Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Journal of Transport Geography 83 (2020) 102636

Contents lists available at ScienceDirect

Journal of Transport Geography


journal homepage: www.elsevier.com/locate/jtrangeo

Relative logistics sprawl: Measuring changes in the relative distribution T


from warehouses to logistics businesses and the general population
Sanggyun Kang1
Department of Logistics Research, The Korea Transport Institute, 370 Sicheong-daero, Sejong-si 30147, Republic of Korea

ARTICLE INFO ABSTRACT

Keywords: The logistics sprawl – warehousing and distribution centers relocating farther away from central urban areas to
Logistics sprawl the urban periphery – has been widely examined in multiple global trade gateways. The spatial shifts have been
Relative centrality attributed to the systematic and geographic restructuring of goods production and distribution systems at the
Warehouse distribution global scale. More recently, growing online shopping sales and demand for instant delivery have put tremendous
pressure on online retail businesses to locate order fulfillment centers in urban areas with direct access to the
consumer markets. Has the trend of sprawling warehouses reversed? Have warehouses been relocating closer to
consumer demand? In this research, I use a relative distribution measure to quantify the spatial distribution of
warehouses relative to the logistics sector, the goods movement sector, and the general population over time.
The main data source is ZIP Code Business Patterns from 2003 to 2016 with the study area of the 64 largest US
metropolitan areas. In 2016, warehouses were most closely distributed relative to logistics sectors (38.6 km),
followed by goods movement sectors (41.2 km), and population (43.5 km). From 2003 to 2016, the distance
between warehouses and related businesses has increased, but the extent varies with respect to metropolitan and
warehouse size. The trend of increasing relative distance from warehouses to related businesses stopped since
the economic recession in 2008. Relative distribution measures are significantly correlated with internal ground
shipment ton-kilometers but only in large trade gateways.

1. Introduction use a relative distribution measure that quantifies the distribution of


warehouses relative to the distribution of various levels of freight de-
The purpose of this research is to use a relative distribution measure mand. Results suggest that in 2016 warehouses were most closely dis-
and quantify the relative distance from all warehouses to the logistics tributed relative to logistics sectors (38.6 km), followed by goods
and goods movement sectors, and the general population over time. The movement sectors (41.2 km), and then the general population (43.5 km).
relocation of warehousing and distribution facilities, farther away from From 2003 to 2016, warehouses moved farther away from urban areas,
central urban areas to the urban periphery, has been widely examined in but the sprawling trend has been halted at the national scale since the
multiple global trade gateways. The spatial shifts have been attributed to economic recession in 2008. Lastly, relative distribution measures are
the systematic and geographic restructuring of goods production and significantly correlated with internal ground shipment ton-kilometers but
distribution systems at the global scale to transport large volumes of only in large trade gateways.
goods quickly and cheaply. More recently, the trend of growing online This paper is organized as follows. The following chapter reviews
shopping sales and demand for instant delivery has put tremendous the recent literature on the logistics sprawl and identifies a research
pressure on online retail businesses to adopt more time-saving logistics gap. Section 3 presents the conceptual framework, relative/absolute
practices and locate order fulfillment facilities strategically in locations distribution measures, definitions of related businesses, data, and the
with direct access to the consumer markets. To examine these spatial study area. Section 4 presents results, and in Section 5, this paper closes
shifts, prior studies have used several spatial measures, such as the with conclusions and discussion.
centrography, decentralization, and concentration measures. However,
they all consider the distribution of warehousing facilities only. 2. Literature review
Warehouses are intermediaries that connect freight flows across goods
producers, distributors, sellers, and consumers. Hence, in this research, I Over the last decade, transportation planners, geographers,

E-mail address: sggkang@koti.re.kr.


1
Sanggyun Kang is the sole author of the manuscript.

https://doi.org/10.1016/j.jtrangeo.2020.102636
Received 11 August 2019; Received in revised form 7 January 2020; Accepted 9 January 2020
Available online 28 January 2020
0966-6923/ © 2020 Elsevier Ltd. All rights reserved.
S. Kang Journal of Transport Geography 83 (2020) 102636

