Radius Restrictions and the Similarity of Neighboring Shopping Centers by Andrew Eckert and Douglas S. West* Department of Economics University of Alberta Edmonton, Alberta T6G 2H4 Email: Andrew Eckert: aeckert@ualberta.ca Douglas West: Douglas.west@ualberta.ca January 2006 Abstract In Canada and the U.S., shopping center developers may impose “radius restrictions” on tenants in their shopping centers, prohibiting tenants in a particular shopping center from opening another store within a certain radius. Whether a radius restriction is imposed on a chain will depend upon the relative bargaining positions of the chain and the developer. This paper presents an empirical analysis of regional shopping center composition in the five westernmost provinces in Canada, using variables that reflect the bargaining power of retail chains and shopping center developers. We find that large, well established chains are more likely to enter neighboring malls, consistent with the hypothesis that whether a chain enters neighboring malls depends upon its bargaining power. As well, we find that a chain is more likely to enter neighboring malls owned by large developers, consistent with large developers trying to keep retail chains out of the malls of small developers. *The authors thank Janene Taylor and Benjamin Atkinson for research assistance, and the University of Alberta for financial assistance. 1 1 Introduction In the United States, it is well known that shopping center developers have imposed “radius restrictions” on some of the tenants in their shopping centers. These restrictions are described in shopping center leases, and usually prohibit a retail tenant in a particular shopping center from opening another store within a certain radius (e.g. five miles) of the center. The restriction might apply to the entire term of the lease, but it will exclude from consideration a retailer’s stores that were already established prior to signing the lease containing the restriction.1 Radius restrictions have attracted the attention of competition authorities.2 Radius restrictions, in economic terms, are vertical restraints, and like other vertical restraints, there are alternative explanations for their use.3 For example, stores in shopping centers confer demand externalities on one another, either because of multipurpose or comparison shopping on the part of consumers. Developers, in making their tenant choices, presumably take into account the strength of these externalities because of the impact they can have on rental income. However, a retail firm will choose its store locations to maximize profits given the demand externalities imposed on its store by established tenants, but will not necessarily be concerned with the effect that its location decision has on other tenants (e.g., the diminished demand externality imposed by the store in one mall if an identical store opens in a neighboring mall). A radius restriction can eliminate this dilution of the demand externality. On the other hand, radius 1 See Kampler and Pomerantz (2003). See Kampler and Pomerantz (2003) and Lentzner (1977). 3 Radius restrictions are similar to the particular vertical restraint of exclusive dealing, in that a retail chain that wishes to lease retail space from one mall is excluded from leasing additional space from a geographically close competing mall. For a discussion of the competitive effects of exclusive dealing restraints, see for example Mathewson and Winter (1987). 2 2 restrictions could be used to make it more difficult for rival centers to attract high quality tenants. It is possible that for certain retail chains, the interests of the developer and the chain will coincide: both the developer and the chain may wish the chain to locate in only one or two malls in a city. However, the interests of the developer and retail chain need not coincide. Some retail chains will see it in their interests to try and locate stores in every mall in a city. The ability of a developer to impose a radius restriction that would prevent this from happening will depend upon the bargaining power of the developer and the retail chain. The purpose of this paper is to carry out a spatial empirical analysis of regional shopping center composition in the five westernmost provinces in Canada using variables that reflect the bargaining power of retail chains and shopping center developers. Our analysis will allow us to identify some important characteristics of chains that tend to appear in neighboring malls, and characteristics of shopping centers associated with chains entering neighboring centers. In terms of results, it is found that certain retail chains are more likely to be present in neighboring malls when both malls are owned by large developers than when one of the malls is owned by a small developer. This result is consistent with large developers trying to keep retail chains out of the malls of small developers. New retail chains are less likely to be in neighboring malls, and this is consistent with the radius restriction clause being imposed on new rather than established tenants. Chains are more likely to be in neighboring malls when the malls are farther apart, but are less likely to be in neighboring malls in large cities. The size of retail chain, as measured by the number of cities containing members of the chain, is also positively associated with whether a 3 chain is in neighboring malls, and this is consistent with bargaining power increasing with the size of the chain. Certain other variables, such as whether a chain sells private label merchandise and is part of a multi-chain firm, also play a role in explaining whether a chain store is in neighboring malls. This paper undertakes a systematic examination of factors that could affect the ability of the developer to exercise market power. Developers with market power would likely exercise it by raising rents to their retail tenants, and their ability to do this could depend on whether their tenants have space in neighboring centers. While there is a variety of factors that can contribute to the explanation of shopping center rents, our empirical work suggests that market power considerations cannot be dismissed. In the next section, the radius restriction literature is briefly reviewed along with the shopping center lease and shopping center similarity literature. There is little consideration in this literature of which party has the power in the shopping center/tenant relationship. Certain variables that may affect or reflect bargaining power, and whether a retail chain will occupy space in neighboring shopping centers, are then discussed. In Section 3, the data sources to be used in the empirical examination are described. Section 4 presents the regression model to be estimated, which essentially involves explaining whether a chain store that is in one shopping center is also in a neighboring shopping center. Section 5 presents summary statistics, while the results of econometric estimation are discussed in Section 6. Section 7 provides a summary and some concluding remarks. 4 2 Radius Restrictions and Shopping Center Similarity 2.1 Radius Restriction Clauses in Shopping Center Leases Radius restriction clauses in shopping center leases have been used for many years in the U.S. Lentzer (1977) has provided the most detailed discussion of these clauses, their impact, their history, and antitrust treatment. Lentzer (1977, 10-11) reported that in a survey of 34 regional shopping centers and five specialty centers in Los Angeles in 1976, 88 percent of the responding landlords reported the use of radius restrictions in their centers for all but the largest and most desirable tenants. These restrictions generally prohibit the retail tenant from opening up another store within a certain radius of the center.4 Eighty-four percent of the respondents indicated a radius restriction that extended between three and five miles from the center. Sixty-seven percent of the respondents imposed the restriction for the duration of the lease. Lentzer (1977, 12) has written that landlords have advanced two arguments in favor of radius restrictions: (1) radius restrictions are necessary to protect percentage rental income paid by the tenant, and (2) the radius restriction “enables the landlord to combat the ‘unfair competition’ of his own tenants that might destroy the economic vitality of the center itself.”5 Without the radius restriction, a tenant’s sales in one center might be reduced by its presence in a neighboring center, thereby reducing or eliminating the percentage rent that the landlord would otherwise have collected. 4 Jones, Addison, and Davidson (2001, 57-64) provide examples of typical radius and exclusive use provisions in shopping center leases. The latter type of provision generally restricts the landlord from renting space in the center to competing retailers. 5 Shopping center leases generally require tenants to pay a fixed minimum monthly rent, and a percentage of sales rent once sales exceed some specified threshold. See Murray (2001) for a detailed discussion of percentage rent provisions in shopping center leases. Lee (1995) provides an economic analysis of the fixed rent/percentage rent shopping center lease. 