An Empirical Analysis of Store Locations
in Planned Regional Shopping Centres*
DRAFT: Please do not quote or cite
Zhen He
Department of Economics
University of Alberta
May 2009
Abstract
In Canada and the U.S., shopping centre leases may contain restrictive clauses that give a
department store veto power over the admission of new tenants and power to influence
the store locations in the shopping centre. Regarding the internal structure of shopping
centres, the interests of shopping centre developers and department stores might not
coincide. In the presence of conflict of interests, the internal composition of shopping
centres would depend upon the relative bargaining positions of the developer and the
department store. In this paper, I investigate the store configurations in an area close to
department stores in planned regional shopping centres. Using maps showing the specific
locations of stores in the shopping centres and data on characteristics of shopping centres,
developers, and department stores collected from 148 planned regional shopping centres
in the five westernmost Canadian provinces in 2007, I examine the relationship between
the characteristics of stores near entrance of a department store and indicators of
bargaining power of developers and department stores. This paper contributes to the
literature in the empirical investigation of the store locations within shopping centres.
* Acknowledgement: I’m grateful to my supervisors Dr. Douglas West and Dr. Andrew Eckert for their
constant support and guidance. I also thank Dr. Robin Lindsey, Dr. Stuart Landon, and Dr. David Ryan for
their valuable comments and suggestions and thank Dr. Benjamin Atkinson for his instruction for using
Arcview. Any remaining errors and omissions are the author's responsibility.
Correspondence: Zhen He, Department of Economics, University of Alberta, 8-14 HM Tory, Edmonton,
AB, Canada T6G 2H4; Tel: 780-4281898; E-mail: zhenh@ualberta.ca.
1.
Introduction
As one of the most important retailing phenomena, shopping centres have been widely
investigated since they appeared in the early 20th century. The theoretical work of Eaton
and Lipsey (1979, 1982) and Stahl (1982a, 1982b) on the agglomeration of firms can be
used to explain the formation of shopping centres: the clustering of stores selling similar
goods can be explained in terms of the comparison-shopping tendencies of consumers
and the locational response of profit-maximizing firms (Eaton and Lipsey, 1979; and
Stahl, 1982a, 1982b), whereas the clustering of stores selling different goods can be
explained in terms of the multipurpose shopping tendencies of consumers and the
locational response of profit-maximizing firms (Eaton and Lipsey, 1982).
Most existing studies of shopping centres have focused on the choice of tenant
mix by the shopping centre owner (Dawson, 1983; and Brown, 1992) and the way in
which shopping centre rents are set (Benjamin et al., 1990, 1992; Gatlaff et al., 1994;
Gerbich, 1998; Gould et al., 2005; and Pashigian and Gould, 1998). Other studies have
investigated the hierarchy of shopping centres (West et al., 1985, 1988; and Ryan et al.,
1990) and shopping centre similarities (West, 1992; and Golosinski and West, 1995). The
locational pattern of stores within shopping centres, however, has been largely
overlooked. This is an interesting issue because the way in which retailers are located is
not random. Within a shopping centre, neighbouring stores generate either positive or
negative demand externalities for each other. When deciding on store locations,
developers of shopping centres would take into account this issue. The location problem
is further complicated by the possibly divergent interests of different parties regarding
store locations within the shopping centre. Although there have been some studies in the
internal composition of shopping centres (e.g. Brown, 1992; Carter and Haloupek, 2002;
and Carter and Vandell, 2005), they are either descriptive or overlook some important
factors. This issue needs more investigation, because shopping centre is the primary
method of retailing for many industries and the internal composition of a shopping centre
plays an important role in its success. This paper will fill this niche and contribute to the
knowledge of shopping centres.
In this paper, I investigate the locational pattern of stores in the proximity of
department stores in planned regional shopping centres. Department stores carry a variety
of comparison and multipurpose shopping goods. They are anchors, attract customers to
visit the shopping centres where they are located, and enjoy substantial rent subsidies
(Pashigian and Gould, 1998 and Gould et al., 2005). Locating non-anchor stores that sell
comparison shopping goods near department stores could facilitate comparison shopping
opportunities in such an area, attract comparison shoppers, and increase the profits of the
non-anchor stores and the rental income of the developer. However, it is also possible
that such a configuration could intensify competition and decrease profits of non-anchor
stores. Thus, the interests of developers regarding store compositions near department
stores are uncertain. With respect to department stores’ interests regarding locations of
non-anchor retailers selling comparison shopping goods relative to them, they are
expected to be ambiguous as well. Department stores might be harmed by the substitution
effect but might also benefit from the market area effect generated by such retailers.
Although the interests of developers and department stores regarding the location pattern
of stores near department stores are ambiguous, their interests might not coincide, given
the fact that the department stores only care about their own profits whereas developers
2
care about the rents they can extract from the entire shopping centre. It is the possible
conflict of interest between developers and department stores that makes the store
configuration close to a department store an interesting issue.
In a planned shopping centre, a developer would be expected to have the power to
decide where a store will be located in its centre. However, shopping centre leases may
contain restrictive clauses that give department stores veto power over the admission of
new tenants (Wunder, 1988) and power to influence the locations of other stores in the
shopping centre (Harvard Law Review Association, 1973; and Mason, 1975). In the
event of a conflict of interest, the locational pattern of stores in an area close to
department stores is expected to depend upon the relative bargaining power between the
developer and the department store.
Using data collected from 148 planned regional shopping centres in 2007 in the
five westernmost provinces in Canada, I examine the relation between variables that
reflect locational pattern of stores close to a department store and variables that indicate
the bargaining power of a developer. The results of estimation show that the fraction of
spots near department stores that are leased to stores selling comparison shopping goods
is bigger in regional shopping centres that have a larger gross leasable area and are
opened earlier, and is negatively related to the number of department stores contained in
a regional shopping centre. A regional shopping centre’s gross leasable area and age and
the number of department stores in a regional shopping centre are expected to reflect a
developer’s bargaining power. Hence, the findings that these variables have explanatory
power in the composition of stores near department stores are consistent with the
hypothesis that the locational pattern of stores in an area close to department stores would
3
depend upon the bargaining power of a developer. In addition, the site-specific variable
indicating whether the observation is the main entrance of a department store and
whether it is on a floor above the ground floor are estimated to have a significant impact
on the locational pattern of stores near a department store as well.
This paper contributes to the literature in the empirical investigation of shopping
centres. While bargaining theory is now well developed 1 , the empirical literature on
bargaining is limited. Most studies concerning bargaining focus on union contract
negotiations and on labour disputes, i.e., strike activity and strike duration (Kennan and
Wilson, 1993; and Morton, Zettelmeyer and Silva-Risso, 2004). This research will
analyze the impact of factors related to the bargaining power of developers and
department stores on store locations within shopping centres.
The remainder of this paper is organized as follows. Section 2 briefly reviews the
existing studies in the literature that are related to the internal composition of shopping
centres. Section 3 presents a discussion of the relevant economic theory. Section 4
describes data sources and the regression models. Section 5 presents the summary
statistics and the results of the econometric estimation. Finally, Section 6 is the
conclusion.
2.
Existing Studies on Shopping Centre Composition
In the shopping centre industry, it has been suggested that developers follow some rules
of thumb when deciding on store locations in their shopping centres. As reviewed by
1
For a detailed survey of the theoretical models of bargaining, see Osborne and Rubinstein (1990), and
Muthoo (1999).
4
Brown (1992, p. 386) and Carter and Haloupek (2002, p. 291-292), these rules include:
keep department stores away from a centre’s main entrances; cluster service shops and
put them close to the entrances; keep pet shops and dry cleaners away from food shops
and keep food shops away from apparel shops; cluster men’s apparel shops, women’s and
children’s shops, and food retailers; separate jewellery stores, record stores, and hardware
stores; and locate fast food retailers, jewellery and electronics in high traffic areas. These
recommended rules are mainly descriptive and lack relevant theoretical discussion. In
addition, some rules are questionable. For example, one of the rules suggests that
jewellery stores be separated. Yet, when shopping for jewellery, one might want to
compare quality, style and price before purchasing. Thus, clustering jewellery stores
would attract comparison shoppers.
Brown (1992) studied the movement of shoppers in 250 malls in Belfast and
found that customers who visit one store are likely to visit another store of this type when
similar types of stores are located close to one another, such as food shops and ladies’
apparel shops. Such an extent of “customer interchange” is “substantially greater than
that between similar, spatially separated shops and contrasting outlet types in close
proximity” (Brown, 1992, p. 398). Brown’s findings support the conventional wisdom.
He also found that service units benefit from customer traffic when located close to
dissimilar shop types.
Carter and Haloupek (2002) extended the spatial economic model of Ingene and
Ghosh (1990) and investigated the dispersion of stores by store type in malls. Using data
collected from nine malls across the US from 1991 to 1992, they found groupings of
5
stores of the same type for the following store types: men’s apparel shops, women’s
apparel shops, and shoe stores. Stores of the other types exhibit dispersion.
Other empirical studies that focus on the internal composition of malls include
Carter and Vandell (2005) and Eckert and West (2008). Carter and Vandell (2005)
assume that the centre of a mall has the highest customer traffic and different types of
retailers bid for location and space. Assuming profit-maximizing behaviour on the part of
mall developers and stores, they constructed a model in which the total profit of a mall is
determined by variables including price per unit of goods sold, quantity of goods sold,
size of a store, customer traffic and the proportion of customer traffic that purchases,
miscellaneous costs (e.g., labour cost, maintenance cost, utility cost, etc.), and rent.