operations researchers, and engineers have extensively examined the and intermodal terminal) that allows direct access to regional consumer
changes in location patterns and location choice decisions of the markets (Andreoli et al., 2010; Bowen, 2008; Cidell, 2010; Kang,
warehousing and logistics industry sectors, as well as potential negative 2018a; van den Heuvel et al., 2013). Factors that push warehouses to
externalities associated with the location decisions. Several review suburban/exurban areas include demand to process greater freight
papers have comprehensively scrutinized multi-dimensional aspects of volumes, demand to build larger facilities, role of a metro area in glo-
the phenomena (Aljohani and Thompson, 2016; Onstein et al., 2018; balized goods supply chains, spatial distribution of high land demand
Wygonik et al., 2015). Onstein et al. (2018) provide a concise and co- and land competition, economic development and land use policies,
herent framework to categorize the prior literature into three research resistance from local residents against locally unwanted land use
streams: supply chain management, transportation, and geography. (LULU), and physical geography of a metro area (Dablanc et al., 2014;
The supply chain management stream focuses on how goods dis- Giuliano and Kang, 2018; Kang, 2018b; Yuan, 2019).
tribution systems, including the spatial distribution of warehousing and
distribution centers, are designed and how the systematic decisions 2.2. Transportation aspects of the logistics sprawl
influence service levels and logistics costs. The transportation stream
focuses on modeling, forecasting, and minimizing freight flows and The logistics sprawl has an implication that such spatial shifts would
logistics costs in conjunction with structuring goods distribution sys- result in increased freight shipment distance and associated negative
tems. The geography stream focuses on analyzing spatial distribution externalities. However, only a few case studies based on establishment
and location decisions of goods distribution systems with respect to surveys have tested the changes in actual shipment distance over time.
various location characteristics (e.g., land rents and accessibility to the In both cases in Tokyo, Japan and Paris, France, freight shipment dis-
consumer markets, labor pool, and transportation nodes). This research tance increased over time (Dablanc and Rakotonarivo, 2010; Sakai
focuses on the transportation and geography streams. et al., 2017; Sakai et al., 2018). A travel model simulation of last-mile
shipment schemes also documented a close association between truck
2.1. Geographical aspects of the logistics sprawl vehicle kilometers traveled and the distance from a regional warehouse
to consumers (Wygonik and Goodchild, 2018). Recent research shows
Logistics sprawl, “the spatial deconcentration of logistics facilities that the relationship between the logistics sprawl and the changes in
and distribution centers in metropolitan areas,” has been examined freight shipment distance is much more complex (Kang, 2018a, 2018b;
throughout the major metropolitan areas in various parts of the world Sakai et al., 2019).
(Bowen, 2008; Cidell, 2010; Dablanc et al., 2014; Dablanc and First, warehouses are not simply moving farther away from central
Rakotonarivo, 2010; Dablanc and Ross, 2012; Giuliano and Kang, 2018; urban areas. Rather, there are varying patterns of location change with
Heitz and Dablanc, 2015; Kang, 2018a, 2018b; Rivera et al., 2014; respect to the characteristics of a warehousing facility. Most spatial
Sakai et al., 2017; Sun et al., 2018; Todesco et al., 2016; van den Heuvel shifts have been driven by larger facilities that have > 100 employees
et al., 2013; Woudsma et al., 2016). In general, the spatial shift has (Kang, 2018b) or larger than 300,000 square feet of building area
been attributed to the systematic and geographic restructuring of goods (Kang, 2018a). If those sprawling facilities accommodate externally-
production and distribution systems – how/where/when goods are driven freight flows outside of the region, the effect of spatial shifts
produced, distributed, and consumed (Hesse and Rodrigue, 2004; would be substantially mitigated. In that perspective, the type of han-
McKinnon, 2009). The restructuring has been driven by a globalized dled goods and functions of warehousing facilities along the goods
economy, advances in information, communication, and logistics supply chain would substantially influence freight shipment distance
management technologies, changes in consumer preference, changes (Giuliano and Kang, 2018; Sakai et al., 2019). A simulation study by
from supply-push to demand-pull production systems that focus on just- Sakai et al. (2019) found that the extent of the logistics sprawl is not
in-time (JIT) operations, rise in e-commerce, and use of centralized directly correlated with the changes in shipment distance; a substantial
distribution, consolidation, and fulfillment centers (Bowen, 2008; increase in negative externalities will occur if the supply of logistics
Dablanc et al., 2011; Dablanc and Ross, 2012; Hesse and Rodrigue, facilities is limited in the urban center, where high freight demand is
2004; Lasserre, 2004; McKinnon, 2009; Movahedi et al., 2009; present.
Rodrigue, 2008). Second, the spatial distribution of freight demand has also changed:
The centrography analysis most widely has been used to examine urban sprawl. This outward migration of population and employment
the spatial shifts – testing the changes in average distance from the has been well documented, and its negative externalities have been
geographical centroid (barycenter) of warehouses and examining the thoroughly examined (Anas et al., 1998; Ewing et al., 2018; Galster
transition in the directional distribution of the activity. Others sug- et al., 2001; Giuliano and Small, 1991; Mieszkowski and Mills, 1993;
gested various spatial measures considering for the spatial character- Newman and Kenworthy, 1998). If the spatial distribution of freight
istics of warehousing and distribution centers (decentralization and demand changes, it will also significantly influence shipment schemes
concentration) and their relative spatial association with other entities, and distance. However, very few spatial measures quantified logistics
such as logistics and freight transportation businesses, and total em- sprawl considering for urban sprawl. One example is the relative de-
ployment. Examples include the Gini index, relative decentralization centralization measures (e.g., the distance between warehouses and
and concentration measures, horizontal clustering location quotient, total employment) suggested by Giuliano and Kang (2018). Compared
and logistics establishments participation index (Bowen, 2008; Cidell, to the change in warehouse distribution with respect to the central
2010; Giuliano and Kang, 2018; Rivera et al., 2014). business district (14.2%), the change with respect to total employment,
Results are mixed: warehouses have been relocated to the urban including all industry sectors, is substantially smaller (6.8%). Similarly,
outskirts farther from the barycenter or the central business district in Sakai et al. (2017) found that increasing truck VMT (vehicle miles
several large trade gateways (Dablanc and Ross, 2012; Giuliano and traveled) is primarily derived from sprawling freight demand rather
Kang, 2018; Heitz and Dablanc, 2015; Heitz et al., 2017; Kang, 2018b; than sprawling warehouses.
Sakai et al., 2015; Woudsma et al., 2016). In several other metro areas,
warehouse location changes have been either insignificant or rather 2.3. Recent trends and research gaps to address
directed toward the urban center (Dablanc et al., 2014; Kang, 2018b).
Multiple papers have examined location factors that explain warehouse After the 2008–2009 economic recession, online shopping sales