5 A survey of retail tenants, whose results are also summarized by Lentzer (1977, 18) indicated that single store and small chain operators do not have sufficient bargaining power to challenge the radius restriction, whereas larger firms were able to negotiate radius restrictions which reduced the size of the territorial restriction and its duration. With increased scrutiny of radius restrictions by the Federal Trade Commission, Lentzer (1977, 23) stated that landlords have reduced radius restrictions to a two to five mile radius, and put them in effect for the first five years of the lease. The latter change suggests that newer retail chains and tenants are more likely to be affected by the radius restriction than long established chains. In addition to the ample evidence of the continuing use of radius restrictions in U.S. shopping center leases, there is some evidence of their use by Canadian developers. For example, in a report on discontinued cases for the years 2000-2002, the Competition Bureau discusses a complaint that Iberville Development was using certain shopping center lease provisions that violated the market restriction provisions of the Competition Act. “Although the 16-km distance used in the radius clauses adopted by Iberville Development Ltd. is higher than the distances usually used in the industry, the Bureau concluded, following an in-depth and detailed study of the facts, that the radius clauses used by Iberville Development Ltd. are not likely to substantially lessen competition in the Sherbrooke area.”6 Some Canadian shopping center developers apparently use radius restrictions in their shopping center leases.7 See Competition Bureau, “Discontinued Cases 2002-2001-2000”, available on the Competition Bureau’s website at http://www.competitionbureau.gc.ca/internet/index.cfm?itemID=1192&1g=e, last accessed on August 26, 2005. 7 Additional evidence of the use of radius restrictions by Canadian shopping centers is provided by a report of a Canadian law firm, Jones, Addison and Davidson (2001). This report contains a discussion of the appropriate wording and drafting of radius restrictions. Radius restrictions also are discussed by the International Council of Shopping Centers (2001), which describes them on page 145 as “the shopping 6 6 2.2 Recent Economic Studies of Shopping Centers There have been some studies of shopping centers in the economics literature, but none that focus on the radius restriction and whether chain stores locate in neighboring centers. Much of the literature tries to explain shopping center rents. Benjamin, Boyle, and Sirmans (1990) construct a model of retail leases that predicts that minimum fixed rents are lower with higher percentage rents and are higher with higher threshold sales levels. Benjamin, Boyle, and Sirmans (1992) find empirical support for a shopping center model’s predictions that developers use tenant characteristics, such as the probability of default and customer traffic-generating potential, to set shopping center rents in a discriminatory fashion. Pashigian and Gould (1998) wish to examine whether shopping center rents reflect demand externalities among stores in the center. They find that the external economies generated by anchor stores are reflected in the lower rents paid by the anchors; the authors report on page 115 that their estimates suggest that “anchors receive a per foot rent subsidy of no less than 72 percent that which nonanchor stores pay.” In a later paper, Gould, Pashigian, and Prendergast (2005) use data on the rent, sales, and contractual provisions of over 2500 stores in large shopping malls in the U.S. to show that (1) the increasing presence of anchor stores in a mall generates higher sales, and consequently higher rents, for non-anchor stores, and (2) “stores with a national name brand, which also tend to generate mall traffic, receive significant rent discounts”.8 The Gould, Pashigian, and Prendergast (2005) study is interesting in that, like the present center landlord’s effort to assure itself that the tenant will devote its energy to the operation of its business at the shopping center, generate the greatest possible sales from the premises, and not be induced by competing landlords.” 8 See Gould, Pashigian, and Prendergast (2005, p. 411). 7 study, it uses store-specific information in the empirical analysis. However, they do not examine whether possible radius restrictions are related to the observed rent differences in their sample. Other papers in the shopping center literature have focused on the shopping center space allocation problem. Brueckner (1993) and Miceli, Sirmans, and Stake (1998) construct models of space allocation in shopping centers when inter-store externalities are present. Neither paper addresses the problem of shopping center allocation of space to specific store brands when those brands might be in neighboring centers. Mejia and Eppli (1999) look at the effects of allocating space to large stores of particular types on the sales per square foot of smaller stores of the same type, and generally find these effects to be positive and significant. In a different but related vein, Eppli and Shilling (1996) use a retail gravity model and data from 38 U.S. regional shopping centers to estimate the retail sales potential of regional shopping centers. They find that actual retail sales at regional shopping centers are largely determined by center size (in square feet), and to a lesser extent by distance to competitors. They conclude that large regional shopping centers may dominate smaller regional centers. There are of course a number of shopping center characteristics (such as store variety and the attractiveness of the shopping center’s store brands) that could be related to center size that are not included in Eppli and Shilling’s analysis. There have been two papers that have compared the store brand compositions of regional shopping centers. West (1992) used regional shopping center data from Calgary and Edmonton to show that malls located in the same city will be more similar (in certain store types) than malls located in different cities, malls owned by the same firm will be 8 more similar than malls owned by different firms, and malls will be more similar in certain store types dominated by multi-chain firms than in store types that are not. Golosinski and West (1995) extended the analysis of shopping center similarity by studying the internal compositions of 97 shopping centers owned by the eight largest shopping center developers in the five westernmost Canadian provinces. They also provided a double moral hazard hypothesis for the similarity of shopping centers owned by the same developer. That is, to reduce the incentive to shirk on the part of each party to the shopping center contract, a multi-center developer will promote the growth of retail chains by offering space in all of its centers to members of the same chain, while retail chains seek locations in the centers owned by a multi-center developer. Golosinski and West found that shopping centers owned by the same developer were more similar than centers owned by different developers and that centers in the same city and same province were more similar than centers in different cities and provinces. Shopping centers owned by large developers tended to be more similar than shopping centers owned by small developers. These results support the double moral hazard explanation for shopping center similarity. While West (1992) and Golosinski and West (1995) considered various factors that might help explain the similarity of shopping centers, neither paper attempted to explain the degree of similarity specifically among neighboring centers, or to consider the role of proximity in determining similarity, other than examining the effect of being in the same city. As will become clear below, the bargaining power of the developer compared to that of the retail chain could be very important in determining whether a chain will be able to acquire space in neighboring shopping centers. 9 2.3 Chain Store Presence in Neighboring Shopping Centers Lentzer (1977) suggests that whether a retail chain will be in neighboring shopping centers depends on the bargaining power of the developer and the retail chain, a suggestion consistent with other available documents and analyses pertaining to radius restrictions. Before proceeding to analyze relative bargaining power further, we must define what is meant by the phrase “neighboring shopping centers”. It will be assumed that each city contains a hierarchy of shopping centers with two important characteristics: (1) a large number of small centers and a small number of large centers, and (2) centers at a given level of the hierarchy contain all of the store types that locate in lower level centers, plus additional store types as well. The shopping center hierarchy that has been analyzed in the past contains at least four levels of centers: neighborhood centers, community centers, regional centers, and the central retail district. Shopping centers can also be planned or unplanned, where a planned center is one that is under the control of a single owner/developer that controls entry into the center. For purposes of this paper, the focus will be on planned regional shopping centers. The reason is that certain store types (e.g. clothing and shoe stores) are more likely to be found in regional rather than lower level shopping centers, and these store types tend to be those for which developers that control entry may wish to impose radius restrictions.9 Once attention is focused on regional shopping centers, there are a number of ways one can define neighboring centers. The fact that lease restrictions are based on a fixed radius leads us to define neighboring centers as centers within such a radius. 