ANOTHER LOOK AT PRICE PROMOTION

ANOTHER LOOK AT PRICE PROMOTION ÃWen Cao wWeekly sales at retail stores exhibit several patterns that the literature onprice promotion does not fully capture.In this paper we develop a simplesymmetric model where duopoly manufacturers distribute through amonopoly retailer to serve consumers with heterogeneous reservationprices.We show that the heterogeneity in consumers’reservation pricescoupled with the retailer’s market power is sufficient to resolve thedeficiency in the literature.We then show that,while pricing patternsunder this model differ significantly from those under a model where theretailer has no market power,the manufacturers’expected profits arethe same in both cases.I.INTRODUCTIONP RICE PROMOTION AT BOTH THE WHOLESALE AND RETAIL LEVELS is a ubiquitous phenomenon.1As Raju et al .[1990]comment,short-term discounts are one of the major competitive weapons in the brand manager’s arsenal.Blattberg et al .[1995]also note that nondurable goods manufacturers spend more money on promotion than on advertising.The most frequently seen price promotions are probably the weekly markdowns at grocery stores.Table I below presents a series of such promotions.2Table I illustrates several prominent and interesting patterns.First,competing brands of the same product are featured in different promotional periods.3That is,in a given week,at most one of the competing brands is featured.For example,Perdue chicken breast was advertised on sale during the week of March 25.Then,during the week of April 1,Tyson chickenr 2011The AuthorThe Journal of Industrial Economics r 2011Blackwell Publishing Ltd and the Editorial Board of The Journal of Industrial Economics.Published by Blackwell Publishing,9600Garsington Road,Oxford OX42DQ,UK,and 350Main Street,Malden,MA 02148,USA.282THE JOURNAL OF INDUSTRIAL ECONOMICS0022-1821Volume LIX June 2011No.2ÃI am grateful to my advisor,Professor Michael R.Baye,for his guidance and encouragement.I also thank two anonymous referees,the Editor,and the participants at the BEPP seminar at Indiana University,Bloomington,U.S.A.for their helpful comments and suggestions.w Author’s affiliation:Department of Decision Sciences and Managerial Economics,Faculty of Business Administration,The Chinese University of Hong Kong,Hong Kong,China.e-mail:claracao@.hk 1Promotions at the wholesale level are called trade deals or trade promotions,and promotions at the retail level are called retail promotions.Among various forms of sales promotions,we concentrate solely on price cuts in this paper.2The data was obtained from weekly sales flyers at the Kroger stores in Bloomington,Indiana,United States.The promotional prices listed in the flyers were typically valid for one week.3This phenomenon is also noted by Lal [1990],who attributes it to collusion between national brands.breast replaced Perdue’s in the flyer.Second,the promotional prices for the same product are similar both across competing brands and across different sales periods,even when the regular prices are different.For instance,Brawny and Scott paper towels had about a $2difference in their regular prices,but their sales prices were both $4.99/pack.Third,the promotional prices represent relatively large discounts from the corresponding regular prices.In our sample above,we see price cuts of 30%–50%off the regular prices.Fourth,each of the brands was sold at the regular price more than half of the times.These patterns naturally invite a search for theoretical explanations.In line with the importance of price promotion,there has been substantial research to rationalize it.Varian [1980]shows that stores may use price promotion to exploit information heterogeneity among consumers.Baye and Morgan [2001]further the analysis by studying the impact of a market for price information on retail pricing strategies on the homogeneous product market it serves.Along a similar line,Narasimhan [1988]analyzes differences in consumers’brand loyalty and shows that firms employ mixed pricing strategy to attract brand switchers.Agrawal [1996],Lal and Villas-Boas [1998],Raju et al .[1990],Rao [1991]and Simester [1997]strengthen this argument of brand loyalty and extend it to more general settings.Conlisk et al .