Evaluating and Managing Brand Repurchase Across Multiple Geographic Retail Markets

Journal of Retailing 89 (4, 2013) 409–422

Evaluating and Managing Brand Repurchase Across Multiple Geographic Retail Markets Raj Echambadi a,1,2 , Rupinder P. Jindal b,1,3 , Edward A. Blair c,∗,1 a

College of Business, University of Illinois at Urbana-Champaign, Champaign, IL 61820, United States Milgard School of Business, University of Washington Tacoma, Tacoma, WA 98402-3100, United States c Department of Marketing and Entrepreneurship, C.T. Bauer College of Business, University of Houston, Houston, TX 77204-6021, United States b

Abstract Many companies manage their business on a geographic basis and evaluate marketing metrics and managers correspondingly. Here, using a multi-level dataset from the U.S. retail gasoline industry, we demonstrate inherent differences in the levels of brand repurchase across territories. Furthermore, we show that the effects of factors that may improve repurchase—customer satisfaction and customers’ relational investments—are moderated by market share at the territorial level. Relational investments have relatively more effect on repurchase in territories where a brand’s market share is higher, while customer satisfaction has relatively more effect in territories where a brand’s market share is lower. These findings imply that one size does not fit all for either evaluating or managing brand performance at a territorial level. © 2013 New York University. Published by Elsevier Inc. All rights reserved. Keywords: Brand loyalty; brand repurchase; brand performance; geographic variation; mixed models; HLM

Introduction The phenomenon of variation in brand performance across geographic retail markets has garnered much recent attention in the marketing literature. For example, Bronnenberg, Dhar, and Dubé (2007) note persistent geographic variation in market shares for leading brands of consumer packaged goods and show that the cross-market variation is generally larger than cross-time variation. Ataman, Mela, and van Heerde (2007) confirm this finding, and Kruger (2007) suggests that the phenomenon is ubiquitous across industries. Mittal, Kamakura, and Govind (2004) likewise demonstrate geographic variation using customer satisfaction as a performance metric. Although such variation in brand performance is of considerable managerial significance, relatively little is known about it, prompting calls for further research in this area (Bronnenberg et al. 2007; Lodish

∗

Corresponding author. Tel.: +1 713 743 4565; fax: +1 713 743 4572. E-mail addresses: [email protected] (R. Echambadi), [email protected] (R.P. Jindal), [email protected] (E.A. Blair). 1 All authors contributed equally in the development of this manuscript. 2 Tel.: +1 217 244 4189; fax: +1 217 244 7969. 3 Tel.: +1 253 692 5885; fax: +1 253 692 4523.

2007). Our paper answers this call by studying geographic variation in repurchase rates for brands in the retail gasoline industry. Our interest in the topic was inspired by a real-life incident that occurred in the gasoline industry. A major oil company obtained data that measured repurchase rates for its brand in each of its marketing territories. These data showed wide variation across markets, and one territory in particular was identified as a low performer. The company made several efforts to improve repurchase in this territory—investing in customer satisfaction programs, promoting ownership of its proprietary credit card, and even replacing territory managers—but the repurchase rate remained low compared with other markets. Then, in a “why didn’t we think of this sooner” moment, someone suggested that perhaps the market had some distinct characteristics, and all brands in the market had low repurchase rates. Further analysis showed that this indeed was the case, which led the company to alter its pattern of local expenditures, its evaluations of local managers, and even the strategic priority attached to the market. This real-life incident illustrates three points. First, many companies manage their business on some geographic basis and evaluate marketing metrics and managers correspondingly. Second, repurchase is a multi-level phenomenon that may be influenced not only by individual level variables such as

0022-4359/$ – see front matter © 2013 New York University. Published by Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.jretai.2013.05.005

410

R. Echambadi et al. / Journal of Retailing 89 (4, 2013) 409–422

customer satisfaction, but also by brand and market level variables. Third, the performance of any given brand—whether repurchase or any other metric—is likely to differ across territorial markets, which implies that customer receptivity to firm strategies may differ across markets. These differences across markets have an obvious managerial implication, that it may be more appropriate to evaluate performance in any given territory by comparing it with the local baseline than by comparing it with other territories. More importantly, these differences suggest that firms may be well advised not to follow a “one size fits all markets” approach in managing performance. Here we argue, and demonstrate, that this is likely to be the case. We consider two levers through which a company might attempt to improve brand repurchase rates: by improving customer satisfaction or by increasing customer relational investments through proprietary credit cards or other loyalty programs (Mägi 2003). Prior literature suggests that both of these variables will, in general, be associated with enhanced repurchase rates. However, we propose that customer satisfaction and relational investments are unlikely to have fully additive effects on repurchase, and that the effect of either variable will be reduced in the presence of the other variable. This implies that even when investing in both variables, firms may wish to choose a primary lever on which to focus. We further propose that the relative efficacy of each lever in raising repurchase for a brand in any given territory will vary with the brand’s market share in that territory, such that relational investments will have relatively more effect in territories where share is higher, while customer satisfaction will have relatively more effect in territories where share is lower. This implies that managers who wish to improve repurchase rates for a brand should place different relative emphasis on satisfaction and relational investments in different territories to the extent the brand’s market share varies across those territories. We conduct our analysis in the context of the U.S. retail gasoline industry, a multi-billion dollar industry that has received relatively little attention in the marketing literature (Ma et al. 2011). It is appropriate to study this phenomenon in the retail gasoline industry because the decisions of firms in this industry are affected by the characteristics of geographic markets (Iyer and Seetharaman 2008) and the gasoline industry considers repurchase rates as a key performance metric (Cindrin and Dolby 1998). We utilize unique data containing information for all major brands across all major geographic markets in the United States, and we model repurchase rates in a multi-level framework with predictor and control variables at the individual, brand, and market levels. We study these questions in a single industry to control for industry-specific effects (Nijssen et al. 2003). Our paper contributes to the marketing and retailing literature in three ways: (a) by demonstrating the multi-level nature of brand repurchase, it advances an understanding of differences in brand performance across geographical retail markets, (b) by exploring the interactions of key decision variables, it adds to our understanding of how to evaluate and manage brand performance when territorial operations are involved, and (c) while most studies that have documented geographic

variation have done so using consumer packaged goods, this research increases our understanding of an important but under-researched industry. The remainder of this paper is organized as follows. First, we provide a conceptual background and develop our hypotheses. Next, we describe the methods used to address those hypotheses. Third, we discuss the results. Finally, we discuss the implications, note the limitations of our research, and offer suggestions for future research. Background and hypotheses A wide variety of factors may affect variations in repurchase rates across individuals, brands and/or markets. Here, for reasons of theoretical and managerial interest, we focus on customer satisfaction and relational investments by customers as brand-related, individual-level antecedents of repurchase that companies might attempt to influence through marketing efforts, and territorial market share as a brand-related, geographic market-level factor that may moderate the relative value of influencing satisfaction and relational programs. We hypothesize that customer satisfaction and relational investments are unlikely to have fully additive effects, and the effect of either variable will be reduced in the presence of the other variable. We further hypothesize that the impact of satisfaction on repurchase will be negatively moderated by territorial market share, while the impact of relational investments on repurchase will be positively moderated by territorial market share, with both effects having managerial as well as theoretical implications. A background discussion of these variables and the rationale for our hypotheses is given below. Our analysis also will control for other individual, brand, and market-level factors that may relate to repurchase; these variables are discussed in Methods section. Effects of customer satisfaction and relational investments on repurchase Customer satisfaction has been shown to be a prominent driver of repurchase intentions across a wide range of studies (see Szymanski and Henard 2001). Scholarly work in the relationship marketing domain also has used relationship commitment as a potential driver of repurchase (Bendapudi and Berry 1997). Commitment has been expressed as an active desire on the part of the consumers to maintain an ongoing relationship (Morgan and Hunt 1994). To build commitment, firms encourage customers to make relational investments by participating in programs such as loyalty programs and proprietary credit cards. These efforts produce a variety of benefits. Relational investments create customer assets that produce higher revenues while lowering marketing costs (Voss and Voss 2008). They have been shown to affect relationship quality thereby affecting loyalty intentions (De Wulf, Odekerken-Schröder, and Iacobucci 2001). They may enable firms to provide preferential treatment and accurately reward customers for their past loyalty (Seiders et al. 2005) and

