of the airport (measured by the density gradient) is dynamic over time and is stronger in the proximity of the more cargo-oriented airports. We find evidence for classifying Memphis as a reinforcing aerotropolis, Louisville as a growing aerotropolis, Los Angeles as a declining aerotropolis, and Newark as an aborted aerotropolis.
II. THE THEORETICAL MODEL
We introduce a simple theoretical framework to guide the interpretation of the results of our empirical exercise. Our model builds on the von Thü nian orthodox framework described by Fujita and Thisse (2013). Space is represented by the real line X = (−∞,∞) with the CBD lying at the origin. The CBD is an exogenous fixed point that corresponds to the airport terminals.7 We adopt a broad definition of CBD, so that it comprises the airport terminals and the surrounding space up to the noise contour line (so that the severe-noise area does not affect the agents’ location choice).8 We define any spatial distance from/ to it as x∈X, with x> 0.
In the wake of the housing problem framework discussed by Glaeser (2008), we model the case of heterogeneous agents. We consider three types of agents competing for land who aim at settling as close as possible to the CBD: (1) a continuum of identical service operators with density na(x) ≥ 0 at x∈X; (2) a continuum of identical firms with density ni(x) ≥ 0 at x∈X; and (3) a continuum of consumers with density nc(x)≥0 at x∈X, where the subscripts “a,” “i,” and “c” denote service operators, commercial firms, and consumers, respectively.
As in the classical tradition, an absentee landlord is assumed. Land is finite and the total area occupied by service operators, firms, and consumers at each x∈X is fixed and normalized to 1 (as done by Cavailhès et al. 2004), that is,
[1]
where Sa(x), Si(x), and Sc(x) stand for the sizes of land plots, and na(x)Sa(x), ni(x)Si(x), and nc(x)Sc(x) denote the total amount of land being used by each type of agent at a location x∈X. The equilibrium is computed by considering any point in the available space at which each agent exhausts her income and maximizes her profits or utility. There is competition for land because each type of agent nurtures a particular interest in settling as close as possible to the CBD. More precisely, service operators need to be close to the airport terminals to provide commercial firms with a full range of services; commercial firms also want to be close to the terminals to have their merchandise delivered as fast as possible; finally, consumers are also attracted by land plots around the airport because we assume a complete information setting in which consumers know that the firms that can hire them are searching for settling as close as possible to the CBD. Any location entails some rental costs and some transportation costs (which increase with distance).
Consumers-workers
We consider the existence of a group of consumers-workers (they supply labor to service operators). Their utility function relies on the consumption of two goods: leisure and land.9 Their source of revenue is the wage they earn for the time they devote to labor. Each consumer-worker’s available time at x∈X is fixed and normalized to 1. Therefore, H(x)=1−L(x), where L(x) stands for labor and H(x) denotes leisure. More precisely, leisure and land consumption report utility to consumer-workers in the following way (á la Cavailhès et al. 2004):
[2]
with 0<α<1. We assume that consumers-workers are hired by service operators and receive a fixed and exogenous wage (w) per unit of labor, which they use to cover land rental and transportation expenses. As explained above, consumers-workers are also interested in settling as close as possible to the CBD, whose access entails incurring transportation costs.10 Thus, consumers-workers’ budget constraint is given by
[3]
where the price of leisure is equal to unity and Rc(x) is either the rent or the purchase price per unit of land paid by consumers-workers settled at distance x. Furthermore, transportation costs to the CBD from distance x are equal to tx, where t> 0 is the cost per unit of distance.11 Consumers-workers choose L(x) and Sc(x) to maximize their utility subject to their budget constraint (which is assumed to be binding). From equations [2] and [3], we can solve a constrained maximization problem that yields consumers-workers’ optimal labor supply and land plot demand:
[4]
and
[5]
where w>tx is assumed to hold. Naturally, L(x)* and 
decrease with w and Rc(x), respectively.12 An increase in transportation costs creates incentives for consumers-workers to increase their labor supply (i.e., working time) to afford them and reduces their land plot demand. Since consumers-workers’ indirect utility function is
their bid-rent function, that is, the highest price they are willing to pay for a unit of land at x∈X, becomes
[6]
which is increasing with w (since consumers-workers have a higher income) and decreasing and strictly convex with respect to x.
Service Operators
The activity of the other two groups of agents is strongly connected. Commercial firms need to deliver their production abroad through the airport and service operators are the one in charge of accomplishing this task.13 Namely, service operators provide commercial firms with a full range of services. The action of delivering merchandise from firms’ premises to the airport (i.e., the CBD) implies the existence of transportation costs to be taken into consideration. For the sake of simplicity, we consider that transportation costs are partially assumed by both commercial firms and service operators and are proportional to their distance from the airport.
