You are on page 1of 22

Grocery Market Pricing and the New Competitive Environment

Purdue University

Purdue University

This article seeks to examine the effect of changes in retail-grocery market conditions upon consumer prices and the response offood retailers to these conditions. We explore these issues for two different product groups. One is composed primarily of branded, dry and frozen groceries (P1) while the second largely represents produce, meat, and perishable products (P2 ). Our statistical model finds that the effect of the of these new market conditions is often markedly sensitive to product type. Competitive conditions affect both price measures, but entry by warehouse and similar grocery formats surprisingly lower perishable prices more than those for staple goods. We also observe that frequent price changes in a market serves to reduce food prices, suggesting that promotional activities in response to new format competition are beneficial for consumers. There is strong evidence of a fast food-supermarket rivalry, but we are unable to identify its specific nature. Operating costs significantly effect prices for staple items, but have virtually no evident influence on perishable.

The Supermarket R e v o l u t i o n - - t h e replacement of small Grocery stores by large, multidepartment grocery stores----came to an end in the 1970s (Marion et al., 1986). Since that

James K. Binkley, Department of Agricultural Economics, Purdue University, West Lafayette, Indiana; and John M. Connor, Department of Agricultural Economics, Purdue University, West Lafayette, Indiana. Purdue Journal Paper 14569.
Journal of Retailing, Volume 74(2), pp. 273--294, ISSN: 0022-4359 Copyright ~ 1998 by New York University. All rights of reproduction in any form reserved.



Journal of Retailing Vol. 74, No. 2 1998

time, supermarkets have come to face heightened rivalry from a proliferation of retail food outlets. The 1987 annual report of Progressive Grocer declared that "...the supermarket industry is moving faster to accommodate changes in consumer shopping and eating patterns." The traditional supermarket design is being supplemented by warehouse stores, supercenters, and combination stores, often incorporating food courts. Investments in such retail formats imply a recognition that conventional supermarkets face a more diverse set of market rivals. With a changing market environment, supermarket pricing practices may be changing as well. The purpose of this study is to examine long-run supermarket pricing across different U.S. markets within this new competitive environment. Pricing practices in the retail grocery industry have long been of interest both from a positive and normative standpoint. In either case, the focus is typically not costs, but environmental factors that can cause differences in price given costs. If such factors exist, then firms with similar costs can charge different prices in accordance with these conditions. This enables the practice of some form of price discrimination if markets are sufficiently segmented as to minimize arbitrage between them. Indeed, a useful way to distinguish between normative (or prescriptive) and positive studies is the nature of the market segmentation implied by the studies. Prescriptive studies are most often concerned with identifying rules for optimal pricing by the grocery firm, usually at the store level. Thus, the emphasis is on factors associated with price differences across product categories. The major factors here are customer demographics and incomes. With this focus, the degree of competition faced by the store is less important since it is seldom viewed as having differential effects on demand by category (a point of importance herein). Competitors are generally considered to be other supermarkets selling similar goods. Only if the question under study involves optimal pricing by the multi-store firm must consideration be given to price levels at stores facing differences in competition. In contrast, positive studies are almost exclusively concerned with pricing under different degrees of firm concentration. Thus, interest is with differences in over-all grocery price across geographic markets that vary in terms of competitive intensity (as measured by market shares or concentration). Demographic factors are occasionally included, but only to the extent they measure differences in demand in different geographic areas. Pricing at the level of product categories is not considered. In short, in one case the emphasis is on product types. In the second, it is market types. This study is a blend of these. Although the central topic is the role of competitive influences on price, we incorporate elements of both market level pricing and category pricing. The principle goal of our paper is to develop and test a model to accomplish two ends. We evaluate the role of traditional competitive factors on market-level retail grocery prices and incorporate the possibility that the changed competitive climate has a significant influence on pricing. We believe the changing retail food landscape, in which grocery-product competition is no longer confined to supermarkets, makes this view more appropriate than previously. This is supported by recent theoretical developments as well as by press reports. To examine these issues, we employ a novel data set, consisting of a large sample of metropolitan areas with widely varying characteristics. A market-level analysis has certain limitations. Factors specific to individual stores cannot be studied. In addition, measures remain aggregations of individual item prices. This

Grocery Market Pricing and the New Competitve Environment


condition remains here despite our distinction among product types, a distinction unlike previous studies at the market level. But, these limitations are dictated by our decision to focus on the changing, market environment of U.S. supermarkets. Our intent is to study the market characteristics that explain long-run average retail prices of two broad classes of grocery goods. We seek to examine what supermarkets are doing, which is not necessarily what they ought to do. Although we can make no assessment of store or industry performance, our results may be useful for those with that purpose. The inclusion of the rising role of restaurants in this study is the first attempt to assess the impact of this significant change in food retailing. This trend has "increasingly disturbed" industry leaders, according to the Progressive Grocer report. Its survey found that two-thirds of store managers rated the competitive threat from fast food outlets as either moderate or serious. There is also credible evidence that the low-priced segment of the restaurant industry, since ultimately it serves much the same purpose as grocery retailers, provides the strongest price competition among foodservice types. In a study of demand for food-away-from-home, Hiemstra and Kim (1995, p. 30) concluded that " food is more of a commodity than most other food-away-from-home and it competes more strongly with food purchased at grocery stores than with other foodaway-from home." Our model allows for foodservice competition, and we indeed find that metropolitan-area grocery prices are affected by competition from the fast food segment of the restaurant industry, though not always in the expected fashion.

LITERATURE REVIEW Positive studies have primarily been within the ken of industrial-organization (IO) economics. IO economists have long considered the question of whether retailing was an imperfectly competitive industry. Although some have reasoned that most retailing, including large-scale grocery retailing, is workably competitive (Adelman, 1948, Stigler, 1950), most early writers agreed with Smith (1937), who judged retailing to be monopolistically competitive. This arises due to consumer search costs and spatial differentiation, a model more formally analyzed by Salop and Stiglitz (1977) and Benson and Faminow (1985). Many other economists believed grocery retailing to be essentially oligopolistic in its pricing behavior (Baumol et al., 1964, Holdren, 1968, and Marion et al., 1979). There are four published cross-sectional, empirical studies of supermarket price indexes in the IO tradition. All measure competitive rivalry with a metropolitan-area sales concentration index, and three of the four also include company market share. The first study used extensive price-check data, generated by grocery retailers operating in 36 cities, to develop a market-basket price index of 94 branded food (excluding meat and produce) items (Marion et al., 1979). Both four-firm concentration (C4) and firm market share were found to be positively related to the index. Cotterill (1986) verified these results, also using subpoenaed price data, for a sample of 35 stores in 18 mostly small, isolated Vermont towns and cities. Cotterill and Harper (1995) further verified the positive concentration-price relationship for a sample of 34 local markets in and around Arkansas. A fourth study, drawing on highly aggregated retail food price indexes published for only 18 large U.S. metropolitan


