Structural Analysis of Manufacturer Pricing in The Presence of A Strategic Retailer

Structural Analysis of Manufacturer
Pricing in the Presence of a

Strategic Retailer
K. Sudhir
School of Management, Yale University, 135 Prospect Street, PO Box 208200, New Haven, CT 06520-8200
[email protected]
Abstract
Consumer goods manufacturers usually sell their brands to

consumers through common independent retailers. Theoretical
research on such channel structures has analyzed the optimal
behavior of channel members under alternative assumptions of
manufacturer-retailer interaction (Vertical Strategic Interaction).
Research in Empirical Industrial Organization has focused on
analyzing the competitive interactions between manufacturers
(Horizontal Strategic Interaction). Decision support systems
have made various assumptions about retailer-pricing rules
(e.g., constant markup, category-profit-maximization). The appropriateness of such assumptions about strategic behavior for
any specific market, however, is an empirical question. This
paper therefore empirically infers (1) the Vertical Strategic Interaction (VSI) between manufacturers and retailer, (2) the Horizontal Strategic Interaction (HSI) between manufacturers simultaneously with the VSI, and (3) the pricing rule used by a
retailer.
The approach is particularly appealing because it can be
used with widely available scanner data, where there is no
information on wholesale prices. Researchers usually have
no access to wholesale prices. Even manufacturers, who
have access to their own wholesale prices, usually have limited information on competitors wholesale prices. In the absence of wholesale prices, we derive formulae for wholesale
prices using game-theoretic solution techniques under the
specific assumptions of vertical and horizontal strategic interaction and retailer-pricing rules. We then embed the formulae for wholesale prices into the estimation equations.
While our empirical illustration is using scanner data without wholesale prices, the model itself can be applied when
wholesale prices are available.
Early research on the inference of HSI among manufacturers in setting wholesale prices using scanner data (e.g.,
Kadiyali et al. 1996, 1999) made the simplifying assumption
that retailers charge a constant margin. This assumption enabled them to infer wholesale prices and analyze competitive interactions between manufacturers. In this paper, we
show that this model is econometrically identical to a model
that measures retail-price coordination across brands.
Hence, the inferred cooperation among manufacturers could
be exaggerated by the coordinated pricing (category management) done by the retailer. We find empirical support for
this argument. This highlights the need to properly model
MARKETING SCIENCE 2001 INFORMS
Vol. 20, No. 3, Summer 2001, pp. 244264
and infer VSI simultaneously to accurately estimate the HSI

when using data at the retail level.
Functional forms of demand have been evaluated in terms
of the fit of the model to sales data. But recent theoretical
research on channels (Lee and Staelin 1997, Tyagi 1999) has
shown that the functional form has serious implications for
strategic behavior such as retail passthrough. While the logit
and linear model implies equilibrium passthrough of less
than 100% (Lee and Staelin call this Vertical Strategic Substitute (VSS)), the multiplicative model implies optimal
passthrough of greater than 100% (Vertical Strategic Complement (VSC)). Because passthrough rates on promotions
have been found to be below or above 100% (Chevalier and
Curhan 1976, Armstrong 1991), we empirically test the appropriateness of the logit (VSS) and the multiplicative (VSC)
functional form for the data.
We perform our analysis in the yogurt and peanut butter
categories for the two biggest stores in a local market. We
found that the VSS implications of the logit fit the data better
than the multiplicative model. We also find that for both
categories, the best-fitting model is one in which (1) the retailer maximizes category profits, (2) the VSI is Manufacturer-Stackelberg, and (3) manufacturer pricing (HSI) is tacitly
collusive. The fact that the retailer maximizes category profits is consistent with theoretical expectations. The inference
that the VSI is Manufacturer-Stackelberg reflects the institutional reality of the timing of the game. Retailers set their
retail prices after manufacturers set their wholesale prices.
Note that in the stores and product categories that we analyze, the two manufacturers own the dominant brands
with combined market shares of about 82% in the yogurt
market and 65% in the peanut butter market. The result is
also consistent with a balance of power argument in the
literature. The finding that manufacturer pricing is tacitly
collusive is consistent with the argument that firms involved
in long-term competition in concentrated markets can
achieve tacit collusion.
Managers use decision support systems for promotion
planning that routinely make assumptions about VSI, HSI,
and the functional form. The results from our analysis are
of substantive import in judging the appropriateness of assumptions made in such decision support systems.
(Structural Models; Horizontal Strategic Interaction; Vertical Strategic Interaction; Retailer Pricing; Promotional Planning; New
Empirical Industrial Organization)
0732-2399/01/2003/0244/$05.00
1526-548X electronic ISSN
STRUCTURAL ANALYSIS OF MANUFACTURER PRICING IN THE PRESENCE OF A STRATEGIC RETAILER
1. Introduction
Manufacturers of most consumer goods sell their
brands to consumers through common independent retailers (e.g., supermarkets and convenience stores for
grocery products, department stores, specialty stores
for consumer electronics, sporting goods, etc.). There
has been substantial theoretical research on such channel structures (e.g., McGuire and Staelin 1983, Choi
1991, Lee and Staelin 1997). This stream of research has
analyzed the optimal behavior of channel members under alternative assumptions about vertical strategic interactions between manufacturers and retailers. Researchers of the new empirical industrial organization
(e.g., Kadiyali et al. 1996, 1999) have studied the competitive interactions between manufacturers (horizontal
strategic interaction). Decision support systems for promotion planning make assumptions about the pricing
rules used by retailers (e.g., Silva-Risso et al. 1999 assume constant margin, Tellis and Zufryden 1995 assume category-profit maximization). The appropriateness of such assumptions for any specific market is an
empirical question. Our goal in this paper is to empirically infer (1) the vertical strategic interaction (VSI) between manufacturers and retailer, (2) the horizontal
strategic interaction (HSI) between manufacturers simultaneously with the VSI, and (3) the pricing rule
used by a retailer.
The approach in this paper is particularly appealing because it can be used with widely available scanner data, where there is no information on wholesale
prices. Researchers usually have no access to wholesale prices. Manufacturers have access to their own
wholesale prices, but usually have only limited information on competitors wholesale prices. In the absence of wholesale prices, we derive formulae for
wholesale prices using game-theoretic solution techniques under the specific assumptions of vertical and
horizontal strategic interaction, demand functional
form, and retailer pricing rules. We then embed the
formulae for wholesale prices into the estimation
equations.1 While our empirical illustration is using
1
It is important to note that the derivation methods used to solve

