Organic label and prots sharing in the French uid milk market
Organic label and prots sharing in the French uid milk market
Organic label and pro…ts sharing in the French ‡ uid milk market Céline Bonnet and Zohra Bouamra-Mechemachey January 2013z Abstract The paper assesses how the value added created by the existence of an organic label in a vertical chain is shared among manufacturers and retailers in the French ‡ uid milk market. Using a struc- tural econometric model of consumer substitution patterns and of vertical relationships, we show that organic label leads to a higher bargaining power in favor of manufacturer. Total margins are heterogeneous among products. They are lower for organic brands (37%) compared to conventional ones (52%).
On the contrary, …rms would bene…t more from organic products than for conventional milk while our results suggest the opposite result for retailers.
Key words: ‡ uid milk market, bargaining power, organic Toulouse School of Economics (INRA, GREMAQ), 21 Allée de Brienne, F-31000 Toulouse France, Tel: +33 (0)5 61 12 85 91, firstname.lastname@example.org yToulouse School of Economics (INRA, GREMAQ), 21 Allée de Brienne, F-31000 Toulouse France, Tel: +33 (0)5 61 12 86 84, email@example.com zWe thank Elodie Moutengou for the helpful work on database and estimations. Any remaining errors are ours. 1
1 Introduction Organic food can be used as a device toward sustainable food supply as it is less harmful to the envi- ronment.
It uses fewer pesticides and arti…cial fertilizers. Moreover, organic food is considered to be healthier and of higher quality by consumers. The market for organic product has continuously increased over time in the US as well as in many European countries. The willingness to pay for organic products has been analyzed in the literature. Consumers are willing to pay more for organic milk in United States (Kiesel and Villa-Boas, 2007). Dhar and Foltz (2005) on US ‡ uid milk data found a more signi…cant consumer valuation for organic milk. Gri¢ th and Nesheim (2008) have shown that the willingness to pay for organic product is quite heterogenous in United Kingdom and product quality is the …rst concerned of consumers of organic products followed by health concerns and environment.
In addition to health and environmental motives, consumers’sociodemographics as well as social features are also key determinants of organic consumption. van Doorn and Verhoef (2011) conduct an experimental analysis in Germany and their analysis suggests that consumers are unwilling to pay a price premium for organic product. Organic products are asociated with higher prosocial bene…ts and the willingness to pay is positively correlated with quality. They do not …nd any correlation between willingness to pay and health perceptions. Finally they show that younger people and women value more the quality of organic product.
Even if organic products were traditionnally sold by specialty shops, conventional retailers has con- tributed to the expansion of organic markets …rst through the listing of organic products in their stores and second by o¤ering organic products under their own retail brand. If consumers’willingness to pay for organic product is well documented, the strategic incentives of food chain to develop such products are less documented. Both large …rms and retailers may have incentive to produce/sell organic products as they can earn higher margins compared to conventional products. For instance, retail margins are in average 25% higher for organic food products sold in a large retail chain in the northeastern United States (Bezawada and Pauwels, 2013).
In this paper, we focus on both …rms and retailers‘incentives to sell organic labeled food products. 2
Incentives to develop such sustainable products for …rms but also for retailers (organic products sold under private retail labels) will depend on their pro…tability and the share of the total margin they can get from their sales. Product di¤erentiation through labeling may lead to a positive total margin, which will bene…t to the whole supply chain. However, depending on the relative market power of retailers compared to …rms, higher prices may not bene…t to upstream …rms.
It may bene…t retailers that can use of their buying power linked to the existence of a large and powerful retail sector (Inderst and Mazzarotto, 2008) and the development of their own privale labels (Berges-Sennou, 2006). On the other hand, when o¤ering organic products, …rms may bene…t from labeling their products because this could help them in increasing their bargaining power with respect to retailers. In this paper, we develop an empirical analysis to assess if organic label helps …rms to increase their bargaining power with respect to retailers and hence, whether …rms or retailers bene…t more from o¤ering organic products (by increasing their margin).
Because the food retailing sector is characterized by the importance of retailers‘store brands, we will also examine if the impact of organic labels di¤ers between national and store brands. The empirical analysis is applied to the French ‡ uid milk market. The global sales of this sector decreased by 10% in the last decade. However, the consumption of organic ‡ uid milk has doubled during the last decade although it remains small. The market share of organic milk sales represented 3.5% in 2001 and reached 7.5% in 2009. This food sector is relevant for the analysis of organic supply chain for two main reasons.
First, milk and dairy products represent 15% of the total French market of organic products, just behind fruit and vegetables with a market share of 17% (Source: Agence Bio, France). More generally, dairy products and more particularly ‡ uid milk is one of the most important organic food product market in the world (along with fresh products). Moreover, ‡ uid milk is one of the main organic food product listed in conventional retailers (with eggs) and more than 80% of the sales of ‡ uid milk (in value) occur in conventional stores (source: Agence Bio, France). Second, the French ‡ uid milk market is characterized by a large market share of store brands.
In France, two major food companies, Lactalis and Sodiaal face the competition of store brands. Retailer brands account for 52% of the market while Lactalis and Sodiaal represent 28% of the ‡ uid milk market.
The analysis explicitly considers the vertical relationships between food companies and retailers. As in Draganska et al. (2010), we consider vertical linear contracting within the supply chain (wholesale price contract) to estimate how the labeling a¤ects the bene…ts and the share of pro…ts between …rms and retailers. We then develop a structural econometric model of the ‡ uid milk vertical chain. Using a ‡ exible demand model, a random coe¢ cient logit model, we infer the substitution patterns between conventional and organic products and national or store brands as well.