Journal of Retailing Vot. 74, No. 2 1998

areas by the Bureau of Labor Statistics, also found that concentration was positively related to food prices (Lamm, 1981). Prescriptive studies fail into the province of those that study supermarket management. Most are category pricing studies that have their foundation in the third degree price discrimination model (Blattberg and Neslin, 1990; Kim et ai., 1995). This model, in common with much IO economics, assumes monopolistic competition. It recognizes that supermarkets have some localized monopoly power due to enterprise reputation and spatial differentiation. This causes consumers to incur search costs and costs of inconvenience and leads to one-stop shopping (Katz, 1984; Bliss, 1988; Holmes, 1989). Except for one early model (Holton, 1957), the price discrimination models demonstrate that retail price margins are greater for products with inelastic demands. The price elasticity of demand incorporates information about consumer buying habits in the trading area. Basing arguments on Becker's (1965) model of household economics, various writers have hypothesized that retail demand elasticity may be related to age, education, income, frequency of product purchase, car ownership, and time of week. Many of these factors reflect differences among households in price-searching effort. Empirical studies have found retail price responsiveness to be related to demographic factors, hut the results are sometimes inconsistent. Nine panel-data studies reviewed by Hoch et al. (1995) found price responsiveness positively associated with age, education level, household size, wealth, car ownership, and single-earner households. In their own study, they found that price responsiveness in 18 grocery product categories was generally positively related to family size, minority ethnic composition, and income, and negatively related to education and household wealth. Different price markups can also arise when prices of selected items are used to create a store price image, or "price signaling." As reported in Dickson and Urbany (1994), a survey of store managers found they "...believed consumers most frequently compare store prices on milk, meat (e.g., ground beef, chicken), produce, and soda" (p. 18). With signaling, markups no longer depend solely on product characteristics. For example, in the absence of signaling, stores might view commodities with limited substitutes, such as milk, to be relatively price inelastic, implying a relatively large markup. However, if managers believe milk prices are of special importance in the store-selection decision, from the store's perspective milk demand will be considered to be highly elastic and carry a low markup. It may possibly become a loss leader. Under category pricing, elasticities are more or less an objective reality. With signaling, elasticities depend upon the store management's subjective views concerning consumer reactions and upon the nature of store competition. Expected outcomes can clearly differ under these two cases. We will consider this in our discussion of empirical results. Price signaling concentrates on consumer choice among stores rather than choice among products within a store. Still, as in most IO studies, the market is viewed as unique, i.e., all supermarkets are in direct competition with one another, but other types of retailers have no effect on their pricing decisions. A more sophisticated view reflecting realistic conditions in the late 1980s recognizes a more graduated set of competitors. The most intense price competition for a given grocery store comes from stores offering the same array of goods in the same trading area (Cassady, 1962). Less intense price rivalry may be generated by neighborhood groceries, convenience stores, warehouse stores, or grocery stores in

Grocery Market Pricing and the New Competitve Environment


adjacent trading areas. Significant, but weak price competition may arise from gasoline stations, drug stores, discount department stores, and food service retailers. Few studies have explicitly incorporated these other, retail rivals in empirical models of supermarket price responsiveness. Hoch et al. (1995) is one exception. They developed four competitive variables to explain store-level price elasticities of 18 branded grocery products. They found that the size of warehouse stores in the trading area increased the elasticity of demand, while the distance from such stores (including those outside the immediate trading area) negatively affected responsiveness of demand. Cotterill and Harper (1995) also found that the presence of warehouse-type stores significantly reduced overall market grocery prices. Competition from alternative retail forms expands possibilities of price discrimination, since different types of consumers may prefer different forms. As Lal and Matutes (1989, p. 532) state, "...multimarket rivalry substantially alters that nature of competition," especially when there are multiple goods. They develop a duopoly model with two goods and two consumer types appropriate to this question. This model is important for our study and we now illustrate it with a stylized case. Consider a supermarket with some degree of spatial monopoly-induced market power selling two goods, a necessity G 1 and a convenience good G2, to two groups of consumers, "rich" and "poor." The poor consumers purchase only G1 and have an elastic demand. The rich purchase both and are not price sensitive. Under these conditions, the optimal price for the GI category would exploit the different demands the store faces: the store would practice second degree price discrimination and charge a lower price to the poor. This is not possible, however, because the two market segments cannot be separated (except imperfectly, e.g., with coupons). Hence, the store will charge the same GI price, determined by both elasticities, to all consumers. The result would be a price between the two that would obtain under price discrimination. Now suppose a new store enters the market. If the entrant is identical to the incumbent, prices for both goods would be expected to fall (at least in the absence of collusion). However, suppose the new store is a low-cost, warehouse store, selling only G1. With lower costs, it will set a G1 price below that of the incumbent All the poor consumers (who consume no G2) will then migrate to the new store. In this case, the direction of the G1 price response by the incumbent supermarket is unpredictable. Attempts to regain poor customers by matching the entrant' s price is not a viable long-run response since it implies pricing below cost. Any higher price will not entice poor consumers back. Hence, the optimal price for G1 is determined purely by the elasticity of G1 demand by the rich. Although this elasticity may be higher than before (given the warehouse penetration), it may still be optimal for the traditional supermarket to increase the price of G 1 if the rich want to avoid the costs of shopping at two stores. The warehouse store has thereby segmented the market in a way that permits the traditional supermarket price to depend solely on the demand exercised by rich customers. As a consequence, the magnitude and direction of the supermarket's G 1 response depends upon three factors: pre-entry level of G1 price, the warehouse price, and the elasticity of demand by the rich, all of which are case-specific. An optimal response might also include measures to increase the demand inelasticity of the high income consumers, such as increasing service levels.


Journal of Retailing Vol. 74, No. 2 1998

That this model appears to capture an important aspect of current food retailing is illustrated in a recent Wall Street Journal (1997) article on supermarket response to supercenter competition. This article notes that rather than lowering prices to new competition, Supermarket chains are ... expanding and remodeling their stores--they are also promoting the quality and freshness of their perishables. Independent supermarkets are pooling their resources to finance better advertising and store improvements. Food retailers say these methods typically have been more effective than price cutting. (p. B 11 )

Similar evidence is provided by a recent study by Messinger and Narasimhan (1997). They found that the expansion in supermarket size and the increase in the number of items carried is associated with higher, not lower, operating costs. Their provisional conclusion is that the observed changes in store type are not to achieve scale economies but to provide one-stop shopping, a response to consumer demand for time-saving convenience. As in the case of category pricing, complications arise in the Lal and Matutes framework when the superrnarket's pricing strategy includes signaling. If pre-entry G1 prices were low (i.e., low relative to no signaling), one would expect these prices, if anything, to rise. Here, signaling with G1 plays into the strength of the warehouse store. To the extent that signaling is continued or adopted, we would also expect G2 price to fall. The incumbent may set a G1 price considerably above the new competitor's, and lower its G2 price. 13y this, it hopes to call attention to G2 goods and attract G2 consumers, who (due to the cost and inconvenience of visiting two stores) then remain to purchase G1, despite the higher price.