R. Echambadi et al. / Journal of Retailing 89 (4, 2013) 409–422

hence may serve to increase trust in the company (Rust, Lemon, and Zeithaml 2004). They also may create social and financial switching costs for customers (Seiders et al. 2005), thereby creating a stickiness to persist with the ongoing relationship. Hence these relationship-specific investments serve to insulate a firm from its competitors by creating incentives for the customer to stick with the brand (Fornell 1992). Overall, the individual-level literature on brand repurchase suggests that there is an element of intrinsic personal motivation required to maintain a relationship with a brand. This intrinsic motivation can derive from a variety of sources, including satisfaction with the brand. The literature also points to relationship-specific investments as elements in maintaining the stability of the relationship. Interaction between the effects of customer satisfaction and relational investments Should a firm focus equally on customer satisfaction and relational investments to improve brand repurchase rates? Or might a firm be better off by choosing a primary lever, even if it invests in both? For example, Fornell (1992) discusses the example of European and Asian airlines focusing relatively more on satisfaction whereas American airlines focus relatively more on participation in loyalty programs. A relevant issue is whether the effects of customer satisfaction and relational investments interact, and if so, the nature of that interaction. A positive interaction—what Voss, Godfrey, and Seiders (2010) describe as complementary effects—would encourage some degree of balance between efforts to build satisfaction and relational investments, to take advantage of the extent to which one variable leverages the other. A negative interaction—that is, substitute effects—might encourage a relative focus on one variable. In considering this issue, there are reasons to expect a negative interaction between customer satisfaction and relational investments. van Doorn and Verhoef (2008) note that the link between satisfaction and loyalty often appears weak or even absent in business markets because “ongoing customer–supplier relationships tend to be characterized by inertia that causes parties to conduct “business as usual” and, in essence, maintain the status quo” (p. 123). The argument, in essence, is that inertial forces will dilute (or even moot) the effects of satisfaction on repurchase. This same argument applies with respect to relational investments by consumers. Such investments create an inertial force that should weaken the effects of satisfaction on repurchase. This argument is supported by Gustafsson, Johnson, and Roos (2005, p. 211) who suggest that various sources of commitment keep customer loyal to a brand even when satisfaction may be low. Hence, we hypothesize: H1. Relational investments negatively affect the customer satisfaction–repurchase relationship such that the impact of customer satisfaction on repurchase is weakened at higher levels of relational investments. A negative interaction implies that even if a firm invests in both customer satisfaction and relational investments to drive

411

repurchase rates, it may wish to choose a primary lever. This leads us to ask: do the effects of customer satisfaction and relational investments on repurchase rates vary across territorial markets, and if so, under what conditions might each variable be relatively more potent in driving repurchase? We suggest that there will indeed be territorial differences, and a brand’s market share within any given territory will play a role. To develop our hypotheses, we first consider the direct effect of territorial market share on brand repurchase, and then argue for its moderating role. Effects of territorial market share on repurchase Customer behavior is influenced by exposure to the firm’s marketing programs at the geographic retail market level (Jayachandran and Varadarajan 1999). Hence, any explanation of geographic variation in repurchase rates should account for brand-related variables that operate at a market level (i.e., are common for all individuals in a particular geographic market). One such variable is territorial market share. Evidence from both the behavioral loyalty literature (e.g., Colombo and Morrison 1989; Raj 1985) and the double jeopardy literature (e.g., Ehrenberg, Goodhardt, and Barwise 1990) point to the role of market share as a brand-level variable associated with brand repurchase rates. From a theoretical perspective, higher territorial market share brands should have higher repurchase rates because of both customer and firm factors (Liu and Yang 2009). On the customer side, higher share in the relevant market serves as a quality cue (Caminal and Vives 1996), engenders higher brand trust (Chaudhuri and Holbrook 2001), and may generate bandwagon effects (Rindfleisch and Inman 1998). On the firm side, higher share is associated with higher distribution levels (Fader and Schmittlein 1993), allows higher advertising expenditures (Bhattacharya 1997), and enables more support for loyalty programs (Liu and Yang 2009). Our specific focus is on territorial market share, defined as the market share of a brand within a geographic territory, and not national market share. We focus on territorial share for two reasons. First, of course, our concern is with geographic variation in repurchase rates, and territorial share is relevant to geographic variation. Second, territorial market share is appropriate because geographic retail markets often vary in their cultural, demographic, economic, and regulatory conditions (cf. Hawkins, Roupe, and Coney 1981; Rosenzweig and Singh 1991; Sorenson and Audia 2000), leading to a spatial association of consumption behaviors (Mittal et al. 2004; Ter Hofstede, Wedel, and Steenkamp 2002). Evidence for these geographic-specific effects comes from the “micro-marketing” literature that has demonstrated that variables such as price elasticity vary across stores within a retail chain because of demographic and competitive conditions in specific trading areas (e.g., Dhar and Hoch 1997; Hoch et al. 1995; Montgomery 1997). Similarly, retailing literature suggests that retail outlets may perform differently due to the demographic characteristics of their local markets (Chan, Padmanabhan, and Seetharaman 2007). Therefore, localized conditions within the specific geographic retail markets are likely to impact repurchasing behaviors.

412

R. Echambadi et al. / Journal of Retailing 89 (4, 2013) 409–422

Hypothesized cross-level moderating effects of territorial market share We now turn to the cross-level interactions between territorial market share (a market-level variable) and customer satisfaction and relational investments (at the individual level). We propose that the effects of satisfaction and relational investments on repurchase rates will be moderated by territorial market share, with the effect of satisfaction being negatively moderated and the effect of relational investments being positively moderated. First, let us consider the interaction of territorial market share and customer satisfaction. Territorial market share should operate as an inertial force in driving repurchase. As already noted, higher share brands in a geographic market are likely to have higher distribution in that market (Fader and Schmittlein 1993), which makes it more convenient for customers to repurchase the brand. This is particularly true in a retailing context, where market share tends to be directly associated with the number of store locations (Michael 2003). Brands with higher territorial market share are also likely to have higher levels of local advertising (Bhattacharya 1997), which reminds customers to repurchase the brand. Again, this is particularly true in a retailing context, where more store locations equates to higher brand visibility. Higher availability and visibility, along with factors such as implied quality (Caminal and Vives 1996) and bandwagon effects (Rindfleisch and Inman 1998), should facilitate buying inertia. If territorial market share operates as an inertial force in driving repurchase, its moderating effect on the satisfaction–repurchase relationship should be similar to the moderating effect of relationship investments stated in H1. That is, the inertial effects of territorial market share should weaken the effects of satisfaction on repurchase. Therefore we hypothesize: H2. Territorial market share negatively affects the customer satisfaction–repurchase relationship such that the impact of customer satisfaction on repurchase is weakened at higher levels of territorial market share. The interaction of territorial market share and relational investments, on the other hand, represents the interaction of two inertial forces. The interaction of such forces may be negative or positive, depending on whether the forces are substitutes or complements. In this case, we argue that market share and relational investments will act as complements. The fact that higher share brands may be more widely available and hence more convenient to buy should enhance the inertial impact of relational investments. For example, in the gasoline industry, possession of a proprietary credit card should encourage repurchase of a brand, and wider availability of that brand’s gas stations should enhance the effect. On a related point, although multiple firms can offer proprietary cards or loyalty programs, customers tend to focus their purchases on one program (Sharp and Sharp 1997), and higher-share brands in a market are likely to win this battle because of higher availability, higher visibility, and the ability to provide more support for relational programs (Liu and Yang 2009). As a result, higher

