Professional Documents
Culture Documents
Retailers choose locations to maximize profit considering location within the city and proximity to other
Journal of Planning Education and Research
OnlineFirst retailers. Major commercial centers form along busy arterial streets with many potential shoppers.
© The Author(s) 2020, Article Reuse Guidelines Traditional commercial cores have small on-street storefronts along these busy commercial roads. The
https://remote-lib.ui.ac.id:2075/10.1177/0739456X20908945
buildings provide a continuum of storefronts, creating a walkable shopping experience. The buildings are
generally small, and stores choose the best location and negotiate with individual property owners for leases.
Planning Research However, there is a lot of inconsistency in uses and quality of the properties. So, developers build strip
malls, indoor malls, lifestyle centers, and mixed-use areas which are operated by one leasing agency that
Retailer Behavior Near Fixed Transit Lines in Los Angeles: A optimizes the retail mix to maximize shopper traffic. While there has been research on clustering uses within
Spatial Autoregressive Probit Model to Evaluate Retail Clustering shopping centers, research on spatial autocorrelations of clustering effects around transit stations has not
been performed.
Craig T. Olwert 1, Chih-Hao Wang 2, Vicente Arellano 3, and David Oulrey 4 Retailers use aesthetics to attract shoppers. Larger malls focus on high-quality aesthetics to make
shopping a leisure activity, increasing length of shopping trips and frequency of visits. However, few studies
examine external aesthetic effects, which should increase foot traffic. Together with the effects of store
Abstract clustering, understanding these built environment effects (i.e., building type and external aesthetics) would
Aggregate studies of retail uses around fixed transit lines and in transit-oriented development are scarce. In
allow planners and decision makers to design or legislate for a better shopping environment and increase the
total, 8,402 retail locations were identified within half mile of twenty-seven stations in Los Angeles County,
USA. Sixteen spatial autoregressive (SAR) probit models help us identify five retail categories (everyday success of stores around transit stations.
uses, opportunists, rivals, hotels, and large footprints) based on how the retail uses self-cluster, cluster with This study is designed to investigate how retail uses cluster or compete around transit stations and each
other uses, choose transit station type preferences, prefer building types and are influenced by other. In total, 8,402 retail uses were recorded across twenty-seven stations on the LA fixed transit system.
gentrification. The results help planners understand that aesthetic improvements and zoning changes are The observations in the survey cover the retails’ store uses, outside physical conditions, building types,
likely to lead to increased restaurant presence and specialized retail clusters, while strip malls further from transit features, and station types. Gentrification status was also computed for the neighborhoods of each
stations beneficially provide locations for everyday uses.
store. To control for the effects of neighborhood qualities, socioeconomic characteristics were obtained
using Census data. Spatial autoregressive (SAR) probit models were estimated for each retail use to account
Keywords for the clustering or competitive effects among the 14 different uses. The SAR models were also used to
retail uses, spatial probit model, fixed transit lines, transit-oriented development, gentrification examine the effects of physical attributes on the retail uses. The model results provide new insights into the
Keywords problem of urban design and land-use planning that can be used to promote commercial activities around
Usos minoristas, modelo espacial Probit, líneas de tránsito fijo, desarrollo orientado al tránsito, transit stations.
gentrificación
Background
1 CaliforniaState University, Northridge, Los Angeles, CA, USA The Los Angeles Transit System
2 CaliforniaState University, Fresno, Fresno, CA, USA
3 New York City Department of Housing Preservation and Development, New York, NY, USA Since the first line opened in 1990, Los Angeles has invested in a fixed transit system to address congestion
4 Willdan Engineering, Elk Grove, CA, USA and pollution concerns. The light rail, subway, and bus rapid transit system has received significant public
support, including two major sales tax initiatives that have raised billions of dollars to continue the system
Corresponding author(s):
Craig T. Olwert, California State University, Northridge, 18111 Nordhoff Street, Northridge, CA 91330, USA. Email: expansion (Metro 2018).
craig.olwert@csun.edu New fixed line stations can change retail through three different mechanisms. First, new stations can
increase consumer presence, increasing sales but possibly changing the retail mix to cater to transit riders.
