Vertical Relationships in the Ready-to-Eat Cereal Market in Boston Benaissa Chidmi, Rigoberto A. Lopez, and Ronald W. Cotterill Selected paper submitted to the European Association of Research in Industrial Economics Meeting, Porto, Portugal, September 1-4, 2005. The authors are Ph.D. candidate and Professors in the Department of Agriculture and Resource Economics, University of Connecticut, Storrs, CT 06269-4021, USA. Contact: Rigoberto.Lopez@uconn.edu, (860)-486-1921. Vertical Relationships and Competition in the Ready-to-Eat Cereal Market in Boston Abstract: In this paper we estimate a discrete choice demand model with random coefficients for 37 brands of ready-to-eat cereals (RTECs) at the supermarket chain level in the Boston area. Then assuming a manufacturer Stacklberg model for vertical pricing, we decompose the market channel price-cost margins (PCMs) for individual brands at four supermarket chains. The results shed light on the share of channel profits accruing to individual RTEC manufacturers and retailers. For instance, Stop & Shop, the leading supermarket chain in Boston, gets greater dollar profits than smaller supermarket chains. In spite of this, smaller supermarkets, especially those with urban locations, charge higher prices due in part to higher retailing costs. Finally, the results also show that both manufacturers and retailers dropped their RTEC prices in 1996 under a regulatory threat provoked by a scandal of supra-normal RTEC prices. Key words: Vertical relationships, discrete choice, manufacturers/retailers, market channel. JEL codes: L110, L130, L660. 1 Vertical Relationships and Competition in the Ready-to-Eat Cereal Market in Boston 1. Introduction Recent empirical literature on vertical relationships between food manufacturers and retailers has focused on contracts and vertical integration, providing structural models to explain the relationships between manufacturers and retailers (Kadiyali et al., 1999; Villas-Boas and Zhao, 2000; Villas-Boas, 2002; and Manuszak, 2001). This paper estimates and decomposes the channel profits for the ready-to-eat cereal (RTEC) market in Boston at the brand/supermarket chain level. It contributes to the existing literature in several ways. First, it is the first study to estimate a discrete choice random coefficient demand system for branded products at the chain as opposed to city market level.1 Second, the study decomposes channel profits accruing to manufacturers and retailers, respectively, instead of attributing all profits to manufacturers as in Nevo’s work (2001). Third, this study pursues chain-wide strategic pricing in a relevant regional market area, the Boston Information Resources Incorporated (IRI) market. Villas-Boas (2002) has estimated a similar model for yogurt at the store level for a few stores. Fourth, this study uses four-week data while prior RTEC brand demand analysis uses either quarterly (Nevo, 2001); Hausman et al., 1994) or weekly observations (Kadiyali et al., 1999; Cotterill and Haller, 1997),2 shedding light on the importance of time aggregation for this particular industry. 1 Cotterill and Dhar (2002) is the only prior chain level demand study and it uses a nested logit model. There is no consensus concerning which time unit is desirable. Quarterly may be too aggregate, while weekly may be too disaggregate to measure strategic pricing moves in a static equilibrium model. 2 2 2. Methodology The methodology used consists of two steps. First, the retail demand for differentiated brands of RTECs is estimated using a random coefficient model. Second, a two-stage pricing model is implemented assuming a Nash-Bertrand competition at each stage. In the vertical market, we assume manufacturer Stacklberg conduct holds; i.e., retailers play Nash-Bertrand when evaluating wholesale prices and maximizing profits, while manufacturers employ the retail reaction functions to their wholesale price change when maximizing profits (Choi, 1991). The demand results are used to compute the total channel price-cost margins and to decompose them into the price-cost margins at the retail and manufacturer stages. 2.1. Demand Side Consider the case where consumers choose a brand of RTEC that maximizes their utility. More specifically, the indirect utility3 of consumer i from buying the brand j is given by U ij = β j + x j β i − α i p j + ζ j + ε ij , i = 1,...n j = 1,..., J (1) where β j represents the store/brand fixed effects, x j are the observed product characteristics of brand j, p j is the price of the brand j , ζ j are the unobserved (by the researcher) product characteristics, and ε i j represents the distribution of consumer preferences about the unobserved product characteristics, with a density f (ε ) . The parameters to be estimated are α i and β i . Note that those parameters are allowed to vary across consumers, therefore taking into account the heterogeneity taste of consumer. 