territorial market share should leverage the effects of relational investments in engendering repeat purchase. We thus hypothesize: H3. Territorial market share positively affects the relational investments–repurchase relationship such that the impact of relational investments on repurchase is strengthened at higher levels of territorial market share. To the best of our knowledge, this is the first time it has been argued in the literature that the relative influence of customer satisfaction and relational investments on repurchase rates will be moderated by geographic retail market factors. If the hypotheses hold, they will have important managerial implications: that one size does not fit all when it comes to managing brand performance across geographic markets, and more specifically, whatever a particular company’s overall mixture may be of seeking to improve brand repurchase through customer satisfaction versus relational investments (which presumably depends on the overall cost effectiveness of these levers), the company would be well advised to allocate relatively more effort to customer satisfaction in territories where it holds a lower market share and relatively more effort to relationship building efforts in territories where it holds a higher share. Method We test our hypotheses with multi-level—individual-, brand-, and geographic market-level—cross-sectional data from the retail gasoline industry. The hierarchical structure of data allows us to estimate a mixed model to account for any effects of markets and brands on brand repurchases of customers. Data We obtained data from two main sources. Data on industryrelated variables such as repurchase, satisfaction, proprietary credit card usage (which we used to operationalize relational investments) and territorial market share came from a leading supplier of data to the retail gasoline industry. The company collected data from a household panel that was balanced to be representative of the entire United States. Our base dataset for industry-related variables has 67,436 individuals purchasing one or more of 18 brands in 226 metropolitan statistical areas, or MSAs in 1998. Data include all major national and regional brands in the country at the time, and all the major population centers in the country (covering 98 percent of the population in MSAs and 80 percent of the total U.S. population). There are 1,249 unique market-brand pairs in the dataset. The average MSA has 5 brands, with a range of 2–10, and the average brand is present in 69 MSAs, with a range of 15–139. Below, we will describe how the relevant variables were operationalized in our base dataset. In addition to these industry data, we obtained data on market-level demographics from the U.S. Census conducted in 2000. The retail gasoline industry provides an apt setting to address the research questions raised in this study because it is a large and highly competitive industry that manages brand marketing

R. Echambadi et al. / Journal of Retailing 89 (4, 2013) 409–422

at a territorial level. Gasoline, the primary product in the industry, is frequently purchased but largely fungible, which makes every purchase occasion an opportunity for competitors to steal a customer. Thus, gasoline companies are very concerned with repurchase rates (Cindrin and Dolby 1998). Decisions in this industry are also affected by characteristics of the geographic markets (Iyer and Seetharaman 2008), which makes it an ideal setting to study geographic variation in repurchase. Description of variables Our base dataset collected data on an individual’s most recent gasoline purchase, and measured tendency to repurchase with a three-level multinomial measure that asks whether the individual always buys the same brand, buys two or three different brands, or buys many different brands. This is the primary measure of brand repurchase, our dependent variable, used by companies in the gasoline industry (as is also true for other measures used in this study). For example, in our introductory example, where a major oil company attempted to improve repurchase rates in a problem territory, this is the repurchase measure the company was tracking. Among independent variables, customer satisfaction is measured using a single-item scale from 1 to 6: overall satisfaction with the last purchase of gasoline. As an indicator of customer relational investments, we used an item that captures whether customers paid for the purchase of gas using a proprietary credit card. Taking and using such cards signals some level of commitment to the brand. Also, purchasing gasoline with such credit cards often provides customers with some sort of incentives to buy that particular brand of gasoline and switching to another brand implies losing out on such rewards. We measured territorial market share of a brand within each geographical market (MSA) as the percentage of customers who reported purchasing that brand out of the total number of customers reporting for the market. These share estimates are part of the regular syndicated data sold by this marketing research company to its clients, and are used by these companies as measures of market share within MSAs. We compared these territorial market share estimates with numbers from another nationwide dataset obtained from a leading gasoline retailer, and found no significant difference based on a chi-square goodness-of-fit test. As another cross-check, we collected data on the number of gas stations operated by each brand in seven MSAs, and found that the market share measure used in our study correlates well with share of stations with a coefficient of 0.82. Control variables We controlled for several other variables at all three levels—individual, brand, and market—that could influence brand repurchase. We controlled for four demographic characteristics at the individual level: age, gender, income, and household size. Age, income, and household size are continuous variables while gender is a binary variable. We also controlled for any effect of inherent customer preferences for brand, price, or locational convenience on brand

413

repurchase. Using binary items, we controlled for whether customers indicated that their purchase had been influenced by brand preference (brand shopper), low price (price shopper), or convenient location of the gas station (locational convenience shopper). Furthermore, we also controlled for any influence of gasoline grade on brand repurchase. Because out of the three grades of gasoline—87, 89, and 93—grade 87 had relatively more customers, it is taken as the base category during model estimation. We controlled for a brand-level positioning variable, qualityorientation of the brand. The retail gasoline market is divided between premium (quality-oriented) and non-premium (priceoriented) brands (Iyer and Seetharaman 2003). Premium brands typically attempt to differentiate themselves on some nonprice attribute whereas non-premium brands generally compete on price. We asked two independent raters, both senior managers in the market research company with long experience in the gasoline industry, to classify each of the 18 brands as quality-oriented or price-oriented and obtained a 100 percent inter-coder agreement with 7 quality-oriented and 11 priceoriented brands. We controlled for the main effect of this variable on repurchase to allow for the possibility that baseline repurchase rates differ between quality- and price-oriented brands. Also, because previous research has shown that quality perceptions of a brand may influence the effectiveness of its marketing efforts (Sivakumar and Raj 1997), we controlled for the interaction of quality-orientation with each of our focal action levers, customer satisfaction and relational investments.4 Finally, we controlled for the possible influence of MSAlevel demographic variables on brand repurchase. As per Mittal et al. (2004), we consider market size (population in thousands), per capita income, per capital vehicle ownership, annualized population growth, average commute time, proportion of male population, proportion of urban population, and proportions of black and Hispanic populations. All of these variables were obtained from U.S. Census data. Descriptive statistics for all variables appear in Tables 1 and 2. Model and estimation We have data at three levels: geographic market (MSA), brand, and individual. The data are structured such that each geographic market has a certain number of respondents purchasing one or more of the brands available in the market. The specific data structure is illustrated in Fig. 1. For such multi-level nested data, hierarchical linear modeling accounts for the lack of independence across different observations and overcomes the limitations of traditional methods (Raudenbush and Bryk 2002). However, because each market in this study has multiple brands available and each brand is available in multiple markets, markets and brands themselves are not nested. Any nation-wide marketing decision implemented by a brand affects customers in all geographic markets

4 We thank an anonymous reviewer for the suggestion to consider these interactions.

414

R. Echambadi et al. / Journal of Retailing 89 (4, 2013) 409–422

Table 1 Means and standard deviations. Variable

N

Mean

Customer-level variables Age (years) 67,436 Female 67,436 Income (000 dollars) 67,436 Household size 67,436 Grade 93 gasoline 67,436 Grade 89 gasoline 67,436 Brand shopper 67,436 Price shopper 67,436 Locational convenience shopper 67,436 Customer satisfaction 67,436 Relational investments 67,436 Brand-level variables Quality-orientation 18 Market-level variables Market size (population in 000) 226 Annual population growth 226 Per capita income (000 dollars) 226 Per capita vehicle ownership 226 Average commuting time (min) 226 Proportion of male population 226 Proportion of urban population 226 Proportion of black population 226 Proportion of Hispanic population 226 Variable cross-classified across markets and brands Territorial market share 1,249

