BUSINESS MANAGEMENT A+ WORK, ON TIME, NO PLAGARIZING; ON TIME

Journal of Marketing Research Vol. XLIX (December 2012), 750–772

© 2012, American Marketing Association ISSN: 0022-2437 (print), 1547-7193 (electronic) 750

PAUL B. ELLICKSON, SANJOG MISRA, and HARIKESH S. NAIR∗

The authors measure the revenue and cost implications to supermarkets of changing their price positioning strategy in oligopolistic downstream retail markets. Their approach formally incorporates the dynamics induced by the repositioning in a model with strategic interaction. They exploit a unique data set containing the price format decisions of all U.S. supermarkets in the 1990s. The data contain the format change decisions of supermarkets in response to a large shock to their local market positions: the entry of Wal-Mart. The authors exploit the responses of retailers to Wal-Mart entry to infer the cost of changing pricing formats using a revealed-preference argument. The interaction between retailers and Wal-Mart in each market is modeled as a dynamic game. The authors find evidence that entry by Wal-Mart had a significant impact on the costs and incidence of switching pricing strategy. Their results add to the marketing literature on the organization of retail markets and have implications for long-term market structure in the supermarket industry. Their approach, which incorporates long-term dynamic consequences, strategic interaction, and sunk investment costs, may be used to empirically model firms’ positioning decisions in marketing more generally.

Keywords: positioning, dynamic games, EDLP, PROMO, pricing, supermarkets, Wal-Mart

Repositioning Dynamics and Pricing Strategy

Large changes to some or all of a firm’s marketing appa- ratuses are referred to as “repositioning.” While common in product markets and extensively discussed in manage- ment textbooks (e.g., Ries and Trout 1981), empirical anal- ysis of repositioning decisions in the academic literature has remained scarce. Perhaps the most visible forms of repositioning are brand related. Recent examples include Domino’s Pizza’s attempt to change its reputation from fast

*Paul B. Ellickson is Associate Professor of Economics and of Mar- keting, Simon School of Business, University of Rochester (e-mail: [email protected]). Sanjog Misra is Professor of Mar- keting, Anderson School of Management, University of California, Los Angeles (e-mail: [email protected]). Harikesh S. Nair is Associate Professor of Marketing, Stanford Graduate School of Business, Stanford University (e-mail: [email protected]). The authors thank the review team, Robert Miller, and Ariel Pakes, as well as seminar participants at Carnegie-Mellon, University of California, Los Angeles, and the 2011 Marketing-Industrial Organization, SICS, and Marketing Science conferences for many helpful comments. The usual disclaimer applies. Ariel Pakes served as guest editor and Steven Ryan served as guest area editor for this article.

delivery to high quality and the repositioning of UPS from shipping to full office solutions. Other examples include adjustments to product lines, such as Hyundai’s recent move into the luxury auto segment in the United States and Kodak’s long-delayed transition to digital imaging. While brand-related changes are common, they are far from the only examples. Apple’s inclusion of third-party retailers can be thought of as repositioning its downstream distribution strategy, and Procter & Gamble’s adoption of value-based pricing in 1992 to reduce trade promotions was a repo- sitioning of the firm’s overall pricing strategy (Ailawadi, Lehmann, and Neslin 2001). St-James (2001) describes several instances of repositioning by firms in U.S. con- sumer product markets, including a detailed history from the 1920s of attempts by Sears Roebuck and Co., Mont- gomery Wards, and JCPenney chain stores to periodically reposition themselves in response to changing consumer tastes and competition.

Repositioning is different from new entry because it is inherently history dependent: Repositioning typically requires incurring costs to undo past product-related deci- sions. Therefore, repositioning costs to incumbents can often be much larger than the cost of entry to new

©

Repositioning Dynamics and Pricing Strategy 751

firms. Repositioning frequently involves complex invest- ments needed to overcome within-firm managerial resis- tance to change, to rework channel relationships, and to educate (and advertise to) consumers about the new positioning—all investments that are large and sunk. The magnitude of repositioning costs has substantive impli- cations for competition and market structure. Low re- positioning costs help constrain market power by enabling competitors to react more quickly to changes in a firm’s product lines and product attributes. When repositioning investments are sunk, they also have commitment value (Dixit and Pindyck 1994), implying that current reposition- ing decisions can affect long-term market outcomes such as the future entry and exit behavior of rival firms. Thus, mea- suring repositioning costs is important to understanding the economic underpinnings of market structure in an industry.

In this article, we examine the repositioning of the pric- ing strategies of U.S. supermarket firms. Our empirical goals are to measure the revenue and cost implications to supermarkets of changing their store-level pricing formats. These pricing formats are broadly split between EDLP (everyday low price) and PROMO (or promotional) strate- gies.1 Whereas EDLP stores charge a low regular price per product with little temporal price variation, PROMO stores are characterized by higher regular prices, punctu- ated by frequent price promotions or sales. A store’s choice between EDLP and PROMO is motivated by both demand- and cost-side considerations. On the demand side, choosing PROMO over EDLP offers an opportunity for supermarkets to intertemporally price discriminate, by using price cycles to sell differentially to consumers of varying price infor- mation, loyalty, stockpiling costs, and valuations (Bell and Hilber 2006; Lal and Rao 1997; Pesendorfer 2002; Salop and Stiglitz 1982; Sobel 1984; Varian 1980). Furthermore, the frequent price variation under PROMO may create an option value to consumers of visiting the store more fre- quently by reducing their average basket size per trip (Bell and Lattin 1998; Ho, Tang, and Bell 1998). On the cost side, EDLP enables retailers to reduce inventory costs, to better coordinate supply chains, and to reduce stock-out risk by smoothing the demand variability induced by fre- quent sales. The choice of EDLP or PROMO is an impor- tant strategic choice retailers face that affects their price image and has significant long-term implications for prof- itability and local market structure (Ellickson and Misra 2008; Lal and Rao 1997).

Several factors may cause a supermarket chain to change its store pricing format in a local market. One first-order factor is response to competitive entry, especially by rivals with a comparative advantage in a particular strategy. Our data, which cover a census of supermarket entry, exit, rev- enue, and format choices in the 1990s, include a period of intense readjustment in response to a one such event: the introduction of Wal-Mart Supercenters, which exclu- sively follow an EDLP strategy. The entries of Wal-Mart Supercenters serve as large shocks to the competitive struc- ture of local markets, inducing a large number of format switches and a host of exits. Our identification of reposi- tioning costs exploits these switches heavily and rests on a revealed preference argument similar to that of Bresnahan

1PROMO is also referred to as “HiLo.”

and Reiss (1991, 1994): If we observe that a firm switched its price positioning, we conclude that the profits (in a present-discounted sense) from the switch were higher than those without it. As we observe revenues, we can further decompose this restriction on profits into a restriction on the costs of the change. Combining this with a structural model of the industry and variation across markets enables us to relate these restrictions to specific market and com- petitive factors. Identification derives from observing both switches to a different price positioning and exits. Intu- itively, if repositioning costs were zero, a firm whose rev- enues do not cover fixed costs under its current pricing strategy could costlessly shift to an alternative pricing strat- egy with higher net payoffs. Observing other firms stay in the market under the alternative pricing strategy reveals that the net payoffs for this alternative are positive. Observ- ing, at the same time, that the focal firm is exiting and not switching thus indicates that the repositioning costs of switching to the alternative strategy are large.

Additional identification derives from the joint distribu- tion of entry, stays, switches, exits, and revenues observed across markets. The level of present-discounted revenues required to make N incumbents switch to a pricing strat- egy compared with that required to make N entrants choose that strategy identifies the extent to which the reposition- ing costs are higher than entry costs. The level of present- discounted revenues required to make N entrants choose a pricing strategy compared with that required to make N incumbents who currently operate under that strategy to exit identifies the extent to which the repositioning costs are sunk.

We then combine this identification strategy with a model of format choice to decompose the effect of repo- sitioning into revenue and cost components. This decom- position is informative to the underlying choice problem. For example, on the revenue side, a move from PROMO to EDLP implies a loss to the supermarket in its price discrim- ination ability, as discussed previously, as well as demand losses from potential consumer antagonism to changes in pricing policies (Anderson and Simester 2010). On the cost side, many of the costs associated with advertising the new positioning, the employee hours involved in updating inven- tory and supply-chain systems for changing pricing strat- egy, and purchase of new pricing and demand-management software to manage promotional activity are sunk. Because the demand-side effects are observed in the long run and the cost-side investments are sunk, these cost–benefit trade- offs involve dynamic considerations. Furthermore, strategic interaction from other supermarkets are likely important. Most retail markets in the United States tend to be concen- trated, with a few (three to five) dominant players control- ling the market, irrespective of its size (Ellickson 2007), implying firms face oligopolistic competition at the local market level.

Accommodating these key considerations, we utilize an empirical framework that treats format change as a dynamic problem with sunk investment. Strategic interactions are accommodated by formulating the model as a dynamic game of incomplete information with entry and exit, in the spirit of Ericson and Pakes (1995). Using the identification strategy presented previously, we propose new ways to infer the structural parameters of the game, exploiting recently developed methods for two-step estimation of dynamic

752 JOURNAL OF MARKETING RESEARCH, DECEMBER 2012

games (Aguirregabiria and Mira 2007; Arcidiacono and Miller 2012). In addition, we demonstrate how to incor- porate revenue information (a continuous outcome) into the estimation procedure in an internally consistent man- ner, while accounting for the dynamic selection induced by the codetermination of these with the discrete choices, by extending the methods that Ellickson and Misra (2012) propose to a dynamic environment. The incorporation of strategic interaction is important to the estimation of repo- sitioning costs. For example, in a competitive market, a supermarket may be reluctant to switch from PROMO to EDLP because it anticipates that price competition may be toughened if a rival firm, currently using PROMO, shifts to EDLP in response to its action. In the absence of this control, the persistence induced on pricing strategy (and observed in the data) by such strategic interaction would be falsely interpreted as repositioning costs. This is the main additional complication that arises when measuring switching costs for firms rather than consumers. This is accommodated in our framework by allowing firms to form beliefs about the reactions of their rivals, which then influ- ence their choice of pricing formats. In our Markov perfect equilibrium (MPE), beliefs and actions are consistent and will be functions of the state variables the firm faces. We are thus able to recover the beliefs of the firms directly from the data for use in estimation, by semiparametrically projecting the observed actions of the firms onto the rele- vant state vector. Our results imply that the cost and revenue effects of

changing pricing formats are large and asymmetric. In par- ticular, for the median store in our data, a change from EDLP to PROMO requires a fixed outlay of approximately $2.3 million borne over a four-year horizon. In contrast, a switch from PROMO to EDLP requires outlays approx- imately six times as large, providing a clear explanation for why EDLP was never uniformly adopted: It is simply too expensive to be viable in most markets. We also find evidence for significant heterogeneity in these costs across markets, holding out scope for geographic segmentation in a given chain’s price positioning strategies. Consistent with existing research (cited in the following section), we find overwhelming evidence that PROMO produces higher revenues. For the median store market, PROMO yields an incremental revenue of approximately $6.2 million annu- ally compared with EDLP. We also find that the entry of Wal-Mart has large and significant effects on the propen- sity to switch pricing formats. It also has a dispropor- tionately asymmetric effect on supermarket revenues, with its entry hurting revenues of EDLP stores approximately twice as much as it does PROMO stores (reducing revenues by $1.47 million compared with $.69 million annually at the median).

Substantively, empirical evidence on the relative attrac- tiveness of EDLP versus PROMO strategies is scarce. In a study of one retailer, Mulhern and Leone (1990) report that sales increased in a switch from EDLP to PROMO. In the strongest evidence available thus far, randomized pricing experiments involving the Dominick’s stores con- ducted by the University of Chicago (Hoch, Drèze, and Purk 1994) find that consumers do not prefer category- by-category EDLP over PROMO (revenues declined when categories, but not stores, were switched from PROMO to EDLP). The literature still lacks a clear accounting of

how these trade-offs change when the long-term economic costs of switching are incorporated. In our data, we find that a switch from EDLP to PROMO increases revenues as well as the probability of store exits, suggesting that format change cost considerations are qualitatively important to an audit of price-positioning strategies.

Our approach is closest in structure to Sweeting (2011), who estimates the dynamic costs radio stations face when changing music formats. Substantively, the question we ask is different in that there is no role for consumer pricing in radio (because radio music is free); furthermore, com- pared with this model, we accommodate new entry and allow incumbent firms to exit, which drives part of the identification. In our model, the margin from staying in the market versus exiting identifies the per-period fixed costs of operation; in contrast, the margin from changing a format, conditional on staying in the market, identifies format-switching costs. Our article is also broadly related to an empirical literature that has documented descrip- tively the effects of Wal-Mart entry on incumbent firms (e.g., Basker and Noel 2009; Ellickson and Grieco 2013; Matsa 2011; Singh, Hansen, and Blattberg 2006), to a recent empirical marketing literature applying static dis- crete games to entry models of supermarket supply (Orhun 2012; Vitorino 2012; Zhu and Singh 2009), and to an ambi- tious recent structural literature that has modeled the entry decisions of Wal-Mart as dynamic (but abstracting from strategic interactions; Holmes 2011) or as jointly deter- mined across geographies (but abstracting from dynam- ics, as in Ellickson, Houghton, and Timmins 2010; Jia 2008). Our focus on measuring dynamic switching costs for firms complements the recent marketing literature that has considered dynamics induced by consumer-side switch- ing costs for demand (Goettler and Clay 2011; Hartmann and Viard 2008) and for firm’s pricing decisions (Dubé, Hitsch, and Rossi 2009). Our empirical exercise can also be thought of as measuring an adjustment cost of changing pricing strategy and is broadly related to the empirical lit- erature measuring the costs to retailers of changing prices (e.g., Levy et al. 1997; Slade 1998).