Based on their model, Carter and Vandell developed the following hypotheses: (1)
non-anchor store rents and sales volume decrease when the store is farther away from the
mall centre, (2) the rents of different types of stores decrease at different rates with
increasing distance from the centre, and (3) non-anchor store size increases with distance
from the centre of the mall. In addition, Carter and Vandell predict that store types with
high sales response to customer traffic (e.g. the food court) and high price per item (e.g.
jewellery stores) will be smaller and locate close to the mall centre, because they are able
to bid for the higher rent per square foot for the location, whereas store types with low
sales response to customer traffic and lower-priced goods relative to their costs (e.g.
family apparel and housewares) will locate away from the centre and near the anchors.
Using data from a sample of 689 leases in eight regional and super-regional shopping
centres of the US, Carter and Vandell tested the above hypotheses and provided
supportive empirical evidence.
6
Carter and Vandell’s discussion is helpful in interpreting the locational pattern of
stores in malls. Unfortunately, they ignore the demand externalities generated by the
comparison-shopping tendencies of consumers and their impact on locational choice.
Family apparel and housewares stores cater to comparison shoppers and provide
merchandise that overlap with that of department stores. Locating these stores close to a
department store is consistent with the incentive of developers to facilitate comparisonshopping opportunities. In addition, based on Carter and Vandell’s proposition, jewellery
stores will locate in the centre of a mall. This proposition is questionable since jewellery
stores are often found in an area close to department stores.
Eckert
and
West
(2008)
analyzed
the
implications
of
the
relative
developer/retailer bargaining for the “radius restrictions”, a restrictive clause that is
imposed by developers on tenants, prohibiting tenants in a particular centre from opening
another store within a certain radius. Using data collected from regional malls in western
Canada, they investigated the tenant mix of neighbouring shopping centres and estimated
the probability that a retail chain appears in both centres as a function of chain
characteristics, mall characteristics, and location characteristics. They provided empirical
evidence that is consistent with the hypothesis that whether a chain enters neighbouring
malls depends on the relative bargaining power of developers and retail chains. Although
Eckert and West investigate another facet of the internal composition of regional malls,
their discussion of demand externalities and bargaining powers sheds light on the issues
analyzed in this paper.
Theoretical studies that focus on the internal composition of malls include those
by Brueckner (1993), Miceli et al. (1998), and Vitorino (2008). Brueckner investigated a
7
developer’s space-allocation problem in the presence of externalities among stores and
proposed that developers allocate space to a store until its marginal revenue from an
additional square foot equals the marginal cost of space minus the marginal increase in
sales enjoyed by all other stores due to the spillover effect. Brueckner’s analysis, later
supported by Miceli et al. (1998), explains why department stores with higher externalitygenerating ability will have a larger space than that of other stores in a shopping centre.
Both studies emphasize the role played by demand externalities in determining the size of
the space that a store type should be allocated in shopping centres. However, their models
say nothing about where a store should be located in the presence of demand externalities.
Vitorino (2008) constructed a model of entry that allows for either positive or
negative externalities among firms and explained the joint entry decisions of department
stores in a regional mall. She broke department stores into three categories: upscale
department stores, mid-scale department stores, and discount department stores, and used
an optimization approach, which includes maximizing the likelihood function subject to
the constraint that the equilibrium conditions given by the economic model are satisfied,
to estimate the impact of strategic effect (through positive or negative demand
externalities) among different types of department stores, mall-specific demographic
variables, and the store-specific characteristics on the profit and entry decision of a
department store.
Using data collected from all US regional malls, Vitorino (2008) found that, at the
median level of the demographic variables, none of the department stores has incentives
to enter a regional shopping centre on its own. This finding is consistent with the
prediction of the model that it is beneficial for two different firms to enter together in the
8
same regional shopping centre in the presence of large and positive demand externalities.
In addition, the results show that (1) mid-scale department stores are strategic
complements, whereas upscale department stores are strategic substitutes, (2) upscale
department stores benefit considerably from the presence of mid-scale department stores,
but not vice versa, and (3) discount department stores have no positive effect on any other
type of department store.
Vitorino also found that several mall-specific demographic variables and storespecific characteristics are important in anchor stores’ profit and entry decisions. For
example, the results show that the purchasing power (measured by the variable
population × income) of the area surrounding a shopping centre has a positive effect for
upscale stores and a negative effect for discount and mid-scale department stores. In
addition, based on the results, a department store’s profits increase when it has a bigger
square footage and is located in a shopping centre that is opened more recently. Upscale
and midscale department stores prefer to anchor larger centres.
3.
Theoretical Discussion
3.1. Planned Regional Shopping Centres and Department Stores
Each city can sustain a hierarchy of shopping centres 2 , where the hierarchy has two
important characteristics: (1) a small number of large centres and a large number of small
centres; (2) centres at various levels differ in the choice of anchor tenants, the
composition of store types, GLA (gross leasable area), and store counts. In addition,
2
A shopping centre hierarchy was originally derived on the basis of Christaller’s (1966) theory of central
places. Eaton and Lipsey (1982) explain the formation of the shopping centre hierarchy based on the
interaction of consumers’ multipurpose shopping behavior and firms’ profit maximizing locational
response.
9
centres at each level contain all of the shop types that can be found in lower level centres,
plus additional ones (West et al., 1985). Based on West et al. (1985), regional shopping
centres have the following characteristics: (1) they have at least one department store, (2)
contain a large variety of store types that include stores selling convenience goods or
providing personal services (e.g. food, drugs, laundries, tailor’s shop), plus stores selling
comparison shopping goods that could be purchased on a multi-purpose shopping trip
(e.g. shoes, clothing, jewelry), and (3) have multiple stores selling similar comparison
shopping goods (e.g. multiple ladies’ apparel shops and shoe shops). In this paper, I study
how the pattern of store locations near department stores in a shopping centre is related to
the relative bargaining power of a developer and a department store. For this purpose, the
focus is on planned regional shopping centres. The reason for focusing on planned
regional shopping centre is because department stores are largely found in such centres
but rarely in centres at lower levels of the shopping centre hierarchy.
Department stores are of vital importance to regional shopping centres, and they
determine the size, character and success of a shopping centre (Vitorino, 2008). Such
influence is expected to come from (1) the department store’s attractiveness to customers,
and (2) its ability to help developers get outside financing for the centre (Harvard Law
Review Association, 1973). Department stores provide a large variety of products and
services, which attract customers to visit the shopping centres where they are located and
increase the sales of the rest of the stores in the centre, as well as increasing their rent
(Pashigian and Gould, 1998 and Gould et al., 2005). Gatzlaff et al. (1994) found that the
rental rates of non-anchor tenants declined by an estimated 25% in response to the loss of
an anchor.
10
Another source of department stores’ influence inside regional malls comes from
their ability to help developers get outside financing (Harvard Law Review Association,
1973). To develop a shopping centre, outside financing is usually critical to a developer.
Mortgage lenders typically require that 60 to 70 percent of total floor area be under longterm lease to firms with low credit risk before financing is approved (Kinnard and
Messner, 1972). Leases with department stores allow developers to easily fulfill the term,
floor area, and credit risk conditions imposed by lenders. At this point, other retailers
cannot help developers as much as department stores.
Because of the above two reasons, department stores are crucial to regional
shopping centres. Developers attract department stores to mall locations by charging
them lower rents (Pashigian and Gould, 1998; and Gould et al., 2005). Specifically,
department stores “receive a per foot rent subsidy of no less than 72 percent that which
other stores pay” (Pashigian and Gould, 1998, p. 115). Furthermore, department stores
usually negotiate some privileges in shopping centres where they locate. Shopping centre
leases may contain restrictive clauses, including “the right of approval or veto clause,
which intends to give an anchor or major tenant veto power over the admission of new
tenants” (Wunder, 1988, p. 30) and power to influence the locations of other stores in the
shopping centre (Harvard Law Review Association, 1973; and Mason, 1975).
3.2. Conflict of Interests
When a department store and a developer negotiate a new agreement, they enter a
bargaining situation. They share a common interest, i.e. both of them wish the customer
traffic for the shopping centre to increase, and hence have an incentive to co-operate.
11
However, their interests are possibly divergent: department stores only care about their
own profits whereas developers care about the rents they can extract from the entire mall.
In this section, the possibly divergent interest regarding store configuration near
department stores between developers and department stores is discussed.
(1) Developers’ Incentive
Developers are expected to have a clear interest in increasing non-anchor stores’3 sales
volume and profit. Firstly, because department stores enjoy substantial rent subsidies,
developers’ income largely depends on non-anchor mall stores’ rental payments 4 .
Secondly, developers would want to increase non-anchor stores’ profit due to the nature
of their rent. Store rent generally consists of two parts: base rent, which is fixed, and
percentage rent, which increases with sales volume after it passes a certain threshold
(Benjamin et al., 1990, 1992). The rents of anchors are largely fixed and independent of
sales volume, whereas the rents of non-anchor stores are based on sales volume (Gould et
al., 2005). Hence, in order to increase rental income, one would expect developers to
have an important incentive to increase the sales volume and profit of non-anchor stores.
Inside a regional shopping centre, due to the attractiveness of department stores to
customers, the area close to department stores would have high customer traffic and
stores located near department stores would be important to developers because of their
potential rental payments. The findings of Yeates et al. (2001) provide supportive
3
Non-anchor stores include all retailers except the department stores that one may find in a planned
shopping centre.
4
Using data collected from 35 malls across the US in 1994, Gould et al. (2005) showed that anchors
occupy over 58% of the total leasable mall space and yet pay only 10% of the total rent collected by the
developer.