for wholesale and retail prices in this paper are different from approaches used in theoretical papers. Theoretical models derive the
optimal wholesale and retail prices (and the resulting market
MARKETING SCIENCE/Vol. 20, No. 3, Summer 2001
scanner data without wholesale prices, the model itself is easily applicable to the situation where wholesale prices are available by using the derived formulae for wholesale prices as additional estimation
equations in the model.
Relationship with Previous Research
For a quick overview of the relative contribution of
this paper, we compare related papers in the literature in Table 1.
Several papers in the new empirical industrial organization (NEIO) tradition have attempted to infer
competitive interactions underlying pricing behavior
among manufacturers (HSI) using national-level data
(for example, see Roy et al. 1994 and Kadiyali 1996).
In these papers, retailer behavior is not modeled. One
justification for this is that the focus is on competition
at the national level. Furthermore, any particular retailers strategic behavior will be lost in the data aggregation across retailers, and hence, is irrelevant for
the level at which the analysis is done.
Because manufacturers do tailor their marketing
mix to local market conditions (Manning et al. 1998,
Nichols 1987), NEIO researchers have recently
shifted their focus to analyzing HSI among manufacturers in local markets using scanner data. But
at this local level, retailer strategic behavior can
have a significant influence on a manufacturers
price-setting behavior. Therefore, retailers interaction with manufacturers (VSI) needs to be accounted for. A further complication when using scanner
data is that wholesale prices are unavailable to the
researcher. As a result, to analyze manufacturer
wholesale-pricing behavior, the wholesale prices
need to be inferred from the data.
Kadiyali et al. (1996, 1999) infer wholesale prices
by making a simplifying assumption about the VSI.
They assume that the retailer is nonstrategic and
charges an exogenous constant margin. This assumpshares) based on the parameters of the demand model (which are
assumed to be completely known with no econometric error). An
econometric model, however, is used to estimate the parameters of
the demand model assuming that the observed market shares and
retail prices are an outcome of optimal firm behavior conditional
on the underlying parameters of the demand model.
245
SUDHIR
Structural Analysis of Manufacturer Pricing
Table 1 Positioning Against Related Research

ManufacturerManufacturer Interaction
Roy et al. (1994); Kadiyali
(1996)
Kadiyali et al. (1996, 1999)
Besanko et al. (1998)
Cotterill and Putsis (2001)
Manufacturer-Retailer
Interaction
Demand Functional
Form
Wholesale
Price Information Available
Estimated
NA
NA
Linear
NA
Estimated
Bertrand assumed
Bertrand assumed
Manufacturer Stackelberg
Vertical Nash
Manufacturer Stackelberg,
vertical Nash, and retailer Stackelberg models
(with Bertrand among
manufacturers)
1. CV estimated
2. Manufacturer Stackelberg, vertical nash, and
retailer Stackelberg models (but with Bertrand
among manufacturers)
Manufacturer Stackelberg,
vertical Nash
Constant margin
Category profit
Category profit
Linear
Logit
Linear, estimates LAAIDs without structural pricing equations
No
No
No
Category profit
No
Yes
Category profit, brand