More generally, we emphasize that repositioning is fun- damentally a dynamic decision both because of the sunk nature of repositioning investments and because current repositioning decisions affect future demand and compet- itive reactions. Thus, repositioning decisions in marketing and industrial organization should formally be thought of as dynamic games. Here, we illustrate how viewing prod- uct markets through this lens enables us to parsimoniously accommodate these dynamic considerations and to struc- turally estimate the benefits and costs of repositioning in real market settings.

We organize the remainder of the article as follows: The next section provides background on the supermarket industry as it appeared in the late 1990s, describes the data set, establishes key stylized facts, and details our approach to identification. Then, we introduce our formal model of retail competition, followed by an outline of our empiri- cal strategy and econometric assumptions. The next section contains our main empirical results, along with a discussion of their broader implications. We conclude with limitations and suggestions for further research.

Repositioning Dynamics and Pricing Strategy 753

INDUSTRY, DATA, AND STYLIZED FACTS

Our data relate to the 1990s, a period of significant change for the U.S. supermarket industry. Conventional supermarket chains faced intense competition from the rise of new store formats and innovative entrants, including club stores (e.g., Sam’s Club, Costco) and limited assort- ment chains (e.g., Aldi, Save-A-Lot). At the forefront was Wal-Mart, which built its first Supercenter (a combina- tion discount store and grocery outlet) in 1988, opened its 200th outlet in 1995, and would operate more than 1000 supercenters by 2001. The basis of the competitive threat from entry lay in the perception that limited service, thin- ner assortments, and EDLP created enormous cost savings and increased credibility with consumers. Together with a limited product assortment, EDLP offered the promise of more predictable demand, reduced inventory and car- rying costs, fewer advertising expenses, and lower menu and labor costs. Larger scale was thought to go hand in hand with lower prices. Much of this perception was driven by the success of Wal-Mart alone, which leveraged tech- nical sophistication in information technology with buying power to squeeze suppliers and tighten margins, staking out a dominant position in the retailing sector and forg- ing an indelible perception as a low-cost leader. Many of the strategic decisions made by the incumbent supermar- ket chains were geared toward competing with Wal-Mart. (A more detailed discussion of reports in the trade press regarding supermarkets’ response to Wal-Mart entry and their choice of EDLP or PROMO in response to that entry is available on request from the authors.)

Although the impact of Wal-Mart on retail competition is undisputed, many observers assumed that the EDLP format would also come to dominate the supermarket landscape, ignoring both the significant sunk investments in reposition- ing necessary to implement it and the offsetting benefits of having frequent promotions. While Wal-Mart has continued its growth in the supermarket industry, it is now understood that the EDLP revolution did not come to pass. Our empir- ical analysis is aimed at understanding why. To do so, we decompose the returns to adopting the EDLP or PROMO format into three components: revenues, operating costs, and repositioning costs. We find that while EDLP pricing provides significant cost savings, it is expensive to imple- ment (i.e., the repositioning costs are significant). More- over, it leads to a significant reduction in revenues relative to PROMO pricing.

Data and Descriptive Results

We now describe our data set and present some key stylized facts in the data that pin down switching costs. We drew the data for the supermarket industry from two primary sources: the Trade Dimensions TDLinx panel database and the 1994 and 1998 frames of the Supermar- kets Plus Database. Trade Dimensions continuously col- lects store-level data from every supermarket operating in the United States for use in its Marketing Guidebook and Market Scope publications, as well as selected issues of Progressive Grocer magazine. The data are also sold to consulting firms and food manufacturers for marketing pur- poses. The supermarket category is defined using the gov- ernment and industry standard: a store selling a full line of food products and generating at least $2 million in yearly revenues. Food stores with less than $2 million in revenues

are classified as convenience stores and are not included in the data set. For the TDLinx panel, Trade Dimensions collects information on average weekly volume, store size, number of checkouts, and several additional store- and chain-level characteristics by surveying store managers and cross-validating their responses with each store’s princi- pal food broker. We use the 1994, 1998, and 2002 frames from this panel. The TDLinx data set does not contain information on pricing format. We obtained the informa- tion on pricing strategy from a second data set, the Super- markets Plus Database, which was only collected in 1994 and 1998, and contains a more detailed set of character- istics. In particular, managers were asked to choose the pricing strategy that was closest to what their store prac- tices on a general basis: EDLP, PROMO, or hybrid. The database defines EDLP as having little reliance on pro- motional pricing strategies such as temporary price cuts. Prices are consistently low across the board, throughout all food departments. The database defined PROMO as the heavy use of specials—usually through manufacturer price breaks or special deals. The hybrid category was included for stores that practiced a combination of the two, presum- ably across separate categories or departments. Because we are interested in the adoption of a pure EDLP positioning, we include hybrid stores in the PROMO category (for addi- tional information on the data set, including a verification of its correlation with actual price variation using indepen- dent scanner data, see Ellickson and Misra 2008).

Markets and market structure. Although there are sev- eral retail channels through which to purchase food for at-home consumption (e.g., supermarkets, mom-and-pop grocers, specialty markets, convenience stores, club stores), we focus on the supermarket channel exclusively, further narrowing our focus to chain supermarkets operating within 276 designated U.S. Metropolitan Statistical Areas (MSAs). Following Ellickson and Misra (2008), who establish that strategic pricing decisions have a strong local component, our unit of observation is a store operating in a local mar- ket, taken here to be a zip code.2 Because we are pri- marily interested in understanding repositioning choices, which only applies to supermarket firms (as opposed to Wal-Mart), the following summary statistics and descrip- tive analysis focus exclusively on this set of firms. Any exceptions are noted explicitly.

Table 1 provides statistics that describe the local markets and firms. Focusing on the first frame of the table, we note that the average market contains approximately 22,000 con- sumers, and the full set ranges in size from unpopulated (i.e., zoned to be purely commercial) to 112,000. There is also substantial variation in both ethnic composition and income levels across markets. Frames two and three in Table 1 sum- marize market structure in the two periods for which we

2A potential concern is the degree to which price decisions are made locally. This issue is discussed in detail in Ellickson and Misra (2008), who document the rich degree of local variation in pricing strategies cho- sen by individual chains. Although several chains maintain a consistent focus (e.g., Food Lion, Winn-Dixie), many choose a diverse mixture of pricing formats. Consistent with our store-level decision model, this diver- sity extends to the repositioning choice. Of the 1145 stores that were part of a chain that switched the pricing format of 3 or more stores, 838 (73%) were owned by chains that did not uniformly switch to a particular focus (i.e., EDLP or PROMO). Notably, the firms that did move in a uniform direction were much smaller on average than those that did not.

754 JOURNAL OF MARKETING RESEARCH, DECEMBER 2012

Table 1 DEMOGRAPHICS AND MARKET STRUCTURE

M SD Range

Market Demographics Population (in 1000s) 22 1502 [0, 112] Per capita income (in $1,000s) 3309 1208 [0, 135] Median rent (in $s) 48707 16302 [0, 1001] Share urban 078 033 [0, 1] Share Hispanic 0075 0143 [0, .979] Share black 0101 0179 [0, .995]

Market Structure (1994) All stores 2058 1086 [1, 16] Chain stores 2008 1078 [0, 14] Share EDLP 0282 0367 [0, 1] Wal-Mart in local market 0002 0024 [0, 1] Wal-Mart in MSA 0100 0300 [0, 1]

Market Structure (1998) All stores 2057 1082 [1, 14] Chain stores 2002 1074 [0, 13] Share EDLP 0281 0371 [0, 1] Wal-Mart in local market 0009 0061 [0, 1] Wal-Mart in MSA 0466 0499 [0, 1]

have pricing data. While the average market contains just over 2 stores, some contain as many as 16. Approximately 28% of stores in the average market choose EDLP, and the remaining 72% offer PROMO. The typical number of stores and the fraction choosing EDLP are both relatively stable over time. The biggest change observed in the data is the number of markets that either contain a Wal-Mart or face one in their surrounding MSA. Both numbers increased by a factor of five over this four-year period, reflecting the dra- matic rollout of the supercenter format that occurred at this time. (The number of supercenters increased from 97 to 487 between 1994 and 1998.)

Table 2 provides summary statistics for all chain super- markets (i.e., excluding Wal-Mart) operating in 1994 and

Table 2 STORE-LEVEL CHARACTERISTICS

All Stores (1994) Exitors Only

M SD Range M SD Range

Store Characteristics (1994) EDLP 0292 0455 [0, 1] 0299 0458 [0, 1] Sales volume 23902 14208 [57, 615] 16607 10507 [57, 615] (in $1,000s per week)

Size (in 1000s of square feet) 3104 1602 [2, 99] 2501 1302 [3, 99] Stores in chain 56804 66708 [10, 2051] 36208 54503 [10, 2051] Average size, stores in chain 3006 1003 [3, 99] 2701 909 [3, 97] Vertical integration 0652 0476 [0, 1] 0599 0490 [0, 1]

All Stores (1998) Entrants Only

M SD Range M SD Range

Store Characteristics (1998) EDLP 0287 0452 [0, 1] 0400 0490 [1, 0] Sales volume 28202 16102 [38, 692] 29704 16909 [38, 691] (in $1,000s per week)

Size (in 1000s of square feet) 3309 16 [2, 190] 3803 1804 [4, 190] Stores in chain 70102 76107 [1, 2316] 70304 75001 [1, 2316] Average size, stores in chain 33005 9067 [2, 110] 3208 1101 [4, 110] Vertical integration 0709 0453 [0, 1] 0758 0428 [0, 1]

1998, along with separate statistics for the new entrants in 1998 and the stores that chose to exit in 1994. Again, sev- eral notable patterns emerge. As in Table 1, the share of stores choosing EDLP is relatively stable across periods. Moreover, the stores that exit were no more likely to be offering EDLP than those in the market as a whole (note, however, that these are unconditional means). In contrast, the stores that were opened in 1998 were disproportion- ately offering EDLP, perhaps reflecting the influence of Wal-Mart or an overall shift in the optimal pricing policy. We further unpack these distinctions subsequently. Most of the other patterns are intuitive. Sales volume and both store and chain sizes all increase over time, as does the percent- age of stores operated by vertically integrated firms, reflect- ing long-term trends toward larger suburban formats and greater consolidation. Stores that choose to exit have lower sales and smaller footprints and are operated by smaller chains. Conversely, stores that just entered are larger; are owned by larger, more often vertically integrated chains; and tend to have higher sales volumes.

Key stylized facts. The identification of repositioning cost is ultimately driven by the firms that choose to switch. We now provide some preliminary descriptive evidence regarding switching behavior. Table 3 summarizes the set of actions taken by the set of incumbent firms that were in operation in 1994. The first frame presents raw counts, the second shows joint probabilities, and the third provides the switching matrix (conditional on a store’s format in 1994, what state did it transition to in 1998?). First, note that the data contain a great deal of switches and a fair number of exits. Both are useful for identification. The switches from EDLP to PROMO (and vice versa) provide the variation necessary to identify switching costs, while the exit choices are instrumental in identifying fixed operating costs (and accounting for continuation values). We make this intuition more precise in the following section. Focusing next on

Repositioning Dynamics and Pricing Strategy 755

Table 3 INCUMBENTS’ DECISIONS

PROMO 98 EDLP 98 EXIT

Counts PROMO 94 9314 494 1673 EDLP 94 836 3180 715

Probabilities PROMO 94 .575 .030 .103 EDLP 94 .051 .196 .044

Transitions PROMO 94 .811 .043 .146 EDLP 94 .177 .672 .151

the joint probabilities, we note that, not surprisingly, stores mostly adhere to their current pricing format. However, as is apparent from the transition matrix, PROMO exhibits the most state dependence: Conditional on choosing PROMO in 1994, 81% of stores continued with PROMO in 1998 (95% if we eliminate stores that exit). In contrast, con- ditional on choosing EDLP in 1994, only 67% of stores adhered to it in 1998 (79% if we eliminate the exits). This suggests that the benefits of switching from EDLP to PROMO are high, the costs of doing so are relatively low, or some mixture of the two is in effect. Finally, we note that, controlling for PROMO being the more dominant strategy, exit rates are slightly higher for the EDLP stores.