SD

45.52 .59 43.73 2.65 .20 .19 .30 .45 .79 4.91 .26

15.78 .49 28.14 1.39 .40 .39 .46 .50 .41 .73 .44

.39

.50

965.60 .01 20.14 .42 24.24 .49 .79 .11 .09

2,148.03 .01 3.07 .04 3.32 .01 .12 .10 .14

18.08

14.07

while a market’s characteristics affect all brands available there. Furthermore, our dependent variable is ordinal in nature. Hence, we estimated a cross-classified random intercept ordered logit model using the GLIMMIX procedure in SAS. We meancentered continuous variables such as per capita income, per capita vehicle ownership, individual age, individual income, household size, market size, customer satisfaction, and territorial market share. The purpose of this mean centering was simply to characterize any interaction effects at the mean levels of the respective variables (Echambadi and Hess 2007). Let Prob(Ymijk = 1) = φmijk

(1)

ordinal, let ηmijk be the log-odds of a customer being in category m, that is,   φmijk ηmijk = log (2) 1 − φmijk The level 1 equation is specified as follows: ηmijk = π0jk + π1jk (age)ijk + π2jk (female)ijk + π3jk (income)ijk + π4jk (household size)ijk + π5jk (grade 93 gasoline)ijk + π6jk (grade 89 gasoline)ijk + π7jk (brand shopper)ijk + π8jk (price shopper)ijk +π9jk (locational convenience shopper)ijk +π10jk (satisfaction)ijk +π11jk (relational investments)ijk +π12jk (satisfaction × relational investments)ijk +m λm Level 2 equations are specified as follows: π0jk = θ0 + β01 (quality orientation)k + γ01 (market size)j +γ02 (annual population growth)j +γ03 (per capita income)j +γ04 (per capita vehicle ownership)j +γ05 (average commuting time)j +γ06 (proportion of male population)j +γ07 (proportion of urban population)j +γ08 (proportion of black population)j +γ09 (proportion of Hispanic population)j +δ01 (territorial market share)jk + b00j + c00k + d0jk (4)

be the probability that customer i in market j purchasing brand k belongs to category m. Because level-1 sampling model is π10jk = θ10 + δ10,1 (territorial market share)jk

Market 1

Brand 1

Customer 111

+δ10,2 (quality orientation)k

Brand 2

Customer 211

Customer 321

Customer 421

Brand 2

Brand 3

Customer 622

Customer 732

Fig. 1. Data structure.

Customer 832

(5)

π11jk = θ11 + δ11,1 (territorial market share)jk +δ11,2 (quality orientation)k

Market 2

Customer 522

(3)

(6)

where πpjk = level-1 coefficients; θ p = model intercept, the expected value of πpjk when all explanatory variables are set to zero; βs = fixed effects of brand-level variables; γs = fixed effects of market-level variables; δs = fixed effects of market-level by brand-level cell-specific variables; λm = ordinal category cutoff differences; b00j = residual random effect of market j;

Table 2 Correlation matrix.

20 21 22 23

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

Brand repurchase Age Female Income Household size Grade 93 gasoline Grade 89 gasoline Price shopper Brand shopper Locational convenience shopper Customer satisfaction Relational investments Territorial market share Market size Annual population growth Per capita income Per capita vehicle ownership Avg commuting time Prop male population Prop urban population Prop black population Prop Hispanic population Quality orientation

1.00 −0.11 −0.06 0.08 0.05 −0.11 −0.04 0.06 −0.27 0.10 −0.19 −0.13 −0.07 −0.01 0.02 0.02 0.02 0.00 0.00 −0.02 0.01 −0.02 −0.07

1.00 −0.11 −0.08 −0.29 −0.06 −0.03 0.03 0.01 −0.04 0.14 0.02 −0.04 0.02 −0.02 0.01 −0.04 0.02 −0.03 0.03 −0.03 0.02 −0.07

1.00 −0.02 0.11 −0.02 0.03 −0.06 −0.03 0.05 0.06 0.00 0.01 0.00 −0.01 −0.01 0.00 0.00 −0.01 −0.01 0.01 −0.01 0.03

1.00 0.28 0.05 0.00 −0.03 0.01 0.05 −0.02 0.10 −0.06 0.18 0.02 0.18 −0.12 0.21 0.00 0.15 0.07 0.09 0.05

1.00 −0.01 0.01 0.02 −0.05 0.01 −0.03 −0.02 0.00 0.03 −0.02 0.00 −0.03 0.02 0.01 0.02 0.01 0.02 −0.01

1.00 −0.24 −0.10 0.12 −0.07 0.05 0.04 0.00 0.06 0.02 0.03 −0.06 0.07 −0.03 0.04 0.06 0.03 0.11

1.00 −0.07 0.05 0.01 0.00 0.03 0.00 0.00 −0.02 −0.01 0.00 0.01 −0.02 −0.01 0.03 0.00 0.04

1.00 0.00 0.06 0.04 −0.15 −0.05 0.03 −0.02 0.00 −0.03 0.02 0.01 0.02 −0.01 0.03 −0.26

1.00 −0.05 0.14 0.10 0.07 0.01 0.00 0.00 −0.01 0.02 −0.02 0.01 0.04 0.00 0.17

1.00 −0.01 −0.04 −0.02 −0.02 0.00 0.00 0.02 −0.02 −0.01 −0.02 −0.01 −0.02 −0.03

1.00 0.05 0.00 −0.03 −0.02 −0.03 0.02 −0.02 −0.04 −0.04 0.03 −0.04 0.03

1.00 −0.01 0.02 0.08 −0.01 −0.01 0.04 0.06 0.05 0.02 0.10 0.15

1.00 −0.14 −0.08 −0.14 0.00 −0.24 −0.01 −0.08 −0.05 0.00 0.13

1.00 0.07 0.40 −0.21 0.63 −0.02 0.38 0.12 0.17 −0.02

1.00 0.22 0.00 0.26 0.31 0.36 −0.04 0.38 −0.22

1.00 0.40 0.35 −0.03 0.40 −0.01 −0.23 0.18

1.00 −0.32 0.05 −0.11 0.00 −0.52 0.28

1.00 −0.11 0.30 0.35 0.18 0.05

1.00 0.14 −0.25 0.20 −0.30

Variable

20

21

22

23

Prop urban population Prop black population Prop Hispanic population Quality orientation

1.00 −0.01 0.41 −0.14

1.00 −0.20 0.22

1.00 −0.28

1.00

R. Echambadi et al. / Journal of Retailing 89 (4, 2013) 409–422

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

Variable

Notes: N = 67,436 for evaluating pairwise correlations between individual-level variables or between individual- and higher-level variables; correlations greater than .008 (absolute value) are significant at the .05 level. N = 1,249 for evaluating pairwise correlations between territorial market share and brand- and market-level variables; correlations greater than .06 (absolute value) are significant at the .05 level. N = 226 for evaluating pairwise correlations between market-level and brand-level variables; correlations greater than .13 (absolute value) are significant at the .05 level.

415

416

R. Echambadi et al. / Journal of Retailing 89 (4, 2013) 409–422

c00k = residual random effect of brand k; d0jk = residual random effect of market-level by brand-level cell jk. Of particular interest are π12jk , δ10,1 and δ11,1 which relate to our hypotheses. π12jk indicates the interaction between customer satisfaction and relational investments (H1). δ10,1 and δ11,1 are the cross-level interaction terms that indicate the extent to which territorial market share moderates the effects of customer satisfaction and relational investments, respectively (H2 and H3). Similarly, δ10,2 and δ11,2 are cross-level interaction terms that indicate the extent to which the quality orientation (control) variable moderates the effects of customer satisfaction and relational investments. We built the model incrementally to balance theory and model parsimony (Kreft and de Leeuw 1998). We evaluated incremental nested models by comparing their Akaike information criteria (AIC) with a chi-square distribution in which the degrees of freedom equal the difference in the number of parameters between the two models. Results were as follows. Results Variation in repurchase rates across geographic markets To begin our analysis, we specified a null model with just an intercept and no random effects. Next, we added random intercept effects at the geographic market level (the b00j term), to allow for variation in baseline repurchase rates across markets. The comparison of these two nested models indicates that adding random intercept effects at the market level significantly improved model fit (AIC(1) = 410,561, p < .01). In other words, there is significant variation in repurchase rates across markets. Out of 226 markets in the dataset, customers in 22 markets were likely to be significantly more brand loyal than customers in a typical market while customers in 21 markets were likely to be significantly less brand loyal. The focal brand in the example with which we opened this paper was found to have a significantly lower rate of repurchase when compared with the brand’s other territories but an average repurchase rate when compared to the local baseline. This finding has an important managerial implication. If there are inherent differences in repurchase rates across territorial markets, for whatever reason, it is more appropriate to evaluate a brand’s performance in any given territory by comparing it to the local baseline than by comparing it with its performance in other territories. To do otherwise is to invite frustration among local managers who know that they are not competing on a level playing field. Tests of hypotheses Moving from the model containing random intercept effects for geographic markets, which we henceforth label as Model 1, we next added (a) the control variables, (b) the main effects of customer satisfaction, relational investments, and territorial market share, and (c) random intercept terms at the brand and market-by-brand levels (the c00k and d0jk terms); in other words, everything except the three hypothesized interaction effects.