9 Empirical affirmation of the hierarchal structure is provided in West, Von Hohenbalken, and Kroner (1985). 10 Because reference is made in Lentzer (1977) to a five mile radius, in this paper we will consider two centers to be neighbors if they are within five miles of each other. The incentive that developers have to impose radius restrictions is related to the drawing power or catchment area of the center. If regional shopping centers were identical, then consumers would have an incentive to patronize the closest center to them in order to minimize transportation costs. A regional center’s catchment area would then consist of the set of points closer to that center than to any other center in the market. Given that regional centers are not identical, some centers will have greater attractive power than others, in part based on their quality level and tenant list, drawing consumers from greater distances. One would expect the sales and profits of the developer of and tenants in more attractive centers to be higher as a result. The presence of a high profile chain in a shopping center can have external effects on the other tenants in the center, in that its presence in the center increases the number of people frequenting the shopping center and doing business with the other tenants. As a result, the decision of a chain to enter a neighboring mall will have external effects on the shopping center and its tenants not taken into account by the chain, and which the shopping center may wish to address with a radius restriction. As well, a radius restriction imposed on a chain that benefits a developer and some of its retail tenants also may not be in the interest of the retail chain since the chain forgoes locations in neighboring centers. For these reasons, the interests of developers and their retail chains may or may not coincide, and the ability of a developer to impose a radius restriction will depend on a number of factors as elaborated below. 11 First, the developer will not necessarily want to impose a radius restriction on all of the retail tenants in its regional centers. Rather, it will wish to impose it on store types that figure most prominently in the consumer’s decision as to which center to patronize. Regional shopping centers contain a variety of different store types, some catering to multipurpose shoppers and selling convenience goods (e.g., grocery and drug stores), and some, catering to multipurpose/comparison shoppers, that sell comparison shopping goods.10 Given that it is the latter type of stores (“C” stores) that largely distinguish regional centers from centers at lower levels of the hierarchy, one would expect comparison shopping store types to be the primary target of radius restrictions.11 Among C stores, there are some that sell brands that tend to be advertised either as private brands for the retailer or as brands that are exclusive to a retailer in a certain geographic area. The combination of offering exclusive brands and attractive brands will make these stores candidates for radius restrictions on the part of the developer. One would also expect newer retail chains to be subject to radius restrictions to a greater extent that older chains. Shopping center developers lease space to retailers for fixed terms, and with lease renewals that may be at the developer’s discretion. This allows the developer the opportunity to replace tenants whose leases expire with new retailers that have greater sales potential. The new retailers may be ones that are selling new products and services that are in high demand, and that can attract consumers to the 10 Golosinski and West (1995) refer to the latter such stores as C stores, and define them as follows: C stores sell merchandise that is (a) highly differentiated in both horizontal and vertical dimensions and for which branding is important, and (b) sufficiently costly that there are perceived positive net returns to search across stores for price and quality. C store types include women’s clothing, shoe and jewelry stores. 11 For example, Cambridge Shopping Centres Limited (1999) states on page 6 that regional malls “derive about 50% of their sales from the fashion and footwear segments…” and on page 8 that “when it comes to categories such as fashion apparel, footwear and accessories, established regional malls will continue to enjoy a competitive advantage.” 12 shopping center. It would be in the interests of the developer to keep such retailers out of neighboring shopping centers.12 Another potential determinant of whether a developer can impose a radius restriction is the developer’s size. As mentioned above, many C store types seek locations in regional shopping centers precisely because of the comparison shopping opportunities that they facilitate.13 Developers that own more than one or two regional shopping centers (“large developers”) may wish to impose a radius restriction on a particular retail chain and can make access to other centers that the developer owns conditional on accepting the restriction.14 However, the restriction may not be imposed in the same way on all C store chains, and there may be differences in whether the restriction is enforced against a neighboring large developer’s as opposed to small developer’s center. Large developers could attempt to use the radius restriction to keep retail chains out of the small developers’ centers, making them less attractive and less profitable. Ultimately, they could fail and exit the market. Of course, the developer will not always be successful at imposing a radius restriction. Certain retail chains may be highly desired as tenants by a developer and have sufficient bargaining power to reject the restriction. In particular, large retail chains will presumably be attractive to developers to the extent that their size signals profitable expansion. In addition, a large chain has less incentive to shirk due to the impact that can 12 See again Lentzer (1997, 23). In a Canadian context, Jones, Addison and Davidson (2001, 58) state that standard exemptions to radius restriction clauses include “existing tenants of the shopping centre” and “major stores and specifically named department stores.” 13 Whether stores selling the same or similar goods will cluster in space depends on the strength of two forces: the competition effect, which generally pulls the stores apart, and the market area effect, which drives them to locate together. See Stahl (1987) for a discussion and analysis of these effects. 14 For certain store types, there may be a much larger number of retail chains than there are shopping centers in a city within which to locate. Competition among retail chains for space in regional shopping centers, especially for chains in store types that consider alternative locations to be unacceptable, could tip the bargaining power balance towards the developer. 13 have on the chain’s reputation, and this too will be attractive to developers. Furthermore, a large attractive chain might exercise bargaining power by threatening not to take space in any of a developer’s centers, or by threatening to withdraw from all of a developer’s centers. Some large retail chains may see it in their interests (and have the bargaining power) to obtain locations in as many regional shopping centers as possible, given that many consumers will patronize the closest regional center, and will not necessarily travel a greater distance to patronize certain retail chains. Besides the size of the chain, two other possible indicators of a chain’s bargaining power are whether it is part of a multi-chain firm and whether it carries private label merchandise. If a chain is part of a multi-chain firm, where each chain is viewed by the developer as a desirable tenant, then the firm may be able to make leasing multiple stores from the developer conditional on the absence of a radius restriction. However, it is also the case that some multi-chain firms (e.g., those that specialize in women’s clothing) would place a high value on acquiring multiple store locations in a shopping center so that the firm can capture the sales of comparison shoppers regardless of which store they patronize. It is then an empirical question whether the developer’s or multi-chain firm’s preferences regarding the radius restriction will be satisfied. If a chain sells private label merchandise, it may also have bargaining power provided the private label is one that is in demand by consumers. As compared to retailers that stock multiple non-exclusive brands, a retailer selling