[1984],Sobel [1984],and Pesendorfer [2002],on the other hand,examine price promotion in intertemporal settings and relate it to intertemporal price discrimination and intertemporal demand effects .While these existing studies cast tremendous insights on understanding firms’price promotion decisions,they cannot satisfactorily account for the patterns revealed in Table I.In particular,models with a common reservation price,such as Varian [1980],Lal and Villas-Boas [1996,1998],and Baye and Morgan [2001],predict that everything is on sale every day,yet the depth of sale can be any amount of a random draw from the equilibrium price distribution.Such predictions contradict some of the main character-istics of weekly sales at Kroger:namely,the alternating featuring of different Table IPrice Promotions Observed in KrogerWeekly FlyersProductBoneless Skinless Chicken Breast Orange Juice (64oz Carton)8Roll Paper Towel BrandPerdue Tyson Minute Maid Tropicana Brawny Scott Regular Price$4.99/lb $4.99/lb $3.75/carton $3.89/carton $9.45/pack $7.29/pack 03/25/2004$1.99/lb Not featured Not featured Not featured Not featured $4.99/pack 04/01/2004Not featured $1.99/lb $1.99/carton Not featured $4.99/pack Not featured 04/08/2004Not featured $1.99/lb Not featured $1.99/carton $4.99/pack Not featured 04/15/2004Not featured Not featured Not featured Not featured Not featured $4.99/pack 04/22/2004$1.99/lb Not featured $1.99/carton Not featured Not featured Not featured 04/29/2004Not featured Not featured Not featured Not featuredNot featured Not featured ANOTHER LOOK AT PRICE PROMOTION 283r 2011The AuthorThe Journal of Industrial Economics r 2011Blackwell Publishing Ltd and the Editorial Board of The Journal of Industrial Economics.284WEN CAObrands in different weeks,use of the same sales price for competing brands, and deep discounts from the regular prices.In addition,in the Krogerflyers we observe that both competing brands have positive probabilities to be priced at their regular and promotional prices,whereas models with asymmetry in brand loyalty,such as Narasimhan[1988]and Raju et al. [1990],prescribe that only the stronger brand has a positive probability of being sold at the regular price.Moreover,many studies in the literature, including Varian[1980],Sobel[1984],Narasimhan[1988],Lal[1990],Raju et al.[1990],and Baye and De Vries[1992],abstract away from the channel setting.But as shown by Lal and Villas-Boas[1996,1998]and Baye and Morgan[2001],strategic interaction between channel members plays an important role in their pricing decisions.In this paper,we employ a parsimonious model to reconcile the above inconsistency between the observed sales patterns and theory,and also to analyze how market structure affects promotional decisions.In our symmetric model,4duopoly manufacturers distribute close substitutes through a common monopoly retailer to serve consumers with hetero-geneous reservation prices.We show that the heterogeneity in consumers’reservation prices(reservation price differential hereinafter)coupled with the retailer’s market power is sufficient to drive all the promotional patterns revealed in Table I.We show that the retailer employs pure-strategy Hi-Lo pricing to exploit the reservation price differential,and it only does so when the size of the low valuation segment is large enough.The competing manufacturers,on the other hand,use mixed strategies in wholesale price cuts to compete for the chances of serving low valuation consumers.We also highlight the effects of channel structure on sales patterns and welfare.A powerful retailer in the channel is able to set stickier retail prices that are not linear functions of the wholesale prices.Thus the Hi-Lo retail pricing will not break down even when the wholesale prices are only one penny apart.5On the other hand,if the retailer has no power in the channel, pricing patterns at both the retail and wholesale levels change significantly. There are more price promotions and deeper discounts to benefit the consumers,but the manufacturers’expected profits remain the same.Our model integrates the extant models of price promotion with added features,which allows us to obtain better datafitting equilibrium pricing patterns than those predicted by the existing studies.For example,our base model can be thought as Narasimhan’s[1988]model with an added monopoly retailer and with reservation price differential between loyal 4The symmetry in our model enables us to obtain more general results that do not rely on any asymmetry in the competing manufacturers’market shares,consumers’brand loyalty level,or relative sizes of different consumer segments associated with competing brands.5Thisfinding differs from the results in Agrawal’s[1996]study,where the Hi-Lo price equilibrium can break down when the two manufacturers price close enough to each other.r2011The AuthorThe Journal of Industrial Economics r2011Blackwell Publishing Ltd and the Editorial Board of The Journal of Industrial Economics.ANOTHER LOOK AT PRICE PROMOTION285 consumers and brand switchers.Our direct selling model can be thought as a general version of Varian’s[1980]and the no-loyalty case of Lal and Villas-Boas’s[1998]models that allows reservation price differential.6While none of these three models predicts both afixed promotional and a regular price as seen in the Krogerflyers,our model does.The rest of the paper is arranged as follows.In Section II,wefirst analyze the base model