location choices as well as location changes. Warehouses have been have been growing by 15% every year (Newtek, 2018; US Census
concentrating in locations that have lower land demand, access to labor Bureau, 2019). This is a significant and consistent growth relative to the
pool, and better transportation infrastructure (e.g., airport, highway, growth of traditional retail sales (3–5% annually). In 2017, U.S.

2
S. Kang Journal of Transport Geography 83 (2020) 102636

consumers spent 453 billion dollars online (8.9% of total retail sales, other related businesses that consume and handle freight shipment, as
gradually increased from 4.5% in 2010) (Cordero and Levy, 2018). This well as the general population as the final consumer demand over time.
trend of growing online shopping sales has significantly influenced the With the centrography measure, an absolute centrality measure pro-
goods supply chains throughout its international, regional, and last- posed by Dablanc and Ross (2012), I quantify the distribution of
mile segments (Tirschwell, 2018). Namely, trade and logistics busi- warehouses with respect to the geographic centroid of warehouses. This
nesses have restructured operational and system-wide aspects of their comparison between absolute and relative measures will help one un-
supply chain to accommodate the ever-growing demand for just-in-time derstand the multi-dimensional aspects of warehouse location shifts.
(JIT) production, online shopping, and instant shipment/delivery. If Further, I also consider the size of freight demand and activity. Large
lead/delivery time becomes more important, the distribution structure shipment volumes are more likely between a large warehouse and a
will be more operationally decentralized and utilize multiple fulfillment large shopping center. The floor area or volume of a warehousing fa-
centers to shorten delivery times (Onstein et al., 2018). Likewise, cility may be the most accurate size measure, but no information is
Amazon.com has put their distribution centers near urban centers available at the national level. As a proxy, employment counts for both
where they have access to cheap land, a low-wage labor pool, high- warehousing and related businesses are used. Population counts are
ways, and the port. The company's distribution centers now have 15% used as a freight demand proxy.
of the U.S. population within the same-day delivery range (Wenger, Eq. (1) presents the relative centrality measure that calculates
2013). Other examples include a substantial increase in the use of air average distance from all warehouses to all related businesses con-
freight, less-than-containerload (LCL) ocean shipment, small, palletized sidering for the size of employment of the two activities.
less-than-truckload (LTL) truck shipment, and order pickup from the n
N i (ei × dij )
retail store. Target, an American retail company, fulfilled approxi- j=1 E
× wj
mately 70% of online orders through their retail branches in 2017 Relative distribution between two activities =
W
(Tirschwell, 2018). Moreover, online fulfillment operations necessitate
(1)
intensive manual labor inputs. Hence, the pressure for facility auto-
mation and relocation to enhance labor pool access has amplified Where
(Phillips, 2018). All these practices have implications for facility re- ei = number of warehousing employment in ZIP Code (i);
location close to the consumer market. E = sum of ei;
The research gap is large. First, most of the logistics sprawl research wj = number of employment in related businesses in ZIP Code (j);
is based on the spatial distribution of warehouses in two periods. Due to W = sum of wj;
the lack of observations, prior studies could not sufficiently address the dij = Euclidean distance from ZIP Code (i) to ZIP Code (j) (i = 1, 2,
following research questions. When exactly did the logistics sprawl …, n; j = 1, 2, …, N).
occur? Are warehouses still sprawling outward? Are warehouses Eq. (2) presents the centrography measure defined in Dablanc and
moving close to consumer demand? This research will address these Ross (2012) and modified to consider the size of warehouse employ-
stated questions. Second, most of the logistics sprawl research has used ment. It calculates the average distance to all warehouses from the
absolute spatial measures that consider the distribution of warehouses geographic centroid (barycenter) of all warehousing employment in a
only, even though warehouses are intermediaries located along the metro area.
supply chains. A measure that accounts for the distribution of ware- n
(ei × di )
houses relative to the distribution of various levels of freight demand is Average distance to barycenter = i=1
E (2)
necessary in order to have more accurate quantification of the logistics
sprawl and its implications for freight activity. In this research, I use a Where,
relative spatial measure to quantify warehouse distribution considering ei = number of warehousing employment in ZIP Code (i) (i = 1, 2,
the distribution of related businesses and examine warehouse spatial …, n);
shifts over time. E = sum of ei;
di = Euclidean distance to ZIP Code (i) from the barycenter of all
3. Research approach warehousing employment; the barycenter is calculated as the mean of
X-Y coordinates with the employment size weight.
3.1. Conceptual framework The Chow test is utilized to examine if there is a structural break in
the trend of sprawling warehouses and if the structural break is sta-
Giuliano and Kang (2018) proposed spatial structure measures that tistically significant across metropolitan areas. As the point of the event
quantify the spatial distribution of warehousing activity in two di- for the structural break, I use late 2008 (the economic recession) when
mensions: (1) centrality/concentration and (2) absolutely by them- the financial and housing real estate markets significantly collapsed.
selves or relative to another activity. A centrality measure considers the The logistics sector, including warehousing businesses, was also influ-
distribution of activity with respect to one or multiple fixed points, enced by the recession (Bonacich and Lara, 2009). The exogenous shock
whereas a concentration measure considers the extent to which the negatively influenced warehouse real estate development; hence, there
activity is clustered in a small number of spatial units (Anas et al., may be a structural break in the trend of warehouse development, post-
1998). An absolute measure quantifies the spatial distribution of ac- recession. I use interaction terms to conduct the Chow test, as follows:
tivity and the activity itself only, whereas a relative measure quantifies Relative distributionit = +
0 1 Yearit + 2 Post + 3 Post Year + uit
the spatial distribution of an activity relative to the distribution of
another activity. For example, warehouse distribution may be calcu- (3)
lated with respect to the distance from the central business district Where,
(absolute centrality) or to the distribution of population (relative cen- t is from 0 for 2003 to 13 for 2016; Post is 0 if t is from 2003 to 2008;
trality). With a relative centrality measure, we can calculate the dis- Post is 1, otherwise.
tribution of warehouses relative to the distribution of potential ship- If β2 and β3 are not statistically significantly different from zero,
ment destinations and have a rough proxy for destination accessibility there is no structural break.
from all warehouses.
In this research, I use both the absolute and relative centrality 3.2. Definition of the warehousing sector and related businesses
measures and compare results. With the relative centrality measure, I
quantify the distribution of warehouses relative to the distribution of First, NAICS 493 Warehousing and Storage is used as the definition