an exclusive private label brand may be in a better position to reject a radius restriction if the developer has a strong preference for having the brand represented in its center. There are, however, certain retail chains that have such strong private brands that they believe that they can 14 attract consumers to their stores even if they are not in every regional center. For such retail chains, the interests of the developer and retail chain can coincide. It is once again an empirical question as to whether, on balance, chains selling private label merchandise are more or less likely to have stores in neighboring regional shopping centers. From what has been reported in the radius restriction literature, the radius used in the restriction can vary. It is likely then that the greater the distance between two regional shopping centers, the less likely a radius restriction will actually apply. It is apparent then that if a chain tends not to be in neighboring centers, that can be consistent with developer bargaining power as well as with a chain with strong brand preference that does not need to be located in all centers to maximize its sales and profits. If a chain tends to be in neighboring centers, that can be consistent with chain bargaining power, as well as with the lack of implementation of radius restrictions for store types or chains for which the developer perceives little net benefit. In the empirical analysis to follow, we will seek to determine whether the conjunction of the results point to greater bargaining power being exercised by retail chains or developers. 3 Data Description 3.1 Shopping Centers This paper uses data on planned regional shopping centers (i.e. malls) collected from the 2002 Canadian Directory of Shopping Centres. Regional shopping centers from the Canadian provinces of British Columbia, Alberta, Saskatchewan, Manitoba, and Ontario are considered. Initially, all shopping centers with a reported store capacity of 75 were included in the data set. It was then required that the actual store count of a center 15 be at least 60 in order for a center to be included in the data set.15 Further, shopping centers were dropped from the data set if the percentage of stores that were chain stores was 50 percent or less. This restriction primarily eliminated malls with a square footage of less than 200,000, and malls lacking a major anchor store. Finally, because one of the objectives of this paper is to explain when chain stores appear in neighboring shopping centers, the data set includes only centers that have at least one other mall in the sample within five miles of their locations. This yields a final sample of 85 shopping centers. Of these, 89 percent have no more than five other shopping centers within five miles, and 61 percent have only one or two other centers within five miles. The malls in the sample produce 109 pairs of shopping centers within five miles of each other. These will henceforth be referred to as neighboring shopping centers. As noted in Section 2, whether a chain is in both of two neighboring centers could depend on the number of centers owned and/or controlled by the developer. Table 1 presents the number of malls owned and/or controlled by each company owning or controlling at least two centers in the sample.16 The final row in the table lists the number of companies that operate only one mall in the sample. Table 1 shows that 62 percent of the sample malls are managed by the five largest companies: Cadillac Fairview Corporation Limited, Ivanhoe Cambridge, 20 Vic Management Inc, Oxford Properties Group Inc, and Morguard Investments Limited. 15 This allows for the elimination of centers from the data set that have high vacancy rates or that have failed to completely report their tenant list to the directory’s publisher. 16 Some malls in the sample are owned by pension funds, but managed by professional management companies. Since the management company is presumably responsible for leasing, malls are classified by management company when owner and manager differ. All companies owning and/or managing malls will be referred to as “developers”. 16 For this study, a developer is defined to be small if it operates one or two sample shopping centers. A developer is large if it operates more than two sample centers. By this definition, 64 of the 85 sample shopping centers (or 75 percent) are considered large. Of the 109 pairs of neighboring shopping centers in the sample, nine are pairs operated by the same large developer, 47 involve two large developers, 46 involve one large and one small developer, and only seven pairs involve two small developers. 3.2 Retail Stores and Chains The 2002 Canadian Directory of Shopping Centres reports the tenant lists of shopping centers in Canada. In the 85 sample malls, there are 10,217 retail stores, where a retail store is defined as a seller of goods or services that might be purchased on a multipurpose shopping trip.17 Of these stores, 7,960, or 78 percent, are members of chains, where a chain is defined as having at least two stores with a common name in the sample.18 There are 930 chains in the data set. The stores in the sample can be divided into a number of categories, according to the goods or services sold by the store. This categorization is made by the 2002 Canadian Directory of Shopping Centers. Not all categories were included in the analysis. First, we excluded categories for which the total number of stores in the sample was 75 or less. Chains in categories with low store counts are unlikely to be found in neighboring centers because of the low store count in the category. Second, because the focus of this paper is on whether radius restrictions are applied to chain stores, store categories for which 17 Certain locations in a mall are occupied by non-retail tenants, such as doctors, lawyers, real estate agents, and financial consultants. These mall tenants are excluded from the store list for each mall. 18 Note that in a small number of instances, a chain operated more than one store in the same mall. Because our interest is in whether a chain is present in neighboring malls, and not in how often it appears in the malls, the second occurrence of a chain in a mall is dropped from the data. 17 attracting a chain store is unlikely to be important were dropped. It is assumed that such categories would be those for which the fraction of stores that are chain stores is less than 50 percent.19 Finally, the restaurants/fast food and specialty food/drink categories were dropped because it seems unlikely that a shopping center would have an incentive to impose a radius restriction on these businesses. Shopping center developers include these types of stores to keep hungry/thirsty shoppers from leaving the mall, but their presence in a particular mall probably does not motivate the choice of mall for most shoppers. Table 2 presents a list of the remaining store type categories used in this study, along with the number of stores of each type in the sample, the percentage of the sample accounted for by each store type category, and the percentage of the category composed of chains. As Table 2 illustrates, the category with the most stores is ladies’ wear (1428 stores), representing 14 percent of all stores in the sample malls. Other large categories include unisex clothing (811 stores), footwear and leather accessories (727 stores), and jewelry (599 stores). The percentage of stores in a category that are chain stores ranges from 53 percent in furniture/home décor to 96 percent in department stores. 4 Econometric Model and Variable Definitions The dependent variable for the regression analysis is Innbri,jk, which is defined for a pair of neighboring malls j and k, and equals one if chain i is in both malls, and zero if the chain is in only one – the variable is not defined if the chain is not in either mall. 19 The categories dropped by these two criteria are: supermarkets, dry cleaners, florists, other groceries, hairstyling, hardware, hobby/craft, specialty merchandise, theatres, toys, draperies and blinds, home appliances, automotive, printing, small department stores, other services, liquor, shoe repair, and video rental. 18 There is therefore one observation for each pair of malls and for each chain that is in at least one mall in the pair. In the econometric specification, the probability that a chain that is in one shopping center of a pair is in both shopping centers is estimated as a function of chain characteristics, mall characteristics, and location characteristics. Specifically, it is assumed that: Prob (Innbri,jk = 1) = Φ(β` Xi,jk), where Xi, jk is a vector of characteristics for observation i,jk, β is a vector of coefficients, and Φ( ) is the cumulative distribution function for a standard normal random variable. Hence, a probit model is estimated. 