3 The indirect utility comes from a quasi-linear utility function. 3 These coefficients can be decomposed into a fixed component and a variable component (changing with consumers’ observed and unobserved characteristics). This decomposition can be expressed as: α i = α + λDi + γvi , (2) β i = β + ϕDi + ρvi , (3) where the Di represents the consumers’ observed characteristics such as demographics variables (e.g., income), and vi denotes the unobserved consumers’ characteristics. Substituting (2) and (3) in (1) yields U ij = β j + x j β + λDi x j + γvi x j − αp j − λDi p j − γvi p j + ζ j + ε ij . (4) Unobserved consumer characteristics vi are assumed to be normally distributed N (0, I ) , where I is the identity matrix; and the observed consumer characteristics Di have an empirical distribution h(D ) , not necessarily a normal distribution. The indirect utility in (4) can be decomposed into two parts: a mean utility given by δ j = β j + β x j − αp j + ζ j and a deviation from that mean, which is a function of the interaction between the observed and unobserved consumer’s characteristics and the price and observed brand characteristics, given by µ ij = λDi x j − λDi p j + γvi x j − γvi p j + ε ij . (5) To complete the model, an outside good is included to give the consumer the possibility not to buy any one of the J brands included in the choice set.4 The utility of 4 The inclusion of the outside good is necessary in order to accomplish with the exhaustiveness of alternatives of the discrete choice model. For a detailed discussion, see Train (2002). For the case at hand, the outside good can include all other brands, or the residual brands not included in the study. 4 the outside good is normalized to be constant over time and equal zero. Given the observed and unobserved consumer characteristics define the set of choice by S ( x j , p j , ζ j ;θ ) = {( Di , vi , ε ij ) : U ijt ≥ U ik ∀k = 0,1,... N } , (6) where θ is a vector that includes all the parameters of the model. The consumer purchases one unit of the brand that yields the highest utility. The global market share of the jth brand corresponds to the probability that the jth brand is chosen. That is, s j = ∫ I {( Di , vi , ε ij ) : U ij ≥ U ik ∀k = 0,1,...N }dH (D ) dG (v ) dF (ε ). (7) Depending on the assumptions regarding D, v , and ε , the integral in (7) can have or not a closed formula. In a general setting, the integral in (7) does not have a closed formula and should be solved numerically (BLP, 1995; Nevo, 2000, Villas-Boas, 2002). The Random coefficients model (RCM) allows for consumer heterogeneity α i and β i as described in (2) and (3). That is, each consumer is different from another consumer in their response to price and brand characteristics. However, the RCM poses two challenges. First, the integral in equation (7) has no closed formula and should be solved numerically.5 Second, information on the distribution of demographics is needed to compute the individual market shares. Intuitively, the integral in (7) is solved based on the choice of the parameters that minimize the distance between the predicted market shares given by equation (7) and the observed market shares. This paper follows Berry (1994) inversion of the market share function that obtains the mean utility valuation δ that equates the predicted market shares with observed market shares. 5 The integral in (7) is solved using the simulation technique proposed by Pakes (1986). 5 Given starting values for θ 2 (parameters that enter non-linearly) in (4) and δ , and the draws from the distributions of D and ν , the integral in (7) is estimated numerically.6 Nevo (2000) proposes to use the smooth estimator that makes use of the extreme value distribution on f (ε ) to integrate the ε s analytically. The predicted market shares are approximated by s j ( p, x, δ , Pns ;θ 2 ) = 1 ns 1 ns s ji = ∑ ∑ ns i =1 ns i =1 exp(δ j + µ ij ) J 1 + ∑ exp(δ m + µ im ) , (8) m =1 where ns is the number of draws from the distributions D andν given by the distribution Pns . The above predicted market shares allow computing the mean utility valuation δ that equates the predicted market shares with observed market shares. This is an iterative step and is solved numerically due to the non-linearity of the inversion of the equation s.t (δ .t ;θ 2 ) = S .t .7 The errors are then computed and interacted with the instruments to form the objective function to be minimized using the General Method of Moments (GMM) estimation technique. The elasticities of the random coefficients model are given by ⎧ p j ns ⎪ ∑ α i s ji (1 − s ji ), ∂s