We then use a Nash bargaining game to model the ‡ uid milk distribution channel and to assess how margins are split. The methodology is the following. Using demand estimates, we recover retail margins. We are then able to estimate the relative bargaining power of …rms with respect to retailers using exogenous cost variables of ‡ uid milk products and next infer the wholesale margins and therefore total margins of the ‡ uid milk channel.
The data used for the empirical analysis consist in a French representative consumer panel data collected by KANTAR in 2009. We show that organic label is only slightly valued by consumers compared to retailer or brand characteristics in the ‡ uid milk market. In addition, the organic characteristics increases the utility of only half of the consumers. Moreover, we show that they are more price sensitive to store brands compared to …rms’ national brand (as in Bonnet and Requillart (2012) in the french soft drink industry). Taking into account the substituability patterns between brands, we get estimates of the total margin for each brand including organic and non organic brands.
These total margins are heterogeneous among products and vary from 12% to 65% depending on the brands but there are lower for organic brands (37%) compared to conventional ones (52%). The lowest margin occurs for retailer brands both for conventional and organic milk and as for national brands, the retailer margin is lower for organic milk. Moreover, if we decompose the total margin into the margin for …rms and for retailers, estimates show that …rm’ s margins are always higher for organic brands compared to non organic brands while we obtain the opposite result for retailers. If manufacturer can bene…t more from organic products, retailers do not.
Accordingly, the bargaining power estimates show that the relative bargaining power of a …rm is substantially higher for its organic brand compared to its conventional one. Furthermore, the bargaining power is in favor of processors on average but there exist a large heterogeneity depending on 4
the retailer and the manufacturer pair. We also …nd the size of retailers in terms of market shares allows to increase their bargaining power against processors. The paper is organized as follows. We …rst present the data. In section 3, we describe the econometric structural model and results are discussed in section 4. Finally, section 5 gives the main conclusions of the paper. 2 Data on ‡ uild milk purchases We use the 2009 data from a French representative consumer panel data of 21,605 households collected by KANTAR. It is a home-scan data set providing detailed information on all the purchases of food products.
Among other the data set provides characteristics of the good purchases (brand, size, organic label product), the store where it was purchased, the quantity purchased as well as its price. The database is composed of 322,755 purchases of ‡ uid milk products. According to our sample, the average household consumption of conventional ‡ uid milk is 72 liters per person per year and 6 liters of organic ‡ uid milk.
We de…ne the ‡ uid milk market as the relevant market in this study. In the data at stake, Lactalis and Sodiaal that are the two main manufacturers on this market represent 28% of purchases on average over the 13 periods1 considered in 2009 while the market share of private labels is around 52%2 (Table 1). We consider the two main brands produced by Lactalis (Lactel and Bridel) and the main brand for Sodiaal (Candia). Moreover, when organic milk is produced under a national brand (NB), we distinguish among the conventional milk brand and the organic milk one. Similarly we consider two private labels (PL) that di¤er by the organic/conventional characteristic.
Other …rms’brands represent only a small market share. Purchases of these other …rms’NB are aggregated in an outside option good that represents 19% of the market. This outside option also includes soya milk. Consumers can thus substitute one of the NB or PL ‡ uid milk brands with an alternative product. 1 We divided the 2009 year into thirteen periods of four weeks. We then compute average price across products and those thirteen periods.
2 Market shares are de…ned as follows. We …rst consider the total market for ‡uid milk. The market share of a given brand at a given retailer is de…ned as the ratio of the sum of the quantities of the brand purchased at the selected retailer during a given period and the sum of quantities of all brands purchased at all of the retailers in the relevant market during the same period. 5
Table 1: General Descriptive Statistics for Prices and Market Shares. Prices (in euros per litre) Market Shares Mean (std) Mean in % Outside Good 19.1 Fluid milk 1.07 (0.30) 80.9 Non-organic products 0.87 (0.15) 92.4 Organic products 1.35 (0.24) 7.6 National brands 1.13 (0.32) 12.9 Private labels 0.92 (0.20) 57.1 We consider purchases that occur in all retailers’stores in 2009.
Retailers are grocery store chains that di¤er by the size of their outlets as well as by the services they provide to consumers. Five main retailers operate in the French retail sector. Among them, three retailer chains are characterized by large outlets while the two other ones have intermediate size outlets. In addition, we de…ne two aggregates: an aggregate of discounters which typically have outlets of small to intermediate size and propose only basic services and an aggregate of the remaining retailers. Taking into account the set of products carried by each retailer we obtain 46 di¤erentiated products that compete on the market.3 The average price over all products and all periods is 1.07 euros per liter (Table 1).
Organic products represent less than 8% of the ‡ uid milk purchases. However, their market shares have increased in time during the last ten years. Organic production has been multiplied by …ve between 1998 and 2009 and consumption has increased by 21% in value between 2008 and 2009. The average price of organic products exceeds the average price of conventional brands (NBs and PLs) by 55% and this price di¤erence is larger for NBs than for PLs. These numbers are in line with the observed price premium observed in the United States with a premium of 60% for NBs and 75% for PLs in the late 1990 ’ s (Glaser and Thompson (2000)).