12
evidence. They investigated the impact of Eaton’s closure on the sales volume and profit
of the other retail stores in the shopping centre. Using data collected from 18 regional
centres in eastern Canada, they found that a store’s profit is related to how far it is located
from a department store and that different store types are not affected equally. Their
results show that, within 100 feet of Eaton’s entrance, merchandise categories most
affected by Eaton’s closure are variety stores, food and ladies wear. In addition, stores
offering fashion accessories, gifts/books/stationary, shoes and men’s wear are also greatly
affected in their sales.
The next question is in regard to the store types that a developer would want to
locate close to department stores. Instead of looking at the specific store category (e.g.
shoe stores and hair salons), I broke non-anchor stores into two types, C stores and M
stores. These terms are defined by Golosinski and West (1995). C stores are defined as
stores that cater to comparison shoppers on a multipurpose shopping trip. C stores sell
merchandise that is both horizontally and vertically differentiated, and customers conduct
comparisons before making a final purchase. Examples of C stores include clothing
stores, shoe stores, and jewellery stores. M stores are defined as stores that cater to
multipurpose shoppers on a multipurpose shopping trip. M stores sell products that are
more frequently purchased, and consumers engage in little search across M stores of a
given type. M stores tend to be the same or very similar. The exception is M stores that
sell services, which vary in price and quality. Examples of M store types include
bookstores, liquor stores, supermarkets, and dry cleaners (Golosinski and West, 1995).
When shopping at C stores like apparel shops, customers may wish to visit
different stores to compare price, quality and style. If comparison shopping goods are
13
carried in both department stores and non-anchor stores, a comparison shopper may wish
to visit stores of both types. Locating C stores in an area close to a department store could
therefore facilitate comparison shopping and increase a developer’s rental income.
However, it might intensify competition in such an area and decrease profits of both the
department store and the retailers. Concerning this matter, developers might want to
locate other types of stores, for example, M stores, near department store.
(2) Department Stores’ Incentive
Regional shopping centres contain a variety of store types. The store types that
department stores would want to have in their proximity are expected to be those that
increase the sales volume and profit of department stores. Department stores provide
products and services that cater to comparison shoppers as well as multipurpose
shoppers. Having C stores in their proximity would enhance comparison opportunities in
this area, generate positive demand externalities, and attract comparison shoppers.
However, it is possible that the department stores’ sales volume and profit could be
diluted by competition coming from these stores. In addition, being close to M stores that
sell merchandise that is also carried by the department stores (e.g. beauty shops,
houseware) could steal customers from the department store as well. Considering this,
department stores might have an incentive to be close to non-retail firms (e.g. doctors,
lawyers, or real estate firms) as their neighbours to avoid direct competition of any sort.
However, it is uncertain whether such stores could generate positive demand externalities
for department stores, because customers of such stores are expected to be “target
shoppers” and might not visit other stores on the same shopping trip. Based on the above
14
discussion, department stores’ interests regarding location pattern of stores in their
proximity is ambiguous.
The interests of department stores and shopping centre developers might be
divergent, and could therefore depend on bargaining power. If this is so, then variables
measuring bargaining power of developers should be significantly related to the
locational configuration around a department store. It is an empirical question as to
whether such a relation exists, and this will be explained in Section 5 and 6.
3.3. Developers/Department Stores Bargaining
As discussed in the previous section, to avoid being close to store types that might
intensify competition and decrease the profits of department stores, department stores
would want to negotiate restrictive clauses that give them the ability to determine store
locations in the shopping centre. Such clauses are vertical restraints in economic terms.
Whether a department store can successfully negotiate such clauses with a developer
depends on the relative developer/department store bargaining power.
A player’s bargaining power is its ability to influence the other players and to
negotiate a bargaining outcome that is favourable to it. In the rest of this section, factors
that are expected to influence the relative developer/department store bargaining will be
discussed in the specific context of shopping centres.
First, a department store might not be successful at imposing on large developers
operating multiple regional shopping centres a restrictive clause that grants it power to
determine the store configuration. A developer’s effort plays an important role in the
success of its shopping centre. However, a developer may engage in an insufficient
15
amount of advertising and centre promotion, underinvest in the maintenance of common
areas as well as in the maintenance of the centre’s structure and operating systems
(Golosinski and West, 1995). Given the potential moral hazard problem, a developer that
operates multiple shopping centres may be more attractive to department stores and
therefore have a more favourable bargaining position than a developer that has a single
centre, as it may have less incentive to shirk on advertising expenditures and centre
maintenance in all of the shopping centres it owns.
Another potential factor is a shopping centre’s GLA (gross leasable area). Centres
with a large GLA can contain more stores, which facilitate comparison and multipurpose
shopping and are attractive to customers and are, thus, attractive to department stores and
giving the developer a more favourable bargaining position. Nevertheless, a smaller GLA
could also give a developer a more favourable position. The reason is because there will
be limited opportunities to serve a given geographic area, and a developer may control
area within a centre with a smaller GLA. There may be more intense competition for a
spot in that centre than a spot in a larger centre or in a centre where tenants have
alternative locations nearby. Given the above discussion, the relation between the GLA of
a regional shopping centre and the shopping centre developer’s bargaining power is
indeterminate5.
The age of a shopping centre could also show a developer’s bargaining power. A
centre opened earlier might have built a reputation and customer loyalty. However,
compared with newer ones, older centres “tend to embody older shopping centre
technology and, in the absence of costly maintenance and renovation, will display more
5
Eppli and Shilling (1996) shows that the aggregate retail sales at regional malls are largely determined by
centre size.
16
signs of deterioration than newer centres” (Ryan et al., 1990, p. 316), which are not
appealing to retailers and customers. Thus, the relation between a developer’s bargaining
power and the age of a centre is ambiguous6.
A developer’s bargaining power would also depend on the competition it faces
from neighbouring centres at the same level of the shopping centre hierarchy. If the
developer owns every neighbouring regional shopping centre, one would expect it to
exercise power to tenants that seek position in this type of shopping centre in this
geographic area. A department store may compromise on locational pattern of stores in
its proximity to seek locations in this area since it has no “outside option7”. Therefore, the
competition a developer faces from its neighbouring centres could affect its bargaining
power.
The population that a shopping centre serves is expected to reflect a developer’s
bargaining power because it signals the potential customer base and purchasing power,
and hence, has implications for a department store’s profits. A department store may
compromise on the composition of stores close to them in order to enter a shopping
centre that serves a large population.
In addition, the number of department stores in a shopping centre may indicate a
developer’s bargaining power as well. There are only four department store chains in
Canada. As the number of department store located in a regional mall increases, there are
6
Vitorino (2008) shows that, everything else held constant, department stores in regional shopping centres
that are constructed more recently have larger profits than those located in older ones. The results of Eppli
and Shilling (1996) show that the impact of the age of regional shopping centre on the centre’s aggregate
retail is negative but is statistically insignificant.
7
According to Muthoo (2001), a player has an outside option if he/she can make a deal with a party outside
the current negotiation.
17
less options available to the developer. In the extreme case, if the developer wants to
have all of the four department stores chains in its mall, it has no outside options. One
would expect that the developer might compromise during the process of negotiations
with the department store chains. Given the above discussion, the relation between the
number of department stores in a centre and a developer’s bargaining power is negative.
For a department store chain, the yearly aggregate revenue of this chain would
imply its customer-drawing ability and be a sign of its bargaining power relative to
developers. In addition, the size of a department store chain, measured by the number of
stores owned by this chain, would also indicate its bargaining power. This is because a
department store chain with more locations would build a stronger brand name than one
with less coverage. A department store chain’s reputation attracts customer traffic and is,
therefore, attractive to developers.
4.
Dataset Description
4.1 Shopping Centres
As discussed in Section 3.1, the focus of this paper is on planned regional shopping
centres with presence of department stores. Data on this type of shopping centre are
collected from the 2008 Canadian Directory of Shopping Centres. Planned shopping
centres that are enclosed, located in the five westernmost provinces of Canada (i.e.
Alberta, British Columbia, Manitoba, Ontario, and Saskatchewan), and have at least one
department store (i.e., Sears, The Bay, Zellers, and Wal-Mart) are initially included in the
dataset. In addition, the actual store counts of a shopping centre are required to be greater
18
than 408. This procedure generates 183 shopping centres, with 35 from Alberta, 35 from
British Columbia, 6 from Manitoba, 94 from Ontario, and 13 from Saskatchewan.
In addition, to be included in the dataset, a shopping centre needs to have a
detailed map showing the specific locations of stores in the shopping centre9. Shopping
centre maps are obtained directly from the website of the centre or the developer where
possible. For shopping centres that have no detailed map available online, I contacted
their developers for the map10. For shopping centres located in Edmonton, I visited the
centre and mapped store locations. The above steps yield a sample of 148 shopping
centres, with 29 from Alberta, 29 from British Columbia, 6 from Manitoba, 74 from
Ontario, and 10 from Saskatchewan. The actual store counts of these centres range from
40 to 589.
In this paper, a firm is defined as the “developer” of a shopping centre if it is
responsible for managing and controlling the centre. Based on this criterion, there are 26
developers in the sample. According to Table 4.1, the number of shopping centres that
are operated by a developer in the sample ranges from 1 to 25. Of these, 54 percent
operate only one regional shopping centre. Ivanhoe Cambridge Inc. is the largest
developer in this sample and owns 25 regional shopping centres.
Data on a shopping centre’s GLA and age are available in the 2008 Canadian
Directory of Shopping Centres. There are shopping centres for which the age or GLA of
the shopping centre is missing in the directory. In such a case, I contacted the developer
8
I set 40 as the threshold of actual store counts for regional shopping centres. Other thresholds will be used
later for robustness check.