j p k ⎪ s j i =1 ηj = =⎨ ∂p k s j ⎪ p k ns α i s ji s ki , − ⎪ sj ∑ i =1 ⎩ 6 7 if j = k , (9) otherwise. The starting values for the mean utility value δ come from the Logit model estimation Berry, Levinsohn and Pakes (1995) suggest using the following contraction mapping δ t +1 = δ t + ln(S ) − ln(s( p, x, δ , Pns ;θ 2 ), where s(.) is the predicted market shares computed by equation (8) and T is the smallest integer such that δ − δ T −1 T t = 0,..., T is smaller than some tolerance level 6 A major advantage of RCM over the traditional discrete choice models (Logit and Nested Logit models) is that by taking into account the consumer heterogeneity taste, it gives another explanation besides the price variation to the variation of market shares across markets, allowing for flexible substitution patterns. 2.2. Supply Side Consider the case where a manufacturer chooses the wholesale price for each brand it sells. Then, each chain retailer chooses the retail price for each brand to maximize his own profits in a horizontal Nash –Bertrand model of competition. The game is solved using backward induction starting from the retailers and going back to the manufacturers’ equilibrium. The rth retailer’s problem is to maximize profits, given by πr = ∑( p j∈S r j − w j − c rj ) s j ( p) M , (10) where S r is the set of brands sold by the rth supermarket, p j is the retail price for brand j, w j is the wholesale price the rth retailer pays for brand j , c rj is the retailer’s marginal cost for brand j , s j ( p ) is the share of brand j , and M is a measure of the market size. The first-order conditions are given by sj + ∑(p m∈S r m − wm − c mr ) ∂s m = 0. ∂p j (11) 7 Repeat the procedure for each supermarket, stack the solutions and write them using an ownership matrix to obtain the RTEC retailers’ PCMs. 8 p − w − c r = −(Tr * ∆ r ) −1 s ( p) , (12) where Tr is the retailer’s ownership matrix with the general element Tr (m, j ) and ∆ r is a matrix of first derivatives of all the shares with respect to all retail prices. The matrix (Tr * ∆ rt ) is the element-by-element multiplication of the two matrices. Now consider the RTEC manufacturer’s problem. Each manufacturer sets the wholesale price w in order to maximize profits, given by πw = ∑ (w j∈S w j − c wj ) s j ( p( w))M , (13) where S w represents the set of brands produced by manufacturer m , and c wj is the manufacturer’s marginal cost for brand j . The first-order conditions are sj + ∑ (w m∈S w m − c mw ) ∂s m = 0. ∂p j (14) Similarly, defining a matrix of manufacturers’ ownership Tw and a matrix of manufacturer’s response ∆ w , and stacking all the manufacturers’ first-order conditions one obtains the RTEC manufacturers’ PCMs: w − c w = −(Tw * ∆ w ) −1 s ( p) . (15) The matrix ∆ w is more complicated to compute than the matrix ∆ r due to the chain rule 8 Tr , the matrix of ownership, is introduced to facilitate the matrix notation of equation (12). It is a matrix ∂sm of 1 and 0. The Tr elements are 1when brands m and j, in in equation (11), are sold by the same ∂p j supermarket and 0 otherwise. 8 effect of wholesale prices on market shares given by ∂s j ( p ( w)) ∂w j = ∂s j ∂p j ∂p j ∂w j . In matrix notation the manufacturers’ response matrix can be written as ∆ w = ∆' p ∆ r , where ∆ p is a matrix of derivatives of all the retail prices with respect to all the wholesale prices. The difficulty lies in estimating ∆ p . Following Villas-Boas (2000), this matrix can be derived by totally differentiating for a given equation j in (11) with respect to all prices and wholesale prices, and solving for the derivatives of all prices with respect to the wholesale prices. That is, N ∂s j k =1 ∂p k ∑[ N + ∑ (Tr (i, j ) i =1 ∂s f ∂ 2 si ∂s ( pi − wi − cir )) + Tr (k , j ) k ]dp k − Tr ( f , j ) dw f = 0 . ∂p j ∂p k ∂p j ∂p j (16) In matrix notation, (16) becomes Gdp − H f dw f = 0 . (17) Solving for the derivatives of all prices with respect to wholesale prices yields ∆ p = G −1 H f . (18) The market channel price-cost margin is the sum of the supermarkets’ and the manufacturers’ price-cost margins given by equations (12) and (15) respectively. p − c r − c w = −(Tr * ∆ r ) −1 s ( p) − (Tw * ∆ w ) −1 s ( p) . (19) 3. Data Sources and Management The data used in the above analysis consists of two kinds of variables: retail sales variables and demographic variables. The sales data were obtained from the Information Resource, Inc. (IRI) Infoscan database at the Food Policy Marketing Center of University of Connecticut. It covers 9 RTEC sales for 37 brands at the four leading supermarkets in Boston (Stop & Shop, Shaw’s, DeMoulas and Star Market) for four-weekly periods between April 1995 and December 1997. One important feature of this period is that it covers significant