20 The variables in the econometric model are divided into three groups: chain store variables, shopping center variables, and location variables. Their definitions and the expected signs of their coefficients are as follows: Chain Store Variables The following variables control for the characteristics of a particular chain: ncities: = the number of census metropolitan areas in which a retail chain has at least one store; multichain: = 1 if the chain is part of a multi-chain firm, and 0 otherwise; 20 In principle, one could jointly estimate an equation determining whether a chain enters a particular pair of malls, and then, given that it has entered a pair, an equation determining whether it enters one mall or both. We have opted against this approach for several reasons. First, such an approach would require a variable that determines whether a chain enters a particular pair, but not whether it enters one or both malls in a pair. Given that the intended analysis in the paper is reduced form, it was not clear to us what such a variable would be. Secondly, because of the limited dependent variables, the two equations would need to be estimated simultaneously. Finally, the specification and estimation of a reasonable model determining whether a chain enters a particular geographic area (one interpretation of a pair of chains) is simply beyond the scope of this paper, but could be an interesting direction for future research. 19 label: = 1 if the chain has a private label, and 0 otherwise; newchain: = 1 if the chain was not among the chains derived from the 1993 Canadian Directory of Shopping Centres (where a chain has at least two stores with a common name in shopping centers with a reported store count of at least 75 stores); growth: = the change in the number of stores in the chain from 1993 to 2002; growth = 0 if newchain = 1. The variable ncities measures whether the chain is a national chain present in a large number of cities, or whether it is a local chain. Because ncities is a measure of the size of a retail chain, and larger retail chains are expected to have greater bargaining power than small local chains, the coefficient on ncities should have a positive sign. The variables multichain and label measure whether a chain is part of a multichain family, and whether the chain has a private label, respectively. The variable multichain was constructed from the 2002 Directory of Retail Chains in Canada, which provides, for each chain listed in the directory that responded to the question, the parent company and subsidiaries. It is anticipated that whether a chain is a member of a multichain firm could have offsetting effects on whether a chain enters both malls of a pair (see Section 2). The expected sign of the coefficient on multichain is therefore ambiguous. The variable label takes on a value of one if a chain sells private label merchandise, and a value of zero otherwise. The expected sign of the coefficient on label is unclear. On the one hand, a chain with a recognizable private label will be more attractive to a mall developer and thus have greater bargaining power than a chain that 20 distributes brands that can be purchased elsewhere. On the other hand, chains with highly attractive private labels may attract consumers from a wide geographic area, and may not desire to be in both neighboring shopping centers. In terms of finding which chains offer private label merchandise, the 2002 Directory of Retail Chains in Canada provides this information for some of the chains in the data set. However, because of the incompleteness of this data, other steps were taken to obtain observations for this variable. First, it was assumed that chains that were not indicated by the directory as having a private label, and which had less than five stores, did not have private labels. Second, some chain stores located in shopping centers in Edmonton, Alberta, were visited to determine whether they sold private label merchandise. Third, internet searches were made of chain store websites to see if private label merchandise was indicated. Fourth, for certain store type categories, the nature of the business was such as to imply that the store would or would not be selling private label merchandise. Finally, all chains for which a private label could not be identified using the above methods were assumed not to have a private label. The variable newchain measures whether the chain is at least ten years old. The anticipated sign of newchain’s coefficient is negative. As discussed earlier, developers are less inclined to impose radius restrictions on older chains. As well, a new chain is less likely to have a widely recognized brand name, and may therefore have less bargaining power. However, it may also have a brand for which a developer would like to have some degree of exclusivity. Finally, new chains may not have had the opportunity yet to expand into neighboring centers. 21 The variable growth is expected to have a positive sign. Chains that have been successfully growing are anticipated to have greater bargaining power than chains whose store counts have been dwindling over time. Finally, to control for systematic differences across store type categories not picked up by chain-specific variables, dummy variables are included for the different store type categories. These variables are expected to pick up differences across categories in the attractiveness of chains to developers, as well as differences in concentration across the various categories. Shopping Center Variables The following variables control for characteristics of the neighboring shopping centers and developers in each pair: bothbig: = 1 if neighboring shopping centers in a pair are operated by big developers as defined earlier; bothsmall: = 1 if neighboring shopping centers in a pair are operated by small developers; bothsame: = 1 if neighboring shopping centers in a pair are operated by the same developer; distance = the distance between two neighboring centers, in miles crst: = 1 if the shopping centers in a pair are across the street from each other; mintenants: = the number of tenants in the smaller of the two malls; maxage: = the age of the older of the two malls downtown: = equals 1 if exactly one mall in the pair is located in its city’s downtown, 22 and 0 otherwise. The variables bothbig and bothsmall are intended to control for differing levels of bargaining power across large and small shopping center developers. The control group is neighboring mall pairs in which one shopping center is operated by a large developer and the other by a small developer. When neighboring malls are owned by small developers, and holding other variables intended to control for shopping center quality constant, it is expected that both malls will have little bargaining power, and therefore it is more likely that a chain will be in neighboring malls. This predicts a positive sign on the coefficient on bothsmall. Alternatively, if the other variables are not adequately controlling for shopping center quality, and if lower quality shopping centers are more likely to be operated by small developers, one would expect a negative sign on the coefficient of bothsmall. Controlling for the qualities of the centers, if a large developer is more likely to use a radius restriction for predatory or anticompetitive purposes to keep a chain out of centers owned by small developers, one would expect a positive sign on bothbig, as it would be more likely that the chain will enter both malls when they are both owned by big developers. This would also be consistent with a collusive outcome in which large developers had agreed (or reached a tacit understanding) that they would not use radius restrictions against each other. If two neighboring malls have the same developer, then the developer has less incentive to keep a chain out of one of the malls, and we would expect a positive sign on bothsame. A chain will likely have less incentive to enter both malls in a pair if the two malls are close together. Further, if the two malls are in close proximity, then the 23 developer’s incentive to impose a radius restriction is increased. To control for the proximity of two malls in a pair, the variables distance, which measures the distance between two malls (in miles), and crst, which equals 1 if two malls in a pair are located across the street from each other, are included in the regression. A positive sign on the coefficient of distance and a negative sign on the coefficient of crst are expected. A chain will be more likely to enter both malls of a neighboring pair when both malls have a large capacity, as measured by the number of tenants. As well, a small shopping center will be less attractive to a chain because it may not offer sufficient opportunities for multipurpose and comparison shopping. For these reasons, a positive sign on mintenants is predicted. The variables maxage and downtown are included to control for shopping center quality and differences in the target demographic of shopping centers. The expected sign on the coefficient of downtown is ambiguous. On the one hand, downtown is a central retail district that can contain stores not found elsewhere (e.g., upscale department stores or jewelers) that draw customers from the entire metropolitan area. Downtown centers will then tend to have some tenants not found in suburban centers. On the other hand, downtown centers also cater to office workers who may wish to shop on a lunch break or after work, and therefore could contain stores found in suburban