Introduction Pagliara and Papa (2011) found higher retail property values near new rail stations. The opening of Los
Angeles’ Expo tripled rail transit trips of those living within 1 km of a station (Spears, Boarnet, and Houston
Los Angeles County has invested heavily in its fixed transit system to reduce congestion and increase 2017) which should impact retailers. Second, new TOD with residential components can increase transit
sustainability for the region. Since Metro opened the light rail Blue Line, connecting downtown Los commuters who walk to stations and shop along their walk. Dill (2008) found transit commuting rates of 26
Angeles (LA) to downtown Long Beach in 1990, the system has been constantly expanding with wide percent in Portland, Oregon area TODs versus 7 to 13 percent for the surrounding areas. Lund (2006) found
public support (Metro 2018; Spears, Boarnet, and Houston 2017). With the increased number of stations and that 35.4 percent of all new TOD residents chose to live there because of access to shopping and services.
the high-cost housing in the LA metropolitan area, transit-oriented development (TOD) has been pursued by Built environments with high walkability (such as in TODs) have generally been shown to increase walking
both developers and planners (Loukaitou-Sideris, Gonzalez, and Ong 2017). The region is encouraging half to the store (Cao, Mokhtarian, and Handy 2009) Third, the new development and higher income residents
of new housing and employment growth to be located near transit (Houston et al. 2015). With more can trigger gentrification as was found for Toronto, Montreal, and Vancouver (Grube-Cavers and Patterson
residents and employers around stations, retailers around the stations are reasonably expected to change to 2015). Kahn (2007) found that walk and ride stations generally experience gentrification. Retail stores will
meet the demands of the new customers who may walk to the stores (Lee and Moudon 2006).
change to accommodate the new residents; small independent stores are likely to be replaced by chain properties paid a 0.9 percent market value premium for each increase in Walk Score. In looking at shopping
stores, especially the new mixed-use buildings. Increased rents frequently drive low-margin independent center rents, Rosiers, Theriault, and Menetrier (2005) identified aesthetics of centers as important in
stores out of business and get replaced by uses serving the new higher income residents. Some of the new determining rent levels, but only measured it by the income levels of the shoppers.
everyday uses that move into the area can be beneficial to the preexisting residents, but other new stores no Building types can offer aesthetic and clustering advantages that attract customers. Larger facilities, such
longer serve their needs (Monroe Sullivan and Shaw 2011). One example is the formation of food mirages, as lifestyle centers, indoor malls, and strip malls, have a more consistent aesthetic controlled by one
where gentrifying areas attract high-end supermarkets that are not affordable to low-income residents landlord. Large shopping centers optimize their store mix to obtain the highest rents (Eppli and Benjamin
(Breyer and Voss-Andreae 2013). 1994). Indoor shopping malls use anchor stores, which pay lower rents, to attract customers to smaller
While case studies have been performed on change in retail uses in gentrifying areas (Loukaitou-Sideris, specialty clothing stores, which pay higher rents (Rosiers, Theriault, and Menetrier 2005). Lifestyle centers,
Gonzalez, and Ong 2017; Monroe Sullivan and Shaw 2011), and the effect on rental rates and property a newer shopping format, focus on a mix of restaurants, specialty retailers and entertainment venues (Little
values of commercial and residential properties around fixed transit systems has been studied (Pagliara and 2006). The lease agreements frequently require enough specialty stores to create a cluster that will attract
Papa 2011; Pan 2013; Pivo and Fisher 2011), no studies look at the aggregate retail mix around transit customers (Little 2006). Mixed used, stand-alone and on-street buildings are generally smaller stores that are
stations. The Los Angeles system is continuing to expand, and recent TOD laws will increase residential not controlled by one landlord, so aesthetics are highly variable and there is no coordinated effort to provide
development. Between the new residents and the new buildings with modified retail spaces, understanding clustering benefits (Eppli and Benjamin 1994).
retail patterns around transit stations is beneficial for planners.