The model employed in this analysis was of the general form:

P =f(R,C,M),
where P is a measure of one of two retail food price categories in a given market (as discussed below). R is a vector of variables explicitly measuring the competitive climate of the retail food market. M is a vector of factors characterizing metropolitan areas, factors which in themselves have no direct effect on competition (i.e., city size) but which can influence market price. C a vector of supply-side or cost factors for sellers. The majority of the variables are measured at the SMA level, but a few refer to somewhat larger grocery marketing regions. The functional form is a simple linear equation. Within R we include a standard measure of concentration, but non-traditional factors, such as the measure of restaurant activity noted above, are incorporated as well. Our primary concern is the effect of the competitive climate on pricing. The city and cost variables, although they can yield

Grocery Market Pricing and the New Competitve Environment


insights into the pricing behavior of retail grocery firms, mainly serve as control variables. The data involves a large variety of cities having wide geographic coverage, which makes cost differences likely.

Dependent Variable and Sample

Supermarket prices were taken from a quarterly commercial publication sold by ACCRA (1988a), the research arm of the U.S. Chamber of Commerce. ACCRA's major activity is collecting, tabulating, checking, organizing and publishing the Inter-City Cost of Living Index. In their words, "This survey ... published quarterly since 1968, is the only generally available source of data on living cost differentials among U.S. urban areas" (p. 1.1). The primary purpose is to provide information on a market basket of goods and services typical of the purchases by mid-management executive households with an eye toward assisting firms in relocating employees to another city. One previously published empirical analysis has employed intercity ACCRA grocery price data (Chevalier, 1995). ACCRA reports prices on 59 individual items, 27 of which are grocery items. Enumerators are personnel from local Chambers of Commerce for whose participation is strictly

TABLE I Ten Highest and Ten Lowest Price Correlations

Items Correlations
.652 .638 .635 .621 .587 .587 .563 .561 .557 .548

A. Highest
Canned tuna - margarine Sugar - Crisco Canned tuna - Crisco Laundry detergent - baby food Crisco - baby food Canned tuna - laundry detergent Coffee - laundry detergent Parmesan cheese - tissue Canned tuna - baby food Canned peas - tissue

B. Lowest
Potatoes - laundry detergent T-bone steak - laundry detergent T-bone steak - coffee M i l k - laundry detergent Hamburger - coffee Potatoes - coffee Milk - canned tomatoes Whole chicken - milk Lettuce - laundry detergent Whole chicken - lettuce -.376 -.357 -.344 -.301 -.281 -.277 -.265 -.259 -.243 -.241


The total number of correlations is 378.


Journal of Retailing Vol. 74, No. 2 1998

voluntary. Thus, cities reporting in one quarter may be absent in the following quarter. ACCRA (1988b) provides very strict guidance concerning the items and methods of data collection. For example, one of the grocery items is described as "18 oz. Kellogg's Corn Flakes or Post Toasties." A minority of items are generic (T-bone steak, iceberg lettuce, etc.) or likely to be store brands (bread, frozen corn). Enumerators are instructed to obtain prices from at least five supermarkets; doing so on Thursday, Friday, or Saturday; using the lowest price at each store, exclusive of coupons. These are then averaged to obtain the price reported for the item for the city. We used 26 of the 27 grocery items, omitting only cigarettes. Because the concentration data are based on 1987 census data, we perforce centered attention on that year. To model long-run prices, we averaged ACCRA price data from the Spring quarter for the three year period 1986-1988. There were 153 cities reporting in all three quarters, and all price measures were based on three year averages for these cities.

Correlations between First Two Principal C o m p o n e n t s and Prices
Price P1 Correlation P2

Canned, 61/2oz., in oil* Babyfood 41/2oz. jar, strained vegetable, lowest price Margarine, 1 lb., stick* Shortening, 3 lb. can, all-vegetable oil* Tissue, 175-count box* Laundry detergent, 42 oz.* Parmesan, grated, 8 oc. canister* Canned peas, 17 oz. can* Frozen orange juice, 12 oz. can* Sugar, 5 lb., can or beet, lowest price Frying chicken, whole, lowest price Eggs, dozen, grade A, large Coffee, 13 oz. can, vacuum* Canned tomatoes, 141/2 oz. can* Bread, white sliced, lowest price Bacon, 12 oz., rashers* Corn Flakes, 18 Oz.* Frozen corn, 10 oz. package, whole kernel, lowest price Canned peaches, 29 oz. can, whole or slices* Cola, 2 liter, excluding deposit* Bananas Milk, ~/2gallon, whole Lettuce Potatoes, 10 lb. sack, white or red, lowest price Hamburger, lowest price T-bone steak, USDA Choice

.297 .272 .293 .270 .258 .258 .255 .253 .250 .249 .233 .220 .193 .178 .169 .145 .113 .107 .097 .083 .092 -.062 -.063 -.087 .074 -.043

.018 -.034 .019 .155 .016 -.234 .013 -.034 .172 .201 -.027 .049 -.207 -.058 -.005 .134 -.069 .080 -.010 -.095 .264 .335 .336 .385 .393 .403

*One or more major brands specified. Items without asterisk may include unbranded or private label products.