More generally, Bezawada and Pauwels (2013) show that the price premium for organic food products sold in a US retail chain ranges from 5 to 182%, with higher premium found for dairy, meat and poultry. We can also note from our data that PLs are approximately 20% cheaper than NBs. 3 From the consumer perspective, a product is the combination of a brand and a retailer. 6
3 Models and methods 3.1 The Demand Model: a random coe¢ cients logit model We use a random coe¢ cients logit model to estimate the demand and the related price elasticities. The indirect utility funtion Vijt for consumer i buying product j in period t is given by Vijt = b(j) + r(j) + ipjt + ilj + "ijt where b(j) and r(j) are respectively brand and retailer …xed e¤ects that capture the (time invariant) unobserved brand and retailer characteristics, pjt is the price of product j in period t; i is the marginal disutility of the price for consumer i, lj is a dummy related to an observed product characteristic (which takes the value of 1 if product j is an organic label product and 0 otherwise), i captures consumer i’ s taste for this organic label and "ijt is an unobserved error term.
We assume that i and i vary across consumers. Indeed, consumers can have a di¤erent price disutility or di¤erent tastes for the organic characteristic. We assume that distributions of i and i are independent and that the parameters have the following speci…cation: i i = + vi where vi = (vi ; vi ) is a 2x1 vector that captures the unobserved consumers characteristics. is a 2 2 diagonal matrix of parameters ) that measures the unobserved heterogeneity of consumers. We assume a parametric distribution for vi denoted by Pv(:) and P is independently and normally distributed with means of ; , and standard deviations of ; .
We can then break down the indirect utility into a mean utility jt = b(j) + r(j) + pjt + lj + jt where jt captures all unobserved product characteristics and a deviation from this mean utility ijt = [pjt; lj] ( vi ; vi ) . The indirect utility is given by Vijt = jt + ijt + "ijt: The consumer can decide not to choose one of the considered products. Thus, we introduce an outside option that permits substitution between the considered products and a substitute. The utility of the outside good is normalized to zero. The indirect utility of choosing the outside good is Vi0t = "i0t. 7
Assuming that "ijt is independently and identically distributed like an extreme value type I distri- bution, we are able to write the market share of product j at period t in the following way (Nevo, 2001): sjt = Z Ajt exp( jt + ijt) 1 + PJt k=1 exp( kt + ikt) ! dP ( 1) where Ajt is the set of consumers who have the highest utility for product j in period t, a consumer being de…ned by the vector ( i; "i0t " iJt).
The random coe¢ cients logit model generates a ‡ exible pattern of substitutions between products that is driven by the di¤erent consumer price disutilities i. Thus, the own and cross-price elasticities of the market share sjt can be written as: @sijt @pkt pkt sijt = ( pjt sjt R isijt(1 sijt) (vi)dvi if j = k pkt sjt R isijtsikt (vi)dvi otherwise: (2) 3.2 Supply models: vertical relationships between processors and retailers We consider the ‡ uid milk vertical channel as a two-tier industry consisting of nf upstream …rms and nr downstream retailers. Each upstream …rm produces a set of goods Gf and each downstream …rm sells Rr products.
We consider the market is composed of J di¤erentiated products where a product is a brand sold in a retailer. The marginal cost of producing a product j is denoted by j while the marginal cost at the retail level is denoted cj. We note pj the retail price of the product j and wj its wholesale price. Retailers’pro…t functions are given by: r (p) = X j2Rr (pj wj cj)Msj(p) (3) where the subscript t is omitted to simplify the notation and M is the total market size. The pro…t of the …rm f from all products sold to retailers is denoted by f : f = X j2Gf (wj j)Msj(p): (4) 8
As in Draganska et al. (2010), we consider that …rms play a two-stage Nash bargaining game. In the …rst stage, each pair of …rms and retailers secretly and simultaneously contracts over the wholesale price of the product j. In the second stage, retailers compete with each other on the …nal ‡ uid milk market and set prices for each product. The game is solved by backward induction. In the second stage, each retailer r maximizes his pro…t r (p). The subgame Nash equilibrium prices of products sold by the retailer r can thus be derived from the …rst order conditions of retailer’ maximization program: sk(p) + X j2Rr (pj wj cj) @sj(p) @pk = 0; 8k 2 Rr (5) Using equation (5), the vector of margins j = pj wj cj for retailer r can be written in matrix notations: r = (IrSpIr) 1 Irs(p) (6) where Ir is a ownership matrix (J J) with element 1 if products j and k are sold by the retailer r and 0 otherwise, Sp is the matrix (J J) of the market share derivatives with respect to retail prices with general element @sj (p) @pk and s(p) is the vector of market shares.
As emphasized by Sha¤er (2001), the main di¢ culty comes from the linkage across negotiations which raises arduous questions and in particular on what each manufacturer knows about their rivals’ contract terms. Indeed, when negotiating, each manufacturer must conjecture the set of terms its rivals have or have been o¤ered. In equilibrium, this conjecture must be correct but out-of-equilibrium beliefs may be important in determining the bargaining outcome. In the cooperative bargaining approach, this problem is solved by assuming that any bargaining outcome must be bilaterally renegotiation proof, i.e.
no processor-retailer pair can deviate from the bargaining outcome in a way that increases their joint pro…t, taking as given all other contracts. Following Marx and Sha¤er (1999) and Sha¤er (2001), we thus assume that bargaining between each retailer-manufacturer pair maximizes the two players’joint pro…t, taking as given all other negotiated contracts. Moreover, we assume that each player earns its 9
disagreement payo¤ (i.e. what it would earn from the sells of their other products if no agreement on this product is reached) plus a share j 2 [0; 1] (respectively 1 j) of the incremental gain from trade going to the retailer (respectively to the manufacturer). We follow Draganska et al. (2010) and we assume that the …rm and the retailer negotiate separatly the di¤erent brands produced by the …rm and that retail prices are not observable when bargaining over the wholesale prices. Then retail prices are considered as …xed when solving for the bargaining solution (cf. Draganska et al. (2010) for a detailed justi…cation of this assumption).