The activity of service operators is provided by using land and all labor supplied by consumers-workers at each location x∈X. Their production function is modeled as the following Cobb-Douglas function with constant returns to scale:
[7]
with 0 <γ<1. Service operators (settled at x∈X) sell their services to commercial firms at a price pa (net of production costs). Revenues earned by service operators are discounted by their transportation costs and equal (pa/tx)Ya(x). Service operators’ production costs comprise labor and land rental expenses and are equal to wL(x)+Ra(x)Sa(x). Therefore, service operators’ profits are
[8]
Service operators choose Sa(x) to maximize profits.14 Using equations [4] and [7], the first-order condition yields
[9]
with w>tx. The plot size increases with consumers-workers’ labor supply and with service operators’ marginal revenue (pa/tx), whereas it decreases with the rental price.15 Competition for land is assumed to extract all profits (zero-profit condition), yielding
[10]
where 
is the bid-rent function for service operators, which decreases with labor’s unit cost w and increases with marginal revenue. In addition, 
is decreasing and strictly convex with respect to x.
Commercial Firms
Finally, commercial firms deliver goods through the airport and by using the services supplied by service operators. Therefore, the activity of commercial firms makes use of land and the services supplied by service operators as inputs. Their production function is modeled as the following Cobb-Douglas function with constant returns to scale:
[11]
with 0<δ<1. Therefore, profits for commercial firms settled at x are
[12]
where revenues are assumed to decrease with transportation costs (tx), and their production costs include the payment for the services provided by service operators (i.e., paYa(x)) and land rental expenses (i.e., Ri(x)Si(x)).
Commercial firms choose Si(x) to maximize profits.16 Using equations [7] and [11], the first-order condition yields
[13]
The plot size increases with service operators’ supply and with commercial firms’ marginal revenue (pi/tx), whereas it decreases with the rental price (the value of 
can be computed by plugging equation [10] into equation [9]).17 As in the case of service operators, competition for land is assumed to extract all profits (zero-profit condition), yielding
[14]
where 
is the bid-rent function for commercial firms, which decreases with the unit price charged by service operators pa and increases with marginal revenue. Finally, 
is decreasing and strictly convex with respect to x.
Land Equilibrium
In the spirit of the von Thünian tradition, the three agents compete for land with an auction mechanism. The land equilibrium is driven by the value each type of agent pegs to a land plot at each possible location x∈X, which is given by their bid-rent functions (i.e., equations [6], [10], and [14]). At the equilibrium, we observe that 
, in other words, R*(x) is the upper envelope of the bid-rent curves and land is assigned to the highest bidder at each location x∈X. As a consequence, land is specialized after the bidding process and no land is vacant (as long as bid-rent functions are positive). Looking at equation [1], it is easy to check that 
holds in a residential area. Equivalently, we observe 
in a services area, and 
in a commercial area.18
From inspection of equations [6], [10], and [14], it can be observed that bid-rent functions are continuous, monotonic, and downward sloping with respect to the distance from the CBD, because agents associate a higher value with the land plots located closer to the airport terminals (i.e., their reservation rent decreases with distance).19 The lemma that follows studies the effect of distance on equilibrium land plots and densities.
Lemma 1
Equilibrium land plots increase with distance, and equilibrium densities decrease with distance.
Equilibrium land plots increase with distance from/to the CBD as the net result of two opposing effects, which can be observed by inspection of equations [5], [9], and [13]. There is a direct negative effect of x on land plot demand functions and an indirect positive effect of x through rental prices 
, 
, and 
, and the indirect effect overcomes the direct effect.20 Therefore, densities after the bidding process decrease with x, a result that clearly matches the empirical evidence.
The lemmas that follow perform a comparative-static analysis that studies the impact of the parameters of the model on 
, 
, and 
. First,we focus on the effect of α, γ, and δ, which determine the land demand elasticity of consumers-workers, service operators, and commercial firms, respectively.
Lemma 2
As α increases, 
moves upward and land demand becomes more inelastic for consumers-workers. In the same way, an increase in γ shifts 
upward, turning service operators’ land demand more inelastic. Finally, 
rises with an increase in δ, and land demand becomes more inelastic for commercial firms.
As a consequence, the amount of land assigned to each type of agent in equilibrium depends on the relative value of these parameters. The upshot is that there may be cases in which residential areas have a strong influence (i.e., when cities are important) and cases in which either commercial firms or service operators prevail as a consequence of the relative magnitude of their land demand elasticities.21 The remaining comparative-static effects, which are as expected, can be easily derived by inspection of equations [6], [10], and [14] and are summarized in the Lemma 3.
Lemma 3.
, and 
fall with an increase in the unit transportation cost (t). An increase in the price (pa) increases the revenues for service operators and the costs for commercial firms, and, therefore, 
increases and 
decreases. An increase in the wage (w) increases the revenues for consumers-workers and the costs for service operators, and, therefore, 
increases and 
decreases. Finally, 
decreases with V and 
increases with pi.