Tables 4 and 5 split these choice and transition pat- terns conditional on the presence or absence of Wal-Mart. In particular, we divide our local zip code markets into two groups: those in which Wal-Mart was present in the surrounding MSA in 1994 and those in which it was not, repeating the analysis of Table 3 for these two subgroups. Several noteworthy patterns emerge. The markets in which Wal-Mart is absent (Table 4) are similar to the full set of markets (not surprising, because they constitute 90% of the overall total). However, the markets in which Wal-Mart is present are quite distinct (Table 5). In particular, firms in these markets are less likely to adhere to PROMO, more likely to adhere to EDLP, and, conditional on switching, much more likely to adopt the EDLP format. Wal-Mart also makes firms more likely to exit. Thus, Wal-Mart does appear to be a disruptive presence, one that pushes its com- petitors toward EDLP or out of the market entirely. This disruption is key to identification, in that it provides a rea- son for firms to change strategies that were ex ante optimal.

Table 4 INCUMBENTS’ DECISIONS (WAL-MART ABSENT)

PROMO 98 EDLP 98 EXIT

Counts PROMO 94 8452 401 1471 EDLP 94 774 2784 622

Probabilities PROMO 94 .583 .028 .101 EDLP 94 .053 .192 .043

Transitions PROMO 94 .819 .039 .142 EDLP 94 .185 .666 .149

Table 5 INCUMBENTS’ DECISIONS (WAL-MART PRESENT)

PROMO 98 EDLP 98 EXIT

Counts PROMO 94 862 93 202 EDLP 94 62 396 93

Probabilities PROMO 94 .505 .054 .118 EDLP 94 .036 .232 .054

Transitions PROMO 94 .745 .080 .175 EDLP 94 .113 .719 .167

We also examine the format decisions of de novo entrants, those firms that entered between 1994 and 1998. Table 6 contains the counts and proportions of their for- mat decisions for the three sets of markets analyzed previ- ously. Note the difference in the split caused by Wal-Mart’s presence. For the full set of markets, the split is 60/40 in favor of PROMO, revealing an overall trend toward EDLP (recall that the proportion in the 1994 data—for all firms—was 70/30). However, there is again a differ- ence between markets with a Wal-Mart and those without: Entrants into markets with a Wal-Mart are 7% more likely to choose EDLP. Some of this is driven by selection (Wal- Mart prefers to enter markets that are amenable to EDLP pricing), but it also reflects that repositioning is costly.

Finally, in Figure 1, we examine how revenues change when switching pricing formats and when Wal-Mart enters the store’s local market between 1994 and 1998. Figure 1 shows the mean revenues (in thousands of dollars per week) across stores and zip codes in 1998, split by pricing strat- egy choice and by Wal-Mart’s presence. We look for a rough estimate of the effect of switching on revenues con- ditional on Wal-Mart’s entry or absence, using a difference- in-difference strategy. We compare the change in revenues between 1994 and 1998 of stores that switched with the change in revenues of stores that stayed with their orig- inal pricing format (see the right-hand side of Figure 1). The noteworthy aspect is the asymmetry associated with Wal-Mart entry in the revenue impact from switching. The effect of switching the pricing format relative to staying is weakly positive and roughly symmetric in markets with no Wal-Mart entry (a gain of roughly $2000 per week when shifting from EDLP to PROMO and approximately $3000 per week from shifting from PROMO to EDLP). However, in markets with Wal-Mart entry, shifting from EDLP to PROMO is associated with a gain of roughly $29,000 per

Table 6 ENTRANTS’ DECISIONS

All Markets Wal-Mart In Wal-Mart Out

Counts PROMO 98 1191 644 547 EDLP 98 795 482 313

Probabilities PROMO 98 .60 .57 .64 EDLP 98 .40 .43 .36

756 JOURNAL OF MARKETING RESEARCH, DECEMBER 2012

Figure 1 CHANGES IN REVENUES BY PRICING FORMAT SWITCH AND BY WAL-MART ENTRY

EDLP 94

PROMO 94

EDLP 98 (Wal-Mart) $273K/wk

$284K/wk

$292K/wk

$289K/wk

$252K/wk

$281K/wk

$302K/wk

$304K/wk

+$2K/wk (Switch Effect | No Wal-Mart)

+$29K/wk (Switch Effect | Wal-Mart)

+$3K/wk (Switch Effect | No Wal-Mart)

- $11K/wk (Switch Effect | Wal-Mart)

EDLP 98 (No Wal-Mart)

EDLP 98 (Wal-Mart)

EDLP 98 (No Wal-Mart)

PROMO 98 (Wal-Mart)

PROMO 98 (No Wal-Mart)

PROMO 98 (Wal-Mart)

PROMO 98 (No Wal-Mart)

week, but a shift from PROMO to EDLP is associated with a loss of roughly $11,000 per week. Although these raw mean comparisons are subject to caveats related to selec- tion, this asymmetry indicates that from a pure revenue perspective, differentiation in pricing policy was a better strategy for incumbent supermarkets to compete with Wal- Mart’s EDLP model. In our structural model, we control for selection in estimating revenue effects. Descriptive conditional policy functions. To further

unpack the dynamics of pricing strategy, we next present several linear probability models characterizing the play- ers’ propensity to choose alternative actions. These can be viewed as descriptive analogs of the structural pol- icy functions that constitute firm strategy. Each descriptive regression explains a store’s discrete choice as a function of market, rival, and own characteristics. We present the coefficients for only a small subset of the included covari- ates to highlight a few patterns, deferring a full analysis to subsequent sections of the article. Column 1 of Table 7 examines a store’s decision to

switch formats (either from EDLP to PROMO or vice versa) as a function of seven key constructs: whether Wal- Mart is present in the local market, whether Wal-Mart is present in the surrounding MSA, whether the store employed the EDLP format in 1994, the share of rival stores employing the EDLP format in 1994, the number

of rival stores, the size of the focal store’s chain, and our own measure of strategic focus. To capture the extent to which chains prefer to concentrate on a single pricing for- mat across stores (e.g., to exploit economies of scale and scope), we defined the variable focus as the squared dif- ference between .5 and the share of EDLP for stores oper- ated by the chain outside the focal market (implying that larger values correspond to chains that tend to use the same strategy in multiple markets). This is intended to capture the scope economies associated with choosing a consis- tent pricing strategy. We use this measure in the descrip- tive regressions, because it is symmetric for share-EDLP or share-PROMO. Turning to the results in column 1, the presence of Wal-

Mart is associated with more switches, and the effect is stronger at the MSA level than at the zip code level (per- haps reflecting the small number of zip codes in which Wal- Mart was present in 1994). As is clear from the switching matrices, EDLP stores are more likely to switch to PROMO than vice versa. The share of rival stores offering EDLP in the local market is also associated with more switching, as is a larger number of competing stores (though the latter effect is not statistically significant). Most notably, we find that larger, more focused chains are less likely to switch. This suggests that switching costs may be heterogeneous and, in particular, higher for larger firms and those whose

Repositioning Dynamics and Pricing Strategy 757

Table 7 DESCRIPTIVE POLICY FUNCTIONS

Dependent Variable

P4Switch � X5 P4Exit � X5 P4Enter � X5 P4EDLP � X5 P4Enter � X5

Wal-Mart in local market 0025 (.016) 0027 (.019) −0141 (.041) 0020 (.104) Wal-Mart in MSA 0017 (.006) 0054 (.008) 0035 (.014) 0053 (.039) EDLP (this store) 0136 (.008) −0023 (.008) Share of EDLP 0010 (.009) −0002 (.010) 0005 (.008) 0252 (.034) 0021 (.005)

(in local market) Number of rival stores 00008 (.0016) 0005 (.002) −0024 (.004) 0009 (.007) 0019 (.003) Total own stores −00000238 (.000004) −0000065 (.00004)

(all markets) Focus −0342 (.035) −0151 (.002)

Notes: All regressions include additional market, store, and chain controls.

reputation is more closely associated with a single pricing strategy (e.g., Food Lion, HEB).

Column 2 examines the decision to exit. Again, Wal-Mart is an important factor in driving stores to exit. In contrast to the switching patterns, EDLP stores are significantly less likely to exit, suggesting that this for- mat, while expensive to adopt, may offer some additional insulation from competitive pressures. Greater competition is associated with more stores exiting, while large, more focused chains are less likely to exit. Column 3 examines entry by supermarket chains. Not surprisingly, firms are less likely to enter local markets that contain a Wal-Mart but more likely to enter local markets in MSAs that have a Wal-Mart. This likely reflects underlying growth patterns, rather than a causal effect. (To reflect this, we incorpo- rate expectations of market growth into the full structural model.) As expected, the effect of competition is nega- tive, and the share of EDLP incumbents is insignificant. Column 4 examines the incumbents’ decision to select the EDLP format, conditional on having decided to enter. The only significant driver here is share EDLP, which is pos- itive (though many of the unreported demographic factors were significant as well). This echoes the patterns of assor- tative matching that Ellickson and Misra (2008) document, in which these patterns persist after accounting for corre- lated unobservables at the market level. Finally, column 5 examines the entry decision of Wal-Mart. Not surprisingly, Wal-Mart proactively targets markets with a large share of EDLP incumbents and prefers markets that already have a large number of stores. (Wal-Mart also tends to enter markets that are closer to its home base of Bentonville, Ark., and in close proximity to a distribution center, which are two of the unreported controls.) The correlation with incumbent store counts likely reflects that Wal-Mart tends to target markets with older, smaller incumbents (which are thus present in larger numbers), rather than a perverse taste for competition.

Identification. The key constructs to be identified are the costs of changing formats, as well as the revenue impact of such changes. We first discuss the revenue side. We observe revenues before and after a change in formats. Thus, the revenue effects are identified directly from these data, con- ditional on being able to account for selectivity induced by the choice of pricing strategy and survival in the market (i.e., not exiting). Stated differently, revenues are observed

only conditional on a chosen pricing strategy and condi- tional on being in the market. Thus, we need some source of independent variation that induces firms either to switch pricing strategy and stay active or to exit. As we explained and documented previously, this variation takes the form of entry by Wal-Mart, a large shock to the profitability of firms that likely causes them to reevaluate their pricing policy and market positioning. However, an identification concern then is that unobservables that induced firms to exit or to change pricing also caused Wal-Mart to enter (or not). To address this concern, we need some exoge- nous source of variation that drives Wal-Mart entry across markets, which can be excluded from a firm’s pricing strat- egy or exit decisions. In our framework, this variation is provided by two sets of market-level variables. The first captures the market’s radial distance from Bentonville, AR. We follow Holmes (2011), who documents convincingly that Wal-Mart followed a systematic strategy of opening its supercenters close to Bentonville and then spreading these radially inside out from the center. Controlling for MSA characteristics, we exclude the distance to Bentonville from supermarket payoffs. This serves as one source of exoge- nous variation driving Wal-Mart entry. The second vari- able represents the distance of a market from the nearest McLane distribution center. These are 22 large-scale distri- bution centers that were operated originally by the McLane company but were acquired in 1990 by Wal-Mart to ser- vice its supercenters.3 In the period from 1990 to 2003, Wal-Mart rolled out supercenters close to these distribution centers, as is evidenced in our data. We geocode the lati- tude and longitude of the distribution centers to calculate the Euclidean distance of each of them to the centroid of each MSA. McLane chose the locations of the distribution centers in the 1980s (to service a preexisting network of convenience stores), and we treat them as predetermined in our analysis of the 1994 and 1998 data.

We now explain how we can use the observed switch- ing matrix, exit behavior, and revenue data to identify the cost side of the model. The key distinction is to separate the switching costs of changing pricing strategies from the fixed costs of per-period operation. Conceptually, these are different constructs, in that the switching costs are sunk

3In May 2003, Berkshire Hathaway acquired McLane Company from Wal-Mart for $1.45 billion.