Grocery Market Pricing and the New Competitve Environment


The final sample was composed of the 95 cities for which we had measures for all other variables. With only one observation per city, standard OLS estimation was used as opposed to a longitudinal method based on panel type data. There are numerous, technically an infinity, combinations of 26 prices that could be used to construct price measures. At one extreme is a single price index, which we investigated at an early stage of the study but for a reason noted presently do not report. At the other is the use of individual prices, an approach we rejected due to the unworkability of 26 models. Between these are indices of prices grouped on the basis of some criterion. We adopted this approach. However, lacking any firm basis for choosing indices, we relied on databased methods. We began by examining price correlations. If supermarkets place similar price markups on all items, then cities with high prices for one item would tend to have high prices for all items. Prices would be highly positively correlated. This is not what we found. The 26 ACCRA grocery items generate 378 unique price correlations. We ranked these, finding the results somewhat surprising given the well-articulated national market for food. In Table 1 we present the ten largest and ten smallest. The largest is an unexpectedly low 0.65; the smallest is -0.38! In all, only 50% were significantly (0.05) positive. This result demonstrates that firms indeed do not uniformly set prices and that a single-index of prices has limitations. There is an indication in the table that when pricing commodities, an important characteristic is whether they are predominantly prepackaged, branded items generally found in the "dry grocery" or "health and beauty aids" departments. In our sample, these prices tend to be positively correlated among themselves, and negatively associated with prices of items sold unbranded, especially fresh items. This is the only pattern discernible in the correlations in Table 1 and in those not presented. This suggests that combining items by categories ("produce," "dairy," etc.) will not serve present purposes. We thus chose to rely upon correlational differences among the items and employed principal components for this purpose. We selected this method because it results in ordered, orthogonal linear combinations of the original variables, a characteristic that makes them useful for our analysis. If there are groups of items with pricing patterns similar within groups, but different across groups, members of different groups may be associated with different principal components. These associations are measured by "loadings" of the original variables on the components. Principal components were constructed for the 26 prices. The first principal component (hereafter P1) accounted for 27 percent of the price variation. A percentage this small confirms that the overall correlation among the prices is not particularly high. The cumulative total variation captured by the first two components is 38%, and the first three account for 47%. From there the cumulative variance increases in small, slowly decreasing increments, further evidence of low price correlations. As expected, the principal components are related to commodity groups that share identifiable characteristics (Table 2). The first component tends to have relatively large positive correlations with prepackaged, branded, dry grocery products listed near the top of the table; the second (P2) primarily with fresh meat, produce, and milk. The pattern for the second is especially strong: its correlation with many of the branded goods is often zero or


Journal of Retailing Vol. 74, No. 2 1998

negative. Importantly it is positively correlated with items previously noted as likely to be used as price signals, the three above-mentioned in particular. This price partitioning is striking, providing evidence that factors governing supermarket pricing differ across items. We used the first two principal components, P1 and P2 to create the two indices. We accomplished this by multiplying each city' s vector of item prices with the corresponding vector of coefficients ("scores") associated with the principal components. These became our dependent variables. We also analyzed ACCRA's general grocery price index. No important differences between the results obtained and those for P1 and results for this measure are not reported.1

Independent Variables
In this section we define the explanatory variables and discuss hypothesized effects.2 This discussion will be mainly confined to the focus of the study, the market competition variables. As already suggested by our example, any such attempt is somewhat difficult. This is certainly true as compared to the case where interest is differences in over-aU prices across markets (e.g., as measured by an index). There, an increase in competition, whatever its nature, would be expected to negatively affect price. A model with two different price types opens the way for differences in response both to competition in general and to specific competitive forms. This problem is compounded by price signaling possibilities (which also has not been an issue in explaining over all price levels). As a consequence, our expectations are somewhat speculative, and as will be seen, some outcomes raise fresh conceptual and empirical questions.

Market Competition
The overriding hypothesis of the study is that competition and market environment affect pricing of P1 and P2 in different ways. Our analysis provides estimates of the relative importance of competition to the two product types as well as the importance of competition relative to cost and city factors. While we expect increases in competition among supermarkets to result in lower prices for both types of commodities, we expect a stronger response from P2. Not only is it weighted toward items used as price signalers, many are also from departments receiving increasing emphasis. However, the difficulties posed by competition from nontraditional sources need be kept in mind, as the earlier example patterned on the Lal and Matutes model illustrated. In general, we expect the response of each price index to competition to depend upon the source of the competition, and the response to a given source to differ for each price. The specific competition variables are mentioned below. C4 is supermarket sales concentration, as measured by four-firm concentration in the SMA (Franklin and Cotterill, 1995). Imperfect market models predict that greater concentration leads to greater price, (Cotterill, 1991; Weiss, 1989), and this is our expectation.3

Grocery Market Pricing and the New Competitve Environment


However, to the extent that competition arises from sources other than supermarkets, the effect of a variable measuring only supermarket competition might weaken. The second variable pertaining to competition is CHAIN, the percent of supermarkets owned by chains. We expect a negative impact, ceteris paribus. When CHAIN is high, there are fewer firms in a market. While this implies less price competition, this should be captured by C4. We believe the negative impact would derive from economies of specialization in chains' management and efficiencies arising from integrated wholesale distribution. These will tend to lower costs and, ceteris paribus, prices. In the case of both C4 and CHAIN, we have no a priori reason to anticipate differences in response in P1 and P2, especially since we expect differences in these to arise mainly in measures of nontraditional forms of competition. There are three variables designed to capture non-supermarket competition. SMALL is the ratio of small store sales to total SMA grocery sales, a measure of the importance of small stores. Its presence can be viewed as a test of whether small stores represent viable competition for supermarkets. If so, as small stores' share of the market increases, prices will fall. However, most researchers believe that price competition from small stores is likely of limited importance in today's market. Moreover, as in the case of CHAIN, the model specification gives SMALL more of a cost interpretation. Thus, since supermarkets clearly have lower costs, we believe a positive effect of SMALL is more likely, but one not likely to be strong. A second measure is WHS, the extent of warehouse-store presence in the market area. 4 Since warehouse stores typically stock fewer items than do standard supermarkets while also providing fewer services, they have lower operating costs and hence lower prices. If supermarkets choose to compete with these directly on price, WHS will carry a negative sign. Furthermore, since warehouse stores concentrate on packaged goods, the effect would be stronger for P1. However, this is the case featured in our stylized example above, where we pointed out that lowering prices against warehouse stores may not be a viable strategy. This makes making specific responses difficult to anticipate. Hence we limit expectations to negative coefficients, while admitting that a stronger effect for P1 is plausible. The third variable is FASTFOOD the average per person expenditures on low-cost restaurants in the SMA. If grocery outlets use price to respond to their avowed concern with competition from fast food outlets, the coefficient on FASTFOOD should be negative. We expect an effect to be stronger for commodities perceived as competitive with food-awayfrom-home, such as those in P2. However, a strong price effect seems unlikely. Many supermarkets are meeting the fast food challenge more directly via selling various prepared meals, often providing on-site dining (Fortune, 1995). What (if any) price implications this carries are unknown. We, therefore, view FASTFOOD as an exploratory variable, though one of special interest to this study. We classify two additional variables under "competition." One is STRSIZE, average store size as measured by square footage. Our view, a view supported by evidence cited above, is that the primary motivation for larger stores is to enhance supermarkets' capability to compete with a wider spectrum of rivals. It does this by permitting new and expanded services, including prepared foods, and more expansive produce, deli, bakery, and other departments. 5 This enhances their competitive position, and probably also raises costs (as


Journal of Retailing Vol. 74, No. 2 1998


Summary of Hypothese Concerning Coefficient Signson Competition Variables

Effect on Supermarket Prices

Independent Variable (Symbol)

Supermarket sales concentration

P1: Dry Grocery


P2: Freshand Refrigerated Foods


(C4) Supermarketsowned by chains (CHAIN) Grocery sales by small stores

(SMALL) Presence of warehouse stores (WHS) Presence of fast-food retailers (FASTFOOD) Size of store (STRSIZE) Pricing turbulence (CHURN)

Negative Positive or zero

Negative* Negative Positive* Negative

Negative Positive or zero

Negative Negative* Positive Negative

*Starreditems indicates we expect a stronger response for that price (i.e., P1 vis-a-vis P2). (as measured by the strength of respective statistical tests.)