In line with the idea suggested by Kasarda (2000) and Kasarda and Lindsay (2011), an aerotropolis appears when the spatial sequence services area–commercial area–residential area arises as the land equilibrium outcome, in a way made clear in Figure 1.22
The Aerotropolis Land Equilibrium
This equilibrium outcome occurs for a certain parameter constellation. Given the stylized nature of the model, parameter choices in Figure 1 are necessarily arbitrary, and the analysis is therefore not exhaustive. However, it reveals some interesting insights that are in line with the empirical evidence we will discuss in Section III. Let pa=2 and pi=3, so that the price charged by commercial firms is higher than the price they pay to service operators. Additionally, let t = 1 and w = 2, so that consumers-workers’ income can cover land rent and transportation expenses for sufficiently low distances (see equations [3] and [4]). The assumption w>tx implies the upper bound for distance x<2.23 Further land plots remain vacant since no agent is willing to make a positive bid to occupy them. Given that 
decreases with consumers-workers’ indirect utility, we assume V = 0.2, which allows for active consumers-workers in the bidding process. Finally, land demand elasticities are α=0.5,γ =0.6,andδ=0.5(the comparative-static effects of the elasticities described in Lemma 2 are also shown in the figure). The following proposition characterizes an aerotropolis land configuration in the light of the previous analysis.24
Proposition 1
An aerotropolis requires a nondegenerated land equilibrium in which both firm types assign a higher value than consumers-workers to land plots located in the proximity of the airport. In addition, the land demand elasticity of service operators and commercial firms has to be moderate so that both of them are active in the bidding process.
This result is illustrated by the empirical evidence, as will be shown in the next section.
III. EMPIRICAL ANALYSIS
We determine from our theoretical framework that the land organization in proximity of an airport depends on the interplay among service operators, commercial firms, and consumers-workers. An aerotropolis appears when the spatial sequence of services area, commercial area, residential area arises as the land equilibrium outcome. In this section, we focus on four U.S. case studies to analyze the presence and evolution of aeropolitan configurations.
In their book, Kasarda and Lindsay (2011) assess the importance of airports on firms’ production and distribution chain and their subsequent impact on the organization of activities around them. They provide a general overview of interesting case studies across the world with a clear focus on the United States. More precisely, they adopt a dynamic perspective to identify four cases in which aerotropolises could be observed:25 (1) the reinforcing aerotropolis, an already existing aerotropolis that experiences a self-reinforcing process yielding a concentration of regional business activities; (2) the growing aerotropolis, an area with an unclear land organization that progressively adopts an aerotropolis-type configuration; (3) the declining aerotropolis, an existing aerotropolis that progressively loses importance due to insufficient infrastructure and available land to expand the airport; and (4) the aborted aerotropolis, an area where an aerotropolis is not formed despite displaying all the relevant features to become an aerotropolis (e.g., available space, good infrastructure, strategic position for land deliveries, culture of just-in-time practices), due to the attraction of a city-downtown preventing any polycentric spatial organization.
Therefore, our empirical strategy consists of selecting a representative airport for each of the aforementioned cases, following Kasarda and Lindsay (2011), and then providing a quantitative assessment of their spatial dynamics. The selected airports are Memphis International Airport (MEM) in Tennessee (Case 1), Louisville International Airport (SDF) in Kentucky (Case 2), Los Angeles International Airport (LAX) in California (Case 3), and Newark Liberty Airport (EWR) in New Jersey (Case 4). All these airports constitute a true reference for the economic activity in their surroundings.
In practical terms, our econometric exercise aims at evaluating airport attractiveness and its evolution over time. We selected a sample using data from the U.S. business census26 corresponding to 2000 and 2010. According to the availability of data, we extracted information concerning establishments belonging to industry sectors (i.e., commercial firms) and service sectors (i.e., service operators).27 Available statistics would also have allowed us to refine the data by providing a finer sectorial organization according to the NAICS classification, but because of our need to focus on the spatial dimension of the sample, we prefer to avoid dealing with too much detailed data for obtaining robust estimation results. Beyond the information about the sector of activity for each firm, our data include the firm’s ZIP code and a rough measure of firm size.
We measure airport attractiveness via the variation of the coefficient associated with an indicator of the distance from/to the airport. Exploiting a parametric estimation method, we regress the gross density of each of the two groups of firms (in each specific ZIP code near the airport) against the distance from/to the airport,28 controlling for the proximity to the city center and other relevant infrastructure.
Selected Airports: Some Descriptive Features
Our empirical exercise deals with the information associated with the spatial organization of the land in the proximity of MEM, SDF, LAX, and EWR. According to our records, this sample of airports does not present important differences in the size of airport runways and facilities, but instead in the whole airport ground (total land): MEM ranks first with 1,668.13 ha, LAX follows with 1,408.08 ha, then SDF with 817.13 ha, and finally EWR with 729.43 ha.29 MEM and SDF are air-express mega-hubs since they are the base of FedEx and UPS, respectively. The e-tailers that normally operate in partnership with FedEx and UPS therefore have strong incentives to settle in their proximity.30 Data released by the Airports Council International rank MEM as the first and the second worldwide cargo airport in 2002 and 2010,31 respectively, LAX was fourth and thirteenth for the same years, SDF was twelfth and tenth, and EWR was eighteenth and twenty-third.
MEM is predominantly a cargo airport.32 In the past, the city’s location on the Mississippi River made Memphis an ideal place for commerce and shipping. Now, airport activities have taken over the historical transportation hubs. The airport in Memphis is often labeled as an aerotropolis because it has been the headquarters for FedEx since 1973. Its impact on the territory is impressive. In 2009, the airport generated around 220,000 jobs, accounting for 34% of jobs in the Memphis area, and its (estimated) total annual output approaches $29 billion. The presence of the airport has also had an impact on the development of the road and rail transportation system: the Memphis area is actually served by several freight railroads and two interstate highways (I40 and I55) plus two other interstate highways under construction.