profit, constant
margin
Logit (VSS), Multiplicative (VSC)
No
Kadiyali et al. (2000)
Cannot separately
identify from reaction to retailer
Sudhir (2001; this paper)
Estimated
tion implies that wholesale prices can be obtained by

scaling down retail prices by a constant factor. These
researchers have found that manufacturer pricing
(HSI) is more cooperative than Bertrand price competition. In this paper, we show that these models
(with retailer assumed to charge a constant margin)
are econometrically indistinguishable from a model
that simply measures the extent of price coordination
across brands by a strategic retailer. This observation
raises the question as to whether the estimated level
of cooperation among manufacturers could just be coordinated pricing (category management) by the retailer. It, therefore, highlights the need to properly
model and infer VSI simultaneously in order to accurately estimate the HSI when using data at the retail level.
In this paper, we investigate competitive behavior
among manufacturers by allowing for other plausible
manufacturer-retailer interaction and infer the interaction that best fits the data. We allow for two types
of VSI between the manufacturers and the retailer: (1)
manufacturers as Stackelberg leaders and (2) vertical
Nash interaction. We also allow for alternative retailer
pricing rules: (1) category profit maximization, (2)
246
Retailer Pricing
Rule
brand profit maximization, and (3) constant margins.

Using game-theoretic analysis, we infer wholesale
prices conditional on observed retail prices, sales, and
other exogenous variables under each of the above
assumptions. This enables us to analyze competition
among manufacturers (HSI) under alternative assumptions about the VSI.
Besanko et al. (1998) (hereafter referred to as BGJ)
focus on the importance of accounting for the endogeneity of pricing decisions by manufacturers and retailers to obtain unbiased estimates for a logit model
using aggregate data. In contrast to our emphasis on
inferring the strategic behavior of market participants, they assume (1) that manufacturers and the
retailer make simultaneous pricing decisions, i.e., they
assume that VSI is vertical Nash, (2) that the retailer
does category management by maximizing total profits from the category, and (3) that the HSI among
manufacturers is Bertrand competition.
Other contemporaneous research has addressed
some of the issues that we discuss in this paper. Kadiyali et al. (2000) measure the balance of power between manufacturers and retailer by measuring the
share of channel profits. Although they model both
SUDHIR
the HSI between manufacturers and the VSI between

manufacturers and retailer, their econometric model
does not permit one to identify the HSI between manufacturers (an issue of key interest in this paper). 2
Another key difference is that their technique is applicable only when researchers have access to wholesale price data.
Cotterill and Putsis (2001) also test for the manufacturer Stackelberg and vertical Nash VSI between
manufacturers and retailers without wholesale price
data. They also test for constant margin and categoryprofit-maximizing behavior by the retailer. They analyze a number of product categories and find support for both vertical Nash and manufacturer
Stackelberg interaction. However, they generally reject the constant margin assumption. In contrast to
our paper, they assume a linear model of demand,
assume that HSI between manufacturers is Bertrand,
and do not test for brand-profit-maximizing behavior
by the retailer.3
In summary, this is the first paper to simultaneously infer both HSI and VSI in a channel. The technique
can be applied even in the absence of wholesale price
data. Methodogically, a key difference with respect to
Cotterill and Putsis (2001) is that they derive their
supply-side pricing equations similarly to theoretical
researchers, by solving for optimal wholesale and re-
We note that Kadiyali et al. (2000) estimate conjectural variations

(CVs) for VSI and, therefore, do not impose the specific restriction
of leader-follower behavior. Estimating CVs, however, does not necessarily make the model more flexible, because the CV measures
reactions as a constant across all periods, while the derived optimal
reaction in a leader-follower model (that we use here) is flexible to
accommodate changes from period to period based on changes in
demand characteristics (a realistic assumption). Further research is
needed on the appropriateness of the two methods. They also test
specific models such as manufacturer Stackelberg, vertical Nash,
etc., but, unlike the author of this paper, they assume that the HSI
is Bertrand when they test those models. We also note that in the
absence of wholesale price data, a CV cannot be estimated for the
VSI. The specific assumption about VSI (leader-follower, vertical
Nash) is necessary to estimate the model without wholesale price
data.
3They also estimate an LA-AIDS model, but without the structural
pricing equations. Hence it is not possible to choose among the
different VSI using the LA-AIDS model.
tail prices in terms of a deterministic demand model.4