Messinger and Narasimhar (1997) find), both of which should lead to higher prices. However, emphasis on service departments will enhance any price-signaling role of service department items, making P2 less likely to rise. The final competitive variable is CHURN, the number of grocery items with large ( 10 percent from the mean) quarterly price changes. We interpret high price turbulence as indication that some supermarkets were employing new price mixes in our 1987 snapshot of SMAs. To counter new forms of retail competition such as supercenters or hypermarkets, traditional supermarkets may have responded with more intensive price promotion. Since greater competitive activity will lower overall prices, we expect the effect of C H U R N to be negative. The specific hypotheses regarding the competition variables are summarized in Table 3.

Retail Cost Factors

The model contained four variables that we classify as cost factors. Generally, we would expect increases in costs to lead to higher prices. However, the cost variables are not individually of interest and serve primarily as covariates. We concentrate on the overall importance of cost relative to the other factor groups on our two product types. 6 Above we stated that, if anything, we expect P2 to be more affected by competitive influences (especially if signaling is important). This suggests a relatively weaker role for costs in the case of P2 than for P1.

Grocery Market Pricingand the New Competitve Environment


Of the four variables, three are explicit measures of retail operating costs. Average retail WAGE reflects the most important operating cost. We used the ACCRA measure RENTthe cost of rental housing in the metro area, since information on commercial rents was not available. ELEC, electricity cost, is measured by average monthly costs for commercial accounts using at least 6,000 kwh. 7 An additional variable that we classify with cost factors is EMPI_/STR, the number of employees per supermarket in the Progressive Grocer region. We were motivated by the large interregional variation in this measure: it ranged between 31 and 60. It is unlikely that such a large variance could be due to differences in wages or store size. Further, these factors are already accounted for in the model. They can reflect intermarket differences in union strength, in local-store practices (e.g., twenty-four hour operation), and other idiosyncratic factors. Such differences are likely to affect costs and hence price, the direction of which is purely an empirical question.

City and Region Factors

Our sample is composed of diverse cities. It is desirable to account for such variation, because supermarket pricing may be influenced by aspects of the market other than local costs and competitiveness. By including measures of these as a second set of covariates, we reduce possible omitted variable bias while controlling for random variation. We included four: CITYSIZE, the market size, as measured by metro-area grocery sales; DENSITY, the population per square mile; the population growth rate GROWTH; and average marketlevel INCOME. Each of these can affect prices in several ways. As an example, for CITYSIZE, buyer price searching may be more costly in large metropolitan areas (although the enthusiastic searcher will have more opportunities). This will discourage supermarket price competition and lead to higher prices. On the other hand, large markets may permit scale economies in wholesale distribution and advertising, lowering costs and prices. Huge urban agglomerations, however, may exceed the size at which these costs are minimized. Similar statements apply to DENSITY and GROWTH. Thus we view any effects of these covariates solely as an empirical issue. In the case of INCOME, average per capita income in the market area, we expect a positive effect. In higher income areas, "upscale" stores with more services are likely to be present. Also, income elasticities will tend to be lower (since food represents a smaller portion of consumption), thus facilitating the exercise of market power (Cotterill, 1986). Finally, the models contained three regional dummies, guided by Larson and Binkley's (1995) results regarding regional taste patterns. These will help capture any price differences due to variations in wholesale prices across regions, as well as other non-measured factors associated with geography.


Journal o f Retailing Vol. 74, N o , 2 1 9 9 8


F Statistics for Variable Groups

Competition Cost City Region All

Dry Groceries P1
3.17 a 4.13 a 2.17 b 5.08 a 6.63 a

Fresh & Chilled P2

5.41a .77 .38 16.12 a 12.43 a

P1 = P2
5.74 a 2.33 c 1.83 14.93 a 12.57 a


a, b, and c represent significance from zero at the 1%, 5%, and 10% level, respectively.


Regression Results Explaining Grocery Price Components Across 95 Cities

Dry Groceries (P1) Variable Coefficient
81.88 0.07 -.19 -11.23 -39.58 -0.17 -0.11 2.07 2.89 -0.34 0.08 -0.01 0.86 -0.46 0.13 -0.85 8.54 -3.81 -2.93 .61 .52

Fresh & Chilled (P2) Coefficient

110.20 ~3.03 .08 -47.00 40.79 -0.36 -0.01 -1,19 -0.43 0.03 -0.02 0.01 0.35 0.44 -0.03 0.30 2.74 12.64 -3.93 .75 .68

5.34 a 1.42 c -1,46 -1,13 -2,58 b -1.53 c -1.55 3.71 a 2.78 a -1.97 c 1.64 c -0.83 0.98 -1.02 1.89 c -1.94 c 3.04 a -2.00 b -1.10

5.42 a -0.52 0.48 -3.57 a 2.01b -2.46 b -0.16 -1.62 -0.31 0,14 -0,34 1.65 c 0.34 0.73 -0.29 0.52 O. 74 5.00 a -1.12

Intercept A. Competition

B. Cost

c. city

D. Region


a, b, and c represent significance from zero at the 1%, 5%, and 10% levels. These levels are based on one-tailed tests for C4, Wage, Rent, Churn, and Elec and on two-tailed for all others.

Grocery Market Pricing and the New Competitve Environment


In Tables 4 and 5 we report the major results of the study. In the first two columns of Table 4 are F statistics for significance tests across the two equations for the four groups of variables. The third column has corresponding values for testing whether coefficients differ across equations. 8 We first discuss these F tests, which test the important hypotheses on variable groups. Then we consider the results for specific variables (Table 5).

