Geographical Review of Japan Vol. 68 (Ser. B), No. 1, 86-93, 1995 Spatial Competition and Marketing Strategy of Fast Food Chains in Tokyo Kenji ISHIZAKI* Key words: spatial competition,marketing strategy, nearest-neighbor spatial associationanalysis, fast food chain, Tokyo distributions of points, while Clark and Evans I. INTRODUCTION (1954) describe a statistic utilizing only a single set of points. This measure is suitable for de One of the important tasks of retail and res scribing the store distribution of competing taurant chain expansion is the development of firms such as retail chains. a suitable marketing strategy. In particular, a The nearest neighbor spatial association location strategy needs to be carefully deter value, R, is obtained as follows: mined for such chains, where store location is often the key to the successful corporate's R=ƒÁ0/ƒÁE (1) growth (Ghosh and McLafferty, 1987; Jones and The average nearest-neighbor distance, ƒÁ0, is Simmons, 1990). A number of chain entries into given by the same market can cause strong spatial com petition. As a result of this, retail and restaurant (2) chains are anxious to develop more accessible and well-defined location strategies. In restau where dAi is the distance between point i in type rant location, various patterns are respectively A distribution and its nearest-neighbor point in presented according to some types of restau type B distribution and dBj is the distance from rants (i.e. general full service, ethnic, fast food point j of type B to its nearest neighbor point of and so on) and stores of the same type are type A. NA and NB are the numbers of points in clustered along roads or in downtown areas each of the two types, and N is the total number (Pillsbury, 1987). It is inevitable that one chain of points (hence N=NA+NB). store will compete with another chain store, The expected mean distance is given by both attempting to scramble for a "good loca tion". However, it is also possible that the segre (3) gation of store distributions appear in conse where nA and nB are the proportions of the total quence of spatial avoidance between chains. points in type A and type B, and ƒÏA and ƒÏB are The purpose of this paper is to analyze spatial the densities of points in each of the two types. competitions of the store distributions in fast R-value less than unity indicates spatial food (hamburger) chains and to examine the clustering while a value greater than unity indi relationships between spatial competition and cates spatial avoidance. location strategies. This method has been applied to the store distributions of retail chains (see Lee, 1979; Lee II. MEASURE OF SPATIAL COMPETITION and Schmidt, 1980), and the spatial patterns of banking offices (Lord and Wright, 1981), who The nearest-neighbor spatial association a suggest that this measure provides some in nalysis described by Lee (1979) is capable of sights into the competitive location strategies measuring the spatial competition between two of firms (Lord and Wright,1981, p. 192). * Graduate student , Tokyo Metropolitan University, Hachioji, Tokyo 192-03, Japan Spatial Competition of Fast Food Chains 87 neighbor analysis. III. SETTING There is, however, a problem known as "edge effects" (Boots and Getis, 1988) in that the The empirical setting is identical to that out nearest-neighbor value depends on delimitation lined in Ishizaki (1990), which examined the of the study area. Tokyo (23 wards) is bounded store distributions of four fast food chains for with physical barriers: the southeastern part of Tokyo in 1987, which consisted of 23 wards. the edge is contiguous to Tokyo Bay and the This data is also updated by using additional southwestern and the northeastern parts are information on chain store locations for 1994, limited along large rivers. There is not any obtained using a telephone directory (Yellow barriers in the northwestern part, but nearest Page) as a source. neighbor distances for those points within the This study area has experienced a rapid study area would exhibit little change even growth of fast food stores since the appearance though stores outside of the study area are of McDonald's in 1971. McDonald's and Mos taken into account, due to a higher density Burger have continuously developed to today, store distribution in the northwestern part. while the expansion of two other chains, Lotte ria and Morinaga Love, have been stable (see IV. ANALYSIS Table 1). In particular, the growth of Mos Burger has outstripped that of McDonald's The results of applications of the nearest during seven years from 1987. One reason for neighbor analysis (Clark and Evans, 1954) and this is that Mos Burger has mainly developed a the nearest-neighbor spatial association analy franchising chain operation which can grow sis are shown by Table 21) and Table 3. rapidly in a short time because less capital in Table 2 shows the clustered store distribution