with a powerful retailer,then deprive the retailer of channel power and solve for the resulted equilibrium,and in the end compare the price promotion patterns from these two models.Section III concludes.II.ANALYSISII(i).The Monopoly Retailer ModelConsider a market with two competing brands,A and B,that are sold through a common retailer.The two brands are close substitutes and are produced at the same constant marginal costs.The retailer enjoys market power,which can be a result of its location convenience or its established consumer base,and serves the market with no capacity constraints.Without loss of generality,we normalize the marginal production costs and retailing costs to zero.The demand side is segmented into two parts:consumers with a high reservation price,V H,and those with a low reservation price V L o V H.The difference in the reservation prices is common knowledge to both the retailer and manufacturers.Each brand enjoys a proportion H,where0o H o1/2, of high valuation consumers(shortened as the high hereinafter),due to brand loyalty or random shopping.The low valuation consumers (shortened as the low hereinafter),of proportion L,0o L o1,always buy the brand with the lower retail price.All consumers are assumed to have unit demand.For completeness,the total number of consumers is normalized to1,so that2HþL51.We model the market interaction as a three-stage game.In thefirst stage, the manufacturers simultaneously set wholesale prices w A R and w B R, respectively,where the superscript R indicates that there exists a retailer with market power in the channel.7After observing the wholesale prices,the retailer makes brand stocking and then retail pricing decisions in the second stage.Consumers shop at the retail store in the last stage of the game.6As pointed out by one reviewer,Lal and Villas-Boas[1998]show in one of the extreme cases that the brand with the smallest trade deal will always be priced at the reservation price in equilibrium.This result is,however,driven by the retailers’pricing power,as compared to the reservation price differential in our case.7In this paper,the analysis focuses on the use of linear wholesale price.However,as pointed out by one reviewer,restricting attention to uniform wholesale prices may not be optimal for a manufacturer.r2011The AuthorThe Journal of Industrial Economics r2011Blackwell Publishing Ltd and the Editorial Board of The Journal of Industrial Economics.The game is solved using backward induction.In the last stage,the high and low shop with the following strategies:the high make purchases if their brand is stocked at a price no higher than V H ;the low buy the brand with the lower retail price,if it is less than or equal to V L .Proposition 1below gives the retailer’s optimal stocking and pricing strategies.Proposition 1.Given the consumers’purchasing strategies,the monopoly retailer always stocks both brands in equilibrium,so long as neither wholesale price is higher than V H .If neither wholesale price is below V L ÀH L ðV H ÀV L Þ,then V H will be charged for both brands.If at least one of the wholesale prices is lower than V L ÀH L ðV H ÀV L Þ,the retailer charges V L for the brand with the lower wholesale price and charges V H for the other brand.Proof .See Appendix.Proposition 1shows that the retailer benefits from its market power in both the retail and wholesale markets.In retail,it can employ pure strategies to fix the price at either V H or V L .Since V L o V H ,we can interpret V H as the regular price and V L as the promotional price.Then in the wholesale market,manufacturers have to offer trade deal prices no higher than the participation threshold,V L ÀH L ðV H ÀV L Þ,to motivate the retailer for retail promotion.At any wholesale price higher than this threshold,the retailer will simply pocket the savings from a trade deal and sell only to the high at the regular price V H .8A caveat is that,regardless of how far the trade deal prices are below the participation threshold,the promotional retail price is always V L .Essentially,the sales price is fixed at the level of V L .The difference between the sale price and the participation threshold can be read as the monopoly rent paid to the retailer.This result is consistent with the empirical findings in Blattberg et al .