رهبری قیمت در یک کانال بازاریابی: مطالعه هم انباشتگی

Building upon a multiple-product channel structure, this paper develops a model to test channel price leadership on the basis of time series observations on retail and wholesale prices and using absence of double marginalisation as a criterion for channel price leadership. The model studies strategic pricing decisions in a two-stage (suppliers–retailers) channel, dealing with several products. Possible long-run relationships between wholesale and retail prices are investigated in relation to three cases. Case 1: suppliers have sufficient power vis-à-vis retailers to enforce double marginalisation; Case 2: retailers do not allow suppliers to enforce double marginalisation; and Case 3: one retailer not only keeps its suppliers from double marginalisation, but is also the horizontal price leader vis-à-vis competing retailers. We explicitly take the time series properties into account to derive the testable implications of strategic price interactions in marketing channels. An attractive feature of our methodology is that price leadership can be tested on the basis of time series on retail prices and wholesale prices only. The procedure for testing the long-run causality implications of the model uses the definition of long-run causality as formulated by Bruneau and Jondeau [Oxf. Bull. Econ. Stat. 61 (4) (1999) 545], but does not use their Wald statistic, which suffers from the undesirable properties of the Wald test when there are nonlinear parameter restrictions. To interpret restrictions on the common stochastic trends of retail and wholesale prices, we show that the common stochastic trends and the deviations from the long-run equilibriums must explicitly be assigned to variables in the channel model. If a common stochastic trend consists of both the retail price and the wholesale price of a product, then the suppliers are able to enforce double marginalisation vis-à-vis the retailers (Case 1). If a common stochastic trend consists solely of a product's wholesale price, then the retailers do not allow suppliers to enforce double marginalisation (Case 2). The opposite situation is a common stochastic trend consisting solely of a product's retail price. In this situation, one of the retailers not only keeps its suppliers from double marginalisation, but is also the horizontal price leader in the retail market for the product (Case 3). The model is applied to a typical multiple product channel allowing various vertical and horizontal interactions between the channel members to co-occur.

The relationship between the fast-moving consumer goods industry and the retail sector is changing as a result of increasing concentration on both sides. Although manufacturers have guaranteed distribution through a large number of retailers, their supremacy is waning. Nevertheless, large companies such as Danone, Heinz, Kraft, Nestlé and Unilever in the food sector are still able to exert substantial channel power because of their strong brands, international market coverage and innovative capacities. Producers and wholesalers can make themselves attractive, sometimes even indispensable, partners for retail companies by offering high product quality, excellent logistical services and competitive prices (Meulenberg, 1997). Consequently, many channel studies in the literature on marketing and industrial organisation have been production-oriented (e.g., Gijsbrechts, 1993 and Tirole, 1988): typical applications have been that retailers are passive decision makers and that manufacturers can influence their retailers' decisions through various incentives, pricing schedules, and cooperation (Choi, 1996). Today, however, the industry is facing a few large and powerful retailers. Ranking all types of firms by revenue, the top players in the global retailing industry in 2001 include Wal-Mart Stores (the USA, the world's largest global food retailer, with 4190 stores owned, $195.27 billion turnover and operating in 10 countries), Carrefour (France, 8926 stores owned, $55.3 billion turnover and operating in 24 countries), Ahold (The Netherlands, 8062 stores owned, $44.8 billion turnover and operating in 18 countries), Metro (Germany, 2169 stores owned, $40.09 billion turnover and operating in 22 countries), Tesco (from the UK top 10 of largest companies, owning 907 stores, $32.38 billion turnover and operating in 9 countries), Ito-Yokado (Japan, 35,600 stores owned, $25.85 billion turnover and operating in 19 countries) and Delhaize “Le Lion” Group (Belgium's largest company with 2310 stores owned, $15.15 billion turnover and operating in 10 countries) (www.supermarketnews.com; Dobson & Waterson, 1999). Retailers are often larger than many suppliers, sell a wide variety of substitutes, and are increasingly influencing which goods are distributed, how they are distributed and at what price (Choi, 1996). These observations suggest that certain suppliers are forced to set their prices largely on the basis of their costs and do