3
S. Kang Journal of Transport Geography 83 (2020) 102636

of warehousing and distribution center businesses.2 NAICS 493 may not Table 1
enumerate all buildings with warehousing functions as defined in Definition of the logistics sectors.
NAICS 493. Rather, it may exclude the warehouses operated as an NAICS Title
auxiliary unit of a business. For example, an establishment, of which
the primary line of business is manufacturing automobile parts, may 481 Air Transportation
481112 Scheduled Freight Air Transportation
operate a warehousing facility as an auxiliary unit. In this case, it is
481212 Nonscheduled Chartered Freight Air Transportation
possible that the warehousing facility is not accounted under the NAICS 481219 Other Nonscheduled Air Transportation
493 definition. Second, the six-digit logistics industry sectors, proposed 483 Water Transportation
in Rivera et al. (2014), is used as the definition of logistics businesses. 483111 Deep-Sea Freight Transportation
Economic geography research has extensively examined the co-location 483113 Coastal and Great Lakes Freight Transportation
483211 Inland Water Freight Transportation
pattern of the logistics sector and related businesses. Several inter-in-
484 Truck Transportation
dustry measures of co-location (input-output links, labor occupation 484110 General Freight Trucking, Local
links, and geographic vicinity measures) have examined the systematic 484121 General Freight Trucking, Long-Distance, Truckload
and geographic linkages between industry sectors (Delgado et al., 2016; 484122 General Freight Trucking, Long-Distance, Less Than Truckload
484220 Specialized Freight (except Used Goods) Trucking, Local
Feser et al., 2005; Kumar et al., 2017; Porter, 2003). Rivera et al. (2014)
484230 Specialized Freight (except Used Goods) Trucking, Long-Distance
identified the geographic and operational linkages between the ware- 488 Support Activities for Transportation
housing and logistics sectors. The logistics businesses have the most 488119 Other Airport Operations
immediate (first-order) connection with warehousing businesses. 488190 Other Support Activities for Air Transportation
Therefore, it is expected that they are geographically most closely lo- 488210 Support Activities for Rail Transportation
488310 Port and Harbor Operations
cated. Table 1 lists all the six-digit logistics sectors.
488320 Marine Cargo Handling
As second-order industry sectors engaged in goods distribution, 488330 Navigational Services to Shipping
NAICS 31–33 Manufacturing, 42 Wholesale Trade, and 44–45 Retail 488390 Other Support Activities for Water Transportation
Trade are utilized. These businesses may or may not have a direct 488410 Motor Vehicle Towing
connection with warehouse operations. They still are mostly involved 488490 Other Support Activities for Road Transportation
488510 Freight Transportation Arrangement
in goods movement because they function as consumers/shippers/ 488991 Packing and Crating
receivers of freight shipment. It is expected that goods movement 488999 All Other Support Activities for Transportation
businesses have a lesser geographic association with warehousing 492 Couriers and Messengers
businesses than the logistics businesses do. Lastly, I apply the general 492110 Couriers and Express Delivery Services
492210 Local Messengers and Local Delivery
population as the ultimate demand/destination of freight movement.
493 Warehousing and Storage
It is expected that the general population is the least likely to spatially 493110 General Warehousing and Storage
co-locate with warehousing businesses. Table 2 includes the defini- 493120 Refrigerated Warehousing and Storage
tions. 493130 Farm Product Warehousing and Storage
493190 Other Warehousing and Storage