centers. The determination of whether a shopping center is considered downtown was made by consulting road maps and shopping center websites. The sign of maxage is uncertain. If older shopping centers are less attractive to consumers or in less desirable neighborhoods, then chains will be less attracted to these centers, and one should expect a negative coefficient on maxage. However, if older 24 centers are more likely to be recently renovated, or are located primarily in popular downtown areas, then chains may desire to be in both malls, and in the absence of radius restrictions, will locate in both. Then too, older centers are also those that contain established tenants for which radius restrictions might have less value for developers. Location Variables The following variables are intended to control for local competition and differences in cities: nother: = the number of other malls within 5 miles of at least one of the malls in a neighboring pair; pop: = the population of the census metropolitan area containing the neighboring pair of malls, in millions; Toronto: =1 for all shopping centers in the Toronto census metropolitan area, and 0 otherwise. In areas in which there are a large number of other sample malls, a chain has many options for its locations, and it is then less likely to be in both malls of a particular neighboring pair. Thus, a negative sign on the coeffieient of nother is predicted. The variable pop is also expected to have a negative coefficient. First, it is anticipated that larger cities with their larger catchment area populations are more attractive locations to chains than smaller cities. This conveys additional bargaining power to shopping centers in large cities, enhancing their ability to impose radius restrictions. Second, shopping centers have a greater incentive to vigorously protect their trade area in larger cities. Finally, nother and pop seem to be related to whether the shopping center pair is located 25 in the Toronto census metropolitan area. Therefore, to ensure that pop and nother are picking up the effects of population and local competition and not other differences between Toronto and the rest of the country, a dummy variable for shopping center pairs in the Toronto census metropolitan area is included. 5 Summary Statistics Summary statistics for the dependent variable and for all explanatory variables (except the category dummy variables) are given in Table 3. For 17 percent of the observations, a chain that is present in a pair of neighboring shopping centers has stores in both centers of the pair. In the sample, four percent of observations involve two small shopping centers, while over half of the observations involve two large centers and eight percent involve the same developer. Very few observations involve shopping centers across the street from each other, and on average neighboring shopping centers are approximately three miles apart. The smaller shopping center in each pair has on average 92 tenants. Just over one-third of the observations involve one downtown shopping center. On average, the oldest mall in a pair is 37 years old. Forty-six percent of the observations are for shopping centers in the Toronto census metropolitan area. The average number of competing malls within five miles of a pair is 3.5, and the average value of pop for each observation is 2.7 million. Regarding chains, approximately half of the observations are for multi-chain firms, and just over half are for chains with private labels. On average, a chain is present in approximately 15 census metropolitan areas across the sample provinces. 26 Approximately one quarter of chains are new, and the average growth of previously existing chains from 1993 to 2002 was approximately 12 stores. Table 4 provides the mean value of the dependent variable for each category of store. Also included in Table 4 are the number of national chains by store type category, and the percentage of stores accounted for by the five largest chains (CR5). For this purpose, a chain is considered national if it is present in all five provinces in the sample. Table 4 indicates a reasonably strong relationship between whether a chain is in both neighboring malls and the concentration of the category – the correlation coefficient between the second and fourth columns is 0.61. The relationship between whether a chain is in both malls and the number of national chains is less clear. This is possibly because a small number of national brands may reflect either a concentrated category or a narrow category (and one for which there is not great demand). 6 Econometric Results Results from the probit estimation are given in Table 5.21 Coefficients on store type category dummy variables are suppressed. Regarding which chains enter into both malls of a pair, probit estimation yields the following results. The coefficient on ncities is positive and significant at the one percent level, indicating as expected that chains that operate in a large number of cities are more likely to enter both malls of a pair than are small local chains. This finding is consistent with the hypothesis that chains with national recognition possess greater bargaining power, and are able to command more favorable lease terms from shopping 21 Reported standard errors are not robust to heteroskedasticity. However, using robust standard errors does not change any conclusions regarding statistical significance. 27 centers. As predicted, new chains are found significantly less likely to be in both malls of a pair than established chains. Conditional on a chain being established, growing chains are more likely to appear in both malls than shrinking chains, which is again consistent with more attractive and successful chains being able to command favorable lease terms. The coefficients on label and mulichain are both negative, although significant only at the ten percent level. The negative coefficient on label may be driven by the fact that, for certain chains, having a private label allows the chain to attract consumers from a wide geographic area, making multiple nearby locations undesirable for the chain. The negative coefficient on mulichain suggests that stores owned by a multi-chain firm tend to locate together, implying that a shopping center developer might be imposing a radius restriction as a condition for a multi-chain firm obtaining multiple store locations in its center. Regarding the characteristics of the shopping centers, the positive and significant coefficient on distance, coupled with the negative and significant coefficient on crst, indicate that a chain is less likely to locate in both shopping centers in a pair when the shopping centers are close together. This result is consistent with basic spatial theory. The positive and significant coefficient on maxage implies that a chain is more likely to be in both malls when at least one of them is old. This result is consistent with at least one of the malls being in an attractive older part of the city, and it might also be the result of older malls being renovated and of higher quality. The coefficient on downtown is not significant at any conventional significance level. The results also show that, controlling for some shopping center characteristics, a chain is more likely to enter both malls of a neighboring pair when they are both operated 28 by major developers, than when one is operated by a major developer and the other is operated by a small developer. This finding is consistent with the hypothesis that large developers use radius restrictions on their chain tenants in a possibly anticompetitive fashion, to reduce the attractiveness of neighboring centers owned by small developers. However, it is possible that the variables included to control for shopping center quality are not sufficient to the task, and that large developers are more likely to operate high quality shopping centers than are small developers. In that case, the positive and significant coefficient on bothbig may simply be picking up the fact that when both shopping centers are of high quality, a chain is more likely to enter both. The coefficient on bothsmall is not significantly different from zero, for reasonable significance levels. This suggests that for mall pairs in which both malls are small, developers are not significantly more or less likely to have the same chain in both malls than for mall pairs in which one mall is operated by a small developer and the other by a large developer. Likewise, the coefficient on bothsame is not significantly different from zero. The coefficient on nother is not significantly different from zero at reasonable significance levels, indicating that we have not identified a significant relationship between whether a chain enters both malls of a given neighboring pair, and the number of other local mall options for the chain. The coefficient on pop is negative and significant, consistent with the interpretation that shopping centers in major cities possess greater bargaining power, and have a greater incentive to vigorously protect their trade area in larger cities. The coefficient on Toronto is not significantly different from zero at standard significance levels. 