not have the opportunity to base their prices on anticipated consumer demand and retailers' reaction functions. Marketing studies analysing channel price formation and price leadership differ not only in their approach to the power structure in the marketing channel but also, and more particularly, in the number of decision makers at the different levels of the marketing channel and the type of demand or price function used. Alternative assumptions have been made about the number of decision makers interacting vertically in the marketing channel. Many groundbreaking studies in the infancy of this exciting research field analyse the interaction between one manufacturer and one retailer in an exclusive dealership, i.e., a bilateral monopoly (e.g., Jeuland & Shugan, 1983, Moorthy, 1985 and Shugan, 1985). Later models consider more actors interacting in the marketing channel, such as one manufacturer supplying two or more retailers (Ingene & Parry, 1995), two manufacturers supplying one common retailer Bandyopadhyay & Divakar, 1999, Choi, 1991 and Zenor, 1994, or even the interaction of several manufacturers and retailers Coughlan & Wernerfelt, 1989 and Lee & Staelin, 1997. Our model considers the case of two retailers and many manufacturers, but can accommodate more retailers too. In the vast body of research on price formation and price leadership in marketing channels what has been investigated is the impact of type of demand function on channel pricing strategy. For instance, Gal-Or (1985) showed that in an exclusive dealer channel it is profitable to be a vertical price leader in the associated Stackelberg game if the demand function slopes downward. Moorthy and Fader (1990) confirmed this result for a linear demand function. Lee and Staelin (1997) argued that the distinction between linearity and nonlinearity of the demand function is not as relevant for understanding channel pricing strategy as the implications of the type of demand function on strategic pricing behaviour such as retail passthrough (i.e., the ratio of retail price reduction to the manufacturer price reduction, see Tyagi, 1999). Some types of demand functions imply a retailer's optimal passthrough of less than 100% (Lee and Staelin, 1997 call this vertical strategic substitutability (VSS)), other functional forms imply optimal retail passthrough of greater than 100% (vertical strategic complement (VSC)) (see also Tyagi, 1999). Empirical studies on manufacturer–retailer interaction in the channel have used linear, logit and multiplicative forms for the demand function (e.g., Sudhir, 2001). In his empirical research, Sudhir (2001, p. 256) concluded that “the logit demand model performed much better in terms of fit compared to the multiplicative demand model”. In the tradition of many studies in this field, our model contains linear demand functions. The objective of this paper is twofold. First, we develop a model to study strategic pricing decisions within channels and to determine the profit maximising long-run relationships for the channel members in a two-stage (suppliers–retailers) channel, dealing with several products. Our model accommodates the three well-known strategic interactions as given by the supplier's price leadership in the sense of a Stackelberg equilibrium (Manufacturer Stackelberg, henceforth referred to as MS), the retailer's price leadership in the sense of a Stackelberg equilibrium (Retailer Stackelberg abbreviated as RS), and simultaneous price-setting by supplier and retailer according to a Nash equilibrium (Vertical Nash, abbreviated as VN). All these strategic interactions have in common that the retailer suffers from double marginalisation (e.g., Tirole, 1988, Chap. 4) compared to the situation in which the retailer is a vertically integrated monopolist. In contrast, we consider versions of our model in which the retailer does or does not allow its suppliers to use anticipated consumer demand and retailer's reaction functions for their own (i.e., suppliers') optimisation purposes when setting wholesale prices, i.e., to enforce double marginalisation. If the retailer does not allow this, he behaves as a vertically integrated monopolist by forcing the suppliers to be price-takers. We shall refer to this strategy as “Retailer Dominance” (RD for short). While both interactions, RD and RS, assume that the retailer is the vertical price leader, they differ in that under RD, suppliers are enforced to be price-takers, whereas under RS the retailer still lacks the power to attract the profits foregone by the externality of double marginalisation. Absence of double marginalisation as a basic criterion for classification of price leadership seems theoretically useful, since double marginalisation implies that economic behaviour is not fully rational. It also seems increasingly relevant for real markets because of the concentration in retailing and the increasing importance of private labels. Therefore, testing RD against the interactions with double marginalisation