In this paper, we recognize that observed market
shares include a demand-side econometric error that
is known to both firms and consumers, and therefore
will affect retail and wholesale passthrough differently under different types of VSI and HSI. By explicitly deriving the equations in terms of observed
market shares, we account for the impact of demandside econometric error on retail and wholesale
passthrough in a manner consistent with the theoretical structure of the specific model being tested, leading to a more complete structural specification of the
econometric model.
Appropriateness of Demand Functional Form
While functional forms of demand have been evaluated in terms of the fit of the model to sales data,
recent game-theoretic research on channels (Lee and
Staelin 1997, Tyagi 1999) has shown that the functional form of demand has serious implications for
strategic behavior such as retail passthrough. While
the logit and linear model implies equilibrium
passthrough of less than 100% (Lee and Staelin call
this vertical strategic substitute (VSS)), the multiplicative model implies optimal passthrough of greater
than 100% (vertical strategic complement (VSC)).
Since passthrough rates on promotions have been
found to be below or above 100% (Chevalier and Curhan 1976, Armstrong 1991), a functional form with
the VSC property may be appropriate for categories
that have retail passthrough greater than 100%, while
a functional form with the VSS property may be appropriate if the passthrough is less than 100%. In this
paper, we therefore treat the appropriateness of functional form of demand as an empirical issue and
choose from the logit (VSS) or multiplicative (VSC)
demand models.5
4It should be noted that in contrast to our interest in inferring the
VSI and HSI, the primary focus of Cotterill and Putsis was to test
specific models in the theoretical literature on channels. Hence their
use of deterministic demand models was guided by the theoretical
literature. Kadiyali et al. (2000) also use the deterministic demand
model approach when they test specific VSI such as the Stackelberg
or Nash interactions.
5We thank an anonymous reviewer for suggesting the empirical
analysis of demand functional forms. A simple explanation for
247
SUDHIR
In summary, we choose from the following strategic interactions and demand models:
(1) Vertical strategic interaction: manufacturer
Stackelberg and vertical Nash;
(2) Horizontal strategic interaction: Bertrand, tacit
collusion. We also estimate a continuous conjectural
variation (CV) parameter6;
(3) Retailer pricing rule: category-profit-maximization, brand-profit-maximization, constant markup;
(4) Demand functional form: logit, multiplicative.
Section 2 describes the model. Section 3 describes
the data, the estimation procedure, and the results.
Section 4 concludes with a discussion of limitations
of the model and suggestions for future research.
2. Model
Figure 1 represents the schematic model of the market
that we analyze: Two manufacturers sell through a
common retailer to consumers. Consumers make
their choices after observing the retail prices. In the
manufacturer Stackelberg models, the manufacturers
choose wholesale prices first; the retailer chooses retail prices after observing the wholesale prices. In the
vertical Nash models, the manufacturers and the re-
passthrough greater than 100% in some categories is that retailers

may be using these categories as loss-leaders to drive traffic to the
store.
6The CV has been used in previous research in the marketing literature (e.g., Kadiyali et al. 1999, Putsis and Dhar 1999) to measure
HSI. As the name suggests, CVs were originally interpreted as measures of a players conjectures about reactions by rivals. This view
has been discredited in the theoretical literature because such conjectures cannot be consistent and are not meaningful in a static
analysis (Tirole 1988, Carlton and Perloff 1994). Empirical researchers, however, have successfully used CVs as a parameter to measure
deviations from Bertrand behavior, without interpreting them as
conjectures about reactions between players. See Bresnahan (1989)
for the distinction between the theoretical and empirical interpretations of conjectural variations. The menu approach based on theoretical models does not face these criticisms leveled against the CV
approach. Briefly, for a strategic variable such as price, zero CV
indicates Bertrand competition. Positive and negative CVs indicate
more cooperative and more competitive outcomes relative to Bertrand competition, respectively. See Kadiyali et al. (Forthcoming)
for a more detailed discussion.
248
Figure 1
A Schematic Model of the Market
tailer choose their prices simultaneously. Manufacturers choose their wholesale prices simultaneously. In
the figure, one-sided arrows represent sequential decisions and double-sided arrows indicate simultaneous decisions. The two main assumptions embedded in the figure are that (1) there are two
manufacturers in the market and (2) that manufacturers sell through one retailer. We justify both of
these assumptions below.
ASSUMPTION 1. There are two manufacturers in the market.
We focus our empirical analysis on the two largest
brands in the yogurt and peanut butter categories.
This is reasonable for our purposes because the two
largest brands have a very large market share (82%
for yogurt and 65% for peanut butter). Other brands
have relatively low market share of less than 10% in
this market. Private labels have a minuscule market
share in these categories in these stores. Hence, the
use of private labels for strategic purposes is not a
serious issue as in the studies by Kadiyali et al. (2000)
and Putsis and Dhar (1999). While it is possible to
extend model development to additional brands,
solving for the optimal vertical strategic reactions becomes computationally cumbersome.
ASSUMPTION 2. Manufacturers sell through one retailer.
The primary effect of this assumption is that we do
not model any effect of retail competition. This may
SUDHIR
be a serious shortcoming when analyzing categories