9
See Graph A.1 in the appendix for a map of a hypothetical regional shopping centre.
10
All of the shopping centre maps are collected from September 2007 to December 2007. The maps
provided by developers were the most recent ones at that time.
19
of the centre to get the information. The population of the trade area that a regional
shopping centre serves is based on the 2006 Census of Canada.
Table 4.1 Developers
Developer With
Multiple Regional Malls
Number of
Regional Malls
Developer with a
Single Regional Mall
Ivanhoe Cambridge Inc.
25
Bayfield Realty Advisors
The Cadillac Fairview Corporation Limited
19
BOSA Development Corp.
Oxford Retail Group
17
Boultbee Realty Limited
Morguard Investments Limited
15
Chez Belle Limited
20 Vic Management Inc.
13
Colliers International
RioCan Property Services
13
Doral Holdings Limited
Redcliff Realty Management Inc.
12
First Gulf Development Corporation
Bentall Retail Services L.P.
9
ICI Shopping Centres
FCB Property Management Services
4
Londonderry Shopping Centre Inc.
T&T Properties
3
Park Royal Shopping Centre Holdings Ltd.
Calloway REIT
2
Shape Property Management Corp.
Darton Property Advisors & Managers Inc.
2
Sterling Vanreal Ltd.
Tanurb Developments Inc.
West Edmonton Mall Property Inc.
4.2 Retailers
Tenant lists of shopping centres in the sample are obtained directly from the website of
centres or developers where possible. For the rest of the shopping centres in the sample,
tenant lists are obtained from the maps that I received from the developers, since those
maps show all tenants located in a shopping centre. In the 148 sample shopping centres,
20
there are altogether 17520 tenants, which include non-retail stores11 like doctors, lawyers,
accountants, real estate firms and insurance companies. Out of the 17520 tenants, 3491
stores have only one location in the sample malls and fall into the category of
independent stores. The remaining 14029 stores are members of chains that have two or
more stores sharing a common name in the sample. Based on the definition of chains,
there are 1356 chains in the data set. The Source by Circuit City is the largest chain in the
dataset, which has 138 stores in the sample malls.
Stores in the sample malls are assigned of store types used by the 2008 Canadian
Directory of Shopping Centres. In some shopping centres, stores with the same name are
put in different categories. To address this problem, a retailer’s website is searched to
find out what products or services are provided by the retailer, which is then used to
determine the store categories. Once the store categorization is made, stores are broken
into C stores, M stores, and other stores by considering the types of merchandise sold by
each store category, relying on the definitions of C store and M store that are discussed in
Section 3.1. Table A.2 in the appendix shows the composition of C stores, M stores, and
other stores, which is proposed by Golosinski and West (1995, p.467) and revised using
the new categorizations of stores made by the 2008 Canadian Directory of Shopping
Centres12.
11
Non-retail stores are included in the dataset, because they are located in an area close to department
stores in some shopping centres and that might be a result of the relative developer/department store
bargaining.
12
There are minor differences between the categorization of stores used by Golosinski and West (1995,
p.467) and that of the 2008 Canadian Directory of Shopping Centre. Because I categorize stores using the
categorization made by the 2008 Canadian Directory of Shopping Centre, to be consistent, I revised the
composition of C stores, M stores, and Other Stores made by Golosinski and West (1995, p.467). For
example, athletic apparel is listed as a category in the 2008 Canadian Directory of Shopping Centre.
Therefore, I added it to the list of C stores suggested by Golosinski and West (1995, p.467). Footwear and
21
To check if the list of C stores proposed by Golosinski and West (1995, p.467) is
well represented in the sample malls, the average fraction of stores that are a particular
type of C stores (i.e. ladies’ apparel, footwear, menswear, etc.) in sample malls is
calculated. C stores of a type are expected to be replicated in a shopping centre to
facilitate comparison shopping. The average store counts of regional shopping centres in
the sample is 117, which means that the average percentage of a type of C store needs to
be at least 1.71% to ensure store replications and comparison shopping opportunity
(117
1.71%
2, i.e. at least two stores of the same type).
Table 4.2 shows the fraction of stores in the sample malls that are a particular type
of C store. According to the table, on average, the types of stores that meet the replication
threshold (i.e. greater than 1.71%) include: athletic apparel, footwear, gift,
jewelry/fashion accessories, ladies’ wear, menswear, and unisex/men’s & Ladies’ wear.
These stores will be defined as C stores in this empirical analysis. Although conceptually
sound, the other C store types proposed by Golosinski and West (1995, p.467) (e.g.
toy/game, camera, computer, children’s wear, electronics, etc.) are not well represented
in the sample centres and are, hence, excluded.
Table 4.2 The Fraction of Stores that is a Type of C Store in the Sample Malls.
Avg. Fraction of Stores in
the Sample Malls
C Store Types
Ladies' Wear
Unisex/Men's & Ladies' Wear
Jewellery/Fashion Access.
12.36%
7.87%
7.69%
leather goods are listed as separate categories rather than a single one. Drapery is not listed as a category in
the 2008 Canadian Directory of Shopping Centre and is therefore not listed as a C store.
22
Footwear
Athletic Apparel
Menswear
Gift
Children's Wear
Furniture & Home Décor
Leather Access. & Luggage
Department/Mass Merchandiser
Electronics
Housewares
Toy/Games
Family Wear
Camera
Hobby/Craft
Sporting Goods
Computer
Home Appliance
Home Improvement
Window Coverings
4.80%
2.24%
2.05%
1.89%
1.67%
1.64%
1.62%
1.56%
1.40%
1.32%
1.04%
1.00%
0.58%
0.21%
0.18%
0.14%
0.13%
0.02%
0.01%
4.3 Department Stores
In this empirical analysis, department store chains include Sears, The Bay, Wal-Mart, and
Zellers. Department store chains mainly differ in site selection, merchandise selection,
target audience, and price level (Vitorino, 2008). Table 4.3 reports the characteristics of
regional shopping centres where a specific type of department store chain is located.
According to the table, in the dataset, compared with regional shopping centres anchored
by other department store chains, shopping centres containing a Bay store, on average,
have the largest number of tenants, the biggest GLA, the biggest share of C stores, and
the smallest share of M stores. Shopping centres containing a Zellers store, in contrast,
23
have a store count and GLA that are below the sample mean. Regional shopping centres
where Wal-Mart stores are located have, on average, the biggest share of M stores.
According to Table 4.3, regional shopping centres where a particular department store
chain is located have differences in the store counts, store composition, and GLA.
Table 4.3 Characteristics of Sample Malls where a Department Store Chain Locates
Department
Store Chain
Sears
The Bay
Wal-mart
Zellers
Sample Mean
Store Counts
138
161
128
115
117
C Stores%
40.9%
41.6%
36.4%
34.8%
37.1%
M Stores %
19.4%
18.6%
22.0%
21.7%
20.9%
GLA
728,931
796,608
654,284
598,208
603,768
Table 4.4 Department Store Chains in the Sample Malls
Department Store
Sears
The Bay
Wal-mart
Zellers
Sample Size
80
67
22
83
AB
14
13
4
15
BC
15
16
4
15
MB
4
2
1
4
ON
43
33
12
42
SK
4
3
1
7
Table 4.4 reports the size of a department store chain and the provincial
breakdown of a department store chain in the sample malls13. Zellers has 83 stores in the
sample malls, which is the most among the four. Sears is second largest with 80 stores.
13
Table 4.4 only reports the size of a department store chain in the sample malls. A large number of WalMart and Zellers stores in the chain or some of them are not located in regional shopping centres. Such
stores are not included. Therefore, the size of the Wal-Mart and Zellers stores may be underestimated.
24
Wal-Mart has only 22 stores, which is the least among the four department store chains.
According to Table 4.4, in the sample, the distribution of the four chains across provinces
follows a similar pattern. All of the chains have the most stores in Ontario, have about the
same number of stores in Alberta and British Columbia, and have the least number of
stores in either Manitoba or Saskatchewan.
5.
Econometric Model
In this part of the paper, the measure of “being close to a department store” will be
discussed. Variables used to reflect the locational patterns of stores in an area close to
department stores in a regional shopping centre will be explained. Then, the econometric
method used to analyze how the locational patterns of stores in department stores’
proximity vary with the relative developer/department store bargaining power will be
presented.
5.1 Variable Definitions
(a) Dependent Variable
In this paper, a store14 is defined as being close to a department store if it is located
within a radius of 100 feet15 from the entrance of the department store, measured by the
Euclidean distance16.
14
Kiosks and carts are excluded because they are movable. For example, kiosks and carts selling Christmas
merchandise appear across shopping centres in November and December, and then are removed in January.
15
Yeates et. al (2001) investigated the impact of Eaton’s closure on the sales of stores located within 100 feet
and within 200 feet of Eaton’s entrance. They found that sales of the stores located 100 to 200 feet from an
Eaton’s entrance are remarkably insulated from the negative impact of Eaton’s closure, which implies that the
influence of a department store may not go beyond 100 feet from its entrance. In this paper, I choose 100 feet as
the benchmark, which may allow me to identify a department store’s main “power zone”. In sensitivity tests, I
25
Two variables are used to reflect the locational pattern of stores near department
stores: ratio_Cr and ratio_Mr, which are defined in equation (1) and (2), respectively.
_
_
_
_
⁄
⁄
_
(1)
_
(2)
where percnt_C100r is the fraction of stores that are C stores in an area within a radius of
100 feet from the entrance of a department store. Percnt_Ccentre is the fraction of stores in
a regional shopping centre that are C stores. Percnt_M100r is the fraction of stores that are
M stores in an area within a radius of 100 feet from the entrance of a department store.