We label this as Model 2. These additions produced significant improvement in model fit (AIC(30) = 26,594, p < .01). Finally, we added the hypothesized interaction effects to create the full model, Model 3. Adding these terms significantly improved the fit (AIC(3) = 563, p < .01) indicating that the interaction effects are jointly significant. The results from Models 1, 2 and 3 appear in Table 3. In constructing these models, we followed Mittal and Kamakura’s (2001) approach to specifying the effects of customer satisfaction to accommodate possible non-linearity. Previous research has shown non-linear effects for satisfaction, and published literature suggests that the observed interaction effects may be spurious if non-linear effects of focal variables are not accommodated (Ganzach 1997). For simple effects, we coded the 6-point satisfaction scale as a dummy variable with the category “extremely dissatisfied” (scale value = 1) set as the base. The rest of the categories were coded relative to that base such that coefficients for the other categories represent changes in log odds relative to the base category. Thus, relative differences between successive pairs of coefficients can be compared to see the nonlinear effect of successive changes in satisfaction rating. To calculate interaction terms, each customer’s satisfaction level was multiplied with the value of the interacting variable—relational investments or territorial market share—for that particular customer. Our discussion will focus on Model 3, the full model. This model shows positive simple effects for all three of our focal variables: customer satisfaction, relational investments (i.e., using a proprietary credit card), and territorial market share. With respect to customer satisfaction, there are no significant effects across levels 2 through 4 (relative to level 1 or each other, based on tests of differences between successive coefficients); however, an increase in satisfaction from level 4 to level 5 increases the log odds of a customer being in a higher category of brand repurchase by .47 (.342 − (−.130)) while an increase from level 5 to level 6 increases the log odds by .52 (.857 − .342). With respect to relational investments, for customers using a proprietary card, the expected log odds of being in a higher category of brand repurchase are .54 more than customers not using a proprietary card. With respect to territorial market share, an increase in territorial market share by one percent point increases the log odds of being in a higher category by .007. Given that the standard deviation for territorial market share is 14.07, an increase of one standard deviation increases the log odds of being in a higher category of brand repurchase by .10. However, the focus of our hypotheses is not on the simple effects of these variables, but on their interactions. Looking at the interaction terms, results support all three hypotheses. Consistent with Hypothesis 1, there is a negative interaction between the impact of customer satisfaction and relational investments on repurchase rates (β = −.177, p < .01). Consistent with Hypothesis 2, the impact of customer satisfaction on repurchase rates is negatively moderated by territorial market share (β = −.002, p < .05). Consistent with Hypothesis 3, the impact of relational investments is positively moderated by territorial market share (β = .006, p < .01).

R. Echambadi et al. / Journal of Retailing 89 (4, 2013) 409–422

417

Table 3 Results from cross-classified random intercept ordered logit model. Model 1 Coeff. 1. Hypothesized interactions Customer satisfaction × Relational investments Customer satisfaction × Territorial mkt. share Relational investments × Territorial mkt. share 2. Basic main effects Customer satisfaction (=2) Customer satisfaction (=3) Customer satisfaction (=4) Customer satisfaction (=5) Customer satisfaction (=6) Relational investments Territorial market share 3. Control variables 3.1. Individual-level control variable Age Female Income Household size Grade 93 gasoline Grade 89 gasoline Brand shopper Price shopper Locational convenience shopper 3.2. Brand-level control variables Quality-orientation Quality orientation × Customer satisfaction Quality orientation × Relational investments 3.3. Market-level control variables Market size Annual population growth Per capita income Per capita vehicle ownership Average commuting time Proportion of male population Proportion of urban population Proportion of black population Proportion of Hispanic population Intercept 1 Intercept 2 AIC AIC Difference

Model 2 SE

Coeff.

−531,877

.018 .022

SE

Coeff.

SE

−.177*** −.002** .006***

.026 .001 .002

−.234 −.348** −.214 .233 .717*** .511*** .009***

.205 .174 .165 .168 .175 .038 .001

−.173 −.284 −.130 .342** .857*** .539*** .007***

.206 .176 .166 .170 .177 .038 .001

.014*** .315*** −.007*** .015** .477*** .241*** 1.032*** −.203*** −.356***

.001 .016 <.000 .006 .021 .021 .019 .017 .020

.014*** .316*** −.007*** .015** .477*** .241*** 1.033*** −.203*** −.353***

.001 .016 <.000 .006 .021 .021 .019 .017 .020

.018 .014 −.037

−.414*** 2.504***

Model 3

<.000 −7.213*** −.017* .853 .023** −.614 .803*** −.805*** −.098 −1.953 1.302 −558,471 26,594***

.086 .023 .044

.024 .041* −.065

.086 .023 .044

<.000 2.087 .009 .644 .009 2.250 .207 .209 .197 1.158 1.158

<.000 −7.218*** −.019** .912 .023*** −.553 .810*** −.817*** −.111 −2.121* 1.140

<.000 2.082 .009 .642 .009 2.246 .207 .208 .196 1.157 1.157

−559,034 563***

All two-tailed tests. * p < .10. ** p < .05. *** p < .01.

To get a sense of the substantive impact of these effects, we compared the probability of a customer being in the highest repurchase category for two hypothetical brands: one with territorial market share 1 standard deviation higher than the average and the other with territorial market share 1 standard deviation lower than the average. Per convention, we assume a customer at base value (i.e., zero) for all categorical variables and mean value for all continuous variables. With respect to customer satisfaction, if territorial market share is one standard deviation above the average, the probability that this customer would fall in the highest repurchase category is 19 percent at satisfaction level 4 (which does not

differ significantly from levels 1 to 3), rising to 26 percent at satisfaction level 5, and to 37 percent at satisfaction level 6 – an overall difference of 18 percent. If territorial market share is one standard deviation below the average, the rise would be from 17 percent to 26 percent to 39 percent, an overall difference of 22 percent. These results are shown in Fig. 2. The effect of satisfaction is substantial in either case, and, as hypothesized, more so for the lower share brand. With respect to relational investments, if territorial market share is one standard deviation above the average, the probability that this customer would fall in the highest repurchase category is 22 percent if the customer did not use a proprietary credit

418

R. Echambadi et al. / Journal of Retailing 89 (4, 2013) 409–422

not fit all when it comes to managing brand performance across territories. In more specific terms, they suggest that whatever a particular company’s overall mixture may be of seeking to improve brand repurchase through customer satisfaction versus relational investments, the company should allocate relatively more effort to customer satisfaction in territories where it holds a lower market share and relatively more effort to relational investments in territories where it holds a higher share. Other results

Fig. 2. Effect of customer satisfaction on brand repurchase of a high share brand versus a low share brand.