where retail competition is severe. Slade (1995) and
Walters and McKenzie (1988), however, have shown
evidence of limited retail competition in some categories. Slade reported, based on her interviews of
grocery chain managers, that the vast majority of
households (over 90%) do not engage in comparison
shopping on a week-to-week basis. Most consumers
visit the same store each week, and hence most competition seems to take place among brands in the
store rather than across stores. Slade also tested this
interview-based evidence statistically on the saltine
cracker market and found that demand at one store
was unaffected by prices in other chains. Walters and
Mackenzie examined the impact of loss-leader pricing
and other promotions on store sales and profits and
concluded that price specials and double-coupon promotions had no significant effect on store traffic. They
found that only one of eight loss-leader categories had
a significant effect on store traffic. In the structural
modeling literature, while analyzing the yogurt category (which is analyzed in this paper), Besanko et
al. (1998) and Vilcassim et al. (1999) have also made
the assumption of no cross-store competition.
To address the issue of whether the assumption is
appropriate for our analysis, we follow Slade in
checking whether a stores sales are affected by prices
at a competing store. For our analysis we used data
from two Every Day Low Price (EDLP) stores, which
happen to be the biggest stores in the market. See
Section 3 for discussion of the data. We ran a simultaneous equations regression with two linear demand
equations for each store. As the two biggest EDLP
stores may also face competition from the promotional (High-Low pricing) stores, we also included data
from a High-Low store. Together these three stores
accounted for 90% of sales in our market. Hence, we
had six equations in our estimation. Each brands
sales in a store were regressed against prices of both
brands at the same store and other stores.7 While
same store price coefficients were significant and of
7In addition to prices of our own and competing stores, we used
displays at the same stores and features at our own and competing
stores (to see if they affected store traffic). We also tested other
functional forms of demand, such as multiplicative and loglinear.
the right sign, other store prices were insignificant in

all six equations. Hence, the assumption seems reasonable for the data that we analyze. Essentially, this
implies that yogurt or peanut butter prices at competing stores have only a very minor effect on the
consumers decision to visit a particular store. However, it is important to recognize that previous research has shown that detecting store competition requires fairly subtle modeling (Bucklin and Lattin
1992). For our purposes, it appears that retail competition is a second-order effect that is not likely to
impact our results.
Demand
As discussed earlier, we consider two functional
forms: the logit demand model with the VSS property and the multiplicative demand model with VSC
property. These functional forms have been used
widely in the marketing literature: The logit demand
model has been used both for modeling household
level data (for example, Guadagni and Little 1983)
and aggregate market share data (for example, Allenby 1989). Recently, the logit model has been
shown to provide similar substantive implications,
whether one uses the model at either the aggregate
(store) or disaggregate (household) level (Allenby
and Rossi 1991, Gupta et al. 1996) under reasonably
realistic conditions.8 The multiplicative demand
model is a fairly popular demand model in the literature (for an example, see the SCAN*PRO-based
model of Christen et al. 1997).
The Logit Model. The utility for brand i in period
t for consumer j is given by
Uijt 0i f fit d dit dp dit pit fp f it p it
pi p it it ijt
Uit ijt
i 1, 2,
(1)
where f it, d it, and p it are the features, displays, and

retail price associated with brand i in period t. 0i is
8A flexible demand model other than the logit is the LA-AIDS
model used by Cotterill et al. (2000). Unfortunately, it is not possible
to derive the exact structural supply-side equations for this model
as we do in this paper.
249
SUDHIR
the intrinsic attraction of brand i, and it is the unobservable (to the econometrician, but observable to
the firm and the consumer) component of utility. U it
is the average utility across consumers for brand i in
period t. We assume that ijt follows an i.i.d. Type-I
extreme value distribution, leading to the logit model.
To allow for the market share of the two inside
brands to expand and contract over different periods
with the choice of the marketing mix, we allow for
nonpurchases (through an outside good) denoted by
i 0. We normalize the utility of the outside good to
zero across periods. That is, we assume U0t 0.9 The
market share for each brand is given by
exp(Uit )
sit
1
exp(U )
2
k1
i 0, 1, 2.
(2)
kt
Therefore, when both brands reduce prices, the total