The equilibrium wholesale price for product j is derived from the bilateral bargaining problem between a …rm and a retailer such that each pair of …rm and retailer maximizes the Nash product: r j (wj) dr j j h f j (wj) df j i(1 j ) (7) where f j (wj) and r j (wj) are respectively the pro…ts of the …rm and the retailer on product j with: f j (wj) = (pj wj cj) Msj(p) = jMsj (p) (8) r j (wj) = wj j Msj(p) = jMsj(p) and df j and dr j denote respectively the payo¤s the …rm (respectively the retailer) can realize outside of their negotiations. The retailer could gain dr j if he delists the supplier’ s product j from his stores but contracts with other suppliers.
Similarly, the …rm could get pro…ts df j from the sales of his other products as well as from the sales products to other retailers if the negotiation fails. If the retail prices are …xed during the negotiation process, the disagreement payo¤s df j and dr j are given by: dr j = X k2Rr fjg kM s j k (p) (9a) df j = X k2Gf fjg kM s j k (p) where the term M s j k (p) is the change in market shares of product k that occurs when the product j in no more sold on the market. Those quantities can be derived through the substitution patterns estimated in the demand model as follows: s j k (p) = Z exp( kt + ikt) 1 + PJt l=1nfjg exp( lt + ilt) exp( kt + ikt) 1 + PJt l=1 exp( lt + ilt) dP ( ): 10
Solving the bargaining problem in equation (7) leads to the following …rst order condition: f j df j @ r j (wj) @wj + (1 ) r j dr j @ f j (wj) @wj = 0: (10) Under the assumption that the matrix of prices for …nal commodities is treated as …xed when the wholesale prices are decided during the bargaining process, we have @ r j (wj ) @wj = Msj(p) and @ f j (wj ) @wj = Msj(p) from equation (8). Equation (10) can thus be written f j df j = 1 ( r j dr j ). Using equations (8) and (9a) the following expression can be derived for the bargaining solution: jMsj(p) X k2Rr fjg kM s j k (p) = 1 j j 2 4 jMsj (p) X k2Gf fjg kM s j k (p) 3 5 : (11) Using equation (11) for all products j, we obtain the matrix of …rms’margins: = nf X f=1 nr X r=1 1 (If SIf ) 1 (IrSIr) (12) where the retail margins of general element = nr X r=1 (IrSpIr) 1 Irs(p) is derived from equation (6), If is the (J J) ownership matrix with element 1 if products j and k are sold by the …rm f and 0 otherwise, and S is the (J J) matrix with market shares as diagonal elements and changes in market shares otherwise: S = 2 6 6 6 4 s1 s 1 2 s 1 J s 2 1 s2 s 2 J .
. . . . . ... .
. . s J 1 s J 2 sJ 3 7 7 7 5 : Equation (12) shows the relationship between the wholesale margin on the one hand and the retail margin on the other hand. This relationship …rst depends on the disagreement payo¤s and thus on the market share changes that are determined by the substitution patterns estimated in the demand model. It also depends on the exogenous parameter j, the relative power of the retailer on the …rm when bargaining over the wholesale price. The higher j, the lower share of the joint pro…t the …rm will get from the bargaining.
Additing equations (12) and (6) yields the total margin of the …rm/retailer pair over product j: p c = 2 4 nf X f=1 nr X r=1 1 (If SIf ) 1 (IrSIr) + I 3 5 (IrSpIr) 1 Irs(p) 11
where I is the (J J) identity matrix. Because we do not directly observe …rms’ marginal production costs as well as retailers’ marginal distribution cost, we are not able to analytically determine the bargaining power parameter j: We rather estimate them specifying the overall channel marginal cost Cj for each product j. We follow the following speci…cation for the total marginal cost: Cjt = !jt + jt where ! is a vector of cost shifters and is a vector of error terms that accounts for unobserved shocks to marginal cost. The …nal equation to be estimated is thus given by: p + 2 41 nf X f=1 nr X r=1 (If SIf ) 1 (IrSIr) + I 3 5 (IrSpIr) 1 Irs(p) + (13) We are then able to get an estimates of for each product.
4 Results on demand and vertical relationships 4.1 Demand results We estimated the demand model using individual data and the simulated maximum likelihood method as in Revelt and Train (1998).
This method relies on the assumption that all product characteristics Xjt = (pjt; lj) are independant of the error term "ijt. However, assuming "ijt = jt+ eijt where jt is a product-speci…c error term varying across periods and eijt is an individual speci…c error term, the independance assumption cannot be hold if unobserved factors included in jt (and hence in "ijt) as promotions, displays, advertising are correlated with observed characteritics Xjt. For instance, we do not know the amount of advertising that …rms invest each month for their brand. This e¤ect is thus included in the error term because advertising might play a role in the choice of ‡ uid milk by households.