SDF has been increasing its importance in the U.S. landscape as a low-fare airport since the mid-1990s.33 At that time, the airport operated as a principal hub for Southwest Airlines. Although the airport has been one of the operation bases for UPS since 1982, its cargo activity improved dramatically in the 2000s with the creation of the UPS Worldport for worldwide deliveries. Since then, the airport has generated an important economic impact. In 2011, it supported around 65,000 jobs that represent roughly 10% of all jobs in the corresponding Metropolitan Statistical Areas (MSAs). Its business expenditures represent more than $7 billion, which generate around $300 million in state and local taxes. Louisville has always served as an important crossroads in the U.S. transportation system: it lies at the crossing point of three major interstate highways (I64, I65, and I71) and along one of the major U.S. railway connections (the Louisville-Portland Railroad). Thus, the airport activities have come to complement these other existing transportation facilities. In this sense, firm location decisions in the area are the result of a clear selection process based on competition among the different transportation options.
LAX was originally known as Mines Field and has been used for general aviation since 1928, with intense industrial activity associated with the aviation services up to the mid-1980s. At that time, according to Giuliano and Small (1991), the airport area was the fifth regional subcenter (for employment), and the territorial distribution of activities around it displayed a typical mixed industrial clustering feature but was not heavily service oriented (as it could be in downtown Los Angeles). Kasarda and Lindsay (2011) reveal that the decline of LAX as a cargo hub is related to a lack of proper infrastructure to meet the demand of different industrial operators and to a constant delay in the distribution of incoming merchandise (stocked in the docks). Statistics of freight deliveries provide further evidence of underinvestment in LAX in a highly competitive environment: the airport managed about 17,771,907 tons of deliveries in 2002, and about 19,070,000 tons of goods transited that airport in 2006. However, that quantity fell to 17,950,000 in 2012, while other airports did not record such an important contraction.34 In any case, the airport is still an important source of revenue and employment in Southern California: 1 of 20 jobs of the region is attributable to LAX, and about 158,000 jobs can be associated with the airport activity in Los Angeles, a number that rises to 408,000 when considering the overall region.35
Finally, EWR is a strong passenger-oriented airport.36 It is located in one of the most densely populated areas in the United States and is near a complex rail and road transportation system. EWR records a significant passenger traffic flow, approaching 35 million passengers in 2011. Concerning the economic impact of the airport activity in its metropolitan region, the results are quite impressive: it keeps around 143,000 jobs (in 2011) of which 20,000 are on the airport premises, accounting for about 30% of the overall jobs created by airport activities in the region (and 4% of the overall jobs in New Jersey). In terms of business activities, the value of its total output represents a bit more than $20 billion in 2011 (again, about 30% of the overall activity in the region).
Econometric Strategy
Our goal is to assess to what extent the airport attracts each type of business and then track the potential variation of that attractiveness across time. Our empirical strategy is straightforward: First, we select a sample of representative sectors capturing the activity of each group of agents (i.e., service operators and commercial firms), and we define their geographical location by looking at the establishment density per unit of space. Then we estimate airport attractiveness, controlling for the distance from/to the city center, the nearest main road (usually an interstate highway), and the nearest railroad.
The idea is to quantify how the real attractiveness of the airport (as a business district) affects firms’ location decisions, even in the presence of other potential alternative transportation infrastructure. Therefore, if the coefficient associated with the distance from/to the airport is statistically significant after controlling for all the other potential effects, we can conclude that the airport is the key element driving firms’ location decisions.37 Furthermore, we also focus on the dynamics of the attractiveness of the airport by looking at the way in which the evolution of the establishment density (by group of activity, according to the classification presented in Appendix B) has been affected by the distance from/to the airport across time.
Our empirical strategy is built on the theoretical and empirical contributions that link the bid-rent function to the density functions in a classical von-Thünen-type framework (see, e.g., McDonald and McMillen 2010). Using the land equilibrium result presented in Proposition 1 as a reference, we are able to identify the land specialization. Our theoretical results favor the adoption of this empirical strategy because the land specialization observed in equilibrium (for each type of agent) takes the form of a density function (for each group of agents) that is dependent on the distance from/to the CBD (as shown in the subsection Land Equilibrium).
The empirical literature exactly tackles this problem by estimating different specifications of density functions. We depart from the traditional linearized negative exponential density to regress 2000 and 2010 densities on distance to the airport, while controlling for proximity to the central city, the nearest main road, and the nearest railroad (as done by Garcia-López 2010 and Garcia-López, Holl, and Viladecans-Marsal 2015).
More precisely, for either commercial firms or service operators, we measure the influence of the distance from/to airports on the intra-metropolitan location of establishments by estimating the following type of density function:
[15]
where ln(Dih) is the natural logarithm of the establishment density (namely, establishment per hectare) for location i, in sector h, dair,i is the distance from location i to the airport, dcc,i is the distance to the city center, droad,i is the distance to the closest interstate highway, drail,i is the distance to the nearest railway, and is the error term. Distances are in kilometers. We estimate the previous equation independently for 2000 and 2010 and for each airport-city.38
The estimated coefficients β1, β2, β3, and β4 are known as density gradients. When they are negative, each one measures the rate at which the density falls relatively to the distance to the airport, the central city, the nearest main road, and the nearest railroad, respectively. In this respect, an aerotropolis structure requires that service operators settle closer to the airport than commercial firms; that is, β1 for service operators should show an absolute value larger than that for commercial firms.