758 JOURNAL OF MARKETING RESEARCH, DECEMBER 2012

and incurred only at the point of a switch, whereas the fixed costs are incurred every period. We identified our switching costs from the margin from changing a pricing format versus staying with the current policy, while the fixed costs affect the propensity to stay with the current pricing policy relative to exiting. To demonstrate this, let vE → P denote the present-discounted payoff from switch- ing from EDLP pricing to PROMO and vP → E denote the present-discounted payoff from switching from PROMO pricing to EDLP. Analogously, vE → E and vP → P are the present-discounted payoff from staying with EDLP and PROMO, respectively. Let vE → Exit and vP → Exit respec- tively denote the present-discounted payoff from exiting. We normalize these to zero. These objects can be recog- nized as the choice-specific value functions associated with each of these six actions. For ease of notation, we suppress the dependence of these functions on the state vector. Let RE and RP denote the per-period revenues from

following EDLP and PROMO, respectively. For the pur- poses of this discussion, assume that these have already been estimated using a selectivity-controlled model from the auxiliary revenue data. Thus, we treat RE and RP as known. Let the fixed costs incurred per period when using EDLP and PROMO, respectively, be 4FCE1FCP5, and let 4RCE → P1RCP → E5 denote the key parameters of interest: the cost of switching from one format to another. Then, we can write the choice-specific values from staying with the current strategy as follows:

vE → E = RE +FCE + 0+ÂƐ6vE4057

vP → P = RP +FCP + 0+ÂƐ6vP40571 (1)

from switching pricing as

vE → P = RP +FCP +RCE → P +ÂƐ6vP4057

vP → E = RE +FCE +RCP → E +ÂƐ6vE40571 (2)

and from exiting as

vE → Exit = 03 vP → Exit = 00

In the preceding equations, Â represents a (fixed) dis- count factor for the supermarkets, vP405 represents the value function conditional on choosing PROMO, and vE405 represents the value function conditional on choos- ing EDLP. The expectation Ɛ4·5 is taken with respect to the state vector at the time of making the deci- sion. (We have suppressed unobservables, because the argument is not changed if we add additive errors.) Following Hotz and Miller (1993), the choice-specific value functions 4vP → E1vE → P1vE → E1vP → P5 are semi- parametrically identified from the observed probabilities of switching, exiting, and staying with current pricing in the data. Then, we can identify the switching costs as

RCE → P = vE → P − vP → P1 and(3)

RCP → E = vP → E − vE → E0(4)

MODEL

In this section, we describe our structural model of supermarket competition and pricing format choice. There

are two types of firms: Wal-Mart and conventional super- markets (e.g., Kroger, Safeway). Supermarket firms are assumed to compete in local markets, taken here to be zip codes, though we allow for some degree of cross-market competition in the case of Wal-Mart. Supermarket firms choose whether to enter a given market and, if so, what pricing format to adopt, either EDLP or PROMO. We also model the entry decisions of Wal-Mart but assume that every Wal-Mart is EDLP, consistent with both the data and its stated business model. After the supermarket firm has entered, its dynamic decisions include whether to con- tinue offering the same format, switch to the alternative (and pay a switching cost), or exit the market entirely. Wal-Marts neither exit nor change formats. For tractability, we assume that firms make independent entry and format decisions across local markets but allow for correlation and economies of scale and scope by allowing fixed operating costs to depend on past choices the firm has made outside these local markets.

The dynamic discrete game unfolds in discrete time over an infinite horizon, t = 11 0 0 0 1�.4 Firms compete in M dis- tinct local geographic markets (m = 11 0 0 0 1M). For ease of notation, we suppress the market subscript in what follows. For each market–period combination, we observe a set of incumbent firms that are currently active in the market. We further assume the existence of two potential supermarket entrants per period, which choose whether to enter the mar- ket in that period and, if so, what pricing strategy to adopt.5

If they choose not to enter, they are replaced by new poten- tial entrants in the subsequent period. Wal-Mart may also choose whether to enter the market each period, and if it does enter, it does so in the EDLP format. Let N denote the total number of firms (both Wal-Mart and the supermarkets) making decisions in each market each period. Within N, the set of active firms are called “incumbents” and the remain- ing firms “potential entrants.” We suppress the distinction between potential entrants and incumbents in the general setup of our model but revisit this when we introduce the empirical framework. Within each market, we index firms by i ∈ I = 81121 0 0 0 1N9. Firm i’s choice in period t is given by dit ∈�i, while the actions of its rivals are denoted d−i

t ≡ 4dit1 0 0 0 1d

i − 1 t 1di + 1

t 1 0 0 0 1dNt 5. The support of �i is discrete and dependent on firm type. For incumbent firms, dit can take three values: [Exit, do EDLP, or do PROMO]. For potential entrants, dit can take three values: [Stay out of the market, Enter with the EDLP pricing format, or Enter with the PROMO pricing format]. For Wal-Mart, dit can take two values: [Stay out of the market or Enter with the EDLP pricing format].

Decisions and payoffs depend on a state vector, which describes the current conditions of the market as well as each firm’s operating status and pricing format. Follow- ing the standard approach in the dynamic discrete choice literature, we partition the current state vector into two

4Surveys of this increasing literature stream include Draganska et al. (2008) and Ellickson and Misra (2011) in the context of static discrete games and Aguirregabiria and Mira (2010) and Ackerberg et al. (2005) in the context of dynamics.

5A normalization on the number of potential entrants of this sort is standard in the dynamic entry literature because it is not identified without additional information.

Repositioning Dynamics and Pricing Strategy 759

components—one that is commonly observed by every- one (including the econometrician) and one that is pri- vately observed by each firm alone—making this a game of incomplete information. We denote the vector of com- mon state variables xt, which includes market demograph- ics such as population, and a full description of each player’s current condition. The key endogenous state vari- ables included in xt are each firm’s current pricing format and whether it is active at the beginning of each period t. In addition to the common state vector, each firm pri-

vately observes a vector Åt4d i t5, which depends on its

current choice and can be interpreted as a shock to the per-period payoffs associated with making that choice, rel- ative to maintaining the status quo.6 Again, following stan- dard practice, we make two additional assumptions: (1) The unobserved state variables enter additively into each firm’s per-period payoff function (additive separability), and (2) Å are also independently and identically distributed (i.i.d.) across time and over players, and conditional on each firm’s choice in period t, Å do not affect the transitions of x (conditional independence with independent private values, CI/IPV). We further assume that Å are distributed Type 1 extreme value (T1EV), with density function g4·5.

Given assumption AS, the per-period (flow) profit of firm i in period t, conditional on the current state, can be decom- posed as çi4xt1d

i t1d

−i t 5+Åt4dit5. The profit function is super-

scripted by i to reflect that the state variables might affect different firms in distinct ways (e.g., own vs. others’ char- acteristics). Assuming that firms move simultaneously in each period, let P(d−i

t � xt) denote the probability that firm i’s rivals choose actions d−i

t conditional on xt. Because Åit is i.i.d. across firms, we can express P(d−i

t � xt) as follows:

P4d−i t � xt5 =

I∏ j6 = i

pj4djt � xt51(5)

where pj4djt � xt5 is player j’s conditional choice probability (CCP). Taking the expectation of çi4xt1d

i t1d

−i t 5 over d−i

t , firm i’s expected current payoff (net of the contribution from its unobserved state variables) is given by

Ïi4xt1d i t5 =

∑

d−i t ∈ D

P4d−i t � xt5çi4xt1d

i t1d

−i t 51(6)

which accounts for the simultaneous actions taken by each of its rivals. We assume that state transitions follow a con- trolled Markov process, F4xt + 1 � xt1dit1d−i

t 5, which we can estimate semiparametrically from the data because all the elements, 4xt + 11xt1d

i t1d

−i t 5 are directly observed. The tran-

sition kernel for the observed state vector is then given by

fi4xt + 1 � xt1dit5 = ∑

d−i t ∈ D

P4d−i t � xt5F4xt + 1 � xt1dit1d

−i t 50(7)

Given the assumption of conditional independence with independent private values mentioned previously, the tran- sition kernel for the full state vector is

f i4xt + 11 Å i t + 1 � xt1d

i t1 Å

i t5 = f i4xt + 1 � xt1dit5g4Å

i t + 150

6This can be interpreted as a shock to either revenues or to costs. We can allow for one but not both; we interpret the Ø as shocks to revenues, which enables us to account for selection on these unobservables when we incorporate revenue data in our estimation procedure.

Next, we construct each firm’s value function, optimal decision rule (strategy), and the conditions for an MPE. Assuming that firms share a common discount factor Â, rational, forward-looking firms will choose actions that maximize expected present discounted profits:

Ɛ

{ �∑ Ò = t

ÂÒ − t6Ïi4xÒ 1d i Ò5+ ÅÒ4d

i Ò57 � xt1 ÅiÒ

} 1(8)

where the expectation is over all states and actions, whose solution is given by the value function

Vi t4xt1 Åt5 = max

dit

{ Ïi4xt1d

i t5+ Åt

+ ÂƐ [ Vt + 14xt + 11 Åt + 1 � xt1dit5

]} 0(9)

Following standard arguments from the dynamic discrete games literature (e.g., Aguirregabiria and Mira 2007), an MPE in this setup implies the following associated condi- tional choice probabilities:

pi4dit � xt5 = exp6vi4xt1d

i t57∑

dit ∈ �i

exp6vi4xt1d i t57

1(10)

where the choice-specific value functions, vi4xt1d i t5, are

defined as follows:

vit4xt1d i t5≡ Ïi4xt1d

i t5(11)

+Â ∫ V̄i

t + 14xt + 15f4xt + 1 � xt1dit5dxt + 10

In Equation 11, the ex ante (or integrated) value function, V̄i

t4xt5, is defined as the continuation value of being in state xt just before Åt is revealed and is computed by integrating Vi

t4xt1 Åt5 over Åt [i.e., V̄ i t4xt5≡

∫ Vi

t4xt1 Åt5g4Åt5dÅt]. Given that the Å are distributed T1EV, Equation 11 reduces to

vi4xt1d i t5 = Ïi4xt1d

i t5(12)

+ Â ∫ {

vi4xt + 11d ∗i t + 15− ln

[ pi4d∗it + 1 � xt + 15

]}

× f i4xt + 1 � xt1dit5dxt + 1 +ÂÃ1

where pi4dit � xt5 is the implied CCP from Equation 10, Ã is Euler’s constant, and d∗it + 1 represents an arbitrary reference choice in period t + 1. (This reference choice reflects the requirement of a normalization for level; for the full deriva- tion of this representation, see Arcidiacono and Ellickson 2011.) Note that by normalizing with respect to exit, which is a terminal state after which no additional decisions are made, the continuation value associated with this refer- ence choice can now be parameterized as a component of the per-period payoff function, eliminating the need to solve the dynamic programming problem when evaluating Equation 12. This simplified representation of the choice- specific value function exploits the property of finite depen- dence, originally developed in the context of single agent dynamics by Altug and Miller (1998) and later extended to games by Arcidiacono and Miller (2012). Avoiding the full solution of the dynamic programming is useful in our setting because our underlying state space is high dimen- sional. Alternative methods would either involve artificial discretization of the state space (to allow transition matrices to be inverted) or a parametric approximation to the value or policy functions.7 The current approach requires neither.

7Another option is to switch to continuous time methods (Arcidiacono et al. 2010).

760 JOURNAL OF MARKETING RESEARCH, DECEMBER 2012

Assuming that firms play stationary Markov strategies, we follow Aguirregabiria and Mira (2007) in representing the associated MPE in probability space, requiring each firm’s best response probability function (Equation 10) to accord with its rivals’ beliefs (Equation 5). While existence of equilibrium follows directly from Brouwer’s fixed point theorem (see, e.g., Aguirregabiria and Mira 2007), unique- ness is unlikely to hold given the inherent nonlinearity of the underlying reaction functions. However, our two-step estimation strategy (described subsequently) enables us to condition on the equilibrium that was played in the data, which we assume is unique. This concludes the discussion of the model setup.

ECONOMETRIC ASSUMPTIONS AND EMPIRICAL STRATEGY

We now introduce the functional forms and explicit state variables that enable us to take the dynamic game described previously to data. Essentially, this involves identifying the exogenous market characteristics that influence profits and specifying a functional form for çi4·5, the deterministic component of the per-period payoff function.

Players

Although our model incorporates the endogenous actions (and state variables) of three sets of players (incumbent supermarkets, potential supermarket entrants, and Wal- Mart), the revenue and cost implications of reposition- ing we are interested in are identified from the actions of incumbent supermarkets. Because we condition on the CCPs of all three classes of players and the structural objec- tive function can be separately factored by type, we are able to recover consistent estimates of the structural parameters of interest without specifying the full structure of the cost and payoff functions for nonincumbents. We use the CCPs outlining the actions of Wal-Mart and the other entrants to capture the beliefs of incumbent supermarkets. We base the inference of switching costs on the likelihood of the actions of the incumbents, conditional on these beliefs. This is use- ful both for reducing the computational burden of estima- tion and in allowing us to remain agnostic regarding these additional components of the underlying structure.

Payoffs

The per-period profit function of incumbent supermar- kets captures the revenues that firms earn in the prod- uct market, the fixed costs of operation, and the fixed costs associated with repositioning. (For potential entrants, it would also include the sunk cost of entry.) Because oper- ating costs are not separately identified from the scrap value of ceasing operation, we normalize the latter to zero. We decompose per-period profits as follows:

çi4xt1d i t1d

−i t 3ä5 = Ri4xt1d

i t1d

−i t 3 ÈR5−Ci4xt1d

i t3 ÈC51(13)

separating the revenues accrued in the product market from the costs associated with taking choice dit. The parameters ä = 4ÈR1 ÈC5 index the revenue and cost functions, respec- tively. Equation 13 is richer than the latent payoff structures often employed in the empirical entry literature because it splits per-period payoffs into revenue and cost components. We are able to do this because we observe revenue data

separately for each supermarket, under its chosen pricing strategy, in each market. The incorporation of the revenue data also serves a useful auxiliary purpose: It enables us to measure all costs in dollars.