As advertising is an appreciable share of ‡ uid milk production costs, it is obviously correlated with prices. To solve the problem that omitted product characteristics might be correlated with prices, we use a control function approach as in Petrin and Train (2010). We then regress prices on instrumental variables, that is input prices, as well 12
as exogenous variables of the demand equation: pjt = Wjt + b(j) + r(j) + lj + jt where Wjt is a vector of input price variables, the vector of parameters associated and jt is an error term that captures the remaining unobserved variations in prices. The estimated error term bjt of the price equation includes some omitted variables as advertising variations, promotions, and shelf displays that are not captured by the other exogenous variables of the demand equation and by the cost shifters. Introducing this term in the mean utility of consumers jt allows to capture unobserved product characteristics varying across time.
Prices are now uncorrelated with the new error term jt +"jht bjt. We then write jt= b(j)+ r(j)+ pjt+ lj+ jt+ bjt (14) where is the estimated parameter associated with the estimated error term of the …rst stage. In practice, we use the price indexes for the main inputs used in the production of ‡ uid milk, that is raw milk, energy and packaging. Cost variables in equation (14) include the price indexes of cardboard, cow milk and gazole as it is unlikely that input prices are correlated with unobserved determinants of demand for ‡ uid milks.4 The ‡ uid milk industry only represents a very small share of the demand for those inputs which justify the absence of correlation between input prices and unobserved determinants of the demand for ‡ uid milks.
These variables are interacted with the manufacturers dummies or private label/national brand dummies because we expect that manufacturers obtain di¤erent prices from suppliers for raw materials and that some characteristics of the inputs (e.g. quality of cardboard) depend on the manufacturers. Estimation results of the price equation (14) are presented Table 7 in the Appendix. We can see that the instruments are not weak since almost all of them are signi…cant. We estimated two models (Table 2). Model 1 is the demand model without controlling for the endogeneity problem of prices whereas model 2 controls for it.5 First, the coe¢ cient of the error term 4 These indexes are provide by the French National Institute for Statistics and Economic Studies.
5 Models were estimated using 100 draws for the parametric distribution that represents the unobserved consumer char- acteristics.
is positive and signi…cant. It means that the unobserved part explaining prices is positively correlated with the choice of the alternative and justify the need to control for endogeneity problem. Comparison of results from model 1 and 2 reveals that the price coe¢ cients would be underestimated (in absolute value) without controlling for the endogeneity problem and the estimates of the parameters of the model are robust to instrumentation. On average, the price has a signi…cant and negative impact on utility. Consumers are more sensitive to the price variations of PLs compared to NBs.
This is consistent with the idea that consumers might have more loyalty with respect to NBs than to PLs. Results suggest that households prefer organic products than non-organic products, since the mean coe¢ cient is positive. However the mean value is very low and the standard deviation is relatively higher, meaning that half of households slightly value this characteristic while the others do not.
The brand …xed e¤ects reveal that the private label products give the highest utility to the households with respect to the other products. This might be explained by the fact that consumers are more sensitive to the level of prices than to the brand they consume when purchasing ‡ uid milk. One reason could be that ‡ uid milk is a quite homogeneous product or at least a not too much di¤erentiated product. From the retailer …xed e¤ect estimation results, the e¤ects of purchasing in one of the seven major retailers on consumer utility are heterogeneous. For the …ve main retailers except for one, the average consumer values more the retailer channel than the hard discount stores (which is the reference for the retailer …xed e¤ect).
This may be due to the characteristics of major retailers that generally o¤er a wider range of products not only for the product at stake but also for the entire range of products sold in their stores. However, hard discounters seem to be more valued by consumers compared to the aggregate of other retailers.
From the demand equation estimation results reported in Table 2, own and cross price elasticities of demand among products can be computed. Average own price elasticities by brands and retailers are given in Table (3). Results show that purchases of ‡ uid milk are more sensitive to changes in own 14
Table 2: Results of the random coe¢ cients logit model. Model 1 Model 2 Mean StD Mean StD Price (pjt) 0.0001 (0.0000) 0.0001 (0.0000) PL -8.7441 (0.0001) -8.9949 (0.0002) NB -3.1223 (0.0001) -3.6815 (0.0002) Organic label (lj) 0.0013 (0.0000) 0.0000 (0.0000) 0.0001 (0.0000) 0.0017 (0.0001) Brand …xed e¤ects B1 0.1471 (0.0000) 0.6776 (0.0002) B2 -1.2851 (0.0001) -0.6333 (0.0002) B3 0.0358 (0.0000) 0.6311 (0.0002) PL 5.5131 (0.0001) 5.6454 (0.0002) Retailers …xed e¤ects R1 0.3115 (0.0000) 0.2803 (0.0000) R2 -0.3318 (0.0000) -0.3575 (0.0000) R3 0.2765 (0.0000) 0.2716 (0.0000) R4 0.1037 (0.0000) 0.1140 (0.0000) R5 0.9707 (0.0000) 0.9564 (0.0000) R6 -0.1806 (0.0000) -0.1335 (0.0000) R7 - - Error term (bjt) 0.8246 (0.0003) Log Likelihood -920,309 -920,181 Number of observations 322,755 322,755 S t a n d a rd e rro rs a re in p a re n t h e s is prices for organic milk compared to conventional milk.
That result is in line with previous studies on ‡ uid milk consumption in the litterature (Glaser and Thompson, 2000; Dhar and Foltz, 2005; Alviola IV and Capps Jr., 2010) for United States and (Jonas and Roosen, 2008) for Germany. Moreover, average own price elasticities for NB and PL products suggest that PL purchases are more sensitive to own price changes compared to NB products, respectively -8.02 and -4.16. Own-price elasticities are more homogeneous across retailers since they vary between -5 for the hard discounter agregate and -5.8 for the aggregate of other retailers on average.