3.3. Data

The primary data source is the ZIP Code Business Patterns (ZBP)
Survey estimates are used to identify the population count by 2010
2003–2016. The ZBP is a subset of the County Business Patterns (CBP)
census tract delineation.
and is based on the Business Register. ZBP Industry Detail datasets
provide the number of establishments at the six-digit NAICS sector level
3.4. Study area
by nine employment size classes by ZIP Code.3 No actual employment
counts are available in the ZBP. There are several methods to approx-
The study area is 64 US metropolitan areas: 58 combined statistical
imate employment counts at the ZIP Code level using county-level
areas (CSAs) and six metropolitan statistical areas (MSAs). An MSA
statistics from the County Business Patterns (Giuliano and Kang, 2018).
consists of one or multiple counties with at least one urban core that has
The estimation method makes use of quadratic programming, and the
a minimum population of 10,000. One or multiple-neighboring MSAs
accuracy of estimation largely depends on the number of county-level
with significant commuting inter-connections comprise a CSA. The 64
samples. In Giuliano and Kang (2018), a California case study, the
metro areas are all consistent with the Commodity Flow Survey (CFS)
authors quantified warehouse employment using 58 county samples in
2012 definition of analysis regions. Based on the same geographic de-
California. The unit of analysis of this research is a metropolitan area,
finition, the Freight Analysis Framework (FAF) provides metropolitan-
and the majority of the metro areas have less than ten counties – the
level freight shipment information by tonnage, dollar-value, and ton-
minimum number of samples necessary to run the program to estimate
mile. Table 3 lists the 64 US metro areas with their short name, full
nine unknown parameters. Therefore, the estimation method is not
name, population statistics, and warehouse statistics. Across metro
appropriate for this research. Hence, the mean for each employment
areas, the 2016 population and percent changes from 2000 vary sig-
size class is used (e.g. 2.5 for the first class, 1–4 employees). This may
nificantly. Kang (2018b) found a non-linear pattern of warehouse de-
bring up the issue of data imprecision, but the ZBP is the only public
centralization (increasing distance from the central business district) by
dataset with the finest unit of spatial analysis, with which warehousing
population size: the pattern of warehouse location changes in large
establishment counts and a proxy for the size can be identified at the
metro areas (2.2 million population) differs significantly from that in
national level. 2003 is the earliest available data year after the revision
small metro areas. Hence, I categorize the metro areas into two groups
of the NAICS definition in 2002, and 2016 is the latest data when this
by population size at 2.2 million. Moreover, the number of warehousing
research is conducted. A more detailed description of the strength and
establishments per million residents as well as changes from 2003 to
weakness of the ZBP datasets is available in Giuliano and Kang (2018).
2016 vary widely across metro areas. Primarily, small metro areas with
The 2000 and 2010 US Census and 2012–2016 American Community
significant trade functions (e.g., Memphis, TN; Savannah, GA) have a
large number of warehousing establishments per resident. Several
2
NAICS, North American Industry Classification System. metro areas gained a significant number of warehousing establishments
3
The nine size classes are 1–4; 5–9; 10–19; 20–49; 50–99; 100–249; 250–499; over time. Yet, according to the preliminary analysis, the changes do
500–999; 1000 or more. not correspond to the extents of spatial shifts.

4
S. Kang Journal of Transport Geography 83 (2020) 102636

Table 2
Relationship between the warehousing sector and related businesses.
Entity Relationship Definition Measure to quantify absolute/relative spatial distribution

Warehouses Self NAICS 493 Warehousing and storage Measure 1: Absolute centrality,
Distance to the geographical centroid (barycenter) of warehouses
Logistics sector businesses Immediate connection Defined in Table 1 Measure 2: Relative centrality,
Distance to logistics businesses
Businesses involved in goods movement Second-order NAICS 31–33 Manufacturing Measure 3: Relative centrality,
NAICS 42 Wholesale trade Distance to goods movement businesses
NAICS 44–45 Retail trade
Consumer demand End of goods movement General population Measure 4: Relative centrality,
Distance to population

Table 3
Sixty-four metropolitan areas included in this research.

(continued on next page)

5
S. Kang Journal of Transport Geography 83 (2020) 102636

Table 3 (continued)

*Green highlights indicate that the metro area has at least ten large warehouses (> 100 employees) in 2003 and 2016 and is included in the
analysis that involves large warehouses.

4. Results all related businesses in a metropolitan area. It is not a measure of


shipment distance but more of relative destination accessibility. Lo-
Finding 1. In 2016, the distance between warehouses and logistics gistics businesses choose warehouse locations that help them optimize
businesses was the shortest (38.6 km), followed by the distance to their operations and reduce costs. Hence, the actual shipment distance
goods movement businesses (41.2 km) and then the distance to the
population (43.5 km). (N = 64 metro areas)
Table 4
The spatial distribution of warehouses is most closely associated Distance to related businesses from all warehouses in 2016 (in kilometers).
with that of logistics businesses and least with that of the general po-
All metro areas N Mean SD T-test (P-value) Median
pulation. According to the t-tests between the means of relative dis-
tribution measures, the differences between relative distribution mea- D to Barycenter (M1) 64 27.4 13.0 – 27.1
sures across metro areas are statistically significantly greater than zero D to Logistics Bs (M2) 64 38.6 14.7 – 38.6
D to Goods mvmt Bs (M3) 64 41.2 14.3 H0: M2 = M3 (0.000) 40.9
(N = 64). Table 4 presents the summary statistics and t-test results.
D to Population (M4) 64 43.5 13.6 H0: M3 = M4 (0.000) 42.7
Note that relative distribution measures account for all warehouses and

6
S. Kang Journal of Transport Geography 83 (2020) 102636

90.00

80.00

70.00
Rela ve distance 60.00

50.00

40.00

30.00

20.00

10.00

0.00

Miami
Boston

Atlanta

Seattle

Denver
Chicago

Houston

Detroit
Dallas

Phoenix

Portland

St. Louis
Minneapolis
Cleveland

Orlando

Pittsburgh

Kansas City
Columbus

Cincinnati
Charlotte
San Francisco

Philadelphia

San Diego

Sacramento
Salt Lake City

Indianapolis
Washington
Los Angeles

San Antonio
Las Vegas
Tampa
New York

Ranked by metro size (le! to right)

D to barycenter D to logis cs bs D to goods mvmt bs D to popula on

Fig. 1. Relative distribution in 2016 (large metro areas).

would be much smaller than the relative distance. Still, results show Table 5 presents the changes in the distance between warehouses
reasonable spatial association across related businesses. Distance to the and related businesses. The changes in distance to logistics businesses,
barycenter is indeed the smallest because it only considers the dis- goods movement businesses, and population are not statistically sig-
tribution of warehouses from their geographic centroid. Figs. 1 and 2 nificantly different. The null hypothesis (ΔM2 = ΔM3 = ΔM4) cannot
show that, in general, relative distances decrease as the size of a metro be rejected. However, the changes in the distance for all relative mea-
area decreases. sures are statistically smaller than the change in distance to barycenter.
This finding implies that not only have warehouses changed location in
Finding 2. From 2003 to 2016, the distance between warehouses and
a way that they are more decentralized from their barycenter, but also
related businesses increased statistically significantly. Across metro
other related businesses and population have shifted spatially. Thus,
areas, the distance from warehouses to their barycenter increased more
relative distribution did not change as significantly as absolute dis-
than other relative distribution measures. (N = 64 metro areas)
tribution did. Another implication is that the absolute measure alone

90.00

80.00

70.00
Relative distance

60.00

50.00

40.00

30.00

20.00

10.00

0.00
Greenville

Albany

El Paso
Milwaukee

Nashville

Greensboro
Jacksonville

Hartford

Fort Wayne
Raleigh

Austin

New Orleans
Grand Rapids

Dayton
Virginia Beach

Oklahoma City
Memphis
Birmingham
Richmond
Buffalo

Tulsa
Fresno

Tucson

Mobile
Savannah
Rochester

Knoxville

Omaha
Charleston
Wichita
Louisville

Corpus Christi

Ranked by metro size (left to right)

D to barycenter D to logistics bs D to goods mvmt bs D to population

Fig. 2. Relative distribution in 2016 (small metro areas).