price drops in the 1990s when the RTEC industry was being questioned on market power (Cotterill, 1999, and Connor, 1999). The sales data collected consists of the following variables: dollar sales, volume (in pounds) sales, and the percent volume sold with any feature. From the RTEC sales data, the market shares and the retail prices were computed for each brand and supermarket. Market shares are obtained by converting volume sales into number of servings sold and dividing by the potential market size. This is done by using the serving weight found on the box of cereals. The potential market size is assumed to be one serving per capita and per day as in Nevo (2001). The real retail prices were computed by dividing the dollar sales of each brand by the number of servings sold and then deflated using the urban consumers CPI for Boston (with CPI=100 for 1981). The analysis is conducted using a set of 37 RTEC brands produced by six manufacturers (Kellogg’s, General Mills, Post-Kraft, Quaker, Ralston and Nabisco) sold in four supermarket channels (Stop & Shop, Shaw’s, DeMoulas and Star market) in Boston market from April 1995 to December 1997 for 5180 observations. Primary data on product characteristics were collected by examining the cereal boxes. The variables collected were the sugar content, the fiber content and the total calories. A dummy variable was created to classify the branded cereals into “Kid cereal” or not. It is assumed that those characteristics did not change since between 1995 and 1997. Besides the sales data, the analysis uses the demographic data to take account of 10 the heterogeneity of consumer taste. This paper uses two demographic variables: the natural logarithm of age and income. Further it is assumed that those variables are jointly normally distributed with mean given by the grocery data and variance-covariance matrix given by the CPS data at Boston level.9 The demand model presented above implies endogeneity of RTEC prices, and, hence, can lead to biased parameter estimates10. This implies that prices are correlated with product characteristics. This study uses a set of instrumental variables to control for retail price endogeneity in a particular supermarket. The set has two subcomponents. The first one consists of the interaction between input prices and brand dummy variables, where input prices included wages in the Boston area and the price of gas, the price of industrial and commercial electricity at the location of manufacturers, the Federal Funds Effective interest rate, and the 3-month Commercial Paper interest rate were interacted with brand dummy variables. The second subcomponent consists of time dummy variables describing the jawboning campaign events that induced price drops (change in conduct) by RTEC manufacturers, as described by Cotterill (1999) and Connor (1999). All the price instruments mentioned above were interacted with the error terms when applying the GMM estimation procedure. The use of GMM technique implies the need for an optimal weighting matrix. This paper follows Hansen (1982) who shows that setting the weighting matrix equal the inverse of an asymptotic covariance matrix is optimal in the sense that it gives parameter estimates with the smallest asymptotic variance. 9 Romeo (2005) shows that knowing the joint distribution for demographics at the city level is sufficient to infer the distribution at the county or zip code levels. 10 This endogeneity comes from the fact that retail prices depend on observed and unobserved product characteristics. Any variation in those characteristics induces a variation in retail prices. 11 4. Empirical Results 4.1. Demand Estimation Results The estimates of the RCM parameters are presented in Table 1. The results of the random coefficients model account for the consumer heterogeneity by allowing the coefficients of RTEC brand prices, sugar, calories, fiber contents and a kid-cereal dummy variable to vary across consumers as a function of the natural logarithm of their age and income, and the unobserved consumer characteristics. As expected, the price coefficient is negative and highly significant, meaning that the mean valuation utility decreases when price increases. The promotion coefficient is positive and highly significant, implying that the promotion increases the mean valuation utility. The brand unobserved characteristics (constant) have a negative and significant effect on the indirect utility. For the average consumer, sugars, fiber as well as the dummy for kids’ cereals have a negative but not significant effect on marginal utility. Stanley and Tschirthart (1991) and Nevo (2000) find a positive sugar coefficient. The negative sign of the coefficient of sugar may be explained by the increased