Group Tests
The most general hypothesis of the study is that the manner in which prices respond to determining factors is not uniform over commodity types, that is, the coefficients in the P 1 and P2 equations differ. The group F statistics in the third column of Table 4 provide considerable support for this hypothesis. For all variable groups except the city factors, the hypothesis of no difference is rejected at a level of significance of at least 10%. The strongest effect is for regions; the least differences are for the cost and aforementioned city variables. Importantly, there is considerable evidence that the nature of retail market competition affects different prices in different ways. The F statistics in the first two columns can be taken as measuring the relative strength of broad classes of factors in pricing of the two groups of items. We had two hypotheses: the P2, "price signaling" commodities are more responsive to competition; and, to the extent that costs matter, they matter more for the dry grocery group P1. As measured by the strength of the F statistics, the two hypotheses are supported. The variables we classified under competition influence both groups: the much larger F-statistic suggests a stronger impact on P2. At the same time, there is no evidence that cost factors play a role in P2 pricing, while for P1 they are statistically more important than are the competitive variables. In both cases, regional differences are strongly in evidence. The fact that P1 is less strongly affected is a reflection that branded goods are priced more uniformly across the U.S. than are produce and other non-branded items.

Competition Variables
We now consider the detailed results in Table 5, first for the seven competition variables. Earlier we described two of these---C4, concentration, and CHAIN, the portion of supermarkets owned by chains--as a subgroup of specifically supermarket competition factors. In neither case is either highly significant, especially for P2. The calculated F statistics for a joint test of the two (not presented) are 1.77 for P1 and 0.83 for P2. It is noteworthy that any perceptible effect of these variables is confined to the commodity group featuring packaged grocery goods. This is especially pertinent for the concentration variable. For P1 it has the expected positive effect, reasonably significant if an (appropriate) one-tailed test


Journal of Retailing Vol. 74, No. 2 1998

is used. This provides the clear suggestion that prices for traditional "center of the store" goods are most sensitive to the effects of concentration. These products are also most likely to be branded, nationally advertised goods manufactured in concentrated markets. For CHAIN, we have the expected negative effect in both models, anemic for P1 and obviously of no impact for P2. We conclude that the extent of chain control is of little importance in determining price differences across markets. SMALL, the measure of competition from small grocery retailers, has little impact. Generally, this was expected. The modest indication of a negative effect on P1 suggests a possible competitive influence, contrary to our expectation. For the warehouse variable WHS, the coefficients are negative in both models, but only significant (< .05) for P2. This outcome would not be expected under an assumption of competition solely via price. Warehouse stores emphasize packaged items of the PI type and stock only a more limited selection of fresh items at best. It implies that supermarkets respond to warehouse store competition by lowering prices more for goods that warehouse stores do not stock. As pointed out earlier, this is quite reasonable within the Lal and Matutes (1989) framework. Given the cost advantage of warehouse stores, supermarkets cannot successfully compete with them simply by reducing prices. Judging from the Wall Street Journal information above, they have not done so, but have also enlarged stores, increased assortments, and expanded specialty departments. In this light, the strong response found for P2 may reflect the role of specialty department items as price signalers. This may, in turn, serve to create an image that the store is the best source of these goods. This role is enhanced when service departments are used to withstand competition, i.e., to attract customers for traditional grocery goods. Obviously, the retailer's goal is to get customers inside the store, for then the availability of P1 goods at lower prices elsewhere is of less importance to the customer. This interpretation is consistent with the estimated coefficients for store size, STRSIZE. The positive and highly significant response for P1, the same as found by Messinger and Narasimhan (1997), and weakly negative effect for P2, are not suggestive of economies of scale. A competitive interpretation seems more reasonable. Drawing customers to larger, more attractive stores reduces their demand elasticity for P 1 goods, since they have already incurred the travel cost, thus permitting higher P1 prices. The negative sign for P2 would not contradict a signaling hypothesis. Taken together, the interpretations for WHS and STRSIZE can explain an observation that puzzled the editors of Progressive Grocer. In the 1987 report quoted earlier they state "We have to wonder how much longer grocers can opt for price-cutting programs while simultaneously adding cost increasing services" (p.5). However, selective application of this strategy, i.e., to certain categories, may be rational under the new competitive environment. The results for FASTFOOD, metropolitan area fast food sales, are problematic. The coefficient for P1 is negative and highly significant. As P1 involves frozen foods, bakery goods, and other prepared foods that may be close substitutes for fast food fare, perhaps this result is not unexpected. In sharp contrast, however, the coefficient is positive for P2, and significant at 0.05. Yet, it is fresh items (e.g., beef) that are more directly competitive with food away from home. While it is conceivable that discriminatory pricing along the lines of that discussed earlier is involved in this result, we strongly doubt that it plays a

Grocery Market Pricing and the New Competitve Environment


major role. Indeed, it is unlikely that the strong results for P1 or P2 are measures of supermarket reaction to fast food competition. An alternative interpretation of the P2 result is reverse causality: high supermarket prices for meat and other fresh items are an incentive for consumers to switch to fast food. Unfortunately, this is not consistent with the negative P1 response. Nevertheless, we are reluctant to dismiss both as spurious effects, given the strong statistical measures. In view of our prior that a price response by supermarkets to perceived restaurant competition is unlikely, we believe the observed results may be crudely reflecting an indirect effect not directly due to price. Our results may be evidence of competition between the grocery and restaurant segments. Uncovering its specific nature will require a more targeted model. The last of the competition variables is CHURN, our measure of the extent of price changing. Here, we find that variation in prices over time has a similar effect on both price measures. In each case its coefficient is negative and reasonably significant. The larger coefficient in the P2 equation may reflect that expanded service departments are playing a key price signaling role as traditional supermarket retailers seek to compete with encroaching warehouse store formats.

Table 4 shows that three of the cost covariates have a perceptible influence on P1. Coefficients for WAGE and RENT are, as expected, positive, with the first highly significant. We have no explanation for the negative effect for EMPL/STR as we had no prior regarding employees per store. We emphasize that this is a ceteris paribus response, in particular, it is the effect with STRSIZE held constant. One interpretation is that given store size, the association between employees with lower P1 prices may be due to more intense use of fixed plant. 9 As expected, the fourth variable, electricity cost, is not found to affect P1. For P2, however, this variable is the only cost predictor variable showing any evidence of a relationship to pricing. Because many P2 items are temperature-controlled, this is reasonable. Overall evidence, however, is that costs play little role in explaining P2. This further suggests that competitive factors are of special importance to these prices. For the city covariates, there is little evidence that pricing practices vary across metropolitan areas with widely different values of CITYSIZE, DENSITY, and GROWTH, for either price measure. We conclude that supermarket pricing follows the same rules across urban areas of different sizes and forms, ceteris paribus. The negative, quite significant coefficient in the P1 equation for INCOME is puzzling. We had expected a positive effect. High income consumers demand a generally higher level of services and have more priceinelastic demands, implying higher prices. We have no explanation for the contrary result. We tested its robustness using various permutations of the model, and the effect persisted, even in a univariate model. It is conceivable that cities with higher income levels have a wider variety of retail establishments, including sources of competition not in our model, and income may be picking up this effect. We leave exploration of this result to further research.