market share of the two inside brands increases vis-a`vis the outside good and vice versa. Note that when
the marketing-mix coefficients are the same for both
brands the above model reduces to a logit model.
It is easy to see that
ln(sit /s0t ) Uit 0i f fit d dit dp dit pit
fp f it p it pi p it it ,
i 1, 2.
Therefore, rewriting the demand equation in terms
of quantities,
ln(qit ) ln(q0t ) 0i f fit d dit dp dit pit
fp f it p it pi p it it ,
i 1, 2. (3)
Equation (3) is the demand-side estimation equation. The term it serves as the error term in the estimation equation. As discussed in Besanko et al.
(1998) and Villas-Boas and Winer (1999), these error
terms can capture the effects of manufacturer advertising and consumer promotions that we do not explicitly model in this paper.
Modeling the different correlations in ijt between the inside goods
and the outside good using a nested logit model would lead to a
richer and more flexible formulation. However, this prevents solving
for the supply-side estimation equations in closed form.
250
The first derivatives of market share with respect

to prices are
s 1t
p1t
st
p t
s1t
p
2t
s2t
p1t
s2t
p2t
1t s1t (1 s1t )
2t s1t s2t
1t s1t s2t
,
2t s2t (1 s2t )
(4)
where it dp dit fp f it pi .
The Multiplicative Model. The multiplicative demand model was
fii fij dii dij
qit a i pitbii p bij
jt f it f jt dit djt .
For estimation, we used the log-log model, by taking the logs of the above equation on both sides and
adding a normal additive error. The derivatives with
respect to price for the multiplicative model are well
known.
Retailer Model
We derive the pricing equations for the retailer under
different types of VSI for the logit model. For the multiplicative model, the equations can be derived similarly.
For the category-profit-maximizing retailer the objective is to maximize category profits in period t.
Since consumers who buy other than the two brands
that we analyze may contribute to retailer profits, we
assume that the outside good gives a constant contribution to retailer profit. Denoting the margin from the
outside good as m0, the retailer objective is given by
Rt [( p 1t w1t )s1t ( p 2t w2t )s2t m0 s0t ]Mt ,
(5)
where p1t and p2t are the retail prices of Products 1

and 2, w1t and w2t are the wholesale prices of Products 1 and 2 set by the manufacturers, and s1t and s2t
are the shares of Products 1 and 2 defined in the demand model (note that s0t 1 s1t s2t is the share
of the outside good), and M t is the size of the market.
The subscript t refers to the period t.
For the retailer maximizing brand profits, the objective is to maximize each brand is profits in period
t without considering the impact of the chosen price
SUDHIR
on the demand and profits from the other brand. The

objective is given by
itR [( p it wit)s it]M t,
i 1, 2.
ing the first-order conditions give us the pricing

equation
(6)
Because the retailer is a follower in the manufacturer Stackelberg games, the retailer does not incorporate manufacturer reactions when choosing retail
prices. In the vertical Nash game also, the retailer
does not incorporate manufacturer reactions, due to
the simultaneous nature of the game. Hence, the firstorder conditions and the pricing equations are identical for both the manufacturer Stackelberg models
and vertical Nash models.
For the retailer maximizing category profits, the
first-order conditions imply
[]
[]
s2t
0,
p it
(7)
Substituting the price derivatives from (4) and solving the first-order conditions give us the retail-pricing
equation
1
s
s
1t 2t
1t
1t s0t
2t s0t
p w 1
p1t
w1t
2t
Retail
price
2t
2t
m.
s1t
s
2t
1t s0t
2t s0t
(8)
For brand-profit-maximizing case, the first-order

conditions for the retailer are
i 1, 2. (9)
Substituting the price derivatives from (4) and solv
(1 s1t ) 2 (1 s2t )
(1 s1t )(1 s2t ) (s1t s2t ) 2
pt
wt
2t
(1 s2t ) 2 (s1t s2t )
1t (1 s 1t )(1 s 2t ) (s 1t s 2t ) 2
2t
(1
s2t
2t
(10)
Wholesale
price
Retail margin
For the constant retailer margin model, the pricing

equation is simply
m.
p w (1 m)
w
w
p1t
w1t
2t
2t
w1t
2t
w1t
2t
(11)
Retail
margin
In Pricing Equations (8), (10), and (11) the first term

represents the input cost (wholesale price) to the retailer
and the second term represents the retailer margin.
For the manufacturer Stackelberg model, the retailers reactions to manufacturers wholesaler prices are
obtained by taking the derivatives of the retail prices
in (8) and (10). It can be shown that (proof in Appendix) for the manufacturer Stackelberg model with
the retailer maximizing category profits, the reactions
are given by
p 1t
w1t
p t
wt
p 1t
w 2t
p2t
w1t
Wholesale
price
Retail margin
[]
2t
Retail
price
Rt
s
0 sit ( p it wit ) 1t 0,
pit
p it
w1t
Wholesale
price
i 1, 2.
p1t
Rt
s
0 sit ( p 1t w1t m0 ) 1t
p it
p it
( p2t w2t m0 )
p w
1t (1 s1t )
1 s
s
p2t
w2t
1t
2t
s1t 10
.
1 s2t
(12)
For the manufacturer Stackelberg model, with the

retailer maximizing brand profits, the reactions are
given by
10
As we stated earlier, the logit demand model leads to vertical

strategic substitutability (VSS) since pi /wi 1.
1t
(1 s1t ) 2 (s1t s2t )
2t (1 s1t )(1 s2t ) (s1t s2t ) 2