is particularly important when assessing retailer power in the marketing channel. By testing for RD against the strategic interactions with double marginalisation, our paper forms an extension of the many marketing studies that focus solely on the three pricing games arising from MS, RS and VN (e.g., Choi, 1991, Choi, 1996 and Lee & Staelin, 1997). We note that Bandyopadhyay and Divakar (1999) have introduced a hybrid power structure, in which a retailer allows some manufacturers of well-known brands to pursue their own pricing policy in the sense of MS, while simultaneously pursuing his own policy for the other brands in the sense of RS. Kadiyali, Chintagunta, and Vilcassim (2000) have recently extended the set of alternative power structures to a continuum of possible channel interactions between manufacturers and retailers. Since we analyse price leadership per product in our study, our model accommodates hybrid power structures as proposed by Bandyopadhyay and Divakar (1999) too. The second, but central, objective of our paper is to develop a procedure for testing price leadership in marketing channels by examining long-run price relationships in a dynamic (i.e., short-run) context to consider the possible existence and nature of a long-run equilibrium between the time series of a product's wholesale price and retail price as generated by a common stochastic trend. If a meaningful long-run price relationship is found, then one can investigate whether or not wholesale prices and/or retail prices respond to changes in the magnitude by which these prices are out of equilibrium. Based on the assumptions that the marginal opportunity costs of the suppliers generate the stochastic trends while the non-price consumer demand shifts are transitory, three cases appear to be of interest. Case 1: both retail and wholesale prices of the same product respond to the deviations from the long-run price equilibrium; Case 2: only the retail price of the product responds to the deviations from the long-run price equilibrium; and Case 3: only the wholesale price of the product responds to the deviations from the long-run price equilibrium. In Case 1, the suppliers have sufficient power vis-à-vis the retailers to enforce double marginalisation, i.e., they take consumer demand and retail behaviour into account when setting wholesale prices for own optimisation purposes under MS, RS or VN. In contrast, in Case 2, suppliers are unable to enforce double marginalisation, as the retailers do not allow them to influence retail prices beyond the opportunity cost fluctuations they face. Lastly, in Case 3, a single retailer not only dominates the suppliers as in Case 2, but is also the horizontal price leader for the product concerned. An important methodological feature of our paper in developing a model to test channel price leadership on the basis of time series observations on retail and wholesale prices is that when deriving the testable implications of hypotheses about price-setting power in marketing channels we explicitly take the common stochastic trends and the deviations from the long-run equilibriums into account. We will show that if the common stochastic trends and the deviations from the long-run equilibriums are not explicitly assigned to variables in the channel model, it is easy to misinterpret how restrictions on the dynamic vertical price model must be interpreted in terms of economic power in the marketing channel. Another methodological feature of our paper is that we propose a test procedure that uses the definition of long-run noncausality as formulated by Bruneau and Jondeau (1999) to derive the testable implications of our three cases on vertical price leadership. Bruneau and Jondeau's (1999) definition can be seen as complementing the earlier concept of Gonzalo and Granger (1995) of common factors on which restrictions were tested for long-run noncausality. However, in contrast to Bruneau and Jondeau (1999) we avoid formulating nonlinear parameter restrictions to be tested by a Wald statistic. The econometric literature contains many warnings of the undesirable properties of the Wald statistic in the case of nonlinear parameter restrictions, see e.g., Dagenais and Dufour (1991), Critchley, Marriott, and Salmon (1996), Gregory and Veall (1985), Lafontaine and White (1986), and, Philips and Park (1988). The empirical illustration we use is a typical multiple-product channel for six vegetable products in the Netherlands. We will show that one of the products fits into Case 1, four of the products are best described by Case 2, and for the remaining product the retailer being analysed appears to be the horizontal price leader in the consumer market too: Case 3. The paper is organised as follows. After this introduction, we formulate the long-run model (Section 2) and then describe its testable implications for the short-run price system (Section 3). The empirical results are presented in Section 4. Section 5 is the concluding discussion of the analysis.