Percnt_Mcentre is the fraction of stores in a regional shopping centre that are M stores.
Ratio_Cr and ratio_Mr show the fraction of C stores and M stores near department stores
relative to the fraction of C stores and M stores for the entire shopping centre,
respectively.
The reason that a ratio is used is because the fraction of stores near a department
store that are C stores (or M stores) could be reflecting the overall fraction of such stores
for the entire mall. To capture the impact of the relative developer/department store
bargaining power on the store composition near department stores, the store composition
for the entire mall should be controlled. However, Percnt_Ccentre and percnt_C100r (or
Percnt_Mcentre and percnt_M100r) could be jointly determined by the relative
also adopt another definition of “being close to a department store” to test if the conclusions are robust to
different measures.
16
There are 58 malls in the sample for which the exact measurement of distance or scale is not available in their
maps. Fortunately, for these malls, the area of stores in squared footage is available either from the 2008
Canadian Directory of Shopping Centres or from their maps. So, the scale of a map can be derived by comparing
the area of a store in the map and its actual area.
26
developer/department store bargaining power. If included as an explanatory variable,
Percnt_Ccentre (or Percnt_Mcentre) would cause an endogeneity problem and lead to biased
and inconsistent estimation. Based on the above considerations, ratio_Cr and ratio_Mr
are used.
A shopping centre may contain more than one department store and a department
store may contain multiple floors and/or multiple entrances on the same floor. Since the
locational pattern of stores close to a department store may vary across entrances and
floors, the observation is set as each entrance of a department store17. A concern is that it
could introduce some correlation between observations within a centre. However, only 11
out of the 148 sample centres (i.e. less than 7.5%) have more than two observations
within a shopping centre. There are not enough observations to estimate the unobserved
group effect within a regional shopping centre. To investigate the difference of location
patterns across entrances, variables controlling for floor and the location of an entrance
are included as explanatory variables.
In the econometric specification, ratio_Cr and ratio_Mr are estimated as a linear
function of variables indicating the bargaining power of developers and department stores
and other determinants that are expected to have an impact on the locational pattern. The
specification of the regression function is as follows: Y = β’X +ε, where Y = ratio_Cr or
ratio_Mr, β is a vector of parameters, X is a vector of independent variables, and ε is a
vector of error terms. OLS (Ordinary Least Squares) is used to estimate the two
regression functions, respectively.
17
For a department store, any entrance facing a corridor that is inside a shopping centre is included in the
analysis. Any entrance that is facing outside (for example, parking lot) is excluded from the analysis unless
there are stores located close to it.
27
(b) Independent Variables
The independent variables used in this econometric model include developer variables,
department store dummies, and the other control variables. Developer dummy variables
for all developers that operate two or more regional shopping centres in this dataset are
included as independent variables. Such dummy variables will capture the idiosyncratic
bargaining strength of a developer. In addition, the fixed effect of a developer across the
shopping centres that it owns could be controlled in this way. GLA 18 is a shopping
centre’s gross leasable area, measured in 100,000 square feet. The variable centrage
measures the age of a shopping centre, which is calculated by subtracting the year that the
centre was opened from 2009. As discussed in Section 3.3, the relation of GLA and
centrage to a developer’s bargaining power are ambiguous.
Population is the number of people, measured by 100,000, living in a regional
shopping centre’s primary trade area. It is predicted to be directly related to a developer’s
bargaining power, because a large population signals the potential purchasing power and
is attractive to department stores. In this paper, a regional shopping centre’s primary trade
area is defined as the intersection of the shopping centre’s Voronoi market area and a 25
mile radius circle around the centre. To estimate the Voronoi market area, the first step is
to plot all 183 regional shopping centres (i.e. including regional shopping centres that
have no detailed map and are excluded from the sample) in the maps of the five
westernmost provinces, using the regional shopping centres’ geographic coordinates (i.e.
longitude and latitude) that are obtained from the Google map. Then, for each regional
18
It is expected that both a centre’s GLA and its store counts could indicate the centre’s bargaining power.
The correlation between a centre’s GLA and store counts is 0.91. Therefore, only one of them, GLA, is
included as an explanatory variable.
28
shopping centre, a Voronoi market area19 is calculated by finding the set of points that are
closer to the regional shopping centre than to any other regional shopping centre located
within a province.
Economically, this approach to calculate trade areas can be justified by the utilitymaximizing consumer’s tendency to minimize transportation cost and visit the nearest
regional shopping centre20, given the assumption that regional shopping centres are the
same and charge the same price. Although regional shopping centres are not in fact
identical, a large fraction of their stores are chain store and they exhibit similarity in store
brands, especially for those located within the same province (Golosinski and West,
1995). Therefore, many consumers are likely to visit the nearest regional shopping centre,
as is assumed in the calculation of Voronoi trade area for shopping centres. The
provincial boundary is chosen instead of the city boundary for the maximum outer
boundary of a Voronoi market area. This is because regional shopping centres are
frequently located in the suburb of a city and might attract customers outside the city.
Using the city limits to bound the Voronoi trade area could lead to an underestimate of
the market area of a regional shopping centre.
After a regional shopping centre’s Voronoi market area is calculated, it is then
intersected with a 25 mile radius circle around the shopping centre. The intersected area
will be considered as the “primary” trade area of the regional shopping centre. The reason
19
This paper follows the methodology of Von Hohenbalken and West (1986) to calculate the Voronoi
diagram. See Von Hohenbalken and West (1986) for a detailed discussion of the Voronoi diagram and its
economic meaning.
20
Based on the discussion of Von Hohenbalken and West (1986), the assumption of the Voronoi diagram
is that stores are homogenous and hence, equally attractive to consumers.
29
that a 25 mile radius threshold is chosen is because customers are not expected to visit a
regional centre on a regular basis if it is located more than 25 miles away21.
Once the primary trade areas of regional shopping centres are calculated, they are
then overlapped with the census tracts. The population of a regional shopping centre’s
trade area consists of the pieces of census tracts that fall into this area22. Graph 1 shows
the Voronoi market areas of regional shopping centres and census tracts in Alberta. The
black dots in the graph are location points of planned regional shopping centres. In Graph
1, Voronoi market areas have bold boundaries and the 2006 census tracts have light
boundaries. The primary trade areas of regional shopping centres in Edmonton area are
the shaded area, shown in Graph 2. At the left corner of Graph 2, there are two centres
and each has a primary trade area with a half-circle shape. In the middle of the graph,
there are a group of centres whose primary trade areas are either fan shaped or polygons.
The variable competition measures the proportion of a regional shopping centre’s
neighbouring centres that are at the same level of the shopping centre hierarchy and are
managed by different developers. Two regional shopping centres are defined as
neighbours when their primary market areas share a boundary segment and the distance
between these two centres is less than 25 miles. This variable is expected to show the
extent of competition that a regional shopping centre faces from its neighbours.
Everything else constant, developers operating more neighbouring centres would have
21
A regional shopping centre’s primary trade area is not well defined in the literature. Vitorino (2008)
defined it to be a circle that has a radius of 20 miles. The ICSC defines it to be a circle that has a radius of 5
to 15 miles for regional shopping centres and 5 to 25 mile for super-regional centres (i.e. regional shopping
centres with 3 or more anchors). The Urban Land Institutes (1985) defines that the primary trade area as a
circle that has a radius of 8 miles for regional centres and 12 miles for super-regional centres. All of these
criterions are based on the U.S. regional shopping centres.
22
All calculations were done using the GIS (geographic information system software) Arcview 9.3.
30
Graph1. Voronoi Market Area of Regional Shopping Centres and Census Tracts
in Alberta
&
&
&&&
&
&
&&
&
&&
&
&
&
&
&
&
&&
&
&
&
&
&
&
&
&
&
Graph2. Primary Trade Areas of Regional Shopping Centres in Edmonton Area
&
&
&&
& & &&
&
&
& &
&
&
&
more control over this region and would have more bargaining strength. The value of
competition ranges from 0 to 1. When competition is 1, it means that all of the
neighbouring centres are owned by other developers. When competition is 0, it means
that the developer owns all of the neighbouring centres. It is defined that competition has
a value of 0 if a regional centre has no neighbouring centres. This variable is expected to
be negatively related to the bargaining power of a developer.
As discussed in Section 3, indicators of department stores’ bargaining power
would include a department store chain’s annual revenue and size. However, including
these variables as explanatory variables would be inappropriate, since the sample only
contains four department store chains and these variables lack sufficient variation. To
solve this problem, department store chain dummies (Sears, Walmart, and Zellers) are
used. I expect them to pick up the systematic differences across department stores
regarding their bargaining positions and the locational pattern in their proximities. The
Bay stores are used as the base group.
The location related variables include downtown, mainEntry, and upperflr. The
variable downtown is a dummy, which equals 1 if the shopping centre is located in
downtown area and 0 otherwise. In the dataset, all shopping centres located in downtown
have multiple floors, which could be explained by the common fact that land in such an
area is more expensive than that in the suburbs. Such a floor plan could have an impact
on store configurations.
MainEntry is a dummy which equals 1 if the observation is the main entrance to
the department store, and 0 otherwise. The main entrance of a department store is defined
as the entrance that faces a main corridor connecting anchors or leading to the main exit
31
of a mall, whereas a side entrance of a department store is defined as the one that faces an
aisle leading to a side exit of a mall. One would expect the area close to a department
store’s main entrance to have more customer traffic than that of a side entrance. As a
result, the locational pattern of stores is expected to be different across entrances. The
dummy variable upperflr indicates whether an entrance of the department store is located
on a level higher than ground level. It is included to capture the difference in locations
pattern of stores across floors.