(1 s2t ) (1 s1t )
(1 s1t )(1 s2t ) (s1t s2t ) 2
2
(13)
251
SUDHIR
For the vertical Nash model, both the retailer and

the manufacturers take their decisions simultaneously. Hence, only the direct effect of the change in
wholesale price on retail price through the retailer
input cost (wholesale price) is accounted for but not
the indirect effect through its impact on the choice of
retail margin. Hence, for both the category-profitmaximizing case and the brand-profit-maximizing
case the impact on retail price for a change in wholesale price is given by
p t
1
wt
0
0
.
1
brand-specific unobservable marginal cost at time t.

Note that it is unobservable to the researcher, but
observable to the manufacturers.
The first-order conditions for the manufacturer are
given by
(wjt cjt )
(14)
w1t
pt
wt
p 1t
w 2t
p2t
w1t
1m
p2t
w2t
0
.
1m
sjt p1t
sjt p 2t
0,
p 1t wit
p 2t wit
i 1, 2;
For the constant margin model, the retailer reaction

to a change in wholesale price is
p 1t
M
sit p 1t
sit p 2t
it
sit (wit cit )
wit
p 1t wit
p 2t wit
st
j i,
w p (w c ) 0,
p t st
t
(17)
where
(15)
1 1
1
for tacit collusion and

Manufacturer Interactions
We now develop the model of manufacturer interactions (HSI) and the supply-side estimation equations.
We allow for two alternative manufacturer interactions: (1) tacit collusion and (2) Bertrand competition.11 The objective function of manufacturer i selling
brand i in period t is given by
j i,
(16)
where w it is the wholesale price for brand i that the

manufacturer charges the retailer and cit is the margin
cost of brand i. F it is the fixed cost to the manufacturer (it can include costs that are not related to the
marginal sales of the brand; e.g., slotting allowances).
Note that 1 for the case of tacit collusion and
0 for the case of Bertrand competition. We define
the marginal cost of brand i as c it i it, where
i is the brand-specific margin cost, and it is the
11After choosing the best-fitting model using Vuongs (1989) test for
these two extreme cases, we also estimate the CVs later. For the CV
estimates, refer to the section Implications of the Identification
Problem.
252
0 1
1
for Bertrand competition. The operator denotes element-by-element multiplication of a matrix.

We can thus solve for the wholesale prices as
wt ct
M
it (w it c it )s it M t (w jt c jt )s it M t F it ,
i 1, 2;
p s
t t
wt p t
st ,
(18)
where the term
p s
t t
wt p t
st
is the vector of margins that manufacturers choose

for their brands.
Note that Equation (18) is purely in terms of observable data and model parameters that we will estimate. In the absence of data on wholesale prices, we
can substitute out the expression for wholesale prices
in (18) into the Retail-Pricing Equation (8), (10), and
(11) to get the appropriate supply-side estimation
equation that accounts for manufacturer and retailer
behavior,
SUDHIR
Table 2 Retail Margins

Model
Category profit maximization
pt
Retailer Margin
s1t
s2t
1t 1t 0 2t s0t
1t (1 s1t )
2t (1 s2t )
Vertical Nash
s1t
(1 s1t )
1t
1t s0t
s2t
2t s0t
m0
s1t
s2t
2t s0t
1t s0t
2t
Vertical Nash
Brand profit maximization
1t
2t (1 s2t )
Constant margin retailer
st
Wholesale Margin
rt .
(19)
Retail
margin
The actual estimation equations for the five models

can be obtained by appropriately substituting into the
above model the expression of rt and p t /w t. We
summarize these in Table 2 and Table 3, respectively,
for quick reference. Because these expressions are in
terms of observed market shares, which include the
demand-side econometric error, we have incorporated demand-side econometric errors into the supplyside estimation equations in a manner consistent with
the theoretical structure of the model. This is a key
difference with respect to Cotterill and Putsis (2001),
who use deterministic demand models in deriving
their supply-side estimation equations, and therefore
do not model the structural effects of demand-side
econometric error in their supply-side model.
Specifically, denoting the i, j element of matrices
pt /wt, s t /pt, and as pw ijt, spijt, and ij, we can
write the two pricing-estimation equations by doing
the appropriate matrix algebra as follows:
pt st
wt p t
Wholesale price (wt )
m0
s1t s2t
2t s0t
1t s0t
2t
Manufacturer cost
ct
ww m
1t
2t
Table 3 Retailer Reactions