worry of consumer on the effect of sugar consumption on weight gain. The calories have a negative effect on marginal utility, illustrating the increased awareness of consumers on their calories intake. The interaction between the unobserved consumer characteristics and the sugar content has a positive and significant effect on the mean utility, while the effect is negative and significant for the calories. The results also show that the price sensitivity increases with the age and decreases with income. Further, the sensitivity for the fiber content increases with age. 12 Table 2 presents the own-price elasticities implied by the RCM. The own price elasticities range from –9.8183 for Ralston Cookie Crisp in Star Market to –3.2811 for Kellogg’s corn Flakes in Stop & Shop. The elasticities were highly significant and were deemed suitable for further results. 4.2. Supply Results This section presents the retailers’ and manufacturers’ PCMs under a double marginalization scenario, given the demand estimates from the previous section. The results are given in Tables 3 and 4 (in $/lb) for the four leading supermarkets and the manufacturers (at the brand level) respectively. For the retailers, the price-cost margins vary from $0.6220/lb for Kellogg Corn Flakes in DeMoulas to $0.8139/lb for Kellogg Corn Pops in Stop & Shop. The remaining supermarkets obtain smaller dollar margins. The highest margins are realized by Star market with an average of $0.7143/lb. For the manufacturers, the margins are higher than the retailers’ PCMs. The manufacturers’ PCMs vary from $1.1621/lb for Ralston Cookie Crisp in DeMoulas to $1.5740/lb for Kellogg’s Raisin Bran in DeMoulas. Figure 1 shows the retail RTEC prices charged to consumer at various supermarkets. Stop & Shop, in spite of obtaining the largest dollar margins, charged prices close to the overall average price ($3.34/lb). On the other hand, Star Market, mostly an urban supermarket with generally small store size, charged the highest prices that were approximately 10% higher than those of other supermarkets ($3.61/lb). Shaw’s supermarket chain charged the lowest prices during the period of this study. The shares of the PCMs accruing to supermarkets are given in Table 5. On average, the share of the manufacturers in Stop & Shop, the leading supermarket in 13 Boston, is the lowest among the four supermarkets. In Stop & Shop, the manufacturer’s share varies between 62.63% for Nabisco and 67.86% for Kellogg. Table 5 also shows that the shares of the price-cost margins accruing to supermarkets are negatively related to the size of the manufacturer. Four Stop & Shop, Shaw’s and DeMoulas these share increase in the following order: Kellogg, General Mills, Post , Quaker, Ralston and Nabisco, while for Star Market the order is Kellogg, General Mill, Post, Ralston, Nabisco and Quaker. In the other hand, manufacturers concede more to bigger supermarkets such as Stop & Shop than to smaller ones. Finally, figures 1, 2 and 3 show share-weighted RTEC prices, PCMs and marginal cost over time. They illustrate several important points. First, supermarket RTEC prices dropped after the jawboning campaign that exposed high RTECs price in the media. Second, supermarket PCMs actually increased after the jawboning campaign, meaning that the price drops by RTEC manufacturers were not fully passed on to the consumers. Last, supermarket marginal cost decreased due to lower RTEC wholesale prices after the jawboning campaign. 5. Conclusion This paper decomposed the channel price-cost margins for ready-to-eat cereals in the Boston area at the supermarket chain and brand level. The random coefficients model is used to estimate the demand for 37 RTEC brands which were used to compute the PCMs for retailers and manufacturers under double marginalization scenario. Overall, the results show that RTEC manufacturers capture most of the channel profits (two-third). The leading supermarket chain captures the highest profits among 14 supermarkets even though they charge one of the lowest prices. Further results show that any cut in RTEC wholesale prices are not fully passed on to consumers. 15 Table 1. RTEC Demand Parameter Estimates and Related Statistics, Boston Market. Variable Price Promotion Constant Sugar Calories Fiber Kid Dummy Estimate -26.3720*** 1.2251*** -8.1029*** -0.1336 -3.6308*** -0.0552 -0.0331 t-Statistic 14.6840 7.8179 11.3300 1.3100 13.7010 1.3597 0.1885 Price Constant Sugar Calories Fiber Kid Dummy -1.0394 0.0810 0.7642*** -2.1721*** 0.0281 -0.8062* 0.4504 0.1303 7.6546 11.9030 0.5226 1.9070 Interaction with Age Price Constant Sugar Calories Fiber Kid Dummy 2.7514 -7.2706** 0.0315 0.7189 1.5242*** 1.0390 0.1409 1.9675 0.0419 0.3449 5.7497 1.4241 Interaction with Age Price Constant Sugar Calories Fiber Kid Dummy -33.4760*** 3.3782*** -0.4749** -2.6385*** -0.1437 -0.1702 10.2980 4.0534 2.0142 5.0415 1.1657 0.6685 Means ( β ’s) Standard Deviation One, two and three asterisks indicate significance at 10%, 5% and 1% levels, respectively. The sample consisted of 6,475 observations. 