5.2 Tests: RESET and the Breusch-Pagan Test
Once the two regression functions are estimated, in addition to the regular t test and F test,
RESET (Ramsey’s regression specification error test) will be conducted to test for
functional form misspecification. In addition, the Breusch-Pagan test will be performed
to test if heteroskedasticity exists.
6. Summary Statistics and Regression Results
6.1 Summary Statistics
The first column of Table 6.1 lists the store types that are most frequently located
within a 100 foot radius from the entrance of a department store in the sample malls,
sorted in descending order, and the second column reports whether these store types are
C Store, M store, Other Store, or Non-Retailer. Column (a) of Table 6.1 reports the share
of each store type and Column (b) of the table reports the fraction of stores for each store
type in the entire mall. According to this table, the four store types that are most often
located in an area near department stores in the sample malls are stores selling ladies’
32
wear, jewellery/fashion accessories, unisex clothing, and footwear. All of them are C
stores. In addition, the fraction of these store types located near department stores is
greater than the fraction of these stores for the entire mall. Among M store types,
hairstyling/esthetic stores are most frequently located near department stores, while
ratio_Mr is the highest for variety/convenience stores. The other store types and nonretail
firms,
with
the
exception
of
restaurant
&
fast
food
stores
and
wireless/telecommunication stores, are seldom located in a department store’s proximity
in the sample.
Table 6.1 The Composition of Stores: Near Department Stores Vs. Whole Mall23
Store Type
(a)
Within 100
Feet
(b)
the Whole
Mall
(c)=(a)/(b)
Ratio
Ladies' Wear
C Store
15.42%
12.36%
1.25
Jewellery/Fashion Access.
C Store
8.79%
7.69%
1.14
Unisex/Men's & Ladies' Wear
C Store
8.79%
7.87%
1.12
Footwear
C Store
5.56%
4.80%
1.16
Hairstyling/Esthetic
M Store
5.47%
3.54%
1.55
Restaurant & Fast Food
Other Store
5.18%
12.95%
0.40
Wireless/Telecommunication
Other Store
3.28%
3.66%
0.89
Optical
M store
2.86%
2.05%
1.39
Menswear
C Store
2.82%
2.05%
1.37
Athletic Apparel
C Store
2.65%
2.24%
1.19
/
2.24%
1.64%
1.36
Variety/Convenience
M Store
2.07%
1.30%
1.60
Drug/Health & Beauty Aids
M Store
2.07%
1.60%
1.30
Gift
C Store
1.99%
1.89%
1.05
Non-Retail
1.99%
5.42%
0.37
Furniture & Home Décor
Non Retail
23
Furniture & Home Décor, Leather Access. & Luggage, and Housewares are not C store in this paper,
because they are not well represented in the sample malls.
33
Beauty Supply
Leather Access. & Luggage
Card/Stationery
Financial
Housewares
M Store
1.91%
2.52%
0.76
/
1.70%
1.62%
1.05
M Store
1.70%
1.18%
1.45
Other Store
1.70%
1.69%
1.01
/
1.66%
1.32%
1.26
Table 6.2 reports the summary statistics for variables used in this empirical
analysis. There are 383 observations in the sample. Ratio_Cr ranges from 0 to 6.095.
There is an observation of Ratio_Cr which has a value of 6.095 and lies outside the range
of the rest of the observation. This outlier (i.e. ratio_Cr = 6.095) will be dropped from the
empirical analysis to test if it will significantly change the result. Ratio_Mr ranges from 0
to 4.454 and has an average value of 1.023.
Regarding developers, the GLA of a shopping centre ranges from 1.52 to 38,
measured by 100,000 square feet. On average, the age of a regional shopping centre is 35
years. The average population in a regional shopping centre’s primary trade area is 1.078,
measured by 100,000 persons. There are two malls in the sample which have less than
10,000 people in their primary trade areas and are therefore substantially lower than that
of the other regional malls in the dataset. One mall has neighbouring shopping centres
located in its close proximity and the other is located in a town with small and relatively
dispersed population. In the regression analysis, the models will be estimated using the
subset of the dataset that excludes the two malls to test if these outliers will significantly
change the result.
Regional shopping centres in the sample face substantial competition from
neighbouring regional centres. As shown in Table 6.2, on average, the fraction of
neighbouring regional centres that are owned by other developers is about 80 percent.
34
Considering department store chains, 31.3 percent of observations are for Sears, which is
slightly higher than Zellers. Only 7.6 percent of observations are for Wal-Mart stores. On
average, shopping centres in the dataset contain two department stores. As to the other
location related variables, only 6.5 percent of observations are in downtown areas, less
than 10 percent of observations are on side entrances, and about 30 percent of
observations are for a floor above the ground floor.
Table 6.2. Summary Statistics (N=383)
Variable
Mean
Std. Dev.
Min
Max
ratio_Cr
1.102
0.722
0
6.095
ratio_Mr
1.023
0.910
0
4.454
GLA
7.587
5.202
1.52
38
centrage
35.337
9.256
13
59
population
1.078
0.666
0.034
2.965
competition
0.799
0.344
0
1
deptnumbr
2.089
0.798
1
4
Sears
0.313
0.464
0
1
WalMart
0.076
0.265
0
1
Zellers
0.292
0.455
0
1
downtown
0.065
0.247
0
1
mainEntry
0.919
0.273
0
1
upperflr
0.300
0.459
0
1
To detect if severe multicollinearity exists, the simple correlation coefficients
between each two of the explanatory variables are examined. The value of the correlation
coefficient between GLA and deptnumbr is 0.52, which has the highest absolute value
35
among the explanatory variables. The correlation coefficient that has the second highest
absolute value is for population and deptnumbr, which is 0.36. The absolute value of the
correlation coefficients of the other explanatory variables is below 0.3. Empirically,
researchers become concerned about multicollinearity if the absolute value of a simple
correlation coefficient exceeds 0.8. Since the value of the correlation coefficient between
explanatory variables is well below 0.8, severe multicollinearity is not diagnosed.
6.2 Regression Results
Table 6.3 reports the results of the two regression functions when ratio_Cr and ratio_Mr
are used as the dependent variable, respectively. When ratio_Cr is the dependent variable,
R2 is 0.12. The estimated coefficient on GLA is 0.015, which means that, everything else
held constant, ratio_Cr is expected to increase by 0.015 units when the gross leasable
area of the regional mall is increased by one unit (i.e. 100,000 square feet). Although the
impact of GLA on ratio_Cr is estimated to be small, it is statistically significant at the 10
percent level. Since a regional shopping centre’s GLA is expected to be related to the
centre developer’s bargaining power, this finding is consistent with the hypothesis that
the locational pattern of stores in an area close to department stores would depend upon
the bargaining power of a developer. The positive sign of the coefficient of GLA is
consistent with the scenario that developers of large regional shopping centres possess
greater bargaining power and are able to locate more C stores near department stores.
The coefficient on centrage is significant at the 5 percent level, which supports
the hypothesis that variables reflecting a developer’s bargaining power have explanatory
power in the locational pattern of stores near department stores. The positive sign of the
36
coefficient of centrage implies that older centres may be attractive to department stores
due to their reputation and customer loyalty. As a result, developers of older regional
centres have a more favorable bargaining position and could locate more C stores near
department stores.
Table 6.3. Regression Results.
Y = ratio_Cr
Y = ratio_Mr
0.015*
(0.009)
0.008**
(0.004)
0.021
(0.060)
-0.101
(0.110)
-0.009
(0.086)
-0.160
(0.141)
-0.142
(0.093)
-0.158***
(0.060)
0.335***
(0.129)
-0.109
(0.170)
0.135*
(0.080)
0.883***
(0.256)
-0.006
(0.011)
-0.011*
(0.006)
0.038
(0.081)
0.064
(0.134)
0.044
(0.126)
0.111
(0.182)
0.172
(0.131)
0.064
(0.082)
-0.446*
(0.239)
-0.092
(0.225)
-0.246**
(0.116)
1.455***
(0.350)
0.12
0.07
Variables
GLA
centrage
population
competition
Sears
WalMart
Zellers
deptNumbr
mainEntry
downtown
uppflr
Constant
R2
*** Indicates significance at the 1 percent level.
** Indicates significance at the 5 percent level.
* Indicates significance at the 10 percent level.
37
The coefficient on competition and population are not statistically significant,
indicating that the impact of competition and population on ratio_Cr is not identified in
this empirical analysis. The estimated coefficient on deptnumbr is -0.158, which means
that ratio_Cr is expected to decrease by 0.158 unit when a regional shopping centre
contains one more department store, all else held constant. Such an impact is statistically
significant at 1 percent level of significance. This finding is consistent with the
hypothesis that developers have more limited outside options as the number of
department stores increases in a shopping centre and have, hence, less bargaining power
and would not be able to locate more C stores near department stores. The coefficients on
the department store dummies Sears, Walmart, and Zellers are negative. However, none
of them is significant at any conventional significance level.
Regarding the location related variables, the coefficient on mainEntry and
upperflr are positive and statistically significant at 1% level and 10% level, respectively,
which means that ratio_Cr is bigger for department stores’ main entrances than for their
side entrances, and is bigger for floors higher than the ground level, holding everything
else fixed. The coefficient on downtown is not statistically significant. Coefficients on
developer dummy variables are suppressed.