Spatial Modeling in Discrete Choice Studies
Factors Effecting Retail Uses Retailers can benefit from clustering near each other as it allows for more convenient comparison shopping
Retailers choose their locations based on the optimal mix of factors, including neighborhood characteristics, (Eppli and Benjamin 1994; Rosiers, Theriault, and Menetrier 2005). In deciding on locations, retailers are
stations types, aesthetics, and building types, to maximize profits. influenced by the presence of other retailers (Larsson and Oner 2014; Oner and Larsson 2014). Some
Retailers want to conveniently serve their customers and consider neighborhood characteristics that retailers self-cluster, as classically suggested by auto malls, while others are competitive and will locate
affect customer visitation and purchasing patterns. While areas with higher incomes and education levels away from each other where they can serve the local customers (Eppli and Benjamin 1994). Similarly,
have more disposable income and a greater likelihood of purchasing experiences, Schuetz (2015) found retailers may benefit or be disadvantaged by being near different types of retailers, which is why landlords
these higher income areas had less retail employment. Areas with higher black and Asian populations are optimize store mix in larger facilities (Eppli and Benjamin 1994; Little 2006).
likely to have different store mixes due to income and cultural differences; Schuetz (2015) found areas with Spatial statistical modeling allows researchers to better understand neighborhood effects that affect an
higher black populations had less retail employment. If TOD includes residential development, the new individual’s decision. Discrete choice models use utility theory to understand individual behavior but do not
potential customers increase population density which Schuetz (2015) found to be a significant predictor of account for spatial autocorrelation and the complex spatial relationships between individuals (Paez and Scott
retail employment. Rosiers, Theriault, and Menetrier (2005) found higher rents were charged in shopping 2007). SAR models have explained behavior among individuals in the same neighborhoods. A SAR model
centers that served larger populations using a normal gravity model, but when accounting for income, lower can be used to study clustering behavior between different retail uses. Larsson and Oner (2014) use a
income areas have lower rents. So, if a TOD starts causing an influx of new middle- and high-income discrete negative binomial model to explain retail location and evaluate proximity to similar and other retail
residents, gentrification may occur. Gentrification of older and originally poorer neighborhoods is likely to activities; however, they did not use SAR for the spatial analysis. SAR can provide new insights into
have significant changes in the neighborhood retail mix (Monroe Sullivan and Shaw 2011). retailers’ clustering behavior with themselves and other uses. The findings can be used to group retailers that
The TOD literature has recognized that station types are important in predicting transit usage and share the same physical environment and provide implications for urban design and land-use planning near
development that occurs around the stations (Atkinson-Palombo and Kuby 2011; Bollinger and Ihlandfeldt transit stations.
1997; Hess and Almeida 2007; Houston et al. 2015; Ratner and Goetz 2013; Zuk et al. 2018). Station types
are a measure of the built environment which influences residential travel behavior (Cao, Mokhtarian, and Data and Methods
Handy 2009). Ratner and Goetz (2013) found that all of Denver’s transit stations had some new retail
Data Collection
development, but most occurred at the downtown stations. Different station types included in the literature
(Atkinson-Palombo and Kuby 2011; Ratner and Goetz 2013) include downtown stations with high-density To better understand retail behavior around transit stations in Los Angeles, retail locations were collected for
commercial uses, urban centers with a smaller commercial core surrounded by high-density residential uses, all stores within half mile of twenty-seven randomly selected stations in the Los Angeles Metro heavy rail,
campus stations with many students that make a unique clientele mix, urban poverty stations with high light rail, and bus rapid transit system as shown in Figure 1. Retail data were collected by five trained
transit ridership and low personal vehicle ownership, and urban and suburban residential neighborhoods that individuals between the May 2014 and May 2015. We collected all retail uses, non-retail stores located in
are not destinations due to little retail development or employment opportunities but the source for many retail locations, and vacant parcels (and parking lots) that could be developed as retail. If a building was
riders. No study has evaluated how retailers respond to station types. primarily office, only the retail uses were mapped. The primary entrance to the store was mapped. For larger
Aesthetics attract shoppers as evidenced by corporate architecture and higher rents for aesthetically indoor malls, the stores were located approximate to their location within the buildings.