Revenues

We parameterize the revenue function, R4xt1d i t1d

−i t 3 ÈR5

as a rich function of both exogenous demographic vari- ables and endogenous decision variables. To capture the heterogeneity of profits across markets, we interact each component of the latter with a full set of variables consti- tuting the former. The demographic (Dm5 variables include population, proportion urban, median household income, median household size, and percentage black and Hispanic. In addition, we shift the intercept with store/firm character- istics zi, which include store size and the number of stores in the parent chain. We can write the actual specification as follows:

R ( xt1d

i t = a1d−i

t 3 ÈR )

(14)

= D′ mÈ

04a5 R + zR

′ i ÈzR +D′

mÈ 14a5 R I4WMMSA4m5 = 15

+D′ mÈ

24a5 R āEDLP−i +D′

mÈ 34a5 R N−i +D′

mÈ 44a5 R FOi4a51

where, āEDLP−i is the share of rival stores choosing the EDLP format, N−i is a count of rival firms, WMMSA4m5 is a dummy for whether Wal-Mart operates in the firm’s MSA, and FOi(a) reflects the focus of the parent chain on the particu- lar pricing strategy measured as a percentage of the chain’s stores adopting strategy a.

Costs

We parameterize the cost term, which we treat as latent, as follows. We assume that all incumbent firms pay a fixed operating cost each period that depends on their current pricing format. In addition, should they choose to switch formats, they incur an additional, one-time repositioning cost. To emphasize the difference between these cost com- ponents, we subset the state vector, xt, into two parts, xt ≡ 4dit − 11 x̃t5, where dit − 1 is supermarket i’s pricing strategy in the previous period (which is part of the state vector), and x̃t is everything in state xt except d

i t − 1. We can express

costs for an incumbent that chooses to stay in the market (the second term in Equation 13) as follows:

Ci4xt1d i t3 ÈC5 = FCi4x̃t1d

i t3 ÈFC5

+ 4dit 6= dit − 15RC i4xt1d

i t3 ÈRC51

where FCi4·5 represents fixed operating costs and RCi4·5 represents repositioning costs (which are only relevant when the firm changes pricing formats). The indicator, 4dit 6= d

i

t − 15, ensures that RC i4xt1d

i t3 ÈRC5 is incurred only

if the pricing strategy chosen today is different from the one chosen in the previous period. This separation clarifies how the identification of the fixed costs separately from the repositioning costs depends on partitions of the state space. The pricing strategy of an incumbent at the beginning of a period is part of the state vector. The difference in out- comes for incumbents when this state changes in a period versus when it does not provides information about repo-

Repositioning Dynamics and Pricing Strategy 761

sitioning costs separate from fixed costs. The specification of FC is

FCi4x̃t1d i t = a3 ÈFC5(15)

= D′ mÈ

04a5 FC + zCi

′ ÈzFC +D′

mÈ 14a5 FC I4WMMSA4m5 = 15

+D′ mÈ

24a5 FC E4āEDLP−i 5+D′

mÈ 34a5 FO FOi4a5

and RC is defined as

RCi4xt1d i t = a3 ÈRC5 = D′

mÈ 4a5 RC + ÈWM

RC I4WMMSA4m5 = 15(16)

+ ÈESRCE4ā EDLP −i 5+ ÈFORCFOi4a50

Finally, incumbent firms that choose to exit receive a scrap value associated with selling their physical assets and residual brand value. Because this is not separately iden- tified from the fixed cost of operation, we normalize this scrap value from exiting to zero. The parameters to be esti- mated are (ÈR1 ÈFC1 ÈRC). In the next section, we present a three-step empirical strategy that delivers estimates of this parameter vector. We first provide a short high-level dis- cussion of our estimation approach and then delve into the specific details.

Estimation Approach

Our estimation strategy is built on the approach intro- duced by Hotz and Miller (1993) in the context of dynamic discrete choice and later extended to games by Aguirregabiria and Mira (2007), Bajari, Benkard, and Levin (2007), Pakes, Ostrovsky, and Berry (2007), and Pesendorfer and Schmidt-Dengler (2008). This approach is typically applied to discrete-choice outcomes. We extend the approach in this literature to incorporate revenue data (a continuous outcome). The key difficulty to overcome is that revenues are observed only conditional on the cho- sen action (staying in the market and choice of pricing). Thus, inference is subject to a complicated selection prob- lem whereby choices are determined in a dynamic game with strategic interaction. We extend methods introduced in Ellickson and Misra (2012) to accommodate the selection in an internally consistent manner to improve inference in the dynamic game.

Our estimation procedure consists of three steps. In Step 1, we obtain consistent estimates of the (nonstruc- tural) CCPs using a flexible, semiparametric approach. In addition, we estimate the transition kernels governing the exogenous state variables (e.g., market characteristics). For these transition kernels, we use a parametric approach because they are already structural objects at this point. We then use both sets of estimates to construct the transitions that govern future states and rival actions, which inform the right-hand side of Equation 12. We also invert the CCPs to construct the choice-specific value functions for each action across firms, markets, and states. We use these objects for estimation of the parameter vector (ÈR1 ÈFC1 ÈRC) in Steps 2 and 3. In Step 2, we use the CCPs obtained from Step 1 to

create a selection correction term for a revenue regression. The correction serves as a control function. Incorporating the control function then enables us to consistently estimate the revenue parameters ÈR using the revenue data. Given estimates of ÈR, we can construct counterfactual revenue

functions that provide the potential revenues to a firm if it chooses any of the available strategies (and not just the one it was observed to choose in the data).

In Step 3, we make a guess of the cost parameters ÈFC, ÈRC, and we combine these with the counterfactual rev- enues constructed from Step 2 to create predicted choice- specific value functions for each incumbent firm across actions, markets, and states. The observed choice-specific value functions implied by the data are available from Step 1, after inverting the CCPs. We then estimate cost parameters ÈFC, ÈRC by minimizing the distance between the “observed” choice-specific value functions and the model- predicted, choice-specific value functions. We construct standard errors that account for the sequential estimation by block bootstrapping the entire procedure over markets. Loosely speaking, the parameters indexing Ri4·5 can be thought of as being estimated from the revenue data (sub- ject to controls for dynamic selection) and the parame- ters indexing both FCi4·5 and RCi4·5 as estimated from the firm’s dynamic discrete choice over actions. Next, we present the specific details of the procedure.

Step 1: Estimating CCPs and transitions. We estimate the CCPs semiparametrically using a second-order polyno- mial approximation in the state variables alongside several additional interactions. We constructed the transition den- sity of the exogenous elements of the state vector (i.e., demographics) using census growth projections, we took firm- and chain-level factors as known. (The exact specifi- cation of the first stage and the full results are available on request.) Thus, at the end of this step, we know the transi- tions conditional on rival’s actions, F4xt + 1 � xt1dit1d−i

t 5, and the CCPs that determine those actions, pi4dit � xt5. Further- more, using Equations 5 and 7, we can compute the joint probability of rivals’ actions, P4d−i

t � xt5, and the transitions that occur after integrating them out, f i4xt + 1 � xt1dit5. Finally, we let d1t denote the option to exit. Given

pi4dit � xt5, we can also invert the CCPs using Equation 10 to recover the observed choice-specific value functions (rel- ative to Exit) as implied by the data for every incumbent firm, action, market, and state as

vi4xt1d i t5 = ln

[ pi4dit � xt5

] − ln

[ pi4d1t � xt5

] 1(17)

where, implicitly, the value from exiting has been normal- ized to zero (i.e., vi4xt1d

1 t 5 = 0 in Equation 10). These

objects are then stored in memory, concluding Step 1. Step 2: Selectivity corrected revenue functions. Next, we

construct the model-predicted analog of Ri4xt1d i t1d

−i t 3 ÈR5.

To deal with selectivity, we approximate expected revenues by a flexible function of the states and actions, R4·5:

Ri4xt1d i t1d

−i t 3 ÈR5 = R4xt1d

i t1d

−i t 3 ÈR5+Çi

t4d i t5+ Åit4d

i t50(18)

Actual revenues, Ri (·), also include two error components: Çi t , representing an unanticipated shock to revenues from

the firm’s perspective, and Åit , which is the same unob- served state variable that appears in the choice model (and, therefore, the source of the selection problem). The dif- ference between Ç and Å is that Ç is unobserved to the firm and the econometrician while making decision dit, whereas Å is known to the firm when making decision dit but is unknown to the econometrician. In the line with Pakes et al. (2005), Ç is an expectation error, while Å is

762 JOURNAL OF MARKETING RESEARCH, DECEMBER 2012

a standard random utility shock. The section problem can be articulated as the fact that revenues are codetermined with choices, and therefore, Ɛ6Åit4d

i t5 � dit7 6= 0. Thus, run-

ning the regression (Equation 18) will give biased estimates of R4·5. However, we can accommodate the selectivity by noting that by construction, Ɛ6Çi

t4d i t5 � dit7 = 01 but that

Ɛ6Åit4d i t5 � dit7 = Ã− ln pi4dit � xt5 6= 0; which follows from well-

known properties of the T1EV distribution. The term à − ln pi4dit � xt5 is a control function that accommodates the fact that from the econometrician’s perspective, unobserv- ables are restricted to lie in a particular subspace when the firm is observed to have chosen strategy dit. Letting Rit4d

i t5

denote the observed revenues to supermarket i when choos- ing strategy dit, we can estimate revenues consistently with the following regression:

R̃4xt1d i t1d

−i t 3 ÈR5 = R4xt1d

i t1d

−i t 3 ÈR5+Çi

t4d i t51(19)

in which

R̃4xt1d i t1d

−i t 3 ÈR5 = Rit4d

i t5−

[ Ã − ln pi4dit � xt5

] (20)

is a selectivity corrected revenue construct that adjusts for the restriction that we only observe revenues for the pricing strategy that was actually chosen. Given consistent esti- mates of the parameters ÈR that index R4xt1d

i t1d

−i t 3 ÈR5, we

are then able to construct the predicted revenues for any choice.8

We can now compute the expected revenues from the firms’ perspective associated with any choice (i.e., the revenue analog of Equation 6). Suppressing the indexing parameters for brevity, these expected revenues are then given by

ri4xt1d i t5 =

∑

d−i t ∈ D

P4d−i t � xt5Ri4xt1d

i t1d

−i t 51(21)

in which the P4d−i t � xt5 are already known from Step 1. By

choosing a functional form that is linear in its parameters for Ri4xt1d

i t1d

−i t 5, expected revenues (Equation 21) can be

constructed directly as a linear function of expected actions. Step 3: Minimum distance estimation of costs. The goal

of Step 3 is to estimate the cost parameters, ÈFC, ÈRC. To understand the approach, recall that we can write the choice specific value function (CSVF) as follows:

vi4xt1d i t5=Ïi4xt1d

i t5+Â

∫ { vi4xt + 11d

∗i t + 15− ln

[ pi4d∗it + 1 �xt + 15

]}

×f i4xt + 1 �xt1dit5dxt + 1+ÂÃ0

In the preceding equation, d∗it + 1 is a reference alternative, here chosen as the option to exit in the next period. Choos- ing to normalize with respect to exit, an action whose con- tinuation value has now been normalized to zero, allows

8In general, researchers should take care in using two-stage approaches with fitted CCPs, particularly in cases with limited data or when the CCPs for the chosen action are naturally small. In such cases, the estimation error can severely affect the quality of the regression results. To check and correct for such bias, we followed an approach outlined in Pakes and Linton (2001) that uses a Taylor-series-based adjustment factor (1/p) to mitigate the bias. Because, in our case, the probabilities are reasonably large (with an IQR = 8061 099), the resultant bias appears negligible. Never- theless, the results reported subsequently are based on this bias-corrected specification. We thank the editor for this suggestion.

the first component in the second term of the CSVF to drop out (i.e., vi4xt + 11d

∗i t + 15 ≡ 05. The remaining component

of the continuation value can now be constructed directly from the data (using the first-stage CCPs and the structural components of the transition kernel) and treated as an off- set term. We construct the empirical analog of this offset term as follows:

�Üln P0 4xt1d i t5 = −Â

∫ ln [ pi4d∗it + 1 � xt + 15

] (22)

×f i4xt + 1 � xt1dit5dxt + 10

We can compute the simulated analog of this future value using Monte Carlo simulation. All that remains is the per-period payoff function Ïi4xt1d

i t5, which has already

been decomposed into its revenue component (constructed from Equation 21) and the contribution from the cost side. Because the parameters that index the revenue functions have already been recovered from Step 2, the expected rev- enues associated with each format (or exit) choice can now be treated as an additional offset term. We can write the model-predicted CSVF as follows:

vi4xt1d i t3 ÈFC1 ÈRC5 = �ri4xt1dit5−Ci4xt1d

i t3 ÈFC1 ÈRC5︸ ︷︷ ︸

Ïi4xt 1d i t5

+ �Üln P0 4xt1d i t51

where �ri4xt1dit5 is available from Step 2, �Üln P0 4xt1d i t5 is

constructed as mentioned previously, and, finally, vi4xt1d i t3

ÈFC1 ÈRC5 is the predicted CSVF for the current guess of the cost parameters ÈFC, ÈRC. We can now recover the cost parameters by minimizing the distance between the model- predicted CSVF and the observed CSVFs from Step 1 (Equation 17):

∥∥vi4xt1dit5− vi4xt1d i t3 ÈFC1 ÈRC5

∥∥0(23)

A concern with this estimator is that the effec- tive instrument in the resulting estimating equations is 6ávi4xt1d

i t3 ÈFC1 ÈRC57/áÈ, which is then correlated with the

“errors,”

Îi4 xt1d i t3 ÈFC1 ÈRC5 = vi4 xt1d

i t5− vi4 xt1d

i t3 ÈFC1 ÈRC51(24)

because the implied estimating equations are implicitly

∑ i

∑ t

ávi4xt1d i t3 ÈFC1 ÈRC5

áÈ Îi4xt1d

i t3 ÈFC1 ÈRC5 = 00(25)

This issue is particularly relevant for the parameters that pertain to endogenous constructs. To address this, we use alternative instruments for these estimating equations. In particular, for the parameters related to Wal-Mart and the strategy choices of the supermarkets, we use functions of the variables (zt) excluded from the focal store’s payoffs (e.g., distance to Bentonville and distance to distribution centers for Wal-Mart; the focus of the chain, store size, and so forth for competing supermarkets), in addition to mar- ket demographics (Dm). Denote these functions h (zt1Dm).