We estimate equation [15] by ordinary least squares. In order to correct possible problems of heteroskedasticity and spatial correlation in our cross-section samples, robust standard errors are clustered by county subdivision.
We study the influence of MEM, SDF, LAX, and EWR on the urban spatial structure of their metropolitan areas (i.e., Memphis, Tennessee–Mississippi–Arkansas; Louisville, Kentucky–Indiana; Los Angeles, California; and New York, New York–New Jersey–Pennsylvania). ZIP codes are our unit of observation: our sample includes 80, 112, 359, and 881 ZIP codes in the areas of Memphis, Louisville, Los Angeles, and New York, respectively.
We use establishment data from the 2000 and 2010 ZIP Business Patterns (ZBP) from the U.S. Census Bureau.39 In particular, we focus on six NAICS industries: 31 Manufacturing, 42 Wholesale trade, 44 Retail trade, 48 Transportation and warehousing, 55 Management of companies and enterprises, and 56 Administration support and other services. Establishments belonging to the first three industries are considered commercial firms, whereas establishments belonging to the last three industries fall under the category of service operators.
In order to use GIS software and obtain additional data, we match ZIP codes from ZBP with ZIP codes from ZCTA maps.40 ZCTAs are statistical geographical units introduced by the U.S. Census Bureau that, in practice, constitute a generalized representation of the U.S. Postal Service ZIP code service areas. Each ZCTA is labeled with the most frequently occurring ZIP code found in the area. Using geography from the year 2000, ZCTA maps provide data on land area in square meters.
Our dependent variable is computed as the number of establishments per hectare. Using ZCTA maps, we obtain tract centroid coordinates and can compute distances in a straight line from each centroid to the airport ZIP code centroid, to the city center ZIP code centroid, and to the nearest railway and interstate highway segment.
We run econometric estimations of equation [15] for each airport in a separate way (see Tables 1–4).41 The four tables share the same format. Columns (1)–(4) show results for year 2000 and columns (5)–(8) for year 2010.
All tables report unconditional results (without control variables) in columns (1) and (5), and gradual conditional results in columns (2) and (6) (adding distance to the central city), columns (3) and (7) (adding distance to the nearest main road), and columns (4) and (8) (adding distance to the nearest railroad). Looking at the differences between the coefficient estimates, we can assess the importance-power of proximity to the airport when we compare the adjusted R2 of unconditional versus conditional results.
Finally, we rank the sectors according to their proximity to the airport using the magnitude of the full conditional estimated coefficients of dair,i. In column (9) we refer to the estimates |β2000| of column (4), while in column (10) we refer to |β2010| of column (8).
This exercise is twofold. On one hand, we are able to sort the cities that follow the aerotropolis configuration in 2000 and 2010, respectively. On the other hand, we control for the evolution of the importance of the proximity from/to the airport in shaping the land organization by comparing the statistically significant coefficient |β2000| versus the statistically significant |β2010|. This second result is gathered in column (11), and we can use this analysis to point out the temporal trends of the spatial configuration.
MEM estimations in Table 1 depict a picture of an aerotropolis land structure. Overall, service operators generally settle closer to the airport premises than commercial firms. The variation in the distribution of the log-density of establishments across time does not present a pattern of relevant variation, even if the magnitude of the coefficient dair,i overall drops. In this respect, the land organization surrounding this airport remains unchanged across time, unveiling an intertemporal consolidation of the aerotropolis structure, that is, a type of reinforcing aerotropolis. This implies that service operators tend to consolidate their positions close to airport terminals and, eventually, organize more efficiently across the territory. In contrast, commercial firms tend to settle farther from the airport premises.
Intrametropolitan Distribution of Establishments and Proximity to Airport: Memphis
Focusing on SDF, the estimations in Table 2 confirm that the airport is becoming a true center of attraction for business activity. Although the magnitude of the estimated coefficient declines with the introduction of further controls (namely, the distances from/to the nearest highway and railway, beyond the distance to the city center), the coefficient associated with dair,i is always highly statistically significant. The land organization follows an interesting dynamics pattern: focusing on the preferred and most constrained estimations (columns (4) and (8)), the transportation sector displays a stronger attractiveness to the airport, and the management sector follows. Furthermore, the growth of the establishment density (column (11)) in service sectors unveils that those establishments progressively clustered around the airport (from 2000 to 2010), while manufacturing sectors are progressively settled far apart. Therefore, the territorial dynamics in the area surrounding SDF presents an interesting case of a growing aerotropolis, which is characterized by a replacement pattern in which service operators replace commercial firms by taking their land plots in the airport premises. When viewed along an intertemporal horizon, the reading of our results is straightforward: the territorial organization in the area surrounding the SDF airport was extremely hybrid in 2000, but it experienced an interesting change over time by converging toward an aerotropolistype structure.42
Intrametropolitan Distribution of Establishments and Proximity to Airport: Louisville
Table 3 refers to LAX, and our estimations confirm the progressive weakening of this aerotropolis-organized space, in other words, it is a declining aerotropolis. Columns (9)–(11) indicate that there is a hybrid land organization structure: locations of service operators tend to merge with those of commercial firms. Nevertheless, transportation services are progressively settling far apart from the airport, whereas manufacturing activities are staying unchanged or progressively approaching (as retail trade). Another evidence of this transforming area is indicated by the loss of statistical significance of the unconditional estimated coefficients of the other sectors, especially the ones belonging to the category of service operators. Although our data cannot identify the causes of this decay, the inclusion of the distance from/to downtown Los Angeles reduces the significance of our estimated coefficients, suggesting that the downtown area becomes more attractive. It seems that, in the presence of poor infrastructure, the urban effect dominates the airport effect, and, in line with Kasarda and Lindsay (2011), this is one of the potential causes of the decline of a few aerotropolis structures in the United States.