pleasing areas. However, studies have not evaluated the impact of street trees and sidewalk quality on
retailer location choices. Forsyth et al. (2008) found that residents walked more miles per day with the
presence of sidewalks, transit, and less litter and graffiti. Pivo and Fisher (2011) found that office and retail
Figure 1. Los Angeles fixed transit system with only the sampled stations labeled.
Source: https://www.metro.net/riding/guide/system-maps/.
The data collected included the name of the store, whether the store is a chain (later verified via web
searches), whether a storefront is vacant, the presence of street trees, whether the sidewalk was wide or
good, and the presence of a drive thru. The collector indicated if they thought the building was new (since
the station was opened) and was then was verified by comparing with the year built from the Los Angeles
County Auditor data. Finally, the store types were collected with an assignment of NAICS categories. The
category assignments were checked for consistency before analysis.
To provide explanatory and control variables of retail locations, the U.S. Census American Community
Survey 2013 five-year average data based on Census tracts of the store location was used. Gentrification
status was computed for each store neighborhood using a methodology similar to that outlined in Freeman
(2005). The variable descriptions and descriptive statitics are shown in Table 1.
Results
A total of 16 SAR probit models were estimated with the results shown in the Appendix. This section will
present the summarized results of all the models. First, the desire of an individual use to cluster will be
discussed. Because the model results demonstrate patterns between individual uses, store groupings are
created to provide a better understanding of the results. Then results will be presented for variables that
planners can influence: building types, urban infrastructure, and station types.
(2)
Equation (2) is then extended to include local proximities, Wxq, which represents the spatially proximate
retail uses (see equation (3)). A 100-meter buffer is used to identify all the neighboring retail uses for the
spatial weight matrix, W, with an average of 60.87 neighboring retail uses at an average distance of 58.64
meters. With the defined W, SAR probit models are estimated using the Bayesian approach developed by
LeSage and Pace (2009), using a large random sample simulated from a posterior density function.
(3)
The modeling results not only determined the self-clustering behavior of retail uses but also identified
Note that the signs of estimated coefficients, ρ and γq, could be either positive or negative, depending upon complimentary (γ > 0) and competing (γ < 0) retail uses. With major consideration given to self-clustering
the nature of business relationships between a retail use and other adjacent uses. A positive sign implies a behavior and complimentary and competing uses, and less consideration to preferred building types and
complementary effect that the same or similar uses would tend to cluster, and a negative sign indicates a station types, five store groupings were designated to help explain retail locational choices (see Table 3).
competing effect that a retail use would avoid clustering.
likely to be located. The urban centers, and especially areas with high ridership, provide guests with a
convenient way to arrive at the hotel.
Large footprint uses include religious, auto industry, furniture, gas stations, and health care uses because
their lot sizes are larger than the other uses except hotels (not shown here). They do not like locating near
restaurants, retailers, and clothing stores or locating in mixed-use, strip malls, indoor malls, or lifestyle
centers. In fact, auto industry and gas stations both prefer stand-alone buildings. However, three of the uses
(religion, auto, and health care) do cluster with themselves; furniture stores do not, similar to Oner and
Larsson’s (2014) finding that household appliance stores like retail diversity. Most of these uses prefer not to
be downtown, where land prices are at a premium, except for gas stations, likely due to the inclusion of
parking lots in this category.