RESET (Ramsey’s regression specification error test) is conducted to test for
functional form misspecification. To implement RESET and test whether important
nonlinearities are missed, the squared and cubed terms of the fitted value are included in
the OLS regression functions as extra explanatory variables. The null hypothesis is that
the coefficients of the squared and cubed terms of the fitted value are equal to zero. The
test result shows that F (2, 356) = 0.42 and the associated P-value is 0.66, indicating that
38
the coefficients of the squared and cubed terms of the fitted value are jointly insignificant.
The result of RESET suggests that evidence of functional form misspecification is not
detected.
In addition, the Breusch-Pagan test for heteroskedasticity is conducted. The
squared OLS residuals are regressed on all of the explanatory variables. The F statistic
for all independent variables is F (23,359) =1.39 with P-value = 0.1118, suggesting that
the test result fails to reject the null hypothesis and that heteroskedasticity does not exist
at the 10 percent level of significance.
When ratio_Mr is the dependent variable, the Breusch-Pagan test for
heteroskedasticity yields F (23,359) =1.67 with P-value = 0.029, which provides
evidence that heteroskedasticity exists at the 5 percent level of significance. The presence
of heteroskedasticity does not cause OLS to be biased or inconsistent. In addition, the R2
is not affected as well. However, the t statistic and F statistic are no longer valid
(Wooldridge, 2006, P.272). In the last two decades, a fashion to deal with
heteroskedasticity is to adjust standard error and test statistic so that they are still valid in
the presence of heteroskedasticity (Wooldridge, 2006, P.272). In this paper, for the
regression where ratio_Mr is the dependent variable, the heteroskedasticity-robust
standard errors are used to construct a heteroskedasticity-robust t statistic and F statistic.
According to Wooldridge (2006), the heteroskedasticity-robust standard errors have the
following form:
∑
39
̂
(3)
where ̂ the ith residual from regressing xj on all other explanatory variables, and SSRj is
the sum of squared residuals from this regression.
When ratio_Mr is the dependent variable, the R2 is 0.07. The heteroskedasticityrobust F statistic that tests the joint significance of all slope variables yields F (23,359)
=1.39 with P-value = 0.1117, which means that the joint impact of the slope variables is
not significantly different from zero at the 10% significance level, although the
coefficient on centrage, mainEntry, upperflr, and constant are individually significant.
The finding that the coefficient on centrage is significant at the 10 percent level coincides
with the hypothesis that the locational pattern of stores is related to the indicators of a
developer’s bargaining power. However, such a relation is not identified between
ratio_Mr and the other indicators of a developer’s bargaining power. The results of
estimation suggest that the interests between developers and department stores regarding
the locations of M stores near department stores might not be substantially divergent. Or,
instead of bargaining over store locations, developers might choose the locational
configuration within the regional shopping centre that maximizes overall centre profit,
and then compensate department stores through rents for having less than ideal stores
located near them.
To test for robustness of results, an alternative measurement of being close to
department stores is used. A store is defined as being close to a department store when
the entrance of the store is located within 100 feet of the entrance of a department store,
measured by walking distance. When a person walks from a department store to another
store in a shopping centre, he/she will presumably choose the shortest possible route.
Here, the “possible route” means that a customer has to follow the pathway in a mall (e.g.
40
he cannot walk through walls etc.). The length of the shortest route is the walking
distance. Using Euclidean distance, 2412 stores are defined as being close to department
stores in the sample mall. When walking distance is used, there are only 2007 stores24. As
the definition changes, the estimated coefficients do not exhibit big differences, which
indicate that the econometric model is robust. When ratio_Cr is the dependent variable,
the variables that are significant in both regressions include GLA, deptnumbr, and
mainEntry. When different measurements of distance are used, there are minor changes
in the level of significance for GLA. With the walking distance, the coefficient on GLA is
more significant. The main changes are for centreage, upperfl, and WalMart. In the
estimation, the coefficient on centreage and upperflr becomes insignificant. In addition,
the coefficient on WalMart becomes significant, although only at the 10 percent level of
significance. The coefficient on WalMart is -0.283, suggesting that ratio_Cr is about
0.283 units smaller for Wal-Mart stores than for the Bay stores.
Regarding the estimation when ratio_Mr is the dependent variable, using walking
distance, the estimated coefficients on mainEntry is significant at the 5 percent level
rather than 10 percent. In addition, the coefficient on upperflr becomes insignificant at
any reasonable levels of significance.
In this paper, regional shopping centres with at least 40 retailers are focused on.
To explore the robustness of the empirical findings, the models were estimated for a
subset of the sample malls that have at least 60 retailers. In addition, alternatives of
dependent variables are used. For example, instead of the ratio, the difference between
the fraction of stores near department stores that are C stores (or M stores) and the
24
See Graph A.1 in the appendix, which discuss the group of stores that are identified as being close to
department stores under the above two definitions.
41
fraction of C stores (or M stores) for the entire shopping centre are studied. The location
patterns of stores near department stores are also compared with the rest of the shopping
centre (i.e. areas that are not close to any department store in the regional shopping centre)
instead of the entire shopping centre. The results of the above estimations show that
changes in the results are minor, which means that the findings of this paper are robust.
7.
Concluding Remarks
Inside a regional shopping centre, the area near department stores is expected to have
high customer traffic due to its attractiveness to customers. Regarding what types of
stores should be located in such an area, the interests of developers and department stores
may not coincide. Given the fact that department stores only care about their own profit
while the developer cares about the rental payment it can extract from the entire centre,
there might be a conflict of interests between their preferred store types for this area. In
the presence of conflict of interest, the locational patterns of stores near department stores
are expected to depend on the relative bargaining power of the two parties.
In this paper, econometric models are constructed to examine the above problem.
Variables indicating the composition of stores in an area close to department stores are
used as the dependent variable. Indicators of bargaining power for developers are
included as explanatory variables. Because there are only four department store chains in
the sample, indicators of bargaining power for department stores are not used as
independent variables in the analysis as they lack sufficient variation. Instead, department
store dummies are included. In addition, location related variables are included as well.
42
The results of estimation show that ratio_Cr is bigger in regional shopping centres
that have a larger gross leasable area and are opened earlier. The results also show that
ratio_Cr is negatively related to the number of department stores contained in a regional
shopping centre. A regional shopping centre’s gross leasable area, age, and the number of
department stores in a regional shopping centre are expected to reflect a developer’s
bargaining power. Hence, the findings that these variables have explanatory power in the
composition of stores near department stores are consistent with the hypothesis that the
locational pattern of stores in an area close to department stores would depend upon the
bargaining power of a developer. In addition, the site-specific variable indicating whether
the observation is the main entrance of a department store and whether it is on a floor
above the ground floor are estimated to have a significant impact on the locational pattern
of stores near a department store as well.
The positive sign of the coefficient on GLA is consistent with the scenario that
developers of large centres possess greater bargaining power because such centres can
provide more comparison and multipurpose shopping opportunities and are attractive to
customers and are, hence, attractive to department stores. The positive sign of the
coefficient on centrage is consistent with the scenario that older centre may have built a
reputation and customer loyalty and have a more favorable bargaining position. As a
developer has more bargaining power, they would be able to locate more C stores near
department stores. The estimated coefficient on deptnumbr is negative, consistent with
the hypothesis that developers have more limited outside options as the number of
department stores increases in a shopping centre and therefore have less bargaining
power and would not be able to locate more C stores near department stores.
43
Ratio_Mr is estimated to be smaller in older centres. However, the relation
between Ratio_Mr and the other indicators of a developer’s bargaining power is not
identified. The results of estimation suggest that the interests between developers and
department stores regarding the locations of M stores near department stores might not be
substantially divergent. Or, instead of bargaining over store locations, developers might
choose the locational configuration within the regional shopping centre that maximizes
overall centre profit, and then compensate department stores through rents for having less
than ideal stores located near them.
To explore the robustness of the empirical findings, the models were estimated for
a subset of the sample malls that have at least 60 retailers. In addition, alternative
variables measuring the locational patterns of stores near department stores and
alternative definitions of being close to a department store are used. The results of the
above estimations show that changes in the results are minor, which means that the
findings of this paper are robust.
This paper investigates how the locational patterns of stores within 100 feet of a
department store’s entrance are determined by the relative bargaining power between
developers and department stores in a planned regional shopping centre. The locational
patterns of stores in the rest of the shopping centre are not investigated. Such an issue
will be explored in future research.
44
REFERENCES:
Benjamin, J.D.; Boyle, G.W. and Sirmans, C.F. "Retail Leasing: The Determinants of
Shopping Center Rents." American Real Estate and Urban Economics
Association Journal, 1990, 18, pp. 302-12.
____. "Price Discrimination in Shopping Center Leases." Journal of Urban Economics,
1992, 32, pp. 299-317.
Brown, S. "Tenant Mix, Tenant Placement and Shopper Behavior in a Planned Shopping
Centre." The Service Industries Journal, 1992, 12, pp. 384-403.
Brueckner, J. "Inter-Store Externalities and Space Allocation in Shopping Centers."
Journal of Real Estate Finance and Economics, 1993, 7, pp. 5-16.
Carter, C.C. and Haloupek, W.J. "Dispersion of Stores of the Same Type in Shopping
Malls: Theory and Preliminary Evidence." Journal of Property Research, 2002,
19, pp. 291-311.
Carter, C.C. and Vandell, K.D. "Store Location in Shopping Centers: Theory and
Estimates." Journal of Real Estate Research, 2005, 27, pp. 237-65.
Christaller, W. Die Zentralen Orte in Suddeutschland. Translated by Baskin, C.W. (1966)
to Central Places in Southern Germany. Englewood Cliffs, N.J.: Prentice-Hall,
1933.
Dawson, J.A. Shopping Centre Development. London: Longman, 1983.