9

9In our implementation, the instrument functions were h(z, D) = x if the relevant covariate x was exogenous or equal to a positive function of excluded variables and demographics if not. Details on the actual instru- ments and functions used are available on request.

Repositioning Dynamics and Pricing Strategy 763

Using these functions, we then define our estimating equa- tions as follows:

∑ i

∑ t

h�zt�Dm�� i�xt�d

i t� �FC� �RC� = 0�(26)

Our cost estimates are then obtained as the (�∗FC, � ∗ RC�

that solve these equations in sample.10

RESULTS

We now discuss results from the estimation of our struc- tural model. We first discuss the estimates from the revenue side and then present the cost side results.

Revenues

We begin by documenting the revenue implications of following an EDLP versus PROMO pricing strategy. We obtain the implied revenues as the selection-corrected pre- dictions from the revenue regression model. Appendixes A and B present the full estimates from the revenue regression for both EDLP and PROMO. These regressions allow for interactions of each of the variables presented in the first

10We thank the editor for suggesting this estimation approach.

Figure 2 COUNTERFACTUAL REVENUES

0 200 400 600 800 1000

Fitted Revenues (PROMO) Fitted Revenues (EDLP)

Change in Revenues (Switch to PROMO) Change in Revenues (Switch to EDLP)

0

0 00

50 10

0 15

0

20 40

60 80

50 0

10 00

15 00

F re

q u

en cy

F re

q u

en cy

F re

q u

en cy

F

re q

u en

cy

20 00

25 00

0 50

0 10

00 15

00 20

00 25

00

200

–150150100500 –100 –50 0

400 600 800 1000

column (named “Variable”), with a full range of market- level demographics presented in the second column (named “Interactions”), and also correct for selectivity using the control function approach outlined previously. Rather than discuss them separately, we present the predicted revenues from this model. We first ask how revenues would appear if every supermarket we observe in 1994 chose EDLP. In Figure 2, we plot a histogram of the predicted EDLP rev- enues (top-right panel). Analogously, we then ask how rev- enues would appear if every supermarket we observe in 1994 instead chose PROMO (plotted in top-left panel). For what follows, these plots and the numbers below are pre- sented in units of 1000s of dollars per week. Comparing the two histograms, we observe that revenues are higher under PROMO. To demonstrate a sense of the differences in dollar terms, in Table 8, we present the 5th, 50th, and 95th percentiles of the distribution of revenues under EDLP and PROMO. Observing first at the 50th percentile, note that the median store market under PROMO earns revenues of approximately $119,720 more per week relative to the median store market under EDLP. Converting to an annual basis, this difference translates to approximately $6.22 mil- lion per year ($119,720 per week×52 weeks). Comparing store markets at the 5th percentile of the revenue distri- bution under both formats, we find that this difference is

764 JOURNAL OF MARKETING RESEARCH, DECEMBER 2012

Table 8 DISTRIBUTION OF ESTIMATED REVENUES

5% 50% 95%

EDLP Intercept 116035 273028 522026 Wal-Mart −48057 −28032 −4021 E(Share of competitors EDLP) −12004 058 43022 Number of competitors −12093 −9058 −2049 Focus of chain (EDLP) −12098 11011 73036 Total revenues (fitted) 185004 364005 617014

PROMO Intercept 115086 295048 510013 Wal-Mart −35013 −20054 −6039 E(Share of competitors EDLP) −15019 −6091 3092 Number of competitors −10006 −5086 −1079 Focus of chain (Promo) −12011 39022 86067 Total revenues (fitted) 255093 483077 720044

approximately $3.68 million annually in favor of PROMO ($70,950 per week× 52 weeks). At the 95th percentile of the revenue distribution under both formats, this differ- ence is approximately $5.37 million annually in favor of PROMO ($103,000 per week× 52 weeks). It is evident that stores earn higher revenues under PROMO, whether large or small, whether in large markets or small markets, and across several competitive conditions. However, our estimates also imply significant heterogeneity across both stores and markets in these effects. In the bottom panels of Figure 2, we plot the distribution

across stores of the change in revenues between the pricing strategy chosen by a store in 1998 versus the alternative strategy. This is analogous to checking the Nash condi- tions in a static model. The bottom-left panel shows how much revenues would have changed if stores that switched to PROMO in 1998 had instead stayed with EDLP (i.e., the distribution of R̂PROMO98 − R̂EDLP98 � switch to [PROMO98]). The bottom-right panel shows how much revenues would have changed if a store that switched to EDLP in 1998 had instead stayed with PROMO (i.e., the distribution of R̂EDLP98 − R̂PROMO98 � switch to [EDLP98]). From the left panel, it is evident that switching to PROMO from EDLP results in an increase in revenues (so the observed switch is revenue enhancing). However, the right panel, indicates that switching to EDLP from PROMO mostly decreases revenues (so the observed switch is revenue reducing). This is consistent with the difference-in-differences approach we presented previously as part of model-free analysis. The only way the model can explain the observed switching into EDLP format given these revenue implications is by postulating a cost saving associated with that format. This is a source of identification in the model. We now discuss the heterogeneity in revenues across

markets and how various factors affect this heterogeneity. Our strategy for summarizing these results is to present his- tograms of effects across markets, which should be read alongside tables containing the 5th, 50th, and 95th per- centiles of each across markets. We organize our discus- sion around four key variables of interest: (1) the revenue implications of Wal-Mart’s presence, (2) the effect of local competition, (3) the effect of the similarity of the chosen pricing strategy with that chosen by local competitors, and (4) economies of scale and scope. Recall from our dis- cussion in the “Econometric Assumptions and Empirical

Strategy” section that we capture these effects by includ- ing the following variables: (1) a dummy for whether Wal- Mart operates in the firm’s MSA (WMMSA5, (2) the num- ber of rival firms in the market (N−i), (3) the share of rival stores choosing the EDLP format (āEDLP−i 5, and (4) the focus of the chain measured as a percentage of the chains’ stores adopting strategy a [FOi4a = EDLP) and analogously FOi4a = PROMO)]. Each of these variables is interacted with a full range of market demographics and included as right-hand-side variables in the revenue regression (see Appendixes A and B). In Figure 3, we plot the distribution across markets of the total effect of each of these variables on revenues under the EDLP format. For example, the top- right panel in Figure 3 contains a histogram of the effect of Wal-Mart on EDLP revenues. Letting m denote a mar- ket, this is essentially a histogram of the Wal-Mart effect in market m, computed as follows:

[ È̂ 14EDLP5 0R + È̂

14EDLP5 1R 4popm5+ È̂

14EDLP5 2R 4hhsizem5

+ È̂ 14EDLP5 3R 4%blackm5+ È̂

14EDLP5 4R 4%urbanm5

+ È̂ 14EDLP5 5R 4%hispm5+ È̂

14EDLP5 6R 4hincm5

] 1

where the È̂R are the estimated coefficients of the inter- actions of the WM variable with market demographics in the revenue regression for the EDLP format reported in Appendix A. The È̂R parameters correspond to Equation 14 in the text. The other histograms in Figure 3 are created analogously for the other variables, N−i, ā

EDLP −i , and FOi

(EDLP). Again, to provide a sense of the heterogeneity, we report the 5th, 50th, and 95th percentiles of these distribu- tions in Table 8.

Observing the Wal-Mart effect in Figure 3, we note that the presence of Wal-Mart in the same MSA as a super- market unambiguously reduces revenues. The net effect for the median EDLP store of Wal-Mart’s presence is approx- imately $28,000 per week ($1.47 million annually). Note that this is a Wal-Mart effect specifically, not a competition effect more generally, because we have already controlled for the effect of the number of stores. There is also signif- icant heterogeneity across markets. From Table 8, for the stores in the 5th percentile, the effect of Wal-Mart entry can be as high as $48,000 per week ($2.5 million per year). These are typically larger, more isolated markets, in which Wal-Mart entry tends to result in especially high substitu- tion. The effect of competition from other supermarkets, as captured by the N−i variable, is also negative, as expected. At the median, the addition of another supermarket into the local market reduces revenues for an EDLP store by $9,580 per week (about $500,000 per year). Observing the effect of the share of other supermarkets in the local area that are also EDLP, we find mixed evidence. In some mar- kets, the effect is negative, suggesting stronger substitution, while in others, the effect is positive. A priori, it is dif- ficult to sign this effect. On the one hand, more EDLP stores in the local area implies stronger substitution and thus lower revenues. On the other hand, the presence of other chains of the same format may induce stores to tacitly soften price competition, enabling them to jointly sustain higher base prices. This can improve the revenue profile. Without detailed price data, it is difficult to drive deeper into these two stories. The main takeaway is that the data

Repositioning Dynamics and Pricing Strategy 765

Figure 3 REVENUE COMPONENTS OF EDLP

Intercept

F re

q u

en cy

F re

q u

en cy

0 500 1000 1500 2000

0 40

0 80

0 12

00 Wal-Mart in MSA

–80 –60 –40 –20 0 20

0 20

0 40

0 60

0

E(Share of Competitors Doing EDLP)

F re

q u

en cy

0 50 100

0 50

0 10

00 15

00

Number of Competing Stores

F re

q u

en cy

–15 –10 –5 0 5 0

20 0

40 0

60 0

Focus of Chain-EDLP

F re

q u

en cy

–50 0 50 100 150 200

0 20

0 60

0

reveal that the cross-store substitution effect does not dom- inate in several markets. Figure 3 also reveals some evidence for economies of

scope and scale on the demand side. In particular, super- markets that have a larger proportion of stores outside the local market doing EDLP also tend to earn more under EDLP. This effect is fairly large. At the median, the economies add approximately $11,100 per week ($577,000 per year) to revenues. These economies may arise from the fact that large chains may commit to doing EDLP across many markets (i.e., a size effect) or from the fact that doing EDLP across many markets may signal a consis- tent price image that has spillovers across markets (i.e., a scope effect). There is also evidence of fairly large size/scope effects (significant mass in the right tail), pre- sumably reflecting the higher revenues earned by the largest chains. Figure 4 presents analogous histograms for these effects

on revenues under PROMO pricing. It is worthwhile to compare the numbers for the effects on PROMO revenues with the effects on revenues under EDLP. The effect on revenues of having a Wal-Mart in the MSA under PROMO is also negative, as expected, but it is significantly lower

than under EDLP. At the median, the effect is a $20,500 reduction in PROMO revenues per week ($1.07 million annually). Comparing this with the effect of Wal-Mart pres- ence for EDLP stores, we observe that Wal-Mart has a 38% larger effect on EDLP supermarket revenues than PROMO ($1.47 million compared with $1.07 million). It is evident that the EDLP positioning of Wal-Mart leads to stronger substitution with other EDLP stores in the local area than with other PROMO stores. Also notable is the evidence for scale and scope economies under PROMO, which con- tributes approximately $39,200 per week ($2.03 million annually) in a median store market. These tend to be higher than those of EDLP. This is not surprising, as communi- cating a coherent and consistent EDLP policy might be more difficult than claiming to have intermittent promo- tions and sales. We also observe that while competition has a negative impact on both formats, the prevalence of EDLP competitors tends to hurt PROMO stores more on average.