16 Table 2. Own-Price Elasticity Estimates for RTECs in Boston Supermarkets RTEC Brand KApple Jacks KComplete Bran KCorn Flakes KCorn Pops Kcrispix Kfroot Loops Kfrosted Flakes Kfrosted Mini Wheats Kraisin Bran Krice Krispies Kspecial K GMcheerios GMCinammon Crunch GMCoco Puffs GMGolden Grahams GMHoney Nut Cheerios GMKix GMLucky Charms GMMulti Gain Cheerios GMTotal GMTotal Raisin Bran GMWheaties GMApple Cinnamon Pbanana Nut Crunch Pcocoa Pebbles Pfruit Pebbles Pgrape Nuts Phoney Comb Praisin Bran Qcap N Crunch Qoat QToasted N Frosted Wheat Bites N Spoon Size R Cookie Crisp R Corn Chex R Rice Chex Stop & Shop -6.6405 -5.3655 -3.2811 -5.6708 -6.1444 -6.1206 -4.6327 -4.5705 -3.9920 -5.6942 -6.4419 -5.0058 -6.1692 -6.0719 -6.6767 -5.5582 -6.4051 -6.2279 -6.9412 -6.5692 -5.1984 -4.5544 -5.6728 -5.3212 -6.4396 -6.2363 -3.7112 -5.8955 -4.1247 -5.4188 -5.5556 -5.9123 -5.3109 -5.0452 -9.0245 -6.5584 -6.5440 Shaw’s DeMoulas -6.2749 -5.2406 -3.4651 -5.5357 -5.9071 -6.4301 -4.8169 -4.5036 -4.335 -5.5619 -6.5383 -4.6451 -5.7531 -5.9009 -6.1482 -5.4489 -6.1278 -5.7973 -6.8014 -5.9602 -5.4267 -4.4745 -5.8485 -4.6780 -5.7366 -5.5027 -3.7289 -5.3765 -3.8369 -5.1140 -5.0589 -5.5711 -5.0439 -4.6056 -7.9934 -5.8920 -5.9151 -6.1003 -5.2411 -3.6698 -6.2584 -6.7430 -5.7442 -4.5219 -4.1984 -3.9813 -5.7505 -7.0539 -5.6274 -6.4313 -6.4782 -6.6970 -5.3830 -7.4510 -6.5970 -6.6132 -6.8404 -5.2357 -4.9020 -5.7377 -5.6519 -6.2701 -6.1688 -3.9438 -5.8610 -3.9477 -5.3418 -4.3969 -5.8291 -5.2597 -4.9987 -7.6441 -6.8352 -6.8656 17 Star Market -6.9988 -5.7329 -3.8211 -6.3551 -6.9214 -6.2090 -4.9636 -4.9135 -4.5659 -5.9265 -6.9246 -5.7413 -6.4880 -6.4021 -7.3264 -5.7054 -7.4505 -6.4199 -7.6776 -7.1392 -5.1982 -5.4208 -6.2264 -5.7781 -6.9037 -6.9147 -3.8775 -6.4101 -4.4766 -6.3534 -6.0631 -6.8210 -5.5465 -5.1703 -9.8183 -7.3917 -7.3653 Simple Average -6.4132 -5.3731 -3.5408 -5.9380 -6.4600 -6.0292 -4.6814 -4.5460 -4.1693 -5.7088 -6.7311 -5.2300 -6.2106 -6.1986 -6.6719 -5.4832 -6.8292 -6.2419 -6.9933 -6.6610 -5.2701 -4.8597 -5.8140 -5.2986 -6.2581 -6.1531 -3.8183 -5.8425 -4.0501 -5.5108 -5.2996 -6.0016 -5.2750 -4.9670 -8.5863 -6.6881 -6.6977 Table 3: Price-Cost Margins for Supermarkets in Boston ($/lb) RTEC Brand KApple Jacks KComplete Bran KCorn Flakes KCorn Pops Kcrispix Kfroot Loops Kfrosted Flakes Kfrosted Mini Wheats Kraisin Bran Krice Krispies Kspecial K GMcheerios GMCinammon Crunch GMCoco Puffs GMGolden Grahams GMHoney Nut Cheerios GMKix GMLucky Charms GMMulti Gain Cheerios GMTotal GMTotal Raisin Bran GMWheaties GMApple Cinnamon Pbanana Nut Crunch Pcocoa Pebbles Pfruit Pebbles Pgrape Nuts Phoney Comb Praisin Bran Qcap N Crunch Qoat QToasted N Frosted Wheat Bites N Spoon Size R Cookie Crisp R Corn Chex R Rice Chex Simple Average Stop & Shop 0.7133 0.6665 0.6572 0.6865 0.6859 0.6880 0.6842 0.7204 0.7203 0.6908 0.6888 0.6825 0.7010 0.6939 0.6966 0.6917 0.7011 0.6995 0.6963 0.6911 0.7145 0.6731 0.6903 0.7524 0.7018 0.6916 0.6736 0.6883 0.7208 0.6849 0.7318 0.7348 0.7279 0.7370 0.7188 0.6969 0.6966 0.6998 Shaw’s DeMoulas 0.6431 0.6213 0.6200 0.6339 0.6320 0.6359 0.6346 0.6512 0.6508 0.6375 0.6340 0.6299 0.6393 0.6385 0.6364 0.6359 0.6388 0.6378 0.6368 0.6310 0.6475 0.6265 0.6367 0.6676 0.6377 0.6325 0.6276 0.6322 0.6488 0.6318 0.6564 0.6570 0.6531 0.6592 0.6448 0.6349 0.6350 0.6391 0.6433 0.6226 0.6220 0.6382 0.6364 0.6336 0.6347 0.6509 0.6514 0.6379 0.6380 0.6341 0.6436 0.6402 0.6399 0.6368 0.6450 0.6413 0.6372 0.6367 0.6488 0.6292 0.6373 0.6664 0.6405 0.6360 0.6287 0.6354 0.6505 0.6339 0.6555 0.6587 0.6535 0.6591 0.6435 0.6409 0.6411 0.6412 18 Star Market 0.7548 0.7057 0.6880 0.8138 0.6665 0.6753 0.6980 0.7432 0.7892 0.6680 0.6733 0.7055 0.6701 0.7294 0.7112 0.6886 0.6736 0.7042 0.6859 0.6791 0.7448 0.7002 0.6979 0.7373 0.7278 0.7344 0.7384 0.6953 0.7997 0.6882 0.7235 0.7194 0.7234 0.7344 0.7992 0.6700 0.6707 0.6488 Simple Average 0.6886 0.6540 0.6468 0.6931 0.6552 0.6582 0.6629 0.6914 0.7029 0.6586 0.6585 0.6630 0.6635 0.6755 0.6710 0.6633 0.6646 0.6707 0.6641 0.6595 0.6889 0.6573 0.6655 0.7059 0.6769 0.6736 0.6671 0.6628 0.7050 0.6597 0.6918 0.6925 0.6895 0.6974 0.7016 0.6607 0.6608 0.6736 Table 4. Price-Cost Margins for RTEC Manufacturers ($/lb) RTEC Brand KApple Jacks KComplete Bran KCorn Flakes KCorn Pops Kcrispix Kfroot Loops Kfrosted Flakes Kfrosted Mini Wheats Kraisin Bran Krice Krispies Kspecial K GMcheerios GMCinammon Crunch GMCoco Puffs GMGolden Grahams GMHoney Nut Cheerios GMKix GMLucky Charms GMMulti Gain Cheerios