Intrametropolitan Distribution of Establishments and Proximity to Airport: Los Angeles
Finally, estimations referring to EWR in Table 4 confirm the importance of the city center (New York City) and the other controls as a point of attraction: when the estimations include these variables, the magnitude of the coefficient of the distance from/to the airport almost always drops or loses most of its statistical significance. As for service operators, only firms in the transportation sector retain a clear interest in settling in the proximity of the airport premises; in all the other cases, we cannot appreciate a truly airport attractiveness effect across time. Furthermore, for wholesale trade (belonging to commercial firms) and administrative support (belonging to service operators) the estimated coefficient of dair,i is not statistically significant at all. This is a hybrid case and thus an aborted aerotropolis, because although the airport exerts a relatively important influence in shaping the structure of the territory, other factors (the proximity to the city center, for instance) play a key role in firms’ location decisions.
To test the robustness of the previous results, we run estimations of equation [15] by exploiting employment data instead of the data of establishments. The results of these estimations are shown in Appendix Tables D1–D4. Overall, estimations with employment data present a lower statistical significance than the estimations with establishment data. Employment data are less precise in capturing the location effects by sector since they do not fully replicate the ZCTA geographical units. Employment data by place of work at the tract level for 2000 come from the 2000 Census Transportation Planning Products (CTPP). For 2010, we rely on the CTPP 5-year small area, which is based on the 2006–2010 American Community Survey (ACS) estimates that provide employment counts at the tract level. We use the 2010 Census Tract Relationship Files to match the 2010 data with the 2000 geography tracts. As a result, our employment estimates for Memphis, Louisville, Los Angeles, and New York use 285, 267, 2,632, and 4,519 observations (tracts), respectively. Because of the statistical composition of employment data, we are not able to deal with the three groups of service operators, as in the case of establishment data. However, the statistically significant results with employment data are not different from those in Tables 1–4, even if the magnitude of the estimated coefficients with employment data is smaller than those derived with establishment data. In this sense, the previous conclusions are confirmed.
Intrametropolitan Distribution of Establishments and Proximity to Airport: New York
IV. CONCLUDING REMARKS
The increasing importance of e-tailers leads to airports being considered as a new type of CBD with sufficient capacity to leverage air commerce into high profits. This paper applies current urban theory to the study of the spatial distribution of activities around airports and provides some insights into the formation of aerotropolises, which are defined as large industrial areas characterized by a high concentration of service activities near certain airports.
Land competition around airports takes place among service operators, commercial firms, and consumers-workers. All agents display an interest in settling close to the airport. As the result of a maximization process, agents associate a value with each unit of land. Finally, location is determined by a comparison of each agent’s bid-rent function. An aerotropolis requires a nondegenerated land equilibrium in which both service operators and commercial firms assign a higher value than consumers-workers to land plots located in the proximity of the airport, and it is characterized by the following spatial sequence: services area, commercial area, residential area. The empirical evidence supports these theoretical results.
Focusing on a selected sample of U.S. airports, our econometric exercise allows us to identify the existence of different spatial structures in the proximity of airports and track their evolution over time. Memphis International Airport turns out to be the one that strictly embeds the typical features of an aerotropolis setting, whereas Louisville International Airport presents territorial dynamics showing a convergence toward an aerotropolis-style territorial organization. The aerotropolis located around Los Angeles International Airport loses importance over time as the urban effect dominates the airport effect and no important airport investments are made. Finally, Newark Liberty Airport exerts a relatively important influence in shaping the structure of the territory, but other elements (mainly New York City center) play a determining role in firms’ location decisions.
A direct implication of this type of analysis concerns policy matters. Once the size of the positive effects associated with being close to the airport has been stated, one can consider the impact of public policies supporting the creation of a specific land structure. The economic contribution of air transportation in terms of employment and income has important effects at a regional level. Consequently, regional governments may be interested in trying to implement the required conditions that allow for the formation of aerotropolises. Some policy recommendations would suggest fostering logistical platforms close to cargo airports, as well as promoting and encouraging partnership between firms and service operators. In such a framework, the intense collaboration among agents would guarantee a sufficiently high level of facilities to allow for the existence of aerotropolises. Of course, this issue would also imply studying the extent to which industrial parks are socially desirable and then determining their optimal size. A welfare analysis examining these questions could be an interesting task to undertake in order to extend and apply the main findings presented in this paper.