[1995]that retailers have less than 100%pass-through of trade deals.Also,when both wholesale prices are below the threshold,the optimal strategy for the retailer is to set one brand on sale to attract all the low while keeping the other brand at the regular price.Given the retailer’s strategies as characterized in Proposition 1,we derive the manufacturers’equilibrium strategies as follows.Proposition 2.Given consumers’and the retailer’s optimal strategies,there exists a symmetric pure strategy Nash equilibrium where both8When all the high valuation consumers shop randomly,the retailer will find it more profitable to stock only one brand if the difference in the wholesale prices is larger than the reservation price differential.The participation threshold becomes a more restrictive V L À2H L ðV H ÀV L Þin such a case.286WEN CAOr 2011The AuthorThe Journal of Industrial Economics r 2011Blackwell Publishing Ltd and the Editorial Board of The Journal of Industrial Economics.manufacturers price at V H ,if and only if L V H ÀV L ðÞþffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiV H ðV H ÀV L Þp V L H .When L >V H ÀV LðÞþffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiV H ðV H ÀV L Þp V L H ,there exists a symmetric mixed strategy Nash equilibrium where each manufacturer randomly chooses a wholesale pricefrom the following distribution:F ðw R Þ¼0if 0 w R <V H H 1ÀðV H Àw R ÞH w R L if V H H H þL w R V L ÀH L ðV H ÀV L Þ1ÀðV H ÀV L ÞH V L L ÀH H þL V H if V L ÀH L ðV H ÀV L Þ<w R V H 1if w R >V H :8>>>>><>>>>>:Proof .See Appendix.Proposition 2states that the manufacturers can use pure or mixed stra-tegies in trade deals;which one to use depends on consumer characteristics.The existence of pure strategy Nash equilibrium is one of the key differences between our model and the models with a common reservation price,such as Varian [1980],Lal and Villas-Boas [1996,1998]and Baye and Morgan[2001].In those models,firms play e -undercut strategy over the entire continuous support,and there exists no pure strategy equilibrium.In our model,however,each brand is guaranteed a demand of H from the high as long as its price is no higher than V H .The intuition behind the pure strategy equilibrium is that the firms face the trade-off between the profit margin and market share.When the reservation price differential is too large,or when the low segment is too small,the loss in profits from the high cannot be recovered by the added profits generated from the low,so the manufacturer offers no sales and prices at V H .When it is profitable for a manufacturer to serve both the high and low,pure strategy is no longer sustainable due to undercutting in competing for the low.As a result,the manufacturers mix between pricing over the interval V H HH þL ;V L ÀH L ðV H ÀV L Þh i and,with probability ðV H ÀV L ÞH L ÀH H þL V H ,charging theregular price V H .Note that this positive probability of charging the regular price is absent in models with a common reservation price due to lack of secure payoffs,although they do capture the competition effect.Let c w R i,i 5A ,B ,denote the equilibrium trade deal prices,that is,c w R i 2V H H ;V L ÀH ðV H ÀV L Þ and c w R ifollows F (w R ).Also denote the probability of such trade deals as a ,and thus a ¼1ÀðV H ÀV L ÞH V L L ÀH H þL VH .9Table II summarizes 9Since a probability should always be less than or equal to 1,we assume that V L L ÀH H þL V H >0.ANOTHER LOOK AT PRICE PROMOTION287r 2011The AuthorThe Journal of Industrial Economics r 2011Blackwell Publishing Ltd and the Editorial Board of The Journal of Industrial Economics.the pricing decisions from the manufacturers and the retailer,together with the corresponding probabilities and quantities sold.The predicted pricing patterns in Table II suggest the following.First,there are only two retail price levels:the regular price V H and sales price V L .With probability (1Àa )2no sales are offered to consumers.In half of the remaining time,brand A is on sale;and in the other half brand B is on sale.These are consistent with the patterns observed from the Kroger flyers.10Second,in contrast to the sticky retail prices,the wholesale prices fluctuate quite a lot.With probability a 2both manufacturers offer trade promotion to the retailer,with probability 2a (1Àa )one of the manufacturers has a trade promotion,and with probability (1Àa )2both firms charge wholesale priceV H .Apart from V H,c w R A and c w R B can be any price within the support of trade deals,and they can change frequently.Third,changes in retail prices are not always driven by changes in wholesale prices.At any time,it is not optimal for the retailer to promote both brands simultaneously.There can be spikes in retail margins for a particular brand,and thus such margins are not a linear function of wholesale prices.11The results in Table II also highlight resolution of the following incompatibilities in the existing literature:(1)The existence of pure strategy Nash equilibrium.In our model,consumers are segmented by reservation prices.There are times when it is optimal to serve only the high,which constitutes the pure strategy equilibrium.Again,extant models with common reservation price predicts no pure strategy equilibrium.