card, rising to 35 percent if they did, a difference of 13 percent. If territorial market share is one standard deviation below the average, the rise would be from 19 percent to 27 percent, a difference of 8 percent. These results are shown in Fig. 3. The effect of relational investments is substantial in either case, and, as hypothesized, more so for the higher share brand. The commercial significance of these differences can be seen as follows. We obtained supplementary data containing 10,603 observations on individuals from 21 MSAs who, within six months of providing the information in the present dataset, had participated in a separate study that captured repurchase rates by asking participants to allocate their next ten purchases across brands. We compared this measure of repurchase intent with our categorical measure, and found that individuals in our highest category had an average repurchase allocation of .76—that is, on average, they expected to buy that brand 7.6 times out of their next ten purchases—compared with .47 and .34 in our middle and bottom categories, respectively. The implication is that moving a customer into the highest repurchase category can be worth 29 percent or more of that customer’s gasoline purchases, which can represent hundreds of dollars per annum (according to U.S. Bureau of Labor Statistics 2011, the average household had 1.9 vehicles and spent $2,132 on gasoline and motor oil in 2010, or more than $1,000 per vehicle). Overall, therefore, the interaction effects found in Model 3 have important managerial implications. In broad terms, they suggest that there are attenuating returns to using both customer satisfaction and relational investments, and that one size does

Table 3 shows two intercepts. Intercept 1 indicates the odds of being in category 1 versus being in categories 2 or 3 and Intercept 2 indicates the odds of being in categories 1 or 2 versus being in category 3. These intercepts can be used to calculate the predicted probabilities for a customer being in a particular category. For example, for a customer with base values of all categorical independent variables and average values of all continuous independent variables, the probability of being in category 1 is 0.20 while being in category 1 or 2 is 0.92. Other results in the model are as follows. Regarding individual-level control variables, the results show that older customers and females are more likely to repurchase a brand in this industry. Results also indicate that as the disposable income of customers increases, they are less likely to repurchase a brand but as their household size increases they are more likely to repurchase a brand. Results indicate that the higher the grade of gasoline purchased, the higher is the repurchase rate. Brand shoppers are more likely to repurchase while price shoppers and locational convenience shoppers are less likely to repurchase. These results seem generally plausible. Regarding brand-level controls, although quality orientation of a brand does not have a direct effect, results show that increasing satisfaction improves repurchase of quality-oriented brands more than of price-oriented brands. Regarding market-level controls, customers living in primarily urban markets and markets with longer average commutes are more likely to repurchase while customers in high per capita income markets and high population growth markets are less likely to repurchase a brand. Customers in markets with relatively higher proportion of black consumers are less likely to repurchase a brand. These other results are not of great theoretical importance, but they potentially would have practical value for managers in the industry. More importantly, they provide evidence that brand repurchase is a multi-level phenomenon that is influenced by variables at the individual, brand and geographic market levels. Robustness checks

Fig. 3. Effect of relational investments on brand repurchase of a high share brand versus a low share brand.

We conducted several tests to check for robustness of results. First, we estimated a cross-classified random intercept multinomial logit model assuming repurchase categories as nominal. This model produces a separate coefficient for each category but the implication of results validated our findings. Next, we simply dichotomized the dependent variable into “always buys the same brand” versus “others,” and estimated a binomial logit

R. Echambadi et al. / Journal of Retailing 89 (4, 2013) 409–422

mixed model. We then estimated a linear mixed model assuming the three levels of brand repurchase as a 3-2-1 interval measure. Finally, we took the average repurchase allocation values observed for each of our repurchase categories in the supplementary dataset described above (.76, .47, and .34), and estimated a linear mixed model with these values in place of the 3-2-1 measure. In all cases, the results validated our findings in that the same factors lead to increased repurchase and the effects have the same pattern of direction and significance. We also tested for potential endogeneity in the relationship between territorial market share and repurchase rates. There is ample theoretical and empirical precedent for treating territorial market share as antecedent to repurchase rates, as in our model (e.g., Colombo and Morrison 1989; Ehrenberg et al. 1990; Raj 1985). Still, we conducted a Granger causality test to address potential endogeneity (cf. Chintagunta and Haldar 1998). We used another supplementary dataset with 26,087 observations from 43 metropolitan areas collected a year after our base dataset, containing the same measures as our base data but with different individual respondents. This supplementary dataset allowed us to calculate repurchase rates and territorial market shares for 353 unique brand-geographic market combinations. We analyzed these figures along with the corresponding values in our base dataset. The Durbin-Watson statistic showed no significant autocorrelation. Results of the analysis showed that whereas lagged values of territorial market share Grangercause repurchase rates (F(1, 351) = 28.43; p < .05), lagged values of repurchase rates do not Granger-cause territorial market share (F(1, 351) = 1.66; ns). Likewise, we tested for potential endogeneity in the relationships between (a) customer satisfaction and repurchase rates and (b) relational investments and repurchase rates. Because the supplementary dataset comes from different respondents, we cannot conduct Granger tests for customer satisfaction and relational investments at the individual level. However, we conducted the tests at the aggregate brand-geographic market level by averaging the measures of variables at this level. Results at the aggregate level indicate that whereas lagged values of customer satisfaction Granger-cause repurchase rates (F(1, 351) = 3.37; p < .10), lagged values of repurchase rates do not Granger-cause customer satisfaction (F(1, 351) = 1.46; ns). Similarly, lagged values of relational investments Granger-cause repurchase rates (F(1, 351) = 13.64; p < .01) but lagged values of repurchase rates do not Granger-cause relational investments (F(1, 351) = 2.53; ns).

Discussion In this paper, we explain the variation in brand repurchase across geographic retail markets. Many companies manage their business on some geographic basis—whether called districts, territories, markets, regions, or zones—and evaluate marketing metrics and managers correspondingly. We contribute to the emerging literature on explaining variations in brand performance across geographic markets, and our findings have important theoretical and managerial implications in this regard.

419

Our findings confirm that brand repurchase is a multi-level phenomenon that is influenced not only by individual level variables such as customer satisfaction and relational investments, but also by brand and market level variables such as territorial market share and demographic characteristics. While the specific variables may be different, these findings are broadly consistent with prior research showing differences in marketing variables across geographic or micro markets (e.g., Bronnenberg et al. 2007; Chan et al. 2007; Dhar and Hoch 1997; Hoch et al. 1995; Iyer and Seetharaman 2003; Mittal et al. 2004; Montgomery 1997). Given that repurchase is influenced by brand and market level factors, any given brand is likely to experience inherent differences in repurchase across territorial markets, and our findings show that this does appear to be the case, at least within the retail gasoline industry. Such differences imply that it may be more appropriate to evaluate a brand’s performance in any given territory by comparing it with the local baseline than by comparing with its performance in other territories. As noted earlier, to do otherwise is to invite frustration among local managers who know that they are not competing on a level playing field. Turning to the question of how to manage repurchase across territories, our findings have two implications. First, we find a negative interaction between the effects of customer satisfaction and relational investments on repurchase rates. This result is broadly consistent with van Doorn and Verhoef (2008) and Gustafsson et al. (2005), and has important managerial implications. Enhancing customer satisfaction and developing commitment through relational investments may entail very different strategies. To improve customer satisfaction, a firm needs to improve the quality of the product or service solutions. To enhance commitment, a firm can either invest in direct relationships with customers or build switching costs vis-à-vis its competitors (Gustafsson et al. 2005). Therefore, if a firm invests in both customer satisfaction and relational investments to drive repurchase rates, it may wish to choose a primary variable on which to focus, with the overall allocation presumably depending on the cost effectiveness of each level. Second, we find that territorial market share moderates the relative influence of customer satisfaction and relational investments on brand repurchase, such that as territorial market share rises, the effect of customer satisfaction becomes smaller and the effect of relational investments becomes larger. In other words, territorial market share and customer satisfaction appear to be substitutable effects whereas territorial market share and relational investments have a complementary relationship. Apart from the managerial importance of these interactions, they are theoretically interesting because they suggest that market share (a) functions as an inertial force on repurchase and (b) is a complement to relational investments. In broad terms, these results imply that just as one size does not fit all when it comes to evaluating brand performance across territories, one size does not fit all when it comes to managing brand performance. In more specific terms, it provides insights for allocating resources towards various company initiatives. As noted earlier, a brand with territorial market share one standard deviation above the mean would improve the probability of a