Building upon a multiple-product channel structure and considering the time-series properties of the variables involved, we have developed a method to detect for different types of price leadership in marketing channels: Case 1: suppliers have sufficient power vis-à-vis retailers to enforce double marginalisation; Case 2: retailers do not allow suppliers to influence retail prices beyond opportunity cost fluctuations; and Case 3: one retailer is not only price leader vis-à-vis the suppliers in the sense of Case 2, but is also the horizontal price leader vis-à-vis the other retailers. To our knowledge, our model is one of the first to recognise that the well-established MS, RS and VN equilibriums imply double marginalisation. Using absence of double marginalisation as a criterion for price leadership in marketing channels seems theoretically meaningful, since double marginalisation implies that economic behaviour is not fully rational. It also seems increasingly relevant for real markets because of the ongoing concentration in retailing and the increasing importance of private labels. Therefore, testing RD against the interaction with double marginalisation is of particular importance when assessing retailer power in the marketing channel. By testing for RD (Cases 2 and 3) against the strategic interactions with double marginalisation (Case 1), which can be further differentiated in MS, RS and VN, our paper forms an extension of the many marketing studies that focus solely on the three pricing games arising from MS, RS and VN (e.g., Choi, 1991, Choi, 1996 and Lee & Staelin, 1997). As far as we know, our study is the first in marketing channel analysis that explicitly takes the time series properties into account to derive the testable implications of strategic price interactions in which the retailer excludes the suppliers from using foresight of consumer and retailer demand behaviour. We did our analysis using time series on retail prices and wholesale prices. Time series observations on non-price demand shift variables were not available, which will often be the case. Fortunately, these data are not required for the analysis of channel price leadership if we can assume that the non-price demand shift is stationary—an assumption that is tested for by the cointegration test as part of our analysis. Moreover, the cointegration test can also be seen as a diagnostic test for the specification of the consumer demand function. The empirical results of our analysis illustrate the usefulness of our model for understanding channel price leadership of a particular retail chain. Marketing channels of four products (iceberg lettuce, cucumber, butterhead lettuce and tomato) are characterised by RD. The wholesale prices of these products are fully dominated by the opportunity costs faced by the producers. For one product (paprika), the retail chain is both vertically and horizontally price leader; it is able to buy at opportunity costs of suppliers and to charge a mark-up that is a given percentage of these costs. This price leadership might result from the fit between the quality image of paprika and the quality strategy of the particular retail chain being investigated. Other paprika buying retailers will follow the prices of the vertical and horizontal price leader. Finally, our analysis indicates that the retail chain was unable to establish price leadership for one product (chicory). A possible explanation might be that market supply was too small to allow the retail chain to take a dominant position while leaving sufficient produce for other buyers in the wholesale market. This study shows that most Dutch vegetable growers were unable to take the lead in a market-driven strategy when selling their produce through the auction system. Although the Dutch horticultural sector has long recognised the importance of a market-driven approach, it has not been able to implement it by using the Dutch auction system that is widely admired abroad. However, at the end of 1996 a majority of the Dutch fruit and vegetable auctions merged into one large cooperative marketing organisation, called “The Greenery”, which aims to create a stronger position in the marketing channel by adequately meeting the needs and wishes of the large retailers. To evaluate the performance of The Greenery, research should be done using the methodology presented in this paper. Our paper raises also some interesting methodological issues. Firstly, in our analysis we used a multiple-product channel framework to derive the hypotheses of interest. However, if there are no cross-price elasticities, i.e., all the B matrices in and are diagonal, then a single-product channel model for each product will suffice. On the basis of the result that rank(Π)=6, we tested whether the six cointegrating relationships between the retail prices and wholesale prices were bivariate. Applying the switching algorithm proposed in Johansen and Juselius (1994), we found that the test statistic strongly rejected the 30 over-identifying restrictions of interest (χ2(30)=238.79, p value≪0.01). Consequently, the B matrices were not diagonal, confirming the importance of considering a multiple-product channel model. Secondly, we recognise that price elasticities have always been important market information for marketing policy. Hence, we may wish to estimate impulse response functions with respect to a VAR or VECM consisting of retail sales, retail prices, merchandising variables and wholesale prices. Unfortunately, in a multiple-product framework such a system becomes rather unwieldy and it becomes difficult to solve the problem of short-run identification. Nevertheless, the approaches reported by Harbo, Johansen, Nielsen, and Rahbek (1998), Kuiper, Pennings, and Meulenberg (2002), Pesaran, Shin, & Smith (2000), Rahbek and Mosconi (1999), Spirtes, Glymour, and Scheines (2001), and Swanson and Granger (1997) are among those offering ways to overcome these problems. The third issue is that, while being able to detect horizontal price dominance (Case 3), in our empirical analysis with one retailer we cannot detect horizontal Stackelberg price leadership, because then the price-leading retailer does not fix the percentage mark-up, but determines it by profit maximisation while taking both the consumer demand function and the reaction functions of the followers into account, so that the retail price will always display error-correcting behaviour. Lastly, an attractive feature of our methodology seems to be that price leadership in marketing channels can be tested on the basis of times series on retail prices and wholesale prices only. However, whereas retail prices can be (and are) collected regularly in the form of scanner data and by household or retail panels, nowadays it is difficult to obtain representative time series of wholesale prices. This drawback seems to have been compounded by increasing coordination in the marketing channel. For this reason, more attention should be paid to the systematic collection of wholesale prices at industry or national level.