(2)The Table IIPrice Promotion Patterns with a Monopoly RetailerRetail Price PromotionBrand A on Sale Brand B on Sale No Sale Wholesale Pricec w R A ;c w R B Ãc w R A ;V H c w R A ;c w R B ÃÃðV H ;c w R B Þ(V H ,V H )Retail Price(V L ,V H )(V L ,V H )(V H ,V L )(V H ,V L )(V H ,V H )Sale Frequency12a 2a (1Àa )12a 2a (1Àa )(1Àa 2)Expected Min.Retail PriceV L V L V L V L V H Price Paid by the LowV L V L V L V L –Price Paid by the High(V L ,V H )(V L ,V H )(V H ,V L )(V H ,V L )(V H ,V H )Quantity Sold11112H Note:Ã:c w R A <c w R B ;ÃÃ:c w R A >c w R B .10This result is also consistent with the observation that Coke and Pepsi are promoted in supermarkets in alternative weeks,as documented by Krishna [1994].With different parameter values for products in different categories,our model can also explain the differences in price promotion frequencies recorded in Krishna’s [1994]study.11This low correlation between wholesale and retail prices is consistent with the empirical findings by Hosken and Reiffen [2004]that wholesale price changes cannot explain the changes in retail prices.288WEN CAOr 2011The AuthorThe Journal of Industrial Economics r 2011Blackwell Publishing Ltd and the Editorial Board of The Journal of Industrial Economics.ANOTHER LOOK AT PRICE PROMOTION289 existence of a regular price and a promotional price.In our model,the high’s reservation price is interpreted as the regular price,and there is a positive probability that this price will be charged.The existing models leading to continuous equilibrium price distributions predict no regular price.Even with the size asymmetry of the loyal segments introduced(as in Narasimhan [1988]and Raju et al.[1990]),the positive probability of charging a regular price can only be obtained for the stronger brand but not for the weaker brand.(3)Deep discounts.In a model with continuous equilibrium price distribution, the discount depth is a random draw from the distribution and can be as little as a penny.With additional assumptions such as the switchers’not being totally indifferent between the brands,discontinuity in the support of the equilibrium distribution function can be obtained.Such discontinuity,however,is driven by the brand switching behavior;andfirms can price close enough to each other to prevent switching.To the contrary,the discontinuity obtained in our model is the result of the reservation price differential,and it is not affected even when the manufacturers price very close to each other.(4)Stickier retail price.The market power enjoyed by the retailer in our model enables it tofix the sales prices at V L.Few of the existing models to our knowledge predicts such afixed promotional price as observed in the Krogerflyers,while our model does.Models that abstract away from the channel setup cannot explain the low correlation between the wholesale and retail prices.II(ii).The Direct Selling ModelIn the above model,the retailer’s monopoly power affects both the wholesale and retail prices.Here we examine a resale price maintenance(RPM)model, which has no strategic pricing by the monopoly retailer,to demonstrate the impact of channel structure on pricing patterns and welfare.Under the assumption of zero retailing cost,the RPM case is equivalent to the case where manufacturers sell directly to the end consumers without a retailer.12,13Following the RPM literature,we assume retail prices are set equal to the corresponding wholesale prices,w A M and w B M,where the superscript M indicates that the manufacturers are the retail price setters.All other setups remain unchanged from those of Section II(i).Given their reservation prices,the consumers’optimal purchasing strategies remain the same.Without the retailer,each manufacturer’s profit now depends solely on the price charged by the competing manufacturer. The following proposition states the equilibria for the direct selling model. 12We thank one reviewer for pointing out the equivalence between RPM and‘‘direct selling’’by competing manufacturers.We will use these two terms interchangeably in the paper.13This direct selling model is similar to the static version of Sobel’s[1984]model,although Sobel’s[1984]analysis focuses on the intertemporal effects of price promotions offered by durable good manufacturers.r2011The AuthorThe Journal of Industrial Economics r2011Blackwell Publishing Ltd and the Editorial Board of The Journal of Industrial Economics.Proposition 3.Given the consumers’optimal purchasing strategies,there exists a symmetric pure strategy Nash equilibrium where both manufac-turers charge V H ,if and only if L V HÀV L ðÞV L H .When L >V HÀV L ðÞV L H ,there exists a symmetric mixed strategy Nash equilibrium,in which eachmanufacturer randomly chooses a wholesale price from the following distribution:G ðw M Þ¼0if 0 w M <V H H H þL 1ÀðV H Àw M ÞH M if V H H H þL w M V L 1ÀðV HÀV L ÞH V L L if V L <w M V H 1if w M >V H :8>>><>>>:Proof .The proof is similar to the one for Proposition 2and is thus omitted.The manufacturers still face