7
S. Kang Journal of Transport Geography 83 (2020) 102636

Table 5
Changes in relative distribution between 2003 and 2016 (in kilometers).
All metro areas N Mean SD T-test 1, P (H0: ΔD = 0) T-test (P-value)

ΔD to Barycenter (ΔM1) 64 4.7 7.8 0.000 –


ΔD to Logistics Bs (ΔM2) 64 3.7 7.0 0.000 H0: ΔM1 = ΔM2 (0.023)
ΔD to Goods mvmt Bs (ΔM3) 64 3.3 6.5 0.000 H0: ΔM1 = ΔM3 (0.010)
ΔD to Population (ΔM4) 64 3.1 6.4 0.000 H0: ΔM1 = ΔM4 (0.007)

Table 6 changes have been very marginal. The trend of increasing distance to
Relative distribution by year by measure (in kilometers). related businesses and barycenter stopped around 2007–2008. Namely,
Year Barycenter Logistics Goods Population
there is a structural break in the trend of warehousing spatial shifts.
businesses movement Similarly, the number of warehousing establishments increased
more significantly (16.5% on average) between 2003 and 2010 than it
2003 22.7 34.9 37.9 40.4 did between 2010 and 2016 (9.7% on average). Results are not pre-
2004 23.1 35.1 37.9
2005 24.0 35.6 38.4
sented here. It is notable that after the recession the extent of changes in
2006 24.7 36.1 38.9 number is far greater than that of spatial shifts. That is, there has been
2007 27.1 38.2 40.9 net growth in the warehousing industry, but the logistics sprawl did not
2008 27.2 38.2 41.0 occur between 2010 and 2016.
2009 27.2 38.2 40.9
Results of the Chow test are in Table 7. Across the three measures,
2010 27.4 38.7 41.4 43.8
Change 2003–2010 4.7 3.8 3.5 3.4 all the interaction terms, as well as the Chow test results, are statisti-
% Change 20.5% 10.9% 9.3% 8.5% cally significant: there is a structural break. Namely, the estimated
2011 27.4 38.5 41.2 coefficients of Post and Post*Year are statistically significantly different
2012 28.0 38.9 41.6 from zero. The model explains approximately 21% of the within-iden-
2013 27.8 38.8 41.4
2014 27.4 38.6 41.2
tity variation over time (within-R2). Across the 64 metropolitan areas,
2015 27.5 38.7 41.2 pre-recession, the annual rate of increasing distance from barycenter
2016 27.4 38.6 41.2 43.5 (1.00 km per year) is approximately 33% higher than that from logistics
Change 2010–2016 0.0 −0.1 −0.2 −0.3 and goods movement businesses (0.72–0.75 km per year) (Step 2). It is
% Change 0.1% −0.2% −0.5% −0.7%
consistent with the previous finding that warehouse location shifts have
Change 2003–2016 4.7 3.7 3.3 3.1
% Change 20.6% 10.6% 8.7% 7.8% been moderately mitigated as related businesses have also spatially
shifted over time. The annual rate of changes (slopes) is similar between
M2 and M3 yet with constants of significantly different sizes.
Finding 4. The total extent and annual rate of changes in relative
would overestimate the extent to which freight destination accessibility
distribution vary between large and small warehouses. There is also a
might diminish due to the logistics sprawl. Similar patterns were
significant structural break in the trend. (N = 25 metro areas)
documented in Sakai et al. (2018). Increasing shipment distance is at-
tributed to sprawling freight demand (e.g., population). Rather, ware- Twenty-five metro areas that have at least ten large warehouses in
house relocation significantly mitigated the negative effect of sprawling both 2003 and 2016 are included, and most of the 25 metro areas are
freight demand. categorized as large metro areas (only four are small). This finding
compares the pattern of relative spatial shifts of large/small warehouses
Finding 3. There is a structural break in the trend of increasing distance
in large metro areas. Table 8 and Fig. 3 present changes in the pattern
between warehouses and related businesses. (N = 64 metro areas)
of relative distribution from 2003 to 2016 by warehouse size.
Table 6 presents the annual average statistics of relative distribu- The relative distances for large warehouses are greater than those
tion. Population data has only three data points (2000, 2010, and for small warehouses. Consistent with the finding in Kang (2018b),
2016). Consistently over time, warehouses are most closely located compared to small warehouses, large facilities are located not only
relative to logistics businesses, followed by goods movement businesses farther from the central business district but also to related businesses
and the general population. Across the measures, most of the spatial and the general population. Relative distribution for large warehouses,
shifts occurred between 2003 and 2010. Between 2010 and 2016, the as well as small ones, have increased statistically significantly from

Table 7
Model estimation and Chow test results, all metro areas, all warehouses.
Variables Measure 1: Distance to Barycenter Measure 2: Distance to Logistics businesses Measure 3: Distance to Goods mvmt businesses

Step 1 Step 2 Step 1 Step 2 Step 1 Step 2

Coef. Sig. Coef. Sig. Coef. Sig. Coef. Sig. Coef. Sig. Coef. Sig.

Year 0.37 ** 1.00 ** 0.30 ** 0.75 ** 0.28 ** 0.72 **


Post 4.99 ** 3.85 ** 3.86 **
Post*Year −0.98 ** −0.72 ** −0.71 **
Constant 23.96 ** 22.30 ** 35.67 ** 34.47 ** 38.54 ** 37.35 **
N 896 896 896 896 896 896
Within R2 0.157 0.219 0.156 0.205 0.156 0.214
Chow test F 33.030 25.840 30.370
P 0.000 0.000 0.000

**P < 0.01, *P < 0.05.

8
S. Kang Journal of Transport Geography 83 (2020) 102636

Table 8
Changes in relative distribution from 2003 to 2016 by warehouse size (in kilometer).
Relative distribution measure Size of warehouse N Mean SD T-test, P (H0: ΔD = 0) Median T-test, P (H0: ΔD large = ΔD small)

Δ Distance to Barycenter Large warehouse 25 5.0 6.7 0.001 5.9 0.004


Small warehouse 25 1.2 3.2 0.065 1.1
Δ Distance to Logistics Bs Large warehouse 25 4.9 5.6 0.000 5.9 0.001
Small warehouse 25 1.8 2.8 0.003 2.6
Δ Distance to Goods mvmt Bs Large warehouse 25 4.9 5.1 0.000 5.5 0.000
Small warehouse 25 1.3 2.7 0.020 1.4
Δ Distance to Population Large warehouse 25 4.8 4.9 0.000 6.0 0.000
Small warehouse 25 1.4 2.6 0.014 1.4

All measures, large warehouses All measures, small warehouses


60.0 60.0
53.5 53.1
55.0 55.0
48.3 46.9
50.0 50.0 45.6 46.7
45.0 45.0
40.0 40.0
35.0 35.0
30.0 30.0
25.0 25.0
20.0 20.0
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016

2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
Barycenter Logis cs bs Barycenter Logis cs bs
Goods bs Popula on Goods bs Popula on

Fig. 3. Relative distance measures, annual average, large (left) and small (right) warehouses.