vestment is required. of McDonald's and Morinaga Love in 1987. Figure 1 shows the distribution of fast food Both of the nearest-neighbor statistics are sig stores in 1987 and 1994. Fast food stores have nificantly different from unity at the 0.05 level mainly been located along railroads or sub or the 0.01 level. Morinaga Love has a particu ways, especially clustering at terminal stations larly clustering tendency because of a concen or shopping cores such as Shibuya, Shinjuku tration to the south of the city center (see Fig. l ). and Ikebukuro. The concentration of stores on This is perhaps because the stores are agglo western Tokyo also implies that some chains merated around the head office of Morinaga have targeted younger customers because of confectionery, the company which established the relatively high density of population aged Morinaga Love. from 15 to 34 in western Tokyo. The nearest-neighbor spatial association It is clearly indicated that all chains have values in Table 2 indicate that there is some expanded ubiquitously in the whole study area. variance in spatial competition among the dif Therefore the study area definition and the dis ferent pairs of fast food chains. Since all near persion of selected fast food chain stores can be est-neighbor spatial association values for considered an ideal setting to apply nearest McDonald's with other chains are significantly different from unity, these spatial associations Table 1. Number of fast food stores in Tokyo are more clustered than random. This implies that McDonald's stores are actively competing with other chain stores, especially strongly competing with Lotteria. On the other hand, the store distribution of Mos Burger is relatively weakly competing with each of chains. These results suggest that the different location strat egy of each chain results in a varied store distri bution and also determines the degree of spatial The figure in parenthesis is the percentage. competition. Ishizaki (1990) recognized the ex 88 K. Ishiaaki Figure 1. Store distributions of four fast food chains in 1987 and 1994 . Spatial Competition of Fast Food Chains 89 Table 2. Nearest-neighbor statistics and nearest-neighbor spatial association values for the fast food stores in 1987 *: Significantly different from unity at the 0 .01 level. **: Significantly different from unity at the 0 .05 level. Nearest-neighbor statistics are on diagonal line. Table 3. Nearest-neighbor statistics and nearest-neighbor spatial association values for the fast food stores in 1994 *: Significantly different from unity at the 0 .01 level. ** Significantly different from unity at the 0 .05 level. Nearest-neighbor statistics are on diagonal line. istence of different location decisions in each of (see Fig. 1). These results implicitly indicates McDonald's and Mos Burger. McDonald's has that location strategy and site selection of each clustered to large shopping cores as a result of chain have changed since 1987. seeking to capture customers shopping or at work. Thus the stores have competed severely V. DISCUSSION with other chains which prefer a higher density of customers during the day. However, Mos It is often necessary for a retail chain to adopt Burger has favored the residential area with a a new strategy, when market conditions, com higher density of night-time activity and also petitors behavior and corporate environment selected sites with low-priced land values. are changed. The successful retail chain may Therefore the pattern of Mos Burger stores ex explore a number of possible strategies (Jones hibits characteristics of spatial avoidance com and Simmons, 1990, p. 388). Location strategy is pared to the patterns of other chain stores. the most important marketing strategy for Table 3 shows that the spatial patterns of fast retail chain, reflecting information from two food stores in 1994 are different from those in areas: firstly, the internal environment such as 1987. The nearest-neighbor spatial association corporate markets, objectives and capital re values represent different conditions of spatial sources and secondly, the external environ competition, while the nearest-neighbor statis ment, including demographic characteristics, tics slightly differ (see Table 2). The nearest consumer expenditure, and competition (Mer neighbor spatial association values for curio, 1984). It seems that demographic charac McDonald's against Lotteria and McDonald's teristics and consumer expenditure in the exter against Morinaga Love are somewhat greater nal environment are similar for each of fast than those in 1987. On the other hand, the food chains because of similar characteristics values between Mos Burger and McDonald's or such as taste and price. However, the distribu Lotteria indicate clustering tendencies. In fact, tion of population during the day or night McDonald's has located stores dispersedly and differs in a large city area like Tokyo.