Cross price elasticities patterns for conventional versus organic products (cf table 8 in the appendix) show that conventional and organic milk are substitutes and assymetric as in Alviola IV and Capps Jr. (2010). We however …nd that organic milk purchases are less sensitive to a change in conventional milk price than conventional milk to a change in the price of organic milk. Actually, when the price of a conventional milk product increases by 1%, the demand for other conventional milk increases by 0.29% while it has only a marginal impact on the demand for organic milk (0.02%). On the contrary, an increase 15
Table 3: Average own-price elasticities of the brands Brands Characteristic Own price Elasticity Retailers Own price Elasticity NB1 C -3.15 (0.47) R1 -5.09 (2.11) NB2 O -5.22 (0.27) R2 -5.31 (2.41) NB3 C -3.73 (0.59) R3 -5.36 (2.39) NB4 C -3.31 (0.41) R4 -5.39 (2.25) NB5 O -5.85 (0.85) R5 -5.33 (2.43) PL1 C -6.00 (0.74) R6 -5.80 (2.19) PL2 O -10.04 (0.50) R7 -5.01 (2.84) C: Conventinal, O: Organic in the price of an organic milk product will have a signi…cant impact on the purchases of conventional milk products (0.34%) and a limited e¤ect on the demand for other organic milk products. Table 9 in the Appendix presents price elasticities for national and store brands.
We …nd that the demand for private labels always increases more than the demand for national brands when the price of a national or store brand product increases. Moreover, as for the organic/conventional characteristic, substitution patterns are assymetric and PL purchases are more sensitive to changes in prices of NB products than vice versa.
4.2 Bargaining power and price-cost margins Using demand estimates, we compute the retail margins using equation (6). We then estimate parameters of equation (13) using retails margins in order to get the estimated bargaining power parameters of the Nash bargaining game. As expected, the total margin is higher for NB products compared to PL products. As shown in Table 4, the highest total margin obtained for NB1 is three to six times higher than the total margin for the PL. The total margin is always non marginally higher for the conventional milk NB com- pared to the organic ones. However, when splitting these margins between retailers and manufacturers, results with respect to the brand and conventional/organic features di¤er for …rms and retailers.
Results suggest that retailers’margins are higher for conventional ‡ uid milk than for organic ones, respectively 29% and 17% on average. Results also show that the margins of retailers are higher for national brands (28%) than for private labels (15%). Those margins vary across retailers and range between 20% and 16
Table 4: Manufacturers and retailers margins Brands Characteristic Manufacturer’ s margins Retailer’ s margins Total margins NB1 C 27.94 (6.14) 34.56 (5.81) 62.51 (6.46) NB2 O 28.46 (3.02) 20.55 (1.53) 49.01 (3.04) NB3 C 35.58 (12.51) 29.57 (5.74) 65.16 (7.79) NB4 C 31.55 (6.90) 32.73 (4.35) 64.28 (8.25) NB5 O 36.26 (7.48) 18.77 (2.97) 55.04 (8.02) PL1 C - 18.55 (2.88) 18.55 (2.88) PL2 O - 11.96 (1.20) 11.96 (1.20) C: Conventional, O: Organic 29%. On the contrary, manufacturers’margins are slighly higher for organic brands, which suggests that manufacturers may bene…t more for their organic brand compared to their conventional milk ones.
To get more insight on the impact of conventional/organic attributes of ‡ uid milk on the relative bargaining power of manufacturers, we give in Table 5 the bargaining power estimates, that is the pro- portions of the joint pro…t that is captured by retailers. The estimated bargaining power of manufacturers is signi…cantly higher for its organic products than for its conventional ones. Indeed, the bargaining power of retailers is lower for the national brand 2 and 5 than for national brands 1, 3 and 4. We can observe the heterogeneity of this parameter across brand-retailer pairs but the bargaining power is globally in favor of manufacturers.
One might ask why retailers should have incentive to o¤er organic products on their shelves as they seem to be able to exert less bargaining power for organic milk and their margins are lower for these products. As emphazised by Bezawada and Pauwels (2013), when o¤ering organic products in their shelves, retailers may not only increase their sales in the ‡ uid milk product category but they can also increase their store pro…ts by enhancing their long term image through the supply of organic products and di¤erentiate their stores from other retailers’channel.
In order to shed light on the determinants of the bargaining power estimates, we regress the bargaining power parameters on manufacturers and retailers’characteristics.
Table 6 presents the results for two models. The …st one (model A) allows for manufacturer …xed e¤ects and retailer …xed e¤ects and allows thus to assess whether signi…cant di¤erences exist across …rms, and for the organic label as well. We …nd 17
Table 5: Brand-retailer estimates of bargaining power. NB1 NB2 NB3 NB4 NB5 R1 0.60 (0.01) 0.38 (0.00) 0.37 (0.00) 0.47 (0.01) 0.39 (0.00) R2 0.48 (0.01) 0.38 (0.00) 0.81 (0.05) 0.47 (0.01) 0.28 (0.00) R3 0.63 (0.02) 0.41 (0.00) 0.40 (0.00) 0.49 (0.01) 0.36 (0.00) R4 0.46 (0.01) 0.43 (0.00) 0.37 (0.00) 0.54 (0.01) 0.28 (0.00) R5 0.55 (0.01) 0.46 (0.01) 0.44 (0.00) 0.62 (0.02) 0.39 (0.00) R6 0.43 (0.00) 0.38 (0.00) 0.34 (0.00) 0.41 (0.00) 0.28 (0.00) R7 0.62 (0.02) 0.44 (0.00) S t a n d a rd e rro rs a re in p a re n t h e s is that the manufacturer 2 has a signi…cant higher bagaining power than the manufacturer 1.