29 Since the magnitudes of estimated coefficients in a probit model are not easily interpreted, the estimated effects of key variables on the probability that a chain that is located in at least one of the malls in a neighboring pair will be located in both have been calculated. These effects are reported in Table 6. For purposes of computation, all dummy variables are initially set equal to zero, and all other variables are set equal to their sample means. For dummy variables, Table 6 reports the effect on the probability that a chain will locate in both malls of switching the variable from zero to one, holding everything else constant. For other variables, the reported effect is that of increasing the variable by one standard deviation. Table 6 shows that, for the most part, while the effects of these variables are statistically significant, they are not large. Switching a mall pair from one large and one small developer to two small developers increases the probability that the chain will locate in both malls by only two percent. Switching a chain (with zero growth) from a previously existing chain to a new chain lowers the probability it will locate in both malls by five percent. The probability of locating in both malls increases by one percent if the growth of the chain since 1993 is increased by one standard deviation (approximately 26 stores), and increases by two percent if the distance between the shopping centers is increased by one standard deviation (1.1 miles). Increasing the population of the city by one standard deviation (approximately 1.9 million) decreases the probability the chain will locate in both malls by three percent. The two variables that seem to have large effects on the probability that a chain will locate in both malls are minten and ncities. Increasing the number of tenants in the smaller of the two malls by 35.33, one standard deviation, increases this probability by 30 nine percent, while increasing the number of census metropolitan areas in which the chain operates by twelve, also one standard deviation, increases this probability by thirteen percent. The estimated coefficients and effects of the store type category dummy variables are reported in Table 7. Because the control category is ladies’ wear, statistical significance indicates that chains in a given category are statistically more or less likely to locate in both malls of a pair than chains in the ladies’ wear category. Likewise, the effects given in the last column are the differences in the probability that a chain locates in both shopping centers, comparing each category to ladies’ wear. According to Table 7, chains in eleven store type categories are statistically more likely (at the five percent level) to enter both malls of a neighboring pair than ladies’ wear chains, while only two categories, department stores and home furnishings, are significantly less likely. The fact that department stores and home furnishing chains are less likely to enter neighboring malls than ladies’ wear chains may reflect the possibility that chains in these categories draw their customers from a wide geographic area, and so entering neighboring malls is simply not attractive to chains in these categories. A common feature of those categories where chains are much more likely to enter neighboring shopping centers than ladies’ wear chains is a high five-firm concentration ratio, as given in Table 4. Of the eight categories with concentration ratios greater than seventy percent, all but two (department stores and books and newsstands) are statistically more likely to enter both shopping centers in a pair than ladies’ wear chains, 31 holding other characteristics constant.22 However, certain categories appear more likely to enter both malls than ladies’ wear chains, despite having a lower CR5. The most notable example is the furniture category, which has a CR5 of only 25 percent. To explore the robustness of our findings, the model was re-estimated on certain subsamples of the data. First, one concern was that some of the results could be driven by the assumptions that any two stores in the same category with the same name should be considered a chain, and chains desire store locations in neighboring malls. To check this, the model was re-estimated using only chains that were present in at least 10 census metropolitan areas. This reduced the sample from 12652 to 7342. In terms of the variables and results reported in Table 5, the only changes to the results of statistical tests were that the coefficients on multichain, label, and maxage were no longer significant at any standard significance level. With respect to the category dummy variables, while there were some changes to rankings of categories and significance of certain coefficients, the general pattern and results were maintained. The most noticeable change is that the coefficient on the furniture dummy becomes negative, although it is not significantly different from zero at a ten percent level. Therefore, it appears that the basic conclusions are maintained when the analysis is restricted to exclude small chains. Second, we were concerned that, despite the exclusion of a number of store type categories, the results were being diluted by the inclusion of certain store type categories for which the imposition of radius restrictions is of little importance to chains or developers. To explore this possibility, the model was re-estimated using only the clothing categories (footwear and leather accessories, menswear, ladies’ wear, unisex 22 Since the largest chain in the books category is Chapters, a big-box retailer with stand-alone stores, the finding that chains in this category are relatively unlikely to enter both malls of a pair may perhaps also be explained by Chapters not desiring to enter neighboring shopping centers. 32 clothing, sporting goods, and children’s wear), for which radius restrictions are more likely to be applied.23 Again, most results remain unchanged from the original specification. There were, however, some differences. First, the coefficient on bothsmall becomes positive and significant at the five percent level. This result is consistent with small shopping centers not having the bargaining power needed to impose radius restrictions and still attract the chain. The coefficients on multichain and label remain negative but become significantly different from zero at the five percent level. The coefficient on pop is no longer significantly different from zero at the ten percent level. Finally, the coefficient on downtown becomes positive and significantly different from zero at the one percent level. This result may indicate that, at least for clothing chains, when neighboring malls are located in different types of districts (e.g., one mall is downtown, the other is a mall in a residential suburb), then because these malls cater to some extent to different clienteles, a chain will be less worried about cannibalizing its business by entering both malls. Further, a shopping center may be less worried about losing business to a neighboring rival center when they operate in different types of areas and attract different shoppers. 7 Conclusions Shopping center developers have included radius restriction clauses in some Canadian and American shopping center leases for many years. These restrictions, which specify a minimum distance around a shopping center over which a tenant cannot open a new store, are largely designed to protect the shopping center rents of the developer and 23 These store types account for 36.4 percent of the stores in the sample malls, and are believed to include the ones that are likely to be key determinants in a consumer’s choice of mall to patronize. 33 demand externalities of the developer’s retail tenants. They are not necessarily consistent with the affected tenant’s interests. The ability of a developer to impose a radius restriction depends on the relative bargaining power of retail tenants and shopping center developers. To examine this problem empirically, an econometric model was specified to describe the common characteristics of those chains that tend to enter both malls in a neighboring pair, and the common characteristics of those shopping center pairs which are most likely to allow a chain to enter both malls of the pair. Whether the results are consistent with a simple bargaining explanation for the use of radius restrictions is considered, as well as the conditions under which one party or the other to the retail lease is likely to have the bargaining power to impose or reject a radius restriction. In terms of chain characteristics, it was found that chains that are most likely to enter both malls of a neighboring pair are large national chains, and chains that have been established since 1993 and growing. This description is consistent with the bargaining hypothesis, since such chains are precisely those that would be expected to be in a strong bargaining position with the developers. Regarding shopping center and locational characteristics, consistent with spatial theory, the results show that a chain is more likely to enter both shopping centers when those centers are located farther apart. A chain is more likely to enter neighboring shopping centers when neither is small, likely reflecting capacity constraints for retail space within a shopping center. Neighboring shopping center pairs in which at least one mall is old are more likely to both contain a particular chain, while shopping centers in higher population cities are less likely to both contain the same chain. 