Journal of Retailing Vol. 74, No. 2 1998

Finally, the F tests showed that there are substantial regional differences in average prices for both types of good, given the effect of all other variables. The individual coefficients identify their sources. Because the excluded region is the Midwest, the coefficients estimate differences from that region. Thus, for P1 commodities, there is substantial evidence that prices in the East exceed those in the Midwest, prices in the South are lower (and afortiori lower than in the East), and little evidence that the Midwest and West differ. For P2, the South unequivocally has the highest prices, while the West has the lowest. In part these differences reflect differing transport costs. For example, it is reasonable for Western areas to have somewhat lower meat and produce prices, dominant components of P2. However, transport costs are unlikely to be the only factor. Why would prices of P2 items in the South exceed those in the East where ther is no location advantage relative to food production (Connor and Shiek, 1997)? Thus, there are evidently other regional differences we have not captured. Specific supermarket chains, for example, tend to be regional firms. Conceivably these chains may vary in overall pricing strategy, e.g., every-day-lowprices versus high-low pricing. Differences in the strategy mix of the larger chains in regions can generate variations in regional prices.


This research examines grocery pricing in the changing market environment in which retail food retailers have been operating due to the increasing role of new store formats such as warehouse and fastfood stores. To evaluate the impact of these new conditions, we derived pricing models using a sample of city data from a wide range of metropolitan areas, with a corresponding variety of market characteristics. We included several variables not previously considered in pricing studies. These variables are intended to measure aspects of what we view as the new, competitive environment facing the retail food industry. The study was conducted under the working hypotheses that food store pricing practices will vary across broad classes of items and this pattern will be shaped by the changing nature of competition. Simple price correlations across different products in the price data base supported this view. Principal components analysis further revealed the presence of two quite different groups of prices: (i) a set of packaged, branded grocery products in the "dry grocery" and "health and beauty aids" departments and (ii) a group of essentially nonbranded, refrigerated products consisting primarily of fresh red meats, milk, and produce items. We derived two price indices from these principal components and used them as dependent variables in our models relating price to market and cost characteristics. Results for the dry grocery component were similar to those for previous studies explaining variation in a general price index. Those for the non-branded or perishable or refrigerated group were quite different, suggesting little, if any, cost factor effects and presenting some unexpected rivalry effects. The evidence is strong that supermarket pricing varies markedly across different kinds of goods. We believe that the large differences observed reflect discriminatory pricing. Discriminatory pricing requires market segments with different demand elasticities. It also requires that the markets can be separated. This is considerably facilitated when retail food markets

Grocery Market Pricing and the New Competitve Environment


have non-identical competitors serving specialized segments, such as that of the working hypothesis of this study. Furthermore, segmentation is likely to enhance the role o f price signaling. Here, the use o f selected prices may generate a store image o f strength and low prices in goods of interest to particular consumer segments. Overall, the results depict a changing market, with the degree of rivalry among supermarkets no longer the only important competitive force shaping supermarket pricing decisions. Our evidence is that serious competition has arisen not only from new formats of grocery retailing--warehouse stores, for e x a m p l e - - b u t also from the restaurant industry. This should not be a surprising outcome in a world in which large changes in the retail landscape are bringing about corresponding changes in consumer shopping behavior. W e end on a cautionary note, with a reminder that our work is exploratory. Our data are from 1987, a snapshot o f a period in which the dominance of the supermarket format was coming to an end. Indeed, some of our specific results, such as those relating to fast food and warehouse stores, hint at a period of disequilibrium that may now have been resolved. Some results suggest the need for further research. Although conducted under the strong guidance of emerging theoretical and empirical results, we note that the unfolding o f our study was in part data driven. Hence, there is a need for further construct development and confirmation studies to establish our findings. Finally, we again remind readers that our purpose was to study how the supermarket industry in fact responded to the new market environment. W e can make no judgement concerning how management should respond.

1. Complete results are available in Binkley and Connor, (1996), which reports our earlier work. 2. All the independent variables are extracted from the U.S. Census Bureau's 1987 City-County Data Book except for the following; C4 and SMALL from Franklin and Cotterill; GROWTH, CITYSIZE and FASTFOOD from the 1987 U.S. Census of Retail Trade; RENT, CHURN, and ELEC from ACCRA; and EMPL/STR, WHS, and CHAIN from Progressive Grocer (1988). All but the latter three are measured at the level of the metropolitan area. The last three refer to the Progressive Grocer region in which the metro area is located. 3. Because the influence of city-wide concentration may depend on city size, in preliminary work we included an interaction term between C4 and a measure of size. Results were qualitatively the same as those for C4 alone. 4. This variable is rather crudely measured as the ratio of warehouse stores to the total number of grocery stores. We note that prices will not be lower due to ACCRA sampling. Even if this measure referred to the individual SMAs, the direct effect may be absent. The ACCRA data is collected only in stores selling all products priced. The latter include meats and produce, items often not carried by warehouse stores. Hence, few warehouse stores are likely to be in the sample. 5. The argument can be made that store expansion is to exploit scale economies, but it is not obvious what these are. A pure scale effect would not include a broadening of the types of goods and services sold. 6. Store pricing models have not placed large emphasis on costs, one reason being that costs of goods sold and unit operating costs are considered to be essentially independent of volume. Most


Journal of Retailing Vol. 74, No. 2 1998

early writers on grocery pricing policies are highly skeptical that operating costs have any effect on gross margins (Cassady, 1962; Holdran, 1968). Leed and German (1973) presented data showing that operating costs in the meat and produce departments are much higher than in other departments, but still downplayed their importance. More recent studies likewise omit operating cost from most models of price determination (Bolton, 1989; Chevalier and Curham, 1976; Lattim and Bucklin, 1983; Shankar and Krishnamurthi, 1996; Tellis and Zufryden, 1995). 7. For several cities with missing information, figures for comparable cities in the region were used. 8. We did this with tests of coefficient equality across the P1 and P2 equations, using a seemingly unrelated regression procedure. Since the explanatory variables are the same in each equation, all estimated coefficients for both were precisely the same as those of OLS. But the SUR procedure is needed for cross-equation tests, since it incorporates the covariance of estimated coefficients across the two equations (and certainly one would expect the errors to be correlated). 9. The sample correlation between these two variables is .6. As an experiment, we dropped STRSIZE from the equation, thus causing its effect to be reflected in the EMPL/STR coefficients. Then the signs for Pl and P2 were reversed and both were significant. 10. We do not regard this sensitivity as a serious problem since EMPL/STR is not a variable of major interest. Much more important is that STRSIZE was not highly sensitive. When EMPL/STR was dropped, the coefficients for STRSIZE did not change signs, and both were significant.