Moreover signi…cant di¤erences across retailers occur. Retailers 4 and 6 have more bargaining power compared to retailer 1 when they negociate with manufacturers while the bargaining power is lower for retailers 2, 5 and 7. Results also con…rm the negative e¤ect of organic labels on retailers’bargaining power. Retailer size and the existence of private labels seem to play a key role on the level of bargaining power (cf. Inderst and Mazzarotto (2008), Kadiyali et al. (2000) and Sudhir (2001)). In the second model (model B), we thus determine the e¤ect of …rms characteristics rather than …xed e¤ects on the level of bargaining power.
The market share of manufacturers, store brands and retailers as well as the dummy for hard discounters do not have any signi…cant impact on the bargaining power of brand-retailer pairs. However, two brands seem to be strong enough to in‡ uence the bargaining power of retailers. On the contrary, the market share of retailers explain positively the bargainaing power estimates. The higher the retailer market share is, the higher the bargaining power of the retailer. The proxy of the ‡ uid milk shelf plays a negative role on the bargaining power estimates. This result can be interpreted as follows.
When the ‡ uid milk shelf in a retailer store is limited, the retailer may have more bargaining power as the sale area devoted to national brands is reduced in favor of private labels. On the contrary, when this area is larger, retailers have to display more national brands in order to o¤er a wider variety to consumers, which increases the bargaining power of manufacturers.
Given the bargaining power estimates, we are then able to recover manufacturer’ s margins thanks to equation (12). The results are given in Table 4. We …nd that the margins of manufacturers are slightly 18
Table 6: Regression of the bargaining power on manufacturers and retailers characteristics. Model A Model B Manufacturers Manufacturer size -0.104 (0.415) M1 - Retailer size 1.174 (0.404) M2 -0.036 (0.008) Private label share -0.856 (0.609) Retailers Assortment depth -0.007 (0.002) R1 Hard Discounter -0.050 (0.033) R2 0.038 (0.014) Brand R3 0.013 (0.014) B1 - R4 -0.027 (0.014) B2 -0.087 (0.012) R5 0.048 (0.014) B3 -0.063 (0.020) R6 -0.073 (0.014) R7 0.039 (0.020) Organic label -0.121 (0.008) Organic label -0.146 (0.009) Constant 0.510 (0.011) Constant 0.534 (0.059) R2 0.46 R2 0.48 s ig n i…c a n t a t 5 % ; s ig n i…c a n t a t 1 0 % M a n u fa c t u re r s iz e is t h e t o t a l m a rk e t s h a re o f e a ch m a n u fa c t u re r in e a ch p e rio d , R e t a ile r s iz e is s im ila rly d e …n e d , a n d t h e a s s o rt m e n t d e p t h is a p rox y o f t h e ‡u id m ilk s h e lf u s in g t h e m e a n s u rfa c e o f e a ch re t a ile r.
higher for organic products than for conventional ones. The total margins, which is the sum of the margins of manufacturers and retailers, is globally lower for organic products (37% on average compared to 52% for conventional products) and for private labels.
5 Conclusion In this paper, we assess how the value added created by the existence of an organic label in a vertical chain is shared among manufacturers and retailers in the French ‡ uid milk market. First of all, our study contributes to the litterature on the consumption of organic products. We show that French consumers slightly value the organic attribute of ‡ uid milk products on average but only half of households (slightly) value positively this characteristic while the others do not. Moreover, cross price elasticity estimates suggest assymetric pattern for organic and ‡ uid milk purchasing such that organic milk purchases are less sensitive to a change in conventional milk price than conventional milk to a change in the price of organic milk.
Second, we estimate relative bargaining power of upstream …rms with respect to retailers using exoge- nous cost variables of ‡ uid milk products and next infer the total margins and how this margin is split 19
into a wholesale margin and a retail margin. Given the substituability patterns between the di¤erent brands that have di¤erent characteristics (NB, PL, conventional milk, organic milk), we show that organic label leads to lower total margins but to a higher bargaining power of manufacturer. Thus, …rms’margins (respectively retailers’margin) are slighly higher (lower) for organic products compared to conventional milk.
Moreover, we show that retailers’margin on their PL is lower for organic milk. Furthermore, the bargaining power is in favor of processors on average but there exist a large heterogeneity depending on the retailer and the product pair.
This study thus give insight on the availability of …rms to countervail the buying power of retailers through the use of organic labeling. It is based on a structural econometric model. This model assume a Nash bargaining between retailers and manufacturers to take into explicitly model the relative bargaining power of each actor. It also assume linear contract between each of the retailer and manufacturer pair with unobservability of retail prices at the time of the negotiation. These assumptions, even if they may appear restrictive allow for the estimation of bargaining power. In future works, we …rst want to specify an econometric model that relax these asumptions.
More particularly, we want to consider non linear contracting and determine how results are changed when retail prices can be observed by …rms. Second, we also want to consider how results are a¤ected when the manufacturer/retailer pair has the possibility to negotiate on the bundle of brands produced by the retailer compared to a separate negotiation product by product. When negotiated on a bundle, the manufacturer may increase his bargaining power not only on the organic brand but also on his non organic brands.