Eaton, B.C. and Lipsey, R.G. "Comparison Shopping and the Clustering of
Homogeneous Firms." Journal of Regional Science, 1979, 19, pp. 421-35.
____. "An Economic Theory of Central Places." Economic Journal, 1982, 92, pp. 56-72.
Eckert, A. and West, D.S. "Radius Restrictions and the Similarity of Neighboring
Shopping Centres," International Journal of the Economics of Business. 2008, 15,
pp. 281-300.
Eppli, M.J. and Shilling, J.D. “How Critical is a Good Location to a Regional Shopping
Centre?” The Journal of Real Estate Research, 1996, 12, pp.459-468.
Gatzlaff, D.H. ; Sirmans, G.S. and Diskin, B.A. "The Effect of Anchor Tenant Loss on
Shopping Center Rents." Journal of Real Estate Research, 1994, 9, pp. 99-110.
Gerbich, G. "Shopping Center Rentals: An Empirical Analysis of the Retail Tenant Mix."
Journal of Real Estate Research, 1998, 15, pp. 283-96.
45
Golosinski, D. and West, D.S. "Double Moral Hazard and Shopping Centre Similarity in
Canada." Journal of Law, Economics, and Organization, 1995, 11, pp. 456-78.
Gould, E.D.; Pashigian, B.P. and Prendergast, C. "Contracts, Externalities, and Incentives
in Shopping Malls." Review of Economics and Statistics, 2005, 87, pp. 411-22.
Harvard Law Review Association, “The Antitrust Implications of Restrictive Covenants
in Shopping Center Leases”, Harvard Law Review, 1973, 86, pp. 1201-1249.
Ingene, C.A. and Ghosh, A. "Consumer and Producer Behavior in a Multipurpose
Shopping Environment." Geographical Analysis, 1990, 22, pp. 70-93.
Kennan, J., and Wilson, R. “Bargaining with Private Information”, Journal of Economic
Literature, 1993, 31, pp. 45-104.
Kinnard, W. N. Jr and Messner, S.D. "Obtaining Competitive Locations for Small
Retailers in Shopping Centers." Journal of Small Business Management, 1972,
10, pp. 21-26.
Mason, J.B. “Power and Channel Conflicts in Shopping Center Development”, Journal of
Marketing, 1975, 39, pp. 28-35.
Miceli, T.J.; Sirmans, C.F. and Stake, D. "Optimal Competition and Allocation of Space
in Shopping Centers." Journal of Real Estate Research, 1998, 6, pp. 113-26.
Morton, F.S.; Zettelmeyer, F. and Silva-Risso, J. “A Test of Bargaining Theory in the
Auto Retailing Industry”, 2004.
http://www.econ.yale.edu/seminars/apmicro/am04/scottmorton040928.pdf#search=%22empirical%20test%20of%20bargaining%20theory%22
Muthoo, A. “Bargaining Theory with Applications”. Cambridge: Cambridge University
Press, 1999.
Muthoo, A. “A Non-Technical Introduction to Bargaining Theory” World Economics,
2001, 1, pp. 145-166.
Osborne, M.J. and Rubinstein, A. “Bargaining and Markets”. San Diego: Academic
Press, 1990.
Pashigian, B.P. and Gould, E.D. "Internalizing Externalities: The Pricing of Space in
Shopping Malls." Journal of Law and Economics, 1998, 41, pp. 115-42.
Ryan, D.L.; V., Hohenbalken B. and West, D.S. "An Econometric-Spatial Analysis of the
Growth and Decline of Shopping Centers." Regional Science and Urban
Economics, 1990, 20, pp. 313-26.
46
Stahl, K. "Differentiated Products, Consumer Search, and Locational Oligopoly." Journal
of Industrial Economics, 1982a, 31, pp. 97-114.
____. "Location and Spatial Pricing Theory with Nonconvex Transportation Cost
Schedules." The Bell Journal of Economics, 1982b, 13, pp. 575-83.
Urban Land Institute. Shopping Center Development Handbook. Washington DC: Urban
Land Institute, 1985.
Vitorino, M.A. “Empirical Entry Games with Complementarities: An Application to the
Shopping Centre Industry”, the Wharton School of Business, University of
Pennsylvania, Working Paper, 2008.
Von Hohenbalken B. and West, D.S. “Empirical Tests for Predatory Reputation”,
Canadian Journal of Economics, 1986, 19, pp. 160-78.
West, D.S. "An Empirical Analysis of Retail Chain and Shopping Centre Similarity."
Journal of Industrial Economics, 1992, 40, pp. 201-21.
West, D.S.; Von Hohenbalken, B. and Kroner, K. "Tests of Intraurban Central Place
Theories." Economic Journal, 1985, 95, pp. 101-17.
West, D.S.; Ryan, D.L. and V., Hohenbalken B. "New Competition in Shopping Centre
Hierarchies: An Empirical Comparison of Alternative Specifications." Journal of
Regional Science, 1988, 28, pp. 329-44.
Wooldridge, J.M. “Introductory Econometrics: A Modern Approach”, Third Edition,
Mason, OH: Thomson South-Western, 2006.
Wunder, G.C., “Restrictive Clauses in Shopping Centre Leases: A Review”, Real Estate
Issues, 1988, 13, pp. 29-34.
Yeates, M.; Charles, A. and Jones, K., “Anchors and Externalities.” Canadian Journal of
Regional Science, 2001, 24, pp. 465-484.
47
APPENDIX
Table A.1. Shopping Centres that have Floors Dropped from the Analysis.
Province
AB
AB
AB
AB
AB
AB
MB
ON
ON
ON
ON
ON
ON
ON
ON
ON
ON
ON
ON
ON
ON
ON
ON
ON
ON
ON
ON
Name
Chinook Centre
Edmonton City Centre East
Kingsway Garden Mall
Northgate Centre
Northland Village
Westbrook Mall
Polo Park Shopping Centre
Billings Bridge Plaza
Bridlewood Mall
Burlington Mall
Cedarbrae Mall
Dixie Outlet Mall
Dufferin Mall
Hillcrest Mall
Hopedale Mall
Kozlov Centre
Lansdowne Place Shopping Centre
Northumberland Mall
Oshawa Centre
Place d'Orleans
Sherway Gardens
St.Laurent Centre
Stone Road Mall
The Promenade
The Toronto Eaton Centre
York Gate Mall
Yorkdale Shopping Centre
48
Floor Dropped
Basement floor
Concourse floor and the 3rd floor.
3rd floor
Upper floor
Upper floor
Professional building
Basement floor
Upper floor
Lower floor
2nd floor
Upper floor
Upper floor
Office floor and basement floor.
Upper and lower floor
2nd floor
Office building.
Lower floor
Upper and lower floor
2nd Floor, executive tower, office galleria.
Upper floor
Upper floor
Lower floor
Upper and lower floor
Basement floor
1st and 4th floor
Upper floor
2nd and 3rd floor
Table A.2 Classification of C, M, Other Stores and Non Retail Stores by Golosinski
and West (1995)
C Stores
Athletic Apparel
M stores
Other Stores
Non-Retail Stores
Beauty Supply
Automotive
Educational/Training
Camera
Book/Newsstand
Business Services
Insurance
Children's Wear
Card/Stationery
Car & Truck Rental
Medical/Dental
Computers
Drug/Health & Beauty Aids
Financial
Non Retail
Department/Mass Merchandiser
Dry Cleaner
Fitness/Recreation Place
Other Non-Retail
Electronics
Fabric/Sewing Access.
Hardware/Paint & Paper
Postal Services
Family Wear
Florist/Nursery
Laundromat
Real Estate
Footwear
Grocery
Other Services
Furniture & Home Décor
Hairstyling/Esthetic
Pet
Gift
Music/Video
Printing
Hobby/Craft
Office Supply
Rental Equipment/Furniture
Home Appliance
Optical
Restaurant & Fast Food
Home Improvement
Photo
Second Hand Merchandise
Housewares
Shoe Repair
Specialty Food & Drink
Jewellery/Fashion Access.
Tailoring/Alterations
Specialty Merchandise
Ladies' Wear
Travel
Theatre/Entertainment
Leather Access. & Luggage
Variety/Convenience
Ticket/Lotto Sales
Menswear
Wireless/Telecommunication
Sporting Goods
Toy/Games
Unisex Clothing
Window Coverings
49
Graph A.1.
Floor Plan for a Hypothetical Regional Shopping Centre
10
12
11
C
4
Sears
8
5
9
B
A
1
2
3
6
7
Zellers
= 100 feet
In this hypothetical regional shopping centre, there are two department stores: Sears and
Zellers. Sears has a single entrance (i.e. point A) and Zellers has two entrances (i.e. point
B and C). As defined in Section 5.1 (b), A and B are the main entrances, which face a
main corridor connecting two anchors, whereas C is the side entrance, which faces an
aisle leading to a side exit of a mall. There are three observations for the above mall.
In this paper, a store is defined as being close to a department store if it is located
within a 100 feet radius from the entrance of the department store, measured by the
Euclidean distance. Based on this definition, the stores near Sears include store 1,2,3,4,
and 5. Stores near the main entrance of Zellers include store 6, 7, 8, and 9. Stores near the
side entrance of Zellers include store 10, 11, and 12.
The alternative measurement defines that a store is close to a department store
when the entrance of this store is located within 100 feet to the entrance of a department
store, measured by walking distance (i.e. the shortest possible route a consumer would
50
choose). Based on this definition, the stores near Sears include store 1, 2, 4, and 5. Stores
near the main entrance of Zellers include store 7, 8, and 9. Stores near the side entrance
of Zellers include store 11 and 12.
51