Another interesting extension of our framework would be to adapt it to study the importance of creating logistic and/or industrial areas in the proximity of transportation hubs such as railway stations or harbors. We have developed an idea by looking at the specific case of an airport, but this type of result can easily be extended to other transportation infrastructure. From a technical viewpoint, such an application would not involve dramatic changes. Even while maintaining the suggested structure and changing the parameters of reference as well as certain interpretations, our conclusions are expected to hold. Nevertheless, there is a general lack of empirical evidence for supporting this type of study. The scarcity of data and statistics is one of the most binding problems for providing a complete analysis of this phenomenon, even in the case of airports. In fact, most of the analysis that remains to be carried out concerns the empirical analysis of those features distinguishing aerotropolis-type configurations from other industrial areas.
Finally, other extensions to undertake a more sophisticated analysis would require, for instance, considering heterogeneous agents within each category or a polycentric model where agents commute to more than one CBD (airport, city center, etc.).
Acknowledgments
We are grateful to the editor, two anonymous referees, H. Overman, and the participants to the NARSC (Ottawa, Canada) and the SAEe (Santander, Spain) conferences for useful comments and suggestions. Financial support from research projects ECO2013-42884-P, ECO2014-52506-R, ECO2014- 52999-R, 2014-SGR327, 2014-SGR1326, 2014-SGR631, XREAP, and XREPP is gratefully acknowledged. All errors are our own responsibility.
APPENDIX A: PROOFS
Proof of Lemma 1
Straightforward. ▪
Proof of Lemma 2
.- .
- . ▪
Proof of Lemma 3
Straightforward. ▪
Proof of Proposition 1
Straightforward. ▪
APPENDIX B: FIRMS SELECTION
Our database is built by extracting information from the U.S. Business Census. We define the sample of firms to be used in this study by focusing on sectors of activity that can easily be identified as industry or service sectors, to match in the best possible way our idea of commercial firms (here, industry) and service operators (here, service). The description for each selected group of sectors, as well as its code according to the NAICS classification system, is provided in Table B1.
The activities included in sectors 55 and 56 deserve a particular attention since they represent a large pro-portion of the service operators included in our sample. Sector 55 (Management of companies and enterprises) comprises (1) establishments that generally hold equity interests in enterprises to control or influence manager decisions, and (2) establishments that manage a company and normally undertake the strategic decision-making role of the company or enterprise. Sector 56 (Administration support and other services) comprises establishments that perform routines to support corporate daily activities such as clerical, office administration, personnel hiring, cleaning, security, or waste disposal services. They specialize in one or more fields and provide their services to clients in a variety of industries and, in some cases, to households.
NAICS Sectors
APPENDIX C: DESCRIPTIVE STATISTICS
In Table C1, we report the statistics for commercial firms and service operators densities in 2000 and 2010 for each of the four selected airports.
Firm Density per Hectare
APPENDIX D: ESTIMATIONS WITH EMPLOYMENT DATA
In order to check the robustness of the results we obtained, we replicate the estimation of equation [15] by using employment data. Selected data are taken from the Census Transportation Planning Products and the American Community Survey. With the dependent variable ln(Dih), we consider the natural logarithm of the employment density (i.e., employment per hectare) for location i in sector h, and we run regressions for 2000 and 2010. Tables D1–D4 show the estimations results. Their structure replicates Tables 1–4 in the main text.
Intrametropolitan Distribution of Employment and Proximity to Airport: Memphis
Intrametropolitan Distribution of Employment and Proximity to Airport: Louisville
Intrametropolitan Distribution of Employment and Proximity to Airport: Los Angeles
Intrametropolitan Distribution of Employment and Proximity to Airport: New York
Footnotes
The authors are, respectively, associate professor, Departament d’Economia and CREIP, Universitat Rovira i Virgili, Reus, Spain; associate professor, Department of Applied Economics, Universitat Autònoma de Barcelona, and, economist, Institut d’Economia de Barcelona, Barcelona, Spain; and associate professor, Department of Applied Economics, Universitat Autònoma de Barcelona, Bellaterra, Spain.
↵1 This type of land organization started becoming more and more visible at the beginning of the 2000s. Kasarda and Lindsay (2011) document this rise and refine the definition of aerotropolis suggested by Kasarda (2000).
↵2 The formation of an aerotropolis is an expression of the self-organizing dynamics of a territory. Another example of this dynamics has been the creation of edge cities in the 1980s. Edge cities are defined as highly dense business and residential centers containing every city function but offering a low-density automobile-oriented land configuration. This is the case of the areas around Route 128 in Massachusetts and the city of Irvine in southern California (Garreau 1991).
↵3 See Fujita and Thisse (2013) for a complete overview of this literature.
↵4 All agents compete for land with an auction mechanism and pay a rent to an absentee landlord once the land auction takes place. Therefore they become actual land owners after the land bidding process.
↵5 Empirical evidence has been presented by Muto (2006).
↵6 See www.azworldairports.com for important nonaviation services provided by the major worldwide airports.