GMTotal GMTotal Raisin Bran GMWheaties GMApple Cinnamon Pbanana Nut Crunch Pcocoa Pebbles Pfruit Pebbles Pgrape Nuts Phoney Comb Praisin Bran Qcap N Crunch Qoat QToasted N Frosted Wheat Bites N Spoon Size R Cookie Crisp R Corn Chex R Rice Chex Simple Average Stop & Shop 1.4831 1.3910 1.3824 1.4609 1.4728 1.4457 1.4242 1.5451 1.5517 1.4511 1.4492 1.4050 1.4148 1.4118 1.4057 1.4199 1.4342 1.4407 1.4358 1.4010 1.4023 1.3734 1.4017 1.3213 1.2400 1.2312 1.2363 1.2283 1.2984 1.1842 1.2629 1.2155 1.2047 1.2505 1.1974 1.1743 1.1749 1.3574 Shaw’s DeMoulas 1.4604 1.3769 1.3853 1.4796 1.4768 1.4409 1.4211 1.5147 1.5345 1.413 1.4328 1.3911 1.3992 1.3844 1.3834 1.3949 1.4338 1.4088 1.4215 1.3727 1.3956 1.3774 1.402 1.327 1.229 1.2227 1.2261 1.2221 1.2894 1.1706 1.243 1.2192 1.2011 1.2193 1.1912 1.2123 1.2115 1.3483 1.5071 1.3822 1.3986 1.4504 1.4276 1.4298 1.4262 1.5275 1.574 1.4221 1.4452 1.4000 1.411 1.4052 1.4442 1.3858 1.4481 1.4174 1.4158 1.3843 1.3865 1.3562 1.3914 1.3234 1.2329 1.2263 1.2370 1.2244 1.3146 1.1889 1.2098 1.2355 1.2011 1.2043 1.1621 1.1867 1.1858 1.3505 19 Star Market 1.4633 1.3821 1.4017 1.5190 1.4763 1.4673 1.4152 1.5162 1.5272 1.4417 1.4325 1.4124 1.4230 1.4115 1.4418 1.4081 1.5107 1.4160 1.4474 1.4299 1.4029 1.3789 1.4085 1.3279 1.2482 1.2418 1.2245 1.2258 1.2902 1.1767 1.2486 1.2303 1.2095 1.2512 1.1853 1.2468 1.2362 1.3642 Simple Average 1.4785 1.3831 1.3920 1.4775 1.4634 1.4459 1.4217 1.5259 1.5468 1.4320 1.4399 1.4021 1.4120 1.4032 1.4188 1.4022 1.4567 1.4207 1.4301 1.3970 1.3968 1.3715 1.4009 1.3249 1.2375 1.2305 1.2310 1.2252 1.2982 1.1801 1.2411 1.2251 1.2041 1.2313 1.1840 1.2050 1.2021 1.3551 Table 5. Supermarkets Shares of the Channel Profit Stop & Shop Shaw’s DeMoulas Star Market Mean Kellogg 32.14 30.68 30.49 30.58 30.97 General Mills 32.96 31.29 31.29 31.20 31.69 Post 35.88 33.86 33.79 34.03 34.39 Quaker 37.00 34.88 34.89 35.21 35.49 Nabisco 37.37 35.16 35.30 35.14 35.74 Ralston 37.32 34.63 35.27 35.02 35.56 Mean 35.45 33.42 33.51 33.53 33.97 20 Stop $ Shop Shaws 21 DeMoulas Star Market Dec 7, 97 Nov 9, 97 Oct 12, 97 Sep 14, 97 Aug 17, 97 Jul 20, 97 Jun 22, 97 May 25, 97 Apr 27, 97 Mar 30, 97 Mar 2, 97 Feb 2, 97 Jan 5, 97 Dec 8, 96 Nov 10, 96 Oct 13, 96 Sep 15, 96 Aug 18, 96 Jul 21, 96 Jun 23, 96 May 26, 96 Apr 28, 96 Mar 31, 96 Mar 3, 96 Feb 4, 96 Jan 7, 96 Dec 10, 95 Nov 12, 95 Oct 15, 95 Sep 17, 95 Aug 20, 95 Jul 23, 95 Jun 25, 95 May 28, 95 Apr 30, 95 Prices ($/lb) Figure 1: Share-Weighted Retail RTEC Prices at Boston Supermarkets, April 1995 to December 1997 3.7 Jawboning Campaign 3.5 3.3 3.1 2.9 2.7 2.5 Stop & Shop Shaws 22 DeMoulas Star Market Dec 7, 97 Nov 9, 97 Oct 12, 97 Sep 14, 97 Aug 17, 97 Jul 20, 97 Jun 22, 97 May 25, 97 Apr 27, 97 Mar 30, 97 Mar 2, 97 Feb 2, 97 Jan 5, 97 Dec 8, 96 Nov 10, 96 Oct 13, 96 Sep 15, 96 Aug 18, 96 Jul 21, 96 Jun 23, 96 May 26, 96 Apr 28, 96 Mar 31, 96 Mar 3, 96 Feb 4, 96 Jan 7, 96 Dec 10, 95 Nov 12, 95 Oct 15, 95 Sep 17, 95 Aug 20, 95 Jul 23, 95 Jun 25, 95 May 28, 95 Apr 30, 95 PCM ($/lb) Figure 2: Estimated Share-Weighted Price-Cost Margins for RTEC at Boston Supermarkets, April 1995 to December 1997 0.74 Jawboning Campaign 0.72 0.7 0.68 0.66 0.64 0.62 0.6 Apr 30, 95 Stop & Shop 23 Shaw's DMoulas Star Market Dec 7, 97 Nov 9, 97 Oct 12, 97 Sep 14, 97 Aug 17, 97 Jul 20, 97 Jun 22, 97 May 25, 97 Apr 27, 97 Mar 30, 97 Mar 2, 97 Feb 2, 97 Jan 5, 97 Dec 8, 96 Nov 10, 96 Oct 13, 96 Sep 15, 96 Aug 18, 96 Jul 21, 96 Jun 23, 96 May 26, 96 Apr 28, 96 Mar 31, 96 Mar 3, 96 Feb 4, 96 Jan 7, 96 Dec 10, 95 Nov 12, 95 Oct 15, 95 Sep 17, 95 Aug 20, 95 Jul 23, 95 Jun 25, 95 May 28, 95 MC ($/lb) Figure 3: Estimated Share-Weighted Marginal Cost for RTEC at Boston Supermarkets, April 1995 to December 1997 3.2 Jawboning Campaign 3 2.8 2.6 2.4 2.2 2 References Azzam A.M. and E. Pagoulatos, 1997. “Vertical Relationships: Economic Theory and Empirical Evidence,” Paper prepared for the June 12-13, 1999 International Conference on Vertical Relationships and Coordination in the Food System. Universita Cattolica Del Sacro Cuore, Piacenza, Italy. Berry, Steven T., 1994. “Estimating Discrete-Choice Models of Product Differentiation,” Rand Journal of Economics, 25 No. 2, pp.242-262. Berry, S., J. Levinsohn and A. Pakes, 1995. “Automobile Prices in Market Equilibrium,” Econometrica, 63, No. 4, pp.841-890. Besanko, D. and M.K. Perry, 1993. “Equilibrium Incentives for Exclusive Dealing in a Differentiated Products Oligopoly,” Rand Journal of Economics, 24, No. 4, pp. 646-667. Bresnahan, T., 1987. “Empirical Studies of Industries with Market Power,” in Schmalensee, R., R.D. Willig eds., Handbook of Industrial Organization, Volume II, Amsterdam: North Holland, pp.1011-1057. Choi, S.C., 1991. “Price Competition in a Channel Structure with a Common Retailer.” Marketing