the same trade-offs between profit margin and market share.But with the rent to the middleman (the retailer)removed,the profitability condition,L >V HÀV L ðÞL H ,becomes less restrictive under the direct selling model.The key differences between the mixed strategyequilibrium in Proposition 3and that in Proposition 2are the participation threshold,which now increases to V L ,and sale frequency.This increase to V L makes the manufacturers price more aggressively to attract the low.As a result,the probability of charging the regular price V H decreases.Let d w M i ,i 5A ,B ,be the equilibrium promotional prices charged by the manufacturers,and let b denote the probability of price promotion,that is,b ¼1ÀðV HÀV L ÞH V L L.Table III summarizes the manufacturers’promotion decisions under direct selling.First,with probability b 2,both firms offer Table IIIPrice Promotion Patterns Under Direct SellingRetail Price PromotionBoth Brands on sale Brand A on sale Brand B on sale No Sale Wholesale Priced w M A ;d w M B d w M A ;V H V H ;d w M B (V H ,V H )Retail Priced w M A ;d w M B d w M A ;V H V H ;d w M B (V H ,V H )Sale Frequencyb 2b (1Àb )b (1Àb )(1Àb )2Expected Min.Retail PriceEmin d w M A ;d w M B E d w M A E d w M B V H Price Paid by the Lowmin d w M A ;d w M B d w M A d w M B –Price Paid by the Highd w M A ;d w M B d w M A ;V H V H ;d w M B (V H ,V H )Quantity sold 1112HNote:(1)E d w M A ¼E d w M B ¼R V L V H H H þL w V H H Lw 2dw ¼V H H L ln V L V H þV L L V H H (2)Emin d w M A ;d w M B ¼R V L V H H H þL w V H H Lw 2ÀÁ2dw ¼ðH þL Þ2V 2L ÀðV H H Þ22L 2V 2L290WEN CAOr 2011The AuthorThe Journal of Industrial Economics r 2011Blackwell Publishing Ltd and the Editorial Board of The Journal of Industrial Economics.price promotions,and the low buy the brand with the lower price.The exactlevels of the sales prices can be any random draw from the interval V H HHþL ;V Lh i and can change frequently as the result of price randomization.Second,with probability(1Àb)2bothfirms charge the regular price,and only the high make stly,with probability2b(1Àb),onefirm offers discount,while the otherfirm charges the regular price.The low buy the brand that is on sale. Note that even under this direct selling model,the positive probability for charging the regular price can still be obtained,although there is no longer a positive probability for charging V L,the promotional price.II(iii).Comparisons of Pricing PatternsSince RPM is illegal per se in the U.S.,we would not expect the patterns in Table III to match what we observe on the market.In fact,they do not;the patterns shown in Tables I and III differ in the following aspects.(1)Price stickiness:At the wholesale level,regardless of whether there is a retailer or not,the promotional prices change frequently due to randomiza-tion by the manufacturers.At the retail level,however,there is a stark difference between the two.When the retailer enjoys market power, consumers only observe two prices at the store–the regular price V H and sales price V L.But under direct selling,the100%pass-through makes it no longer optimal to keep the price at V L.Thus the retail prices,especially the sales prices,are much stickier in the model with a monopoly retailer. (2)Discount depth:Since thefixed retail sales price under the monopoly retailer model,i.e.,V L,is the upper bound of the interval of price randomization under direct selling,the retail discount depth is shallower in the former case on average.In addition,manufacturers must offer lower wholesale prices to motivate the monopoly retailer to pass on the savings,so the wholesale discounts are also deeper in such cases.Therefore,the retailer’s profits increase with its(bargaining)power in the channel.(3)Probability for both brands on sale:With probability b2,consumers observe both brands on sale at the store under direct selling.However,it is not optimal for a monopoly retailer to promote both brands at the same time.This implies that competition between the brands isfiercer under direct selling. (4)Average price and consumer surplus:In both models,sales are offered from time to time.Given w A M V L,w B M V L,and b4a,it is easy to see that at both wholesale and retail levels,the frequency of price promotions is lower and discounts are smaller under a monopoly retailer than they are under direct selling.In fact,Table II indicates that the dominant retailer extracts all the surplus from the low byfixing sales price at V L.As a result, consumer surplus is smaller in the model with a monopoly retailer.It is the monopoly retailer’s market power that drives the differences in pricing patterns between the two models.But contrary to what is commonlyANOTHER LOOK AT PRICE PROMOTION291r2011The AuthorThe Journal of Industrial Economics r2011Blackwell Publishing Ltd and the Editorial Board of The Journal of Industrial Economics.。