Costs

Next, we discuss the results on the cost side of the model. We organize the discussion along similar lines to the rev- enue side, presenting histograms of totals first and then

766 JOURNAL OF MARKETING RESEARCH, DECEMBER 2012

Figure 4 REVENUE COMPONENTS OF PROMO

Intercept

F re q u en cy

0 500 1000 1500

0 5

00 1

00 0

2 00

0 3

00 0

Wal-Mart in MSA

F re q u en cy

-50 - 40 - 30 - 20 - 10 0

0 1

00 0

2 00

0 3

00 0

E(Share of Competitors Doing EDLP)

F re q u en cy

- 40 - 30 - 20 - 10 0 10 20 30

0 1

00 0

2 00

0 30

00 4

00 0

Number of Competing Stores

F re q u en cy

- 15 - 10 - 5 0 0

5 00

1

00 0

1 50

0 Focus of Chain-PROMO

F re q u en cy

- 200 - 150 - 100 - 50 0 50 100

0 5

00 1

50 0

2 50

0 3

50 0

of individual effects across markets. Complete estimation results appear in Appendixes A and B. Figure 5 presents histograms of the total fixed costs

incurred by incumbent supermarkets under EDLP and PROMO. Analogously, Table 9 presents the 5th, 50th, and 95th percentiles of these costs distributions. How should we interpret these costs? First, note that the fixed costs are estimated relative to the value of exiting, which has been normalized to zero. A negative fixed cost estimate indicates that the scrap value from exit was higher than incurring the fixed cost from continued operation under that particu- lar pricing format. Second, the revenue data are in $1000s per week. Thus, the fixed costs should be interpreted in the same units. At the same time, the discrete-choice model is estimated for data on two periods (1994 and 1998) that are separated by four years. Thus, the one-time switching costs should be thought of as borne over the four-year window. In Figure 5 and Table 9, we observe that the median

fixed cost is $293,750 per week under EDLP ($15.3 mil- lion annually) and $550,200 per week ($28.6 million annu- ally) for PROMO stores. Note that this should not be com- pared directly with the median revenues reported in Table 8,

Figure 5 ESTIMATED FIXED COSTS

Fixed Costs-EDLP

F re

q u

en cy

0 200 400 600 800 1000

0

10 00

25 00

Fixed Costs-PROMO

F re

q u

en cy

0 200 400 600 800 1000

0

20 00

4

00 0

Repositioning Dynamics and Pricing Strategy 767

Table 9

DISTRIBUTION OF ESTIMATED COSTS

5% 50% 95%

EDLP Intercept 201.20 349.04 510.47 Wal-Mart -128.93 -102.75 -84.44 E(Shaie of competitors EDLP) -24.33 -3.65 21.05 Focus of chain (EDLP) -226.71 -179.66 -138.66 Total fixed costs (for Nonswitchers) SI.12 293.75 465.87

PROMO Intercept 433.89 538.61 665.55 Wal-Mart 17.17 27.05 35.60 E(Shaie of competitors' EDLP) 15.65 37.69 61.78 Focus of chain (PROMO) -54.01 -7.48 26.90 Total fixed costs (for Nonswitchers) 440.S3 550.17 677.98 Switching cost (EDLP to PROMO) -30.36 11.13 112.13 Switching cost (PROMO to EDLP) 357.59 477,34 563.36

because the median store market for the revenue distribu- tion is not the same as the median store market for the cost distribution.

Table 9 also reports the switching cost estimates. We esti- mate the mediati cost of switching from EDLP to PROMO as $11,100 per week, which works out to be approximately

$2.3 million over a four-year horizon. We estimate the median cost of switching from PROMO to EDLP to be much larger, $477,300 per week, which works out to be approximately $99.3 million over a four-year horizon. This implies that the cost to the median EDLP store of switch- ing to PROMO is approximately 42 times higher than the cost to the median PROMO store to switch to EDLP. One issue is that the median PROMO store is different from the median EDLP store. To obtain a comparison, holding store type fixed, we also compute for each store the ratio of switching from PROMO to EDLP to the estimated cost of switching from EDLP to PROMO, The mean is 6.3, sug- gesting that the switch from PROMO to EDLP is approx- imately 6 times more costly for the average firm in the distribution.

To understand the relative comparison of the fixed costs to the switching costs, note that the scale of the fixed costs and the switching costs are expected to be different The fixed costs are scaled in relation to the revenue from staying versus exiting, while tire switching costs are scaled in rela- lion to the present-discounted revenues from staying versus exiting. This aspect along with the fact that tire model must rationalize that there are a large number of switches from EDLP to PROMO, but few from PROMO to EDLP, implies large, asymmetric switching costs.

Figure 6

COST COMPONENTS OF EDLP

I § Ll_ О

Intercept

200 400 600

E(Share of Competitors Doing EDLP)

800

Wal-Mart in MSA

Focus (EDLP)

Ì I I s

U- 9.

I' 8 s¡ Ş ? §

u- Я

-100 -50 -400 -350 - 300 -250 -200 -150 -100

Switching Cost

768 JOURNAL OF MARKETING RESEARCH, DECEMBER 2012

We now explore heterogeneity in fixed and switching costs across stores and markets. Analogous to the revenue results, we report in Figure 6 histograms across markets of the effect of Wal-Mart, the share of competitors doing EDLP, and the “EDLP focus” of the supermarket on fixed costs. We also report the distribution of estimated costs of switching to EDLP across markets. Figure 7 reports the same constructs for the costs of doing PROMO. Examining Figure 6, we find that the presence of Wal-

Mart in the supermarket’s MSA reduces fixed costs of oper- ation for the EDLP format. One interpretation of this result is that the presence of Wal-Mart lowers the costs of mar- keting an EDLP price positioning in a local market. For example, the presence of a Wal-Mart drives traffic into the local market, which reduces the costs of doing week-by- week advertising. Another interpretation is that the entry of Wal-Mart effectively educates consumers in the local market about the value of an EDLP positioning. The trade press reports anecdotal evidence consistent with this phe- nomenon. For example, when Wegmans (a regional super- market chain in the northeast) moved from PROMO to EDLP in anticipation of Wal-Mart’s entry, the supermar- ket made large investments in advertising and public rela- tions to justify this repositioning to consumers, while also investing in reeducating and retraining their workforce (at stores and warehouses) to be attuned with the new strategy (several more examples that have been documented in the

Figure 7 COST COMPONENTS OF PROMO

Intercept

F re

q u

en cy

300 400 500 600 700 800 900

0 20

00

4 00

0

Wal-Mart in MSA

F re

q u

en cy

0 20 40 60

0

2 00

0

50 00

E(Share of Competitors Doing EDLP)

F re

q u

en cy

0 50 100

0

20

00

4 00

0

Focus (PROMO)

F re

q u

en cy

- 150 - 100 - 50 0 50 100 150

0

2 00

0

5 00

0

Switching Cost

F re

q u

en cy

- 50 0 50 100 150 200

0

20 00

4 00

0

industry trade press are available on request in an unpub- lished appendix). Finally, Wal-Mart may reduce costs for everyone by putting pressure on suppliers and improving the overall distribution channel. This impact would be felt irrespective of pricing strategy. We also observe that the effect of the chain’s “focus” is

to reduce fixed costs, which is essentially another manifes- tation of a scope or scale economy on the cost side. The more the chain tends to do EDLP across the United States, the lower are its operating costs of running an EDLP super- market in a local market. This is intuitive and in line with expectations. Table 9 shows that the effect of these scope economies

on the cost side are significantly large: At the median of the distribution, the net effect of chain focus on EDLP posi- tioning is approximately $179,660 per week. We conjecture that EDLP cost savings are directly linked to a widespread adoption of the practice by the chain, and this is reflected in this estimate. A broad takeaway from these results is that scope economies in retailing operate over both revenues and costs. Considering the results on the PROMO side in Figure 7,

we observe that the presence of Wal-Mart in the local MSA tends to increase the fixed costs of doing PROMO. This likely reflects Wal-Mart’s overall impact, making it rela- tively more difficult to convincingly communicate the value

Repositioning Dynamics and Pricing Strategy 769

of PROMO and consequently increasing service, adver- tising, and other costs to help maintain the positioning. The effect of focus is also consistent with the results for PROMO from the revenue side: A strong focus on PROMO across the country leads to some scale or scope economies in local markets, but there is significant heterogeneity in how this plays out across stores.

We can summarize the results of the structural model as follows: Doing PROMO provides higher margins and revenues. At the median, PROMO pricing provides incre- mental revenue of approximately $6.2 million compared with EDLP. Whereas EDLP does offer lower fixed costs of operation, the cost of switching from PROMO to EDLP is estimated to be approximately six times larger than that of switching from EDLP to PROMO. Furthermore, competing with Wal-Mart under EDLP lowers revenues by much more than under PROMO. The 1990s were predicted by some as the decade of the EDLP format. These results add to our understanding of why EDLP adoption has been much more limited than predicted.

Robustness and Simulations

Common unobservables. Next, we discuss the robust- ness of our estimates to the presence of common, market- level unobservables. A potential concern involves selection issues arising from unobservables common across firms in a market, which might drive firms’ actions. This is a recurring but difficult to solve problem in the entry liter- ature: Markets that are more attractive due to an unob- served (to the econometrician) reason attract more firms (or more firms of a particular type), which, when not controlled for, leads to a counterintuitive finding that firms prefer to enter markets with more competition. To assess whether this is an issue, we reestimate our revenue regressions with market-level (MSA) random effects. Although we find that these random effects are useful in predicting revenues, the incremental impact on predicted counterfactual revenues is not large. The correlation between the predicted counterfac- tual revenues from the random effects specification and our chosen specification is greater than .9. However, including unobservables into revenues greatly complicates the struc- ture of the dynamic choice problem if these unobservables are included in the players’ information sets. Treating these unobservables as persistent shocks would make our first- stage CCP estimates inconsistent, and this inconsistency would transmit to the remaining stages as well. Given the qualitatively similar results on the revenue front and the econometric issues in treating these shocks as unobserved information, we choose to retain our simpler specification of the revenue functionals. What if Wal-Mart was everywhere? Finally, we use our

model and estimates to ask how the distribution of pric- ing formats would look if Wal-Mart expanded significantly across the United States. Our goal is to informally ver- ify that at the estimated parameters, the model predicts, consistent with actual evidence, that there will not be sig- nificant en masse switching by supermarkets into EDLP, as some had predicted in the early 1990s. We simulate a simple counterfactual by assuming that the industry state includes the presence of Wal-Mart in all markets. We then forward-simulate the markets from this initial state, allow- ing for entry, exit, and strategy changes based on the model estimates. We report the distribution under the steady state. Our steady state results suggest that adding Wal-Mart to

every market beginning from the 1998 conditions does push more market participants toward EDLP, but not overwhelm- ingly. In particular, the overall effect is on the order of an 18.8% increase in EDLP adoption across the entire United States (34.01% of active supermarkets chose to be EDLP in the counterfactual steady state, compared with 28.7% in the data). At the same time, market structure is also pushed toward higher concentration, because exits in the steady state are approximately 16.67%, versus the 15.4% observed in the data. Overall, the effect of Wal-Mart is to move the share of EDLP higher and to increase the exit rate. The simulations indicate that even when blanketed by Wal-Mart entry, it is unlikely that the U.S. market would tip toward EDLP.

CONCLUSIONS

This article makes three contributions. First, we draw attention to three salient features of repositioning decisions in marketing: that they involve long-term consequences, require significant sunk investments, and are dynamic in their impact. We illustrate that positioning decisions can be empirically analyzed as dynamic games to measure struc- tural constructs such as firm’s repositioning costs. Second, we cast empirical light on an age-old question in the mar- keting of consumer packaged goods: the costs and benefits of using EDLP versus PROMO. Despite the significant interest in this topic, a full accounting of the long-term costs and benefits of these strategies remains lacking in the literature. Our estimates add to the evaluation of either strategy and also identify the sources of heterogeneity in the relative attractiveness of either across markets. This increases understanding of the economics of the super- market industry and the determinants of long-term mar- ket structure. Third, we illustrate how observed switches combined with auxiliary postgame data (e.g., revenues, prices, sales) are useful in cleanly articulating the costs and benefits of repositioning in an environment with strategic interaction.

Our modeling approach has limitations and is based on assumptions that further research might aim at relax- ing. In particular, we highlight three potential avenues of improvement. First, the stage game could be extended to accommodate additional structural elements to better shed light on the source of revenue differences. For exam- ple, Beresteanu, Ellickson, and Misra (2007) employ a Bertrand-Nash stage game that allows them to recover price–cost margins and evaluate changes in consumer sur- plus. Although we do not have the appropriate data to afford such a specification here, other researchers might consider this option in the future. Second, we have focused on local market drivers of pricing strategy but incorpo- rated scale and scope economies to capture dependencies of strategies across markets in a limited way. A significant, but challenging, extension to the current work would be to fully accommodate the joint choice of pricing strategy across markets. This would require the solution of a daunt- ing dynamic network game, which is outside the scope of our current analysis. Finally, extending our framework to allow for rich layers of persistent unobserved heterogeneity would be an important direction forward. This is an active area of frontier econometric research. We intend to address some of these extensions in further study. We hope our cur- rent work spurs further interest in the dynamics of pricing and repositioning in the field.