420

R. Echambadi et al. / Journal of Retailing 89 (4, 2013) 409–422

base case customer falling into the highest repurchase category by 18 percent (from 19 percent to 37 percent) by increasing customer satisfaction from a level of 4 to 6 on the 6 point scale used here. This compares with 13 percent gain (from 22 percent to 35 percent) by inducing use of a proprietary credit card. A brand with territorial market share one standard deviation below the mean would experience corresponding gains of 22 percent and 8 percent. The higher numbers for satisfaction do not necessarily indicate that firms should focus resources on customer satisfaction versus relational investments; that decision will also depend on any given firm’s starting values for satisfaction and card usage and the relative cost of building those numbers. However, whatever a particular company’s overall mixture may be of seeking to improve brand repurchase through customer satisfaction versus relational investments, the comparative effect sizes imply that the company should allocate relatively more effort and resources to customer satisfaction in territories where it holds a lower market share and relatively more effort and resources to simultaneously building relational investments in territories where it holds a higher share. While our focus has been on intra-brand variation across territories, the results also have implications for inter-brand variation. What if a brand holds a systematically high or low market share? Presumably, just as a company should allocate relatively more resources to relational investments in territories where it holds a higher share, a brand that holds systematically high share should place relatively more emphasis on relational investments than a brand that holds systematically low share. Likewise, a brand that holds systematically low share should place relatively more emphasis on customer satisfaction. These points may be reinforced by the fact that low share brands often have a more homogeneous customer base, allowing them to suit specific customer tastes, while the greater heterogeneity that accompanies a larger customer base can make it difficult for high share brands to maintain an extremely high level of customer satisfaction (Fornell 1992, p. 9). Also, a high share brand is better able to cover the fixed costs of relational programs such as loyalty programs or proprietary credit cards. Limitations Although this research contributes to our understanding of brand repurchase and how to evaluate and manage it when territorial operations are involved, the empirical analysis has some limitations. First is the issue of generalizability. Gasoline is a strong-satiation product category in which consumers buy in response to need and derive little marginal utility from additional consumption. Voss et al. (2010) suggest that such categories are highly conducive to routinized buying behaviors. Also, gasoline retailing features characteristics such as a fungible product, frequent purchases, and brand exclusivity at the station (store) level. In product categories where demand is more discretionary and routinization of purchasing less likely, the negative interaction between the effects of customer satisfaction and relational investments on likelihood of repurchase may be weaker. Also, at an individual level, this interaction may depend on the source of

dissatisfaction and the nature of commitment. Relational investments may be associated with what Ganesan et al. (2010) label as calculative commitment, affective commitment, or both. In cases where dissatisfaction stems from perceived opportunism or a betrayal of trust by the brand, calculative commitment may buffer the effects of dissatisfaction on repurchase, but affective commitment may exacerbate them (cf. Ganesan et al. 2010). We would expect the moderating effects of territorial market share to generalize in terms of valence, but not necessarily in terms of size. For example, given the importance of spatial convenience in gasoline retailing, territorial market share may be relatively more important than in other industries. Furthermore, whereas in gasoline retailing the development of relational investments primarily takes the form of proprietary credit cards with some use of continuity promotions, in other marketing contexts, it may manifest more in the form of reward programs or communications (De Wulf et al. 2001), and this may lead to changes in effect size. For example, the complementary interaction between the effects of territorial market share and relational investments on likelihood of repurchase might be even stronger in the airline industry with its strong reward programs. A second limitation pertains to some of our measures. Some key measures were single-item self-reports by respondents. However, as pointed out by Mittal and Kamakura (2001), most commercial surveys use single-item measures. Also, our dependent repurchase measure is not clearly prospective (vs. retrospective), which raises questions about causal sequencing. However, supplementary data allowed us to test for endogeneity at the brand-geographic level, and those tests indicate that satisfaction, relational investments, and territorial market share cause repurchase, not vice versa. A third issue pertains to our definition of territories. Our results are obtained at an MSA level of analysis. In territories defined at a lower level such as ZIP codes, demographic variables might play an even larger role. It also should be noted that the retail gasoline industry has undergone changes since the time at which these data were collected, with mergers, acquisitions, and shifting market shares. We have no reason to believe that these changes would affect the hypothesized pattern of effects, which are primarily driven by theory—for example, the idea that relational investments are likely to be more effective in markets where a brand holds higher market share—but specific effect sizes may well be different. Future research This research suggests several avenues for further study. One promising area for future research is further examination of market-level drivers of the variation in repurchase across geographic markets. What causes some markets to have higher repurchase rates than other markets? We have some findings in this regard, but they should be viewed as a first step. A related area for future research would be to conduct longitudinal analyses of such variation in repurchase across geographic markets. Some authors contend that behavioral loyalty might evolve over time (East and Hammond 1996). It would be

R. Echambadi et al. / Journal of Retailing 89 (4, 2013) 409–422

interesting to explore if markets vary in repurchase over time, and if so, what factors contribute to such changes. Also, of course, researchers would do well to replicate this study in different settings, considering possible differences between brands of retail stores and brands of products, or between strong and weak satiation categories. Such replication will provide useful information about the generalizability of our findings. References Ataman, M. Berk, Carl F. Mela and Harald J. van Heerde (2007), “Consumer Packaged Goods in France: National Brands, Regional Chains, and Local Branding,” Journal of Marketing Research, 44 (1), 14–20. Bendapudi, Neeli and Leonard Berry (1997), “Customer’s Motivations for Motivating Relationships,” Journal of Retailing, 73 (1), 15–37. Bhattacharya, C.B. (1997), ““Is Your Brand’s Loyalty Too Much, Too Little, or Just Right?: Explaining Deviations in Loyalty from the Dirichlet Norm,” International Journal of Research in Marketing, 14 (5), 421–35. Bronnenberg, Bart, Sanjay K. Dhar and Jean-Pierre H. Dubé (2007), “Consumer Packaged Goods in the United States: National Brands, Local Branding,” Journal of Marketing Research, 44 (1), 4–13. Bureau of Labor Statistics (2011), “Economic News Release,” (accessed June 22, 2012), [available at http://www.bls.gov/news.release/cesan.nr0.htm]. Caminal, Ramon and Xavier Vives (1996), “Why Market Shares Matter: An Information-Based Theory,” Rand Journal of Economics, 27 (Summer), 221–39. Chan, Tat, V. Padmanabhan and P.B. Seetharaman (2007), “An Econometric Model of Location and Pricing in the Gasoline Market,” Journal of Marketing Research, 44 (4), 622–35. Chaudhuri, Arjun and Morris B. Holbrook (2001), “The Chain of Effects from Brand Trust and Brand Affect to Brand Performance: The Role of Brand Loyalty,” Journal of Marketing, 65 (2), 81–93. Chintagunta, Pradeep and Sudeep Haldar (1998), “Investigating Purchase Timing Behavior in Two Related Product Categories,” Journal of Marketing Research, 35 (1), 43–53. Cindrin, Nick and Petrina Dolby (1998), “Store Wars,” Oil & Gas Investor, II, Quarter, 21–5. Colombo, Richard A. and Donald G. Morrison (1989), “A Brand Switching Model with Implications for Marketing Strategies,” Marketing Science, 8 (1), 89–99. De Wulf, Kristof, Gaby Odekerken-Schröder and Dawn Iacobucci (2001), “Investments in Consumer Relationships: A Cross-Country and CrossIndustry Exploration,” Journal of Marketing, 65 (4), 33–50. Dhar, Sanjay K. and Stephen J. Hoch (1997), “Why Store Brand Penetration Varies by Retailer,” Marketing Science, 16 (3), 208–27. East, Robert and Kathy Hammond (1996), “The Erosion of Repeat-Purchase Loyalty,” Marketing Letters, 7 (2), 163–71. Echambadi, Raj and James D. Hess (2007), “Mean-Centering Does Not Alleviate Collinearity Problems in Moderated Multiple Regression Models,” Marketing Science, 26 (3), 438–45. Ehrenberg, Andrew S.C., Gerald J. Goodhardt and T. Patrick Barwise (1990), “Double Jeopardy Revisited,” Journal of Marketing, 54 (3), 82–91. Fader, Peter S. and David C. Schmittlein (1993), “Excess Behavioral Loyalty for High-Share Brands: Deviations from the Dirichlet Model for Repeat Purchasing,” Journal of Marketing Research, 30 (4), 478–93. Fornell, Claes (1992), “A National Customer Satisfaction Barometer: The Swedish Experience,” Journal of Marketing, 56 (1), 6–21. Ganesan, Shankar, Steven P. Brown, Babu L. John and Dixon H. Ho (2010), “Buffering and Amplifying Effects of Commitment in Business-to-Business Relationships,” Journal of Marketing Research, 47 (2), 361–73. Ganzach, Yoav (1997), “Misleading Interaction and Curvilinear Terms,” Psychological Methods, 2 (3), 235–47. Gustafsson, Anders, Michael D. Johnson and Inger Roos (2005), “The Effects of Customer Satisfaction, Relationship Commitment Dimensions, and Triggers on Customer Retention,” Journal of Marketing, 69 (4), 210–8.