References Alviola IV, P. A. and Capps Jr., O. (2010). Household demand analysis of organic and conventional ‡ uid milk in the united states based on the 2004 nielsen homescan panel. Agribusiness, 26(3):369 –388. Berges-Sennou, F. (2006). Store loyalty, bargaining power and the private label production issue. European Review of Agricultural Economics, 33(3):315 –335. 20
Bezawada, R. and Pauwels, K. (2013). What is special about marketing organic products? how organic assortment, price, and promotions drive retailer performance. Journal of Marketing, 77(1):31 –51.
Dhar, T. and Foltz, J. D. (2005). Milk by any other name . consumer bene…ts from labeled milk. American Journal of Agricultural Economics, 87:214 –228. Draganska, M., Klapper, D., and Villas-Boas, S. B. (2010). A larger slice or a larger pie? an empirical investigation of bargaining power in the distribution channel. Marketing Science, 29(1):57 –74. Glaser, L. K. and Thompson, G. D. (2000). Demand for organic and conventional beverage milk. Gri¢ th, R. and Nesheim, L. (2008). Household willingness to pay for organic products. CEPR Discussion Papers 6905, C.E.P.R. Discussion Papers.
Inderst, R. and Mazzarotto, N. (2008). Buyer power in distribution. Jonas, A. and Roosen, J. (2008). Demand for milk labels in germany: organic milk, conventional brands, and retail labels. Agribusiness, 24(2):192 –206. Kadiyali, V., Chintagunta, P., and Vilcassim, N. (2000). Manufacturer-retailer channel interactions and implications for channel power: An empirical investigation of pricing in a local market. Marketing Science, 19(2):127 –148. Marx, L. M. and Sha¤er, G. (1999). Predatory accommodation: Below-cost pricing without exclusion in intermediate goods market. RAND Journal of Economics, 30(1):22 –43.
Petrin, A. and Train, K. (2010). A control function approach to endogeneity in consumer choice models. Journal of Marketing Research (JMR), 47(1):3 –13. Revelt, D. and Train, K. (1998). Mixed logit with repeated choices: Households’choices of appliance e¢ ciency level. Review of Economics and Statistics, LXXX(4):647– 657. Sha¤er, G. (2001). 21
Sudhir, K. (2001). Structural analysis of manufacturer pricing in the presence of a strategic retailer. Marketing Science, 20(3):244. van Doorn, J. and Verhoef, P. C. (2011). Willingness to pay for organic products: Di¤erences between virtue and vice foods.
International Journal of Research in Marketing, 28(3):167 –180. 6 Appendix 6.1 Results on price equation Table 7: Results on price equation. Coe¢ cient (Standard Error) Cow milk -0.002 (0.001) Cow milk PL 0.001 (0.001) Cow milk OL 0.002 (0.002) Diesel -0.005 (0.001) Diesel PL 0.005 (0.002) Cardboard 0.023 (0.010) Cardboard Manuf2 -0.022 (0.012) Cardboard PL -0.0261 (0.016) Brand Fixed E¤ects 12.88 (0.000) Retailer Fixed E¤ects 17.87 (0.000) Organic label 0.437 (0.108) R-squared 0.988 Number of observations 590 s ig n i…c a n t a t 5 % ; s ig n i…c a n t a t 1 0 % 6.2 Results on price elasticities 22
Table 8: Own and cross price elasticities for conventional and organic products. Own-price elasticities Aggregated cross-price elasticities Conventional Organic Conventional -4.06 0.29 0.02 Organic -7.24 0.34 0.02 O w n -p ric e e la s t ic it ie s fo r c o n v e n t io n a l a n d o rg a n ic m ilk a re c o m p u t e d a s t h e av e ra g e ov e r p ro d u c t s a n d t im e p e rio d s o f ow n -p ric e e la s t it ic it ie s o f c o n v e n t io n a l o r o rg a n ic p ro d u c t s re s p e c t iv e ly. A g g re g a t e d c ro s s p ric e e la s t ic it ie s a re c o m p u t e d a s t h e g lo b a l ch a n g e e it h e r fo r o t h e r c o n v e n t io n a l m ilk s ( t h e t h ird c o lu m n ) o r o t h e r o rg a n ic m ilk ( t h e fo u rt h c o lu m n ) w h e n t h e p ric e o f a a c o n v e n t io n a l p ro d u c t ( t h e t h ird row ) o r t h e p ric e o f a n o rg a n ic p ro d u c t ( t h e fo u rt h row ) va rie s .
Table 9: Own and cross price elasticities for national brand and store brand products..
Own-price elasticities Aggregated cross-price elasticities National brands Private labels National brands -4.16 0.04 0.27 Private labels -8.02 0.12 0.63 O w n -p ric e e la s t ic it ie s fo r c o n v e n t io n a l a n d o rg a n ic m ilk a re c o m p u t e d a s t h e av e ra g e ov e r p ro d u c t s a n d t im e p e rio d s o f ow n -p ric e e la s t it ic it ie s o f c o n v e n t io n a l o r o rg a n ic p ro d u c t s re s p e c t iv e ly. A g g re g a t e d c ro s s p ric e e la s t ic it ie s a re c o m p u t e d a s t h e g lo b a l ch a n g e e it h e r fo r o t h e r c o n v e n t io n a l m ilk s ( t h e t h ird c o lu m n ) o r o t h e r o rg a n ic m ilk ( t h e fo u rt h c o lu m n ) w h e n t h e p ric e o f a a c o n v e n t io n a l p ro d u c t ( t h e t h ird row ) o r t h e p ric e o f a n o rg a n ic p ro d u c t ( t h e fo u rt h row ) va rie s .