Building Types
Everyday uses are store types that people need on a regular basis and visit frequently. This being the case,
these retailers locate near residents and each other to promote foot traffic. Restaurants, grocery, laundry, and It is good for planners to understand what store types are likely to locate in different building types (see
Table 4).
barbers best fall in this category, which is similar to Lee and Moudon’s (2006) finding that more walking
occurs in neighborhoods where grocery, restaurants, and retail uses are nearby. Oner and Larsson (2014),
using factor analysis, also identified a category “mix of mostly small-scale stores selling common goods”
which fits the everyday store grouping. Of these four uses, restaurants are the most different, because they
like to cluster with themselves, whereas the other uses do not like self-clustering. Popkowski Leszczyc,
Sinha, and Sahgal (2004) found that single-purpose shoppers do frequent the closest grocery stores leading
to a location strategy of dispersing grocery stores. They all prefer being in strip malls and mixed-use
buildings. They have no preference for station types because they want to be everywhere; the only exception
is that restaurants prefer campus locations.
Opportunists include FIRE and arts uses. They have few preferences. People infrequently purchase from
these store types, so opportunists do not pay for premium locations. The arts retailers co-locate with gas
stations, an unpleasant next-door neighbor but not near high ridership stations with higher rents. FIRE uses
are located at most station types, but slightly prefer suburban stations, which have less office space
alternatives. Opportunists have some quality requirements: they do not like being near auto uses.
The rivals are retail and clothing stores. They both like to cluster with themselves so much that they
exclude the other, consistent with Larsson and Oner’s (2014) finding that clothing stores do not like retail
diversity in their immediate neighborhood. Arentze, Oppewal, and Timmermans (2005) also found that of
their three categories, the clothing and shoe category did not benefit from the presence of grocery and other-
goods categories. However, retail and clothing stores tend to behave similarly. They like to cluster around
barbers, perhaps due to a luxury effect; customers willing to spend at barbers, salons, and spas are more
likely to continue the beautification process with new clothes or other goods. However, they do not like
locating near health care facilities. They both prefer indoor malls, with retailers preferring lifestyle centers
and clothing stores preferring mixed used opportunities, but neither likes being in stand-alone buildings.
They also like wide sidewalks, a measure of improved aesthetics. Retailers are strongly attracted to
downtown stations while clothing stores are more broadly spread with a dislike for campus and industrial
areas. Part of the retailers’ downtown preference is due to the large number of jewelers in the downtown Los
Angeles Jewelry District.
Hotels are not a true retail use, but influence retail uses and compete for commercial locations. Hotels do
not like to cluster but do like being near restaurants so that their guests can have access to convenient food.
They do not like wide sidewalks or mixed-use, strip malls, indoor malls, or lifestyle centers. But they do like
to locate in urban centers and industrial areas. Industrial areas have larger lots that allow building new
suburban style hotels with easy auto access and surface parking, also an area where wide sidewalks are not
Of the large footprint stores, only furniture did not like stand-alone buildings. Hotels dislike the four
building types, suggesting they prefer on-street and stand-alone formats. Stand-alone hotels have surface
parking which provides easy auto access.
Mixed use buildings (MIXU) frequently have office and residential uses above retail uses. Everyday uses
like mixed used buildings which provide good access to many potential customers. Rivals also like this
setting. This building type is most likely to have vacancies.
Strip malls (STRI) are preferred by everyday uses and chains. The auto-oriented setup allows for
frequent shoppers to pick-up all types of goods. Shoppers like the convenient parking (Eppli and Benjamin
1994) and multi-purpose shoppers like the opportunity to stop at a few conveniently located stores
(Popkowski Leszczyc, Sinha, and Sahgal 2004). Retailers, health care, and FIRE uses also locate in strip
malls, but do not prefer strip malls to on-street locations.
While indoor malls (INDO) are preferred by rivals, lifestyle centers (LIFE) are preferred by restaurants
and retailers. This is consistent with retail differences expected with lifestyle centers (Little 2006) and the
indifference for the arts for lifestyle, where arts dislike most other building types, indicates the change to
include more entertainment in the lifestyle center mix. Note that some of the indoor malls included swap
meets which mostly contain independent stores. Only one regional mall was in the sample (Westfield
Topanga Mall).
Acknowledgments
The authors thank Jonathan Caballeros and Marisol Zamar for their data collection work.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this
article: This research was funded by a California State University, Northridge, College of Social and Behavioral
Sciences research Grant.