Table 9
Model estimation and Chow test results by warehouse size.
Variables Measure 1: Distance to Barycenter Measure 2: Relative distance to Logistics businesses Measure 3: Relative distance to Goods mvmt businesses

Large warehouses Small warehouses Large warehouses Small warehouses Large warehouses Small warehouses

Coef. Sig. Coef. Sig. Coef. Sig. Coef. Sig. Coef. Sig. Coef. Sig.

Year 1.069 ** 0.237 ** 0.951 ** 0.241 ** 0.977 ** 0.241 **


Post 6.349 ** 0.433 5.816 ** 0.198 5.901 ** 0.235
Post*Year −1.151 ** −0.124 −0.995 ** −0.071 −1.024 ** −0.114
Constant 29.086 ** 26.633 ** 42.181 ** 40.198 ** 45.536 ** 43.130 **
N 350 350 350 350 350 350
Within R2 0.232 0.111 0.298 0.259 0.313 0.170
Chow test F 17.330 0.960 19.820 0.590 21.830 1.870
P 0.000 0.384 0.000 0.557 0.000 0.155

2003 to 2016 (t-tests in Table 8). The change in relative distribution for interaction terms, as well as Chow test results, are not statistically
large warehouses is significantly greater than that for small warehouses significant. Small warehouses have steadily decentralized from related
across all four measures. Between 2003 and 2016, across all measures businesses at the rate of 0.24 km per year.
for large warehouses, distance increased approximately by 4.9 km,
Finding 5. Relative distribution measures have a significant association
whereas the change was 1.2–1.8 km for small warehouses.
with internal ground shipment ton-kilometers. (N = 64 metro areas)
The Chow test is conducted to examine if there is a structural break.
Table 9 presents the results. When large warehouses are considered, The Freight Analysis Framework Version 4 (FAF4) datasets are
across the three measures, all the interaction terms, as well as Chow test utilized to examine the relationship between relative distribution
results, are statistically significant. There is a structural break. The measures and freight shipment ton-kilometers (converted from miles).
model explains approximately 23–31% of the within-identity variation The base year of FAF4 is 2012. Based on the Commodity Flow Survey
over time (within R2). Pre-recession, across the 25 metro areas, the (CFS), the FAF provides more detailed freight shipment information at
annual rate of increasing distance from barycenter for large warehouses the metropolitan level. The FAF uses CFS' definition of metropolitan
(1.07 km per year) is approximately 12% greater than that from lo- regions and provides tonnage, dollar value, and ton-mile measures of
gistics and goods movement businesses (0.95–0.98 km per year). When freight shipment by commodity type by mode of transportation and by
small warehouses are considered, across the three measures, all the type of shipment flows. Shipment distance may have been the most

9
S. Kang Journal of Transport Geography 83 (2020) 102636

Table 10 logistics businesses, followed by goods movement businesses and then


Bivariate correlation between ton-kilometers and population, urban area, and the general population. This finding confirms that warehouses have an
relative distribution measures. immediate connection with logistics businesses operationally and spa-
Ton-kilometers (million) 64 metro areas tially. Second, from 2003 to 2016, the relative distance between
warehouses and all related businesses as well as the general population
Corr. Sig. increased significantly at the national level. However, the extent was
26–50% less than when the spatial shifts were quantified by an absolute
Population 0.893 ⁎⁎
Urban area (mi2) 0.815 ⁎⁎ measure with respect to the barycenter of warehouses. This implies that
Absolute distance to barycenter 0.513 ⁎⁎ not only warehouses but also related businesses and population have
Relative distance to logistics bs 0.572 ⁎⁎ decentralized significantly. Third, the extent to which relative dis-
Relative distance to goods mvmt bs 0.585 ⁎⁎
tribution changes varies with respect to warehouse size. Fourth, the
Relative distance to population 0.627 ⁎⁎
spatial shifts of warehouses occurred mostly before the economic re-
⁎⁎
P < .01. cession in 2008. After the recession, the location shifts occurred only in
large metro areas at a much slower rate (0.2 km per year) than it was
prior to the recession (0.8–1.0 km per year). Fifth, freight activity, in
relevant, but no information is available in FAF. I include all com- terms of internal ground shipment ton-kilometers, is significantly cor-
modity types and consider truck, rail, trucking portion of air, and related with relative distribution measures. More specifically, relative
multiple modes and mail (USPS and couriers) as the mode of ground distance to population is the most significantly correlated with the
transportation. Excluded modes are water, pipeline, and other/un- freight activity. It may be derived from the broad correlation across
known. All internal shipment flows that originate from and destined to population distribution, urban area size, relative distance measures,
the very metro area are included. All other domestic and foreign flows and shipment ton-kilometers. It may not be completely ignorable as a
are excluded. Bivariate correlation coefficients are measured between spurious association but may be applied as a rough proxy for destina-
ground shipment ton-kilometers in a million and the size of a metro tion accessibility for freight shipment when freight shipment informa-
area in terms of population and urban area as well as relative dis- tion is not publicly available. Further research is necessary to examine
tribution measures in 2012. if the measure can trace shipment distance consistently over time.
Results are in Table 10. Across 64 metro areas, ground shipment Which factor explains the relative spatial shifts? First, the 2008
ton-kilometers are significantly associated with the population (0.893) economic recession influenced commercial real estate, logistics, and
and urban area (0.815). This correlation is not surprising because the freight transport businesses greatly over the following 5–6 years in
population and urban area are direct proxies for freight demand and terms of how they operate global supply chain systems as well as how
shipment distance. Population and urban area are also highly correlated they develop, own, lease, or rent warehousing space (Hoffman, 2008).
with each other (0.957) (Anas et al., 1998). Despite the lesser magni- On one hand, the commercial real estate businesses have maintained
tude, relative distribution measures are also significantly correlated the oversupply of property at the minimum level. On the other, shippers
with ground shipment ton-kilometers. Distance to the population has have become more cautious and efficient in managing inventory levels
the highest correlation (0.627) followed by goods movement, logistics (lean inventory) due to consumers' weakened purchasing power and
businesses, and distance to the barycenter. This result is consistent with changed shopping patterns (online shopping). However, even with the
the previous findings that warehouses are spatially distributed closer to lean inventory strategy, due to lack of sufficient purchasing power, total
logistics businesses and farther from the population. This proxy for private inventory levels have continued to increase 30% to $2.5 trillion
destination (population) accessibility explains shipment ton-kilometers in 2015 from $1.9 trillion in 2009 (Cassidy and Hutchins, 2016).
quite well. Second, online shopping and e-commerce have supported the ware-
housing and distribution industry. However, how they utilize ware-
5. Conclusion and discussion house space has completely changed in that more warehouse demand is
now close to consumer markets (Bonney, 2017). Expanding online
In this research, I examined the spatial distribution of warehouses shopping sales has boosted the development of large-scale distribution
relative to the distribution of logistics businesses, goods movement centers as well as smaller, modernized facilities with high ceilings
businesses, and the consumer demand from 2003 to 2016 in 64 major (10,000–30,000 square meters) in central urban areas (Mongelluzzo,
US metropolitan areas. Warehouses are defined as the establishments in 2019). All this business-oriented evidence supports the patterns of re-
NAICS 493 Warehousing and Storage. Logistics businesses – im- lative logistics sprawl documented in this research.
mediately connected with the warehousing businesses – are defined as a Several metropolitan areas merit further research. For example, in
combination of six-digit subsectors of NAICS 48–49 transportation and Atlanta, Los Angeles, New York, and Philadelphia metropolitan areas,
warehousing, as presented in Rivera et al. (2014). Goods movement large/labor-intensive warehouses have moved away from related
businesses – the second-order association with warehousing establish- businesses at a significant rate, whereas in Cincinnati, Indianapolis, and
ments – are defined as NAICS 31–33 manufacturing, 42 wholesale Salt Lake City metropolitan areas, large warehouses have moved ra-
trade, and 44–45 retail trade. The consumer demand – the ultimate pidly close to related businesses including the consumer demand. In this
destination of goods supply chains – is defined as the US Census po- case, a detailed case study may be a better format to verify the reasons
pulation. Warehouse and related business employment, as well as po- and pattern for the reverse sprawl.
pulation, are used as a proxy for the size of demand/activity. As the
main data sources, ZIP Code Business Patterns 2003–2016, US Census Acknowledgment
2000 and 2010, and American Community Survey 2012–2016 are
employed. Findings provide empirical evidence regarding these re- This work was supported by the Center for Transportation Equity,
search questions: When did the logistics sprawl occur? Are warehouses Decisions and Dollars (CTEDD), a Tier-1 University Transportation
still sprawling outward? Are warehouses moving close to the consumer Center funded by the U.S. Department of Transportation and housed at
demand? Do relative distribution measures provide any implications for the University of Texas at Arlington.
freight activity?
Findings indicate that, at the national level, the logistics sprawl has References
slowed down quite substantially since 2009. More findings are as fol-
lows. First, in 2016, warehouses are most closely located relative to Aljohani, K., Thompson, R.G., 2016. Impacts of logistics sprawl on the urban environment