Science, 10, No. 4, pp. 271-296. Cotterill, R. W. 1999. “Jawboning Cereal: The Campaign to Lower Cereal Prices.” Agribusiness: An International Journal, 15, No. 2, pp. 197-205. Cotterill, Ronald W. and Haller, Lawrence E. 1997. An Economic Analysis of the Demand for RTE Cereal: Product Market Definition and Unilateral Market Power Effects. University of Connecticut Food Marketing Policy Center Research Report No. 35 (September). Connor, J.M., 1999. “Breakfast Cereals: The Extreme Food Industry,” Agribusiness: An International Journal, 15, No. 2, pp. 257-259. Gal-Or, E., 1991. “Vertical Restraints with Incomplete Information,” Journal of Industrial Economics, pp. 503-516. Gilligan, T.W., 1986. “The Competitive Effects of Resale Price Maintenance,” Rand Journal of economics, 17, No. 4, pp.544-556. Hausman, J., G. Leonard, and J.D. Zona, 1994. “Competitive Analysis with Differentiated Products.” Annales D’Economie et de Statistique, 34: 159-180. 24 Hastings, J.S., 2002. “Vertical Relationships and Competition in Retail Gasoline Markets: Empirical Evidence from Contract Changes in California,” University of California Energy Institute Power, Working Paper No. 84. Kadiyali, V., N. Vilcassim and P. Chintagunta, 1999. “Product Line Extensions and Competitive Market Interactions: An Empirical Analysis,” Journal of Econometrics, 88, pp.339-363. Klein, B. and K.M. Murphy, 1997. “Vertical Integration as a Self-Enforcing Contractual Arrangement,” American Economic Review, 87, pp. 415-420. Kuhn, K.U., 1997. “Nonlinear Pricing in Vertically Related Duopolies,” Rand Journal of Economics, 28, No. 1, pp.37-62. Ma, L.Y., 1997. “An Econometric Analysis of Competition in a Differentiated Product Industry: The U.S. Ready-to-Eat Cereal Industry,” Ph.D. Dissertation. Department Of Agricultural & Resource Economics, University of Connecticut. Manuszak, M. D., 2001. “The impact of Upstream Mergers on Retail Gasoline Markets,” Working Paper, Carnegie Mellon University. Mathewson, G.F. and R.A. Winter, 1984. “An Economic Theory of Vertical Restraints,” Rand Journal of Economics, 15, No. 1, pp.27-38. McFadden, D., 1973. “Conditional Logit Analysis of Qualitative Choice Behavior,” Frontiers of Econometrics, P. Zarembka, eds., New York, Academic Press, pp.105142. McFadden, D. and K. Train, 2002. “Mixed MNL Models of Discrete Response,” Journal of Applied Econometrics, 15, No. 5, pp.447-470. McGuire, T.W. and R.Staelin, 1983. “An Industry Analysis of Downstream Vertical Integration,” Marketing Science, 2, No. 2, pp.161.191. Messinger, P.R. and C. Narashiman, 1995. “Has Power Shifted in the Grocery Channel?,” Marketing Science, 14, No. 2 pp.189-223. Nevo, A., 1998. “Identification of the Oligopoly Solution Concept in a Differentiated Products Industry,” Economics Letters, 59, pp.391-395. Nevo, A., 2000. “A Practitioner’s Guide to Estimation of random Coefficients Logit Models of demand,” Journal of Economics and management Strategy, 9, No. 4, pp.513-548. 25 Nevo, A., 2001. “Measuring Market Power in the Ready-To-Eat Cereal Industry,” Econometrica, 69, No. 2, pp.307.342. Rey, P. and J. Tirole, 1986. “The Logic of Vertical restraints,” American Economic Review, 76, pp.921-939. Romeo, C., 2005. “Estimating discrete joint probability distributions for demographic characteristics at the store level given store level marginal distributions and a citywide joint distribution.” forthcoming Quantitative Marketing and Economics. Scherer, F.M., 1979. “The Welfare Economics of Product Variety, An Application to the Ready-To-Eat Cereals Industry,” Journal of Industrial Economics, 28 (December), pp. 113-134. Schmalensee, R., 1978. “Entry Deterrence in the Ready-to-Eat Breakfast Cereal Industry,” Bell Journal of Economics, 9 No. 4, pp. 305-327. Shaffer, G., 1991. “Slotting Allowances and resale Price Maintenance: A Comparison of Facilitating Practices,” Rand Journal of Economics, 22, No. 1, pp.120-135. Shaffer, G. and D. P. O’Brien, 1997. “Nonlinear Supply Contracts, Exclusive Dealing, and Equilibrium Market Foreclosure,” Journal of Economics and Management Strategy, 6, pp.755-785. Sudhir, K., 2001. “Structural Analysis of Manufacturer Pricing in the Presence of a Strategic Retailer,” Marketing Science, 20, No. 3, pp.244-264. Tirole, J. 1988. The Theory of Industrial Organization, Cambridge: The MIT Press. Villas-Boas, J.M. and Y. Zhao, 2000. “The Ketchup Marketplace: Retailer, Manufacturers and Individual Consumers,” Working Paper, University of California Berkeley. Villas-Boas, B.S., 2002. “Vertical Contracts Between manufacturers and Retailers: An Empirical Analysis,” Working Paper, University of California Berkeley. Waterson, M., 1993. “Vertical Integration and Vertical Restraints,” Oxford Review of Economic Policy, 9, No. 2, pp. 41-57. 26