421

Hawkins, Del I., Don Roupe and Kenneth A. Coney (1981), “The Influence of Geographic Subcultures in the United States,” In Advances in Consumer Research, Vol. 8, Monroe Kent ed. Provo, UT: Association for Consumer Research, 713–7. Hoch, Stephen J., Byung-Do Kim, Alan Montgomery and Peter Rossi (1995), “Determinants of Store-level Price Elasticity,” Journal of Marketing Research, 32 (1), 17–29. Iyer, Ganesh and P.B. Seetharaman (2003), “To Price Discriminate or Not: Product Choice and the Selection Bias Problem,” Quantitative Marketing and Economics, 1 (2), 155–78. and (2008), “Too Close to be Similar: Product and Price Competition in Retail Gasoline Markets,” Quantitative Marketing and Economics, 6 (3), 205–34. Jayachandran, Satish and Rajan Varadarajan (1999), “Marketing Strategy: An Assessment of the State of the Field and Outlook,” Journal of the Academy of Marketing Science, 27 (2), 120–43. Kreft, Ita and Jan de Leeuw (1998), Introducing Multilevel Modeling, Thousand Oaks, CA: Sage Publications. Kruger, Michael W. (2007), “How Geographic Variation Persists: Comments on “Consumer Packaged Goods in the United States: National Brands, Local Branding,” Journal of Marketing Research, 44 (1), 21–2. Liu, Yuping and Rong Yang (2009), “Competing Loyalty Programs: Impact of Market Saturation, Market Share, and Category Expandability,” Journal of Marketing, 73 (1), 93–107. Lodish, Leonard M. (2007), “Another Reason Academics and Practitioners Should Communicate More,” Journal of Marketing Research, 44 (1), 21–2. Ma, Yu, Kusum L. Ailawadi, Dinesh K. Gauri and Dhruv Grewal (2011), “An Empirical Investigation of the Impact of Gasoline Prices on Grocery Shopping Behavior,” Journal of Marketing, 75 (2), 18–35. Mägi, Anne W. (2003), “Share of Wallet in Retailing: The Effects of Customer Satisfaction, Loyalty Cards and Shopper Characteristics,” Journal of Retailing, 79, 97–106. Michael, Steven C. (2003), “First Mover Advantage through Franchising,” Journal of Business Venturing, 18, 61–80. Mittal, Vikas and Wagner A. Kamakura (2001), “Satisfaction, Repurchase Intent, and Repurchase Behavior: Investigating the Moderating Effect of Customer Characteristics,” Journal of Marketing, 38 (February), 131–42. Mittal, Vikas, Wagner A. Kamakura and Rahul Govind (2004), “Geographic Patterns in Customer Service and Satisfaction: An Empirical Investigation,” Journal of Marketing, 68 (3), 48–62. Montgomery, Alan L. (1997), “Creating Micro-marketing Pricing Strategies Using Supermarket Scanner Data,” Marketing Science, 16 (4), 315–37. Morgan, Robert M. and Shelby D. Hunt (1994), “The Commitment-Trust Theory of Relationship Marketing,” Journal of Marketing, 58 (3), 20–39. Nijssen, Edwin, Jagdip Singh, Deepak Sirdeshmukh and Hartmut Holzmuller (2003), “Investigating Industry Context Effects in Consumer-Firm Relationships: Preliminary Results From a Dispositional Approach,” Journal of the Academy of Marketing Science, 31 (1), 46–60. Raj, S.P. (1985), “Striking a Balance Between Brand Loyalty and Brand Popularity,” Journal of Marketing, 49 (Winter), 53–9. Raudenbush, Stephen W. and Anthony W. Bryk (2002), Hierarchical Linear Models: Applications and Data Analysis Methods, 2nd ed. Thousand Oaks, CA: Sage Publications. Rindfleisch, Aric and Jeffrey Inman (1998), “Explaining the Familiarity-Liking Relationship: Mere exposure, Information Accessibility, or Social Desirability,” Marketing Letters, 9 (1), 5–19. Rosenzweig, Philip M. and Jitendra V. Singh (1991), “Organizational Environments and the Multinational Enterprise,” Academy of Management Review, 16 (2), 340–61. Rust, Roland T., Katherine N. Lemon and Valarie A. Zeithaml (2004), “Return on Marketing: Using Customer Equity to Focus Marketing Strategy,” Journal of Marketing, 68 (1), 109–27. Seiders, Kathleen, Glenn B. Voss, Dhruv Grewal and Andrea L. Godfrey (2005), “Do Satisfied Customers Buy More? Examining Moderating Influences in a Retailing Context,” Journal of Marketing, 69 (4), 26–43.

422

R. Echambadi et al. / Journal of Retailing 89 (4, 2013) 409–422

Sharp, Byron and Anne Sharp (1997), “Loyalty Programs and Their Impact on Repeat-Purchase Loyalty Patterns,” International Journal of Research in Marketing, 14 (5), 473–86. Sivakumar, K. and S.P. Raj (1997), “Quality-Tier Competition: How Price Changes Influence Brand Choice and Category Choice,” Journal of Marketing, 61 (3), 71–84. Sorenson, Olav and P.G. Audia (2000), “The Social Structure of Entrepreneurial Activity: Geographic Concentration of Footwear Production in the United States, 1940–1989,” American Journal of Sociology, 106, 424–62. Szymanski, David M. and David H. Henard (2001), “Customer Satisfaction: A Meta-Analysis of the Empirical Evidence,” Journal of the Academy of Marketing Science, 29 (1), 16–35.

Ter Hofstede, Frenkel, Michel Wedel and Jan-Benedict E.M. Steenkamp (2002), “Identifying Spatial Segments in International Markets,” Marketing Science, 21 (2), 160–77. van Doorn, Jenny and Peter Verhoef (2008), “Critical Incidents and the Impact of Satisfaction on Customer Share,” Journal of Marketing, 72 (2), 123–42. Voss, Glenn B., Andrea L. Godfrey and Kathleen Seiders (2010), “How Complementarity and Substitution Alter the Customer Satisfaction-Repurchase Link,” Journal of Marketing, 74 (6), 111–27. Voss, Glenn B. and Zannie G. Voss (2008), “Competitive Density and the Customer Acquisition-Retention Tradeoff,” Journal of Marketing, 72 (4), 3–18.