10
S. Kang Journal of Transport Geography 83 (2020) 102636

and logistics: taxonomy and review of literature. J. Transp. Geogr. 57, 255–263. Lasserre, F., 2004. Logistics and the Internet: Transportation and location issues are
https://doi.org/10.1016/j.jtrangeo.2016.08.009. crucial in the logistics chain. J. Transp. Geogr. 12 (1), 73–84. https://doi.org/10.
Anas, A., Arnott, R., Small, K.A., 1998. Urban spatial structure. J. Econ. Lit. 36 (3), 1016/S0966-6923(03)00027-9.
1426–1464. McKinnon, A., 2009. The present and future land requirements of logistical activities.
Andreoli, D., Goodchild, A., Vitasek, K., 2010. The rise of mega distribution centers and Land Use Policy 26, S293–S301. https://doi.org/10.1016/j.landusepol.2009.08.014.
the impact on logistical uncertainty. Transp. Lett. 2 (2), 75–88. https://doi.org/10. Mieszkowski, P., Mills, E.S., 1993. The causes of metropolitan suburbanization. J. Econ.
3328/TL.2010.02.02.75-88. Perspect. 7 (3), 135–147.
Bonacich, E., Lara, J.D.D., 2009. Economic Crisis and the Logistics Industry: Financial Mongelluzzo, B., 2019, October 4. New space won’t reverse climb in US warehouse rents:
Insecurity for Warehouse Workers in the Inland Empire. No. WP 2009–13. UCLA report. JOC.Com. https://www.joc.com/international-logistics/distribution-centers/
Institute for Research on Labor and Employment, pp. 25. new-space-won%E2%80%99t-reverse-climb-warehouse-rents-report_20191024.html.
Bonney, J., 2017, November 8. CBRE: E-commerce’s warehousing impact will survive Movahedi, B., Miri-Lavassani, K., Kumar, V., 2009. Transition to B2B E-marketplace en-
recession. JOC.Com. https://www.joc.com/international-logistics/distribution- abled supply chain: readiness assessment and success factors. Int. J. Technol. Knowl.
centers/cbre-e-commerce-warehousing-impact-will-survive-recession_20171108. Soc. 5 (3), 75–88. https://doi.org/10.18848/1832-3669/CGP/v05i03/55999.
html. Newman, P., Kenworthy, J., 1998. Sustainability and Cities: Overcoming Automobile
Bowen, J.T., 2008. Moving places: the geography of warehousing in the US. J. Transp. Dependence. Island Press Books.
Geogr. 16 (6), 379–387. https://doi.org/10.1016/j.jtrangeo.2008.03.001. Newtek, 2018, June 20. Trends in Ecommerce and What They Mean for Your Business.
Cassidy, W.B., Hutchins, R., 2016, June 29. High US inventories could be the new normal. Retrieved from Forbes.com website: https://www.forbes.com/sites/thesba/2018/
JOC.Com. https://www.joc.com/economy-watch/us-economy-news/some-signs- 06/20/trends-in-ecommerce-and-what-they-mean-for-your-business/#
indicate-long-awaited-destocking-under-way_20160629.html. 216b1b7524c3.
Cidell, J., 2010. Concentration and decentralization: the new geography of freight dis- Onstein, A.T.C., Tavasszy, L.A., van Damme, D.A., 2018. Factors determining distribution
tribution in US metropolitan areas. J. Transp. Geogr. 18 (3), 363–371. https://doi. structure decisions in logistics: a literature review and research agenda. Transp. Rev.
org/10.1016/j.jtrangeo.2009.06.017. 1–18. https://doi.org/10.1080/01441647.2018.1459929.
Cordero, M., Levy, S., 2018, February. E-Commerce Retail Sales Hit $453.5 Billion in Phillips, E.E., 2018, September 4. E-commerce driving need for more warehouse workers.
2017, as Brands Invest In Omnichannel. Retrieved February 18, 2019, from CBRE.US Wall Street J Retrieved from. https://www.wsj.com/articles/e-commerce-driving-
website: http://www.cbre.us/real-estate-services/real-estate-industries/retail- need-for-more-warehouse-workers-1536055205.
services/research-and-insights/us-marketflash-e-commerce-2017. Porter, M., 2003. The economic performance of regions. Reg. Stud. 37 (6–7), 549–578.
Dablanc, L., Rakotonarivo, D., 2010. The impacts of logistics sprawl: how does the lo- Rivera, L., Sheffi, Y., Welsch, R., 2014. Logistics agglomeration in the US. Transp. Res. A
cation of parcel transport terminals affect the energy efficiency of goods’ movements Policy Pract. 59, 222–238. https://doi.org/10.1016/j.tra.2013.11.009.
in Paris and what can we do about it? Procedia Soc. Behav. Sci. 2 (3), 6087–6096. Rodrigue, J.-P., 2008. The Thruport concept and transmodal rail freight distribution in
https://doi.org/10.1016/j.sbspro.2010.04.021. North America. J. Transp. Geogr. 16 (4), 233–246. https://doi.org/10.1016/j.
Dablanc, L., Ross, C., 2012. Atlanta: a mega logistics center in the Piedmont Atlantic jtrangeo.2007.08.003.
Megaregion (PAM). J. Transp. Geogr. 24, 432–442. https://doi.org/10.1016/j. Sakai, T., Kawamura, K., Hyodo, T., 2015. Locational dynamics of logistics facilities:
jtrangeo.2012.05.001. evidence from Tokyo. J. Transp. Geogr. 46, 10–19. https://doi.org/10.1016/j.
Dablanc, L., Diziain, D., Levifve, H., 2011. Urban freight consultations in the Paris region. jtrangeo.2015.05.003.
Eur. Transp. Res. Rev. 3 (1), 47–57. https://doi.org/10.1007/s12544-011-0049-2. Sakai, T., Kawamura, K., Hyodo, T., 2017. Spatial reorganization of urban logistics system
Dablanc, L., Ogilvie, S., Goodchild, A., 2014. Logistics sprawl: differential warehousing and its impacts: case of Tokyo. J. Transp. Geogr. 60, 110–118. https://doi.org/10.
development patterns in Los Angeles, California, and Seattle, Washington. Transp. 1016/j.jtrangeo.2017.03.001.
Res. Rec. 2410 (1), 105–112. https://doi.org/10.3141/2410-12. Sakai, T., Kawamura, K., Hyodo, T., 2018. The relationship between commodity types,
Delgado, M., Porter, M.E., Stern, S., 2016. Defining clusters of related industries. J. Econ. spatial characteristics, and distance optimality of logistics facilities. J. Transp. Land
Geogr. 16 (1), 1–38. https://doi.org/10.1093/jeg/lbv017. Use 11 (1). https://doi.org/10.5198/jtlu.2018.1363.
Ewing, R., Hamidi, S., Tian, G., Proffitt, D., Tonin, S., Fregolent, L., 2018. Testing Sakai, T., Kawamura, K., Hyodo, T., 2019. Evaluation of the spatial pattern of logistics
Newman and Kenworthy’s theory of density and automobile dependence. J. Plan. facilities using urban logistics land-use and traffic simulator. J. Transp. Geogr. 74,
Educ. Res. 38 (2), 167–182. https://doi.org/10.1177/0739456X16688767. 145–160. https://doi.org/10.1016/j.jtrangeo.2018.10.011.
Feser, E., Sweeney, S., Renski, H., 2005. A descriptive analysis of discrete US industrial Sun, B., Li, H., Zhao, Q., 2018. Logistics agglomeration and logistics productivity in the
complexes. J. Reg. Sci. 45 (2), 395–419. USA. Ann. Reg. Sci. 61 (2), 273–293. https://doi.org/10.1007/s00168-018-0867-4.
Galster, G., Hanson, R., Ratcliffe, M.R., Wolman, H., Coleman, S., Freihage, J., 2001. Tirschwell, P., 2018, January 11. E-commerce Reshaping Supply Chains End-to-End.
Wrestling sprawl to the ground: defining and measuring an elusive concept. Hous. Retrieved February 18, 2019, from Joc.com website: https://www.joc.com/
Policy Debate 12 (4), 681–717. https://doi.org/10.1080/10511482.2001.9521426. technology/e-commerce-reshaping-supply-chains-end-end_20180111.html.
Giuliano, G., Kang, S., 2018. Spatial dynamics of the logistics industry: evidence from Todesco, P., Weidmann, U., Haefeli, U., 2016. Logistics Sprawl in the Region Zurich. 21.
California. J. Transp. Geogr. 66, 248–258. https://doi.org/10.1016/j.jtrangeo.2017. Monte Verità, Ascona.
11.013. US Census Bureau, April 11, 2019. Sales for U.S. Retail Trade Sector at $5,046.9 Billion.
Giuliano, G., Small, K.A., 1991. Subcenters in the Los Angeles region. Reg. Sci. Urban Retrieved July 28, 2019, from The United States Census Bureau website: https://
Econ. 21 (2), 163–182. https://doi.org/10.1016/0166-0462(91)90032-I. www.census.gov/newsroom/press-releases/2019/retail-trade.html.
Heitz, A., Dablanc, L., 2015. Logistics spatial patterns in Paris: rise of Paris Basin as lo- van den Heuvel, F.P., de Langen, P.W., van Donselaar, K.H., Fransoo, J.C., 2013. Spatial
gistics megaregion. Transp. Res. Rec. 2477 (1), 76–84. https://doi.org/10.3141/ concentration and location dynamics in logistics: the case of a Dutch province. J.
2477-09. Transp. Geogr. 28, 39–48. https://doi.org/10.1016/j.jtrangeo.2012.10.001.
Heitz, A., Dablanc, L., Tavasszy, L.A., 2017. Logistics sprawl in monocentric and poly- Wenger, Y., 2013, August 23. Proposed Baltimore warehouse fits Amazon’s growth.
centric metropolitan areas: the cases of Paris, France, and the Randstad, the Baltimore Sun Retrieved from. http://articles.baltimoresun.com/2013-08-23/news/
Netherlands. REGION 4 (1), 93. https://doi.org/10.18335/region.v4i1.158. bs-md-ci-amazon-site-20130823_1_online-retailer-amazon-same-day-delivery-
Hesse, M., Rodrigue, J.-P., 2004. The transport geography of logistics and freight dis- distribution-center.
tribution. J. Transp. Geogr. 12 (3), 171–184. https://doi.org/10.1016/j.jtrangeo. Woudsma, C., Jakubicek, P., Dablanc, L., 2016. Logistics sprawl in North America:
2003.12.004. methodological issues and a case study in Toronto. Transp. Res. Procedia 12,
Hoffman, W., 2008, October 5. DC demand caving in. JOC.Com. https://www.joc.com/ 474–488. https://doi.org/10.1016/j.trpro.2016.02.081.
economy-watch/dc-demand-caving_20081005.html. Wygonik, E., Goodchild, A.V., 2018. Urban form and last-mile goods movement: factors
Kang, S., 2018a. Warehouse location choice: a case study in Los Angeles, CA. J. Transp. affecting vehicle miles travelled and emissions. Transp. Res. Part D: Transp. Environ.
Geogr. https://doi.org/10.1016/j.jtrangeo.2018.08.007. 61, 217–229. https://doi.org/10.1016/j.trd.2016.09.015.
Kang, S., 2018b. Why do warehouses decentralize more in certain metropolitan areas? J. Wygonik, E., Bassok, A., Goodchild, A., McCormack, E., Carlson, D., 2015. Smart growth
Transp. Geogr. https://doi.org/10.1016/j.jtrangeo.2018.10.005. and goods movement: emerging research agendas. J. Urban. Int. Res. Placemak.
Kumar, I., Zhalnin, A., Kim, A., Beaulieu, L.J., 2017. Transportation and logistics cluster Urban Sustain. 8 (2), 115–132. https://doi.org/10.1080/17549175.2013.875058.
competitive advantages in the U.S. regions: a cross-sectional and spatio-temporal Yuan, Q., 2019. Does context matter in environmental justice patterns? Evidence on
analysis. Res. Transp. Econ. 61, 25–36. https://doi.org/10.1016/j.retrec.2016.07. warehousing location from four metro areas in California. Land Use Policy 82,
028. 328–338. https://doi.org/10.1016/j.landusepol.2018.12.011.

11

You might also like