770 JOURNAL OF MARKETING RESEARCH, DECEMBER 2012

Appendix A REVENUE REGRESSION ESTIMATES

PROMO EDLP

Variable Interactions Estimate SE Estimate SE

Intercept Constant −14304917 3804107 207606 3706958 pop 00018 00003 00016 00003 hhsize 2904688 1309660 −2703737 1308402 p_black −4205423 2005609 −9107737 2601777 p_urban 3806491 1506274 3506342 1504588 p_hisp 8602794 3200506 −10504643 4204421 h_inc 00019 00004 00023 00004 size 701859 01294 705947 01838 Tstores 00099 00012 −00021 00017

Wal-Mart Constant 304924 1809370 4000459 2608277 pop −00001 00001 00000 00002 hhsize 01696 700774 −905970 1006311 p_black −300332 1201695 205743 1600800 p_urban −800267 803187 −3204061 1200374 p_hisp −3608025 1402069 3905778 2909141 h_inc −00001 00002 −00007 00004

E(Comp EDLP) Constant 4504532 2303660 −1108019 3509686 pop 00000 00002 −00005 00003 hhsize −1600171 804788 604146 1304419 p_black 1802148 1206664 405995 1909579 p_urban −1502756 102790 701724 1104828 p_hisp 2101111 2004637 12901107 3203871 h_inc 00002 00002 −00001 00005

Focus PROMO/EDLP Constant 18502156 4204260 −12008199 4505139 pop −00006 00003 −00007 00004 hhsize −7104293 1507550 5306338 1607937 p_black 4504121 2308604 1208066 2606348 p_urban −900338 1508865 2703170 1703471 p_hisp −4304603 3404132 12007215 3808835 h_inc 00006 00004 −00007 00006

Number of Competing Stores Constant 304682 404740 109628 602802 pop −00001 00000 00000 00000 hhsize −09573 103945 −308233 202772 p_black −204410 202300 1506963 307920 p_urban −105155 109349 −204647 205457 p_hisp 800419 305968 1702897 805114 h_inc −00001 00000 00000 00001

Appendix B COST REGRESSION ESTIMATES

PROMO EDLP

Variable Interactions Estimate SE Estimate SE

Intercept Constant −4008906 3101996 48605116 2904313 pop 00004 00003 00011 00003 hhsize 3101818 1106570 −6807441 1006749 p_black −1609209 2201849 1803064 1805656 p_urban 14709501 1206642 2306120 1208810 p_hisp 7803481 2804861 12102934 3705116 h_inc 00023 00003 00027 00003 size 400016 00642 302505 00654 Tstores −00475 00014 −00329 00012 VI 505494 108091 1508291 107537

Wal-Mart Constant 3601164 3102078 7100421 2409721 pop −00005 00002 00003 00002 hhsize −701406 1101883 −1609365 900943 p_black 4302456 2001492 206052 1509359 p_urban −1001861 1509323 −1506697 1004204 p_hisp −4408020 3102732 108502 2003777 h_inc 00002 00003 00002 00003

E(Comp EDLP) Constant 11904247 2102418 10007033 1806769 pop 00006 00002 00008 00002 hhsize −3703200 802631 −2206546 609649 p_black −03581 1504459 2501018 1301668 p_urban −3505122 905463 −2202237 709338 p_hisp 7703049 1909969 3906135 1703118 h_inc 00001 00003 −00003 00002

Focus PROMO/EDLP Constant 13108787 3604711 −22309241 3209406 pop −00002 00003 −00001 00003 hhsize −8400079 1306877 9000610 1108279 p_black 7602809 2607638 −8203544 2109314 p_urban −1008516 1404771 5407418 1307870 p_hisp 08312 3608522 −14500459 4007833 h_inc 00010 00004 −00015 00004

Switching Costs Constant 76808435 2702031 14608000 2504979 pop 00003 00002 −00008 00002 hhsize −6304806 1003650 −500595 906423 p_black −14909769 1806848 4101803 1701722 p_urban −1809121 1109473 −3609124 1100398 p_hisp −4905052 2601670 5206532 3008164 h_inc −00001 00003 00006 00002 WalMart in MSA −10702976 409895 2907865 407857 E(Comp EDLP) −1705035 401388 −06812 405161 Focus −11409448 700153 −12305959 608326

Repositioning Dynamics and Pricing Strategy 771

REFERENCES Ackerberg, Daniel, C. Lanier Benkard, Steven Berry, and Ariel Pakes (2005), “Econometric Tools for Analyzing Market Out- comes,” in Handbook of Econometrics, Vol. 6A, J. Heckman and E. Leamer, eds. Amsterdam: North-Holland, 4171–4279.

Aguirregabiria, Victor and Pedro Mira (2007), “Sequential Esti- mation of Dynamic Discrete Games,” Econometrica, 75 (1), 1–53.

and (2010), “Dynamic Discrete Choice Structural Models: A Survey,” Journal of Econometrics, 156 (1), 38–67.

Ailawadi, Kusin, Donald R. Lehmann, and Scott A. Neslin (2001), “Market Response to a Major Policy Change in Marketing Mix: Learning from P&G’s Value Pricing Strategy,” Journal of Mar- keting, 65 (January), 44–61.

Altug, Sumru and Robert A. Miller (1998), “The Effect of Work Experience on Female Wages and Labour Supply,” Review of Economic Studies, 65 (1), 45–85.

Anderson, Erin T. and Duncan Simester (2010), “Price Stickiness and Customer Antagonism,” Quarterly Journal of Economics, 125 (2), 729–65.

Arcidiacono, Peter, Patrick Bayer, Jason Blevins, and Paul B. Ellickson (2010), “Estimation of Dynamic Discrete Choice Models in Continuous Time,” ERID Working Paper No. 50, Duke University.

and Paul B. Ellickson (2011), “Practical Methods for Esti- mation of Dynamic Discrete Choice Models,” in Annual Review of Economics, Vol. 3, K. Arrow and T. Bresnahan, eds. Palo Alto, CA: Annual Reviews, 363–94.

and Robert A. Miller (2012), “CCP Estimation of Dynamic Discrete Choice Models with Unobserved Hetero- geneity,” Econometrica, 79 (6), 1823–67.

Bajari, Patrick, C. Lanier Benkard, and Jonathan Levin (2007), “Estimating Dynamic Models of Imperfect Competition,” Econometrica, 75 (5), 1331–70.

Basker, Emek and Michael Noel (2009), “The Evolving Food Chain: Competitive Effects of Wal-Mart’s Entry into the Super- market Industry,” Journal of Economics and Management Strat- egy,18 (4), 977–1009.

Bell, David R. and Christian Hilber (2006), “An Empirical Test of the Theory of Sales: Do Household Storage Constraints Influ- ence Consumer and Store Behavior?” Quantitative Marketing and Economics, 4 (2), 87–117.

and James Lattin (1998), “Shopping Behavior and Con- sumer Preference for Retail Price Format: Why “Large Basket” Shoppers Prefer EDLP,” Marketing Science, 17 (1), 66–88.

Beresteanu, Arie, Paul B. Ellickson, and Sanjog Misra (2007), “The Dynamics of Retail Oligopoly,” working paper, Simon Graduate School of Business, University of Rochester.

Bresnahan, Timothy F. and Peter C. Reiss (1991), “Empirical Models of Discrete Games,” Journal of Econometrics, 48 (1/2), 57–82.

(1994), “Measuring the Importance of Sunk Costs,” Annales d’Economie et de Statistique, 34, 181–217.

Dixit, Avinash and Robert Pindyck (1994), Investment Under Uncertainty. Princeton, NJ: Princeton University Press.

Draganska, Michaela, Sanjog Misra, Victor Aguirregabiria, Pat Bajari, Liran Einav, Paul Ellickson, et al. (2008), “Discrete Choice Models of Firms: Strategic Decisions,” Marketing Let- ters, 19 (3/4), 399–416.

Dubé, Jean-Pierre, Günter J. Hitsch, and Peter E. Rossi (2009), “Do Switching Costs Make Markets Less Competitive?” Jour- nal of Marketing Research, 46 (August), 435–45.

Ellickson, Paul B. (2007), “Does Sutton Apply to Supermarkets?” RAND Journal of Economics, 38 (1), 43–59.

and Paul L.E. Grieco (2013), "Wal-Mart and the Geogra- phy of Grocery Retailing," Journal of Urban Economics, forth- coming.

, Stephanie Houghton, and Christopher Timmins (2010), “Estimating Network Economies in Retail Chains: A Revealed Preference Approach,” NBER Working Paper 15832.

and Sanjog Misra (2008), “Supermarket Pricing Strate- gies,” Marketing Science, 27 (5), 811–28.

and (2011), “Estimating Discrete Games,” Mar- keting Science, 30 (6), 997–1010.

and (2012), “Enriching Interactions: Incorporat- ing Outcome Data into Static Discrete Games,” Quantitative Marketing and Economics, 10 (1), 1–26.

Ericson, Richard and Ariel Pakes (1995), “Markov-Perfect Indus- try Dynamics: A Framework for Empirical Work,” Review of Economic Studies, 62 (1), 53–82.

Goettler, Ronald and Karen Clay (2011), “Tariff Choice with Con- sumer Learning and Switching Costs,” Journal of Marketing Research, 48 (August), 633–52.

Hartmann, Wesley R. and V. Brian Viard (2008), “Do Frequency Reward Programs Create Switching Costs? A Dynamic Struc- tural Analysis of Demand in a Reward Program,” Quantitative Marketing and Economics, 6 (2), 109–137.

Ho, Teck-Hua, Christopher S. Tang, and David R. Bell (1998), “Rational Shopping Behavior and the Option Value of Variable Pricing,” Management Science, 44 (12), 145–60.

Hoch, Steven J., Xavier Drèze, and Mary E. Purk (1994), “EDLP, Hi-Lo, and Margin Arithmetic,” Journal of Marketing, 58 (October), 16–27.

Holmes, Thomas (2011), “The Diffusion of Wal-Mart and Economies of Density,” Econometrica, 79 (1), 253–302.

Hotz, V. Joseph and Robert A. Miller (1993), “Conditional Choice Probabilities and the Estimation of Dynamic Models,” Review of Economic Studies, 60 (3), 497–529.

Jia, Panle (2008), “What Happens When Wal-Mart Comes to Town: An Empirical Analysis of the Discount Retail Industry,” Econometrica, 76 (6), 1263–1316.

Lal, Rajiv and Ram Rao (1997), “Supermarket Competition: The Case of Every Day Low Pricing,” Marketing Science, 16 (1), 60–81.

Levy, Daniel, Mark Bergen, Shantanu Dutta, and Robert Venable (1997), “The Magnitude of Menu Costs: Direct Evidence from Large U.S. Supermarket Chains,” Quarterly Journal of Eco- nomics, 112 (3), 791–825.

Matsa, David A. (2011), “Competition and Product Quality in the Supermarket Industry,” Quarterly Journal of Economics, 126 (3), 1539–91.

Mulhern, Francis J. and Robert P. Leone (1990), “Retail Promo- tional Advertising: Do the Number of Deal Items and Size of Deal Discounts Affect Store Performance?” Journal of Business Research, 21 (3), 179–94.

Orhun, A. Yesim (2012), “Spatial Differentiation in the Supermar- ket Industry: The Role of Common Information,” Quantitative Marketing and Economics, forthcoming.

Pakes, Ariel and Oliver Linton (2001), “Nonlinear Methods for Econometrics,” working paper, London School of Economics.

772 JOURNAL OF MARKETING RESEARCH, DECEMBER 2012

, Michael Ostrovsky, and Steven Berry (2007), “Simple Estimators for the Parameters of Discrete Dynamic Games (with Entry/Exit Examples),” RAND Journal of Economics, 38 (2), 373–99.

, J. Porter, Kate Ho, and Joy Ishii (2005), “Moment Inequalities and Their Applications,” working paper, Depart- ment of Economics, Harvard University.

Pesendorfer, Martin (2002), “Retail Sales: A Study of Pricing Behavior in Supermarkets,” Journal of Business, 75 (1), 33–66.

and Philipp Schmidt-Dengler (2008), “Asymptotic Least Squares Estimators for Dynamic Games,” Review of Economic Studies, 75 (3), 901–928.

Ries, Al and Jack Trout (1981), Positioning: The Battle for Your Mind. New York: Warner Books, McGraw-Hill Inc.

Salop, Steven and Joseph E. Stiglitz (1982), “The Theory of Sales: A Simple Model of Equilibrium Price Dispersion with Identical Agents,” American Economic Review, 72 (5), 1121–30.

Singh, Vishal, Karsten Hansen, and Robert Blattberg (2006), “Market Entry and Consumer Behavior: An Investigation of a Wal-Mart Supercenter,” Marketing Science, 25 (5), 457–76.

Slade, Margaret (1998), “Optimal Pricing with Costly Adjust- ment: Evidence from Retail-Grocery Prices,” Review of Eco- nomic Studies, 65 (1), 87–107.

Sobel, Joel (1984), “The Timing of Sales,” Review of Economic Studies, 51 (3), 353–68.

St-James, Yannik (2001), “Retail Brand Repositioning: A His- torical Analysis,” in Proceedings of Conference on Historical Analysis and Research in Marketing, Vol. 10, 164–75.

Sweeting, A. (2011), “Dynamic Product Positioning in Differen- tiated Product Industries: The Effect of Fees for Musical Per- formance Rights on the Commercial Radio Industry,” working paper, Duke University.

Varian, Hal R. (1980), “A Model of Sales,” American Economic Review, 70 (4), 651–59.

Vitorino, Maria Ana (2012), “Empirical Entry Games with Com- plementarities: An Application to the Shopping Center Indus- try,” Journal of Marketing Research, 49 (April), 175–91.

Zhu, Ting, and Vishal Singh (2009), “Spatial Competition and Endogenous Location Choices: An Application to Discount Retailing,” Quantitative Marketing and Economics, 7 (1), 1–35.

Copyright of Journal of Marketing Research (JMR) is the property of American Marketing Association and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.