↵7 The force driving land competition (and ultimately, land assignment and land specialization) is the attraction exerted by the CBD. Introducing explicit agglomeration economies in our model would reinforce the attractiveness of land plots located in the proximity of airports. However, this would complicate unnecessarily the analysis without providing any additional insight.
↵8 Alternatively, we could simply argue that in accordance to the findings in the literature, the existing severe-noise area around airports is relatively small and is becoming less and less important (see McMillen 2004b; Federal Aviation Administration 2005). This is due to the recent dramatic gains in aircraft quietness, as reported by both McMillen (2004b) and Brueckner and Girvin (2008). Therefore, for the sake of simplicity, the severe-noise area around the airport terminals is not formally included in the analysis (since this would imply assuming an inverted U-shape bidrent function for consumers-workers, which would complicate the analysis substantially without providing any additional insight).
↵9 Labor enters as a choice variable because it is an input for the service operators’ production function.
↵10 Consumers-workers are interested in reducing transportation costs even if this is done at the expense of higher land rents.
↵11 For the sake of simplicity, we assume identical unit transportation costs t across agent types.
↵12 Consumers-workers’ leisure demand increases with wages. Higher wages yield a lower labor supply.
↵13 We assume that the final consumers-workers of commercial products are settled abroad, and, therefore, we do not model them.
↵14 Alternatively, following Glaeser (2008), the optimization problem can be equally solved by defining a measure of labor intensity
and maximizing πa(x)/Sa(x) with respect to .↵15 It is easy to check that the second-order condition always holds.
↵16 Alternatively, following Glaeser (2008), the optimization problem can_be equally solved by defining a measure of land intensity
and maximizing πi(x)/Si(x) with respect to .↵17 It is easy to check that the second-order condition always holds.
↵18 Note that the precise value of
, and can be computed by plugging equation [6] into equation [5], equation [10] into equation [9], and equation [14] into equation [13], respectively.↵19 Given that bid-rent functions decrease with x, they become negative at a certain distance. Naturally, land becomes vacant when all of them become negative.
↵20 The net effect of x on
, and can be easily computed. More information is available from the authors upon request.↵21 Of course, a more sophisticated analysis would require considering heterogeneous agents within each category and a polycentric model where agents commute to more than one CBD (e.g., airport, city center).
↵22 By computing the crossing point between the bid-rent functions of commercial firms and service operators (i.e., equations [10] and [14]), we get a parametric measure (xia) of the boundary separating both agent types, that is,
This point approaches the CBD as transportation costs increase. The reason is that commercial firms evaluate the opportunity costs derived from settling at a certain distance from the CBD, concluding that their valuation of closer locations to the terminals increases as the costs to access the airport increase.
↵23 It could be argued that the restriction w>tx should concern only consumers-workers, given that it is required to ensure
(see equation [5]). In any case, the aforementioned restriction constitutes a relevant upper bound in the considered land equilibrium configuration (see Figure 1), given that consumers-workers occupy the last land plots before land is vacant.↵24 Note that degenerated land configurations in which a certain agent type obtains no land in the bidding process are not possible in our setting, given that consumers-workers supply labor to service operators and that service operators supply services to commercial firms.
↵25 Although their identification criteria are mostly qualitative, they have the advantage of providing a sensitive and realistic interpretation of some current (and future) industrial agglomerations in selected areas.
↵26 See www.census.gov/econ/cbp/.
↵27 Further information on the composition of the sample by sector is provided in Appendix B.
↵28 The gross density is the ratio between the number of establishments and the total area. In the literature, it is usually selected as a density measure when one needs to embed the idea of competition for land among agents.
↵29 These measures have been obtained by referring to the layer package file (airport hub size) in the GIS software.
↵30 For instance, Barnesandnoble.com, Planetex.com, and Toysrus.com are located close to MEM; whereas Nike.com and Guess.com are close to SDF.
↵31 Data source: Airports Council International (www.aci.aero/).
↵32 See www.mscaa.com/.
↵33 According to the information collected at www. flylouisville.com/.
↵34 Data source: Airports Council International (www.aci.aero/).
↵35 Data source: Los Angeles World Airports, city of Los Angeles (www.lawa.org/).
↵36 Refer to www.panynj.gov/airports/newark-liberty.html/.
↵37 Airport facilities may enhance synergies among firms.
↵38 In Table C1 (Appendix C), we present some descriptive statistics featuring the density of commercial firms and service operators for the four airports in 2000 and 2010. The values extracted from our database suggest that the existing market structure involves local firm competition inside each group of firms, even if the market became more concentrated passing from 2000 to 2010. Unfortunately, we are not able to control for the firm size, and, therefore, we cannot be more precise about this point.
↵39 See www.census.gov/econ/cbp/.
↵40 Both available at www.census.gov/.
↵41 The complete and detailed results of the estimations in Tables 1–4 are available from the authors upon request.
↵42 In this respect, Kasarda and Lindsay (2011) discuss an interesting particular case: the location of the main warehouse of Zappos.com in the proximity of SDF in 2002. This new arrival represented a big push for the creation of an aerotropolis, given that other firms attracted by Zappos’s business moved to the area.
This article requires a subscription to view the full text. If you have a subscription you may use the login form below to view the article. Access to this article can also be purchased.
Log in using your username and password
Purchase access
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.