Adelman, M.A. (1948). "Effective Competition and the Anti-Trust Laws," Harvard Law Review, 62: 1927. American Chamber of Commerce Researcher's Association. (1988a). Inter-City Cost of Living Index. Indianapolis, IN:ACCRA. . (1988b). A CCRA Cost of Living Index Manual. Indianapolis, IN: ACCRA. Baumol, William J. et al. (1964). "Oligopoly Theory and Retail Food Pricing," Journal of Business, 37: 346-369. Becket, Gary. (1965). "A Theory of the Allocation of Time," Economic Journal, 75:493-517. Benson, Bruce L. and Merle D. Faminow. (1985). "An Alternative View of Pricing in Retail Food Markets," American Journal of Agricultural Economics, 67: 296-305. Binkley, James K. and John M. Connor. (1996). Concentration Demographics, and Conundrums: Metropolitan--Area Food Prices. Department of Agricultural Economics Staff Paper SP-96-15, Purdue University. Blattberg, R.C. and S. Neslin. (1990). Sales Promotion. Englewood Cliffs, NJ: Prentice-Hall. Bliss, C. A. (1988). "Theory of Retail Pricing," Journal of Industrial Economics, 36: 375-390. Bolton, Ruth N. (Spring, 1989). "The Relationship Between Market Characteristics and Promotional Price Elasticities," Marketing Science, 8: 153-169. Cassady, Ralph. (1962). Competition and Price Making in Food Retailing. New York: Ronald Press. Chevalier, Judith A. (1995). "Do LBO Supermarkets Charge More? An Empirical Analysis of the Effects of LBOs on Supermarket Pricing," Journal of Finance, 50:1095-1112. Chevalier, Michel and Ronald C. Curham. (Fall, 1976). "Retail Promotions as a Function of Trade Promotions: A Descriptive Analysis," Sloan Management Review, 18:19-32. Connor, John M. and William A. Schiek. (1997). Food Processing: An Industrial Powerhouse in Transition (Second Edition). New York: John Wiley & Sons.

Grocery Market Pricing and the New Competitve Environment


Cotterill, R.W. (1991 ). Food Retailing: Mergers, Leveraged Buyouts, and Performance. Food Marketing Policy Center Research Center, Report 14. University of Connecticut (1986). "Market Power in the Retail Food Industry: Evidence from Vermont." The Review of Economics and Statistics, 68: 379-386. Cotterill, Ronald W. and C. David Harper. (1995). Market Power and the Demsetz Quality Critique: An Evaluation for Food Retailing, Res. Report No. 29. Storrs, CT: Food Marketing Policy Center, Univ. of Connecticut. Dickson, Peter R. and Joe E. Urbany. (1994). "Retailer Reactions to Competitive Price Changes," Journal of Retailing, 70: 1-21. Franklin, Andrew and Ronald Cotterill. (1993). An Analysis of the U.S. Grocery Retailing Industry. Storrs, CT: Food Marketing Policy Center, University of Connecticut. Hiemstra, S.J. and W.G. Kim. (1995). "Factors Affecting Expenditures for Food Away From Home in Commercial Establishments By Type of Eating Place and Meal Occasion," Hospitality Research Journal, 19(3): 15-31. Hoch, Stephen J., B. Kim, A.L. Montgomery, and P. E. Rossi. (1995). "Determinants of Store-Level Price Elasticity," Journal of Marketing Research, 32:17-29. Holdren, Bob R. (1968). The Structure of a Retail Market and the Market Behavior of Retail Units. Ames, IA: The Iowa State University Press. "How Grocers Are Fighting Giant Rivals." The Wall Street Journal, March 27, 1997. Holmes, T.J. (1989). "The Effects of Third-Degree Price Discrimination in Monopoly," American Economic Review, 79: 244-250. Holton, Richard H. (1957). "Price Discrimination at Retail: The Supermarket Case," Journal of Industrial Economics, 6: 28-36. Katz, M.L. (1984). "Price Discrimination and Monopolistic Competition," Econometrica, 52: 14531471. Kim, Byung-Do, Robert C. Blattberg, and Peter E. Rossi. (1995). "Modeling the Distribution of Price Sensitivity and Implications for Optimal Retail Pricing," Journal of Business and Economic Statistics, 13: 291-303. Knebel, Donald E. et al. (1991). "Predation to Coerce Competitors' Acceptance or Predator's Price Leadership: The Supermarket Model," Antitrust Law and Economics Review: 65-74. Lal, Rajiv and Carmen Matutes. (1989). "Price Competition in Multi-market Duopolies," Rand Journal of Economics, 20: 516-537. Lamm, R. McFall. (September, 1981). "Prices and Concentration in the Food Retailing Industry," Journal of Industrial Economics, 30: 67-77. Larson, Ronald B. and James K. Binkley. (1994). "Examining Regional Food Consumption Patterns." Paper presented at the annual meetings of the American Agricultural Economics Association, San Diego, CA. Lattin, James M. and Randolph E. Bucklin. (August, 1989). "Reference Effects of Price and Promotion on Brand Choice Behavior," Journal of Marketing Research, 26:299-310. Leed, Theodore W. and Gene A. German. (1973). Food Merchandising: Principles and Practices. New York: Chain Store Age Books. Marion, Bruce W. et al. (I 979). The Food Retailing Industry: Market Structure, Profits and Prices. New York: Praeger. Marion, Bruce W. and the NC-117 Committee. (1986). The Organization and Performance of the U.S. Food System. Lexington, MA: Lexington Books. Messinger, Paul R. and Chakravarthi Narasimhan. (1997). "A Model of Retail Formats Based on Consumers' Economizing on Shopping Time," Marketing Science, 16: 1-23. Progressive Grocer. (1987). 54th Annual Report of the Grocery Industry. Stamford, CT: MacLean Hunter Media, Inc., April.


Journal of Retailing Vol. 74, No. 2 1998

(1988). 1988 Marketing Guidebook. Stamford, CT: Maclean Hunter Media, Inc. Salop, Steven and Joseph Stiglitz. (1977). "Bargains and Ripoffs: A Model of Monopolistically Competitive Price Dispersion," Review of Economic Studies, 44:493-510. Shankar, Venkatesh and Lakshman Krishnamurthi. (1996). "Relating Price Sensitivity to Retailer Promotional Variables and Pricing Policy: An Empirical Analysis," Journal of Retailing, 72: 249-272. Smith, Henry. (1937). Retail Distribution. London: Oxford University Press. Stigler, George J. (1950). Five Lectures on Economic Problems. New York: McMillan. Tellis, Gerard J. and Fred S. Zufryden. (1995). "Tackling the Retailer Maze: Which Brands to Discount, How Much, When and Why?" Marketing Science, 14:271-299. Weiss, Leonard. (1989). Concentration and Price. Cambridge, MA: MIT Press. "What's for Dinner?" Fortune, May 15, 1995, p. 50-64.