34 Finally, it is found that, after controlling for other factors, a chain is more likely to enter both malls when they are both operated by large developers than when one developer is large and the other is small, although this effect on the probability of entering both shopping centers is not large. This finding is consistent with the hypothesis that large developers use radius restrictions in an anticompetitive fashion, particularly against independent malls and small chains of malls. However, it could also suggest that the malls of large developers and the malls of independents differ in a way uncontrolled for by our explanatory variables. Examining where shopping center tenant data are consistent with radius restrictions being imposed is one way of attempting to assess the relative bargaining power of shopping center developers and retail tenants. It may not be the only way, but its value lies in relying on data that are reasonably easy to obtain. Other strategies for assessing relative bargaining power will be explored in future research. 35 References Benjamin, J.D., Boyle, G.W., and Sirmans, C.F. “Retail Leasing: The Determinants of Shopping Center Rents,” AREUEA Journal, Vol. 18 (1990), 302-317. Benjamin, J.D., Boyle, G.W., and Sirmans, C.F. “Price Discrimination in Shopping Center Leases,” Journal of Urban Economics, Vol. 32 (1992), 299-317. Brueckner, J.K. “Inter-Store Externalities and Space Allocation in Shopping Centers,” Journal of Real Estate Finance and Economics, Vol. 7 (1993), 5-16. Cambridge Shopping Centres Limited. Annual Report (1999). Eppli, M.J. and Shilling, J.D. “How Critical Is a Good Location to a Regional Shopping Center?,” Journal of Real Estate Research, Vol. 12 (1996), 459-468. Golosinkski, D. and West, D.S. “Double Moral Hazard and Shopping Center Similarity in Canada”, Journal of Law, Economics, and Organization, Vol. 11 (1995), 456478. Gould, E.D., Pashigian, B.P, and Prendergast, C. “Contracts, Externalities, and Incentives in Shopping Malls”, Review of Economics and Statistics, Vol. 87 (2005), 411-422. International Council of Shopping Centres. Shopping Center Leasing. New York: International Council of Shopping Centers (2000). Jones, D.A., Addison, S.M., and Davidson, P.T. “Fast Effective Pain Relief…McLean & Kerr LLP’s Update on Over-the-Counter Commercial Lease Remedies,” mimeo, March 30, 2001. Kampler, N.L. and Pomerantz, H.B. “Challenging Radius Restrictions in Shopping Center Leases”, American Bar Association 14th Annual Real Property Symposium, April 3, 2003. 36 Lee, K. “Optimal Retail Lease Contracts: The Principal-Agent Approach”, Regional Science and Urban Economics, Vol. 25 (1995), 727-738. Lentzner, J. “The Antitrust Implications of Radius Clauses in Shopping Center Leases”, University of Detroit Journal of Urban Law, Vol. 55 (1977), 1-71. Mathewson, G.F., and Winter, R. A. “The Competitive Effects of Vertical Agreements: Comment,” American Economic Review, Vol. 77 (1987), 1057-1062. Mejia, L.C. and Eppli, M.J. “The Effect of Merchandise Space Allocation on Retail Sales in Enclosed Shopping Centers”, Journal of Shopping Center Research, Vol. 6 (1999), 23-40. Miceli, T.J., Sirmans, C.F., and Stake, D. “Optimal Competition and Allocation of Space in Shopping Centers”, Journal of Real Estate Research, Vol.16 (1998), 113-126. Murray, J.C. “Percentage Rent Provisions in Shopping Center Leases: A Changing World?,” Real Property, Probate and Trust Journal, Vol. 35 (2001), 1-58. Pashigian, B.P. and Gould, E.D. “Internalizing Externalitites: The Pricing of Space in Shopping Malls,” Journal of Law and Economics, Vol. 41 (1998), 115-142. Stahl, K. “Theories of Urban Business Location”, in E.S. Mills (ed.), Handbook of Regional and Urban Economics, Vol. II, Amsterdam: Elsevier Science Publishers B.V., 1987. West, D.S. “An Empirical Analysis of Retail Chains and Shopping Center Similarity”, Journal of Industrial Economics, Vol. 40 (1992), 201-221. West, D.S., Von Hohenbalken, B., and Kroner, K. “Tests of Intraurban Central Place Theories”, Economic Journal, Vol. 95 (1985), 101-117. 37 Table 1: Developers Developer Cadillac Fairview Corporation Limited Ivanhoe Cambridge 20 Vic Management Inc. Oxford Properties Group Inc. Morguard Investments Limited Bentall Retail Services Brookfield Properties Ltd. O & Y Enterprise National Retail Services Omers Realty Management Corp. Revenue Properties Company Limited Number of developers with only one center in sample 38 Number of Centers 20 14 9 5 5 4 4 3 2 2 17 Table 2: The size of store type categories and their chain store counts Category Number of Stores 1428 811 727 Percentage of all stores in sample malls 14.0 7.9 7.1 Percentage of stores that are chains 86 88 83 Ladies’ wear Unisex clothing Footwear and leather accessories Jewelry Gifts Menswear Electronics Sporting goods Telecommunications Variety/convenience Furniture/home décor Optical Cosmetics/toiletry Children’s wear Housewares Cards/stationery Department stores Financial Photo/camera Music/instruments Travel Fabric and sewing Books and newsstands Drugs/health and beauty aids 599 336 321 280 248 219 218 216 5.9 3.3 3.1 2.7 2.4 2.1 2.1 2.1 66 71 82 90 89 89 77 53 208 196 194 194 167 150 145 140 128 110 109 93 2.0 1.9 1.9 1.9 1.6 1.5 1.4 1.4 1.3 1.1 1.1 0.9 72 87 86 81 90 96 89 84 91 80 59 86 86 0.8 86 39 Table 3: Summary statistics (N=12652) Variable Innbr bothsmall bothbig bothsame crst multichain mintenants ncities label nother pop newchain growth downtown maxage Toronto distance Mean 0.17 0.04 0.52 0.08 0.02 0.50 92.26 15.39 0.55 3.53 2.71 0.27 11.75 0.36 37.33 0.46 3.24 St. Dev 0.37 0.20 0.50 0.28 0.14 0.50 35.33 11.98 0.50 2.54 1.88 0.45 25.50 0.48 8.07 0.49 1.19 40 Table 4: Mean values of Innbrit, the number of national chains, and CR5, by store type category Category Cards/stationery Drug/health and beauty aids Photo/camera Electronics Music/instruments Books and newsstands Footwear and leather accessories Jewelry Department stores Telecommunications Variety/convenience Financial Optical Sporting goods Ladies' wear Cosmetics/toiletry Unisex clothing Menswear Furniture/home décor Gift Fabric and sewing Housewares Children's wear Travel Mean( Innbr) Number of National Chains 0.49 2 1 0.46 0.36 2 0.32 3 0.25 2 2 0.23 9 0.19 0.18 6 0.17 4 0.16 2 0.16 2 0.15 4 0.15 1 0.15 6 0.15 21 0.15 1 0.13 12 0.11 4 1 0.11 0.10 2 0.10 1 0.09 1 0.08 2 0.07 2 41 CR5 87 87 76 76 85 80 34 35 93 51 50 73 52 58 17 64 24 38 25 33 57 50 48 65 Table 5: Probit results (dependent variable Innbrit), N=126521 Variable bothsmall Coefficient (St Dev) 0.12 (0.08) bothbig 0.10*** (0.03) bothsame 0.02 (0.06) crst -1.11*** (0.18) multichain -0.06* (0.04) mintenants 0.01*** (0.0004) ncities 0.04*** (0.002) label -0.07* (0.04) nother -0.01 (0.01) pop -0.11*** (0.04) newchain -0.30*** (0.05) growth 0.002*** (0.001) downtown 0.02 (0.03) maxage 0.005** (0.002) Toronto 0.16 (0.13) distance 0.08*** (0.02) constant -2.90*** (0.13) Pseudo R2 0.19 1 Coefficients on store type dummy variables are not included. *** Indicates significance at the 1 percent level. ** Indicates significance at the 5 percent level. * Indicates significance at the 10 percent level. 42 Table 6: Effects of selected variables* Variable Effect Variable Effect bothbig 0.02 growth 0.01 mintenants 0.09 distance 0.02 ncities 0.13 newchain -0.05 pop -0.03 * For dummy variables, the reported effect is the effect on the probability that a chain will locate in both malls of switching the variable from zero to one. For other variables, the reported effect is that of increasing the variable by one standard deviation. 43 Table 7: Store type dummy variables: estimated coefficients and effects Store Type Category Drugs/health and beauty aids Photo/camera Cards/stationery Coefficient (Standard Error) 0.88 (0.12)*** Effect1 0.68 (0.10)*** 0.61(0.10)*** 0.19 0.17 0.27 Electronics 0.51 (0.08)*** 0.13 Telecommunications 0.46 (0.09)*** 0.12 Furniture/home décor 0.41 (0.12)*** 0.10 Financial 0.34 (0.10)*** 0.08 Music/instruments 0.29 (0.11)*** 0.07 Variety/convenience 0.27 (0.10)*** 0.06 Optical 0.20 (0.10)** 0.05 Footwear and leather 0.13 (0.06)** 0.03 accessories Books and newsstands 0.12 (0.13) 0.02 Fabric and sewing 0.09 (0.16) 0.02 Jewelry 0.06 (0.07) 0.01 Menswear 0.06 (0.08) 0.01 Gifts 0.05 (0.09) 0.01 Unisex clothing 0.05 (0.06) 0.01 Cosmetics/toiletry -0.001 (0.10) 0.00 Sporting goods -0.03 (0.09) -0.01 Children’s wear -0.10 (0.11) -0.02 Travel -0.13 (0.14) -0.02 Housewares -0.22 (0.11)** -0.04 Department stores -0.24 (0.10)** -0.04 1 The reported effect is the estimated difference in the probabilities that a chain will enter both malls, comparing each category to the control category of ladies’ wear. *** Indicates significance at the 1 percent level. ** Indicates significance at the 5 percent level. * Indicates significance at the 10 percent level. 44