Retailer Reaction
Model
pt
1 s s
1t
2t
s1t
1 s2t
(1 s1t ) 2 (1 s2t )
(1 s1t )(1 s2t ) (s1t s2t ) 2
(1 s2t ) 2 (s1t s2t )

1t (1 s1t )(1 s2t ) (s1t s2t ) 2
2t
Vertical Nash
10 01
Vertical Nash
10 01
1 0 m
w
1t
(1 s1t ) 2 (s1t s2t )
2t (1 s1t )(1 s2t ) (s1t s2t ) 2
(1 s2t ) 2 (1 s1t )
(1 s1t )(1 s2t ) (s1t s2t ) 2
0
1m
253
SUDHIR
p1t 1 {[12 s2t ( pw11t sp 12t pw12t sp22t )

22 s1t ( pw21t sp12t pw22t sp 22t )]/Dett }
r 1t 1t ,
p2t 2 {[21 s1t ( pw22t sp 21t pw21t sp11t )
we therefore derive the actual estimation equation

that is fit on the data. Then we derive the estimation
equation for the retailer coordination model and
show that the two equations are equivalent.
Substituting the retailer reaction to wholesale prices for the constant-margin model in Equation (15), the
estimation equation becomes
pt ct
11 s2t ( pw12t sp21t pw11t sp 11t )]/Dett }
1
s
t I
1m
pt
st (1 m).
Hence,
r 2t 2t ,
pt ct (1 m)
where
st .
(21)
is
pt 0 t
( pw21t sp12t pw22t sp 22t )

1221 ( pw11t sp 12t pw12t sp 22t )
(20)
Therefore, our estimation equations for the logit

model are the demand equations in (3) and the pricing equations in (20). Note, however, that in the category-profit-maximization case the quantity m0 in retailer margin cannot be separately identified from the
cost intercept (i ). Hence, the intercept estimate in the
pricing equation in the category-profit-maximization
model is the sum of the brand-specific cost and the
per-unit profit from the outside good.
Equivalence of the Constant-Margin Retailer
Model and Retailer Coordination Model. In this section we demonstrate that the constant-margin retailer
model that measures HSI using the CV approach
(that has been used in previous studies) is econometrically identical to an alternative model that ignores
manufacturers and just measures the degree of coordination in pricing across brands by the retailer using the CV approach. To do this, we first show that
the margin parameter (m) in the estimation equation
for the constant-margin retailer is not identified, and
254
Therefore, the final estimation equation in this case
Dett 11 22 ( pw11t sp 11t pw12t sp21t )
( pw21t sp11t pw22t sp 22t ).
st
I
p t
s
t t I
pt
st ,
(22)
where (1 m) and t t(1 m). Of course,

m cannot be identified from this estimation equation.
This estimation equation generalizes the equation derived in Vilcassim et al. (1999) for their linear demand
model to a nonlinear demand model.
We now show that the supply-side estimation equation for a model measuring the degree of coordination in pricing across brands using the CV approach
is econometrically identical to (22). Here we parameterize the level of coordination in retail pricing
across brands.
The objective function of a retailer maximizing profits from brand i in period t is given by
itR ( pit wit)s it M t,
i 1, 2.
(23)
The first-order conditions are given by
sit
sit pjt
itM
0 sit ( p it wit )
0,
p it
pit
pjt p it
i 1, 2;
j i.
Denoting 21 p1t /p2t and 12 p2t /p1t for all

t, we can write the above equation in matrix form as
st
st
I ( p t wt ) 0,
pt
(24)
where
SUDHIR
1
21
12
1
and the operator denotes element-by-element multiplication of a matrix and I is the identity matrix. We
can interpret the ij terms as the measure of coordination in pricing across brands by the retailer. Therefore,
pt wt
s
t I
p t
st .
(25)
If we express wt in terms of factor prices and an

error term, then the estimation Equation (25) is econometrically identical to (22). Hence, the ambiguity about
what the estimates of the s mean, i.e., whether they
measure the extent of retail-pricing coordination or the
degree of cooperation among manufacturers when retailers are nonstrategic and charge a constant margin.
3. Empirical Analysis
Data
We do our analysis on the yogurt and peanut butter
categories on two EDLP stores from two different regional chains in a suburban market.12 These stores
happen to be the two largest in this market. The data
are for a period of 104 weeks during 19911993. In
the yogurt market, we analyze competition between
the two largest brands, Dannon and Yoplait. Together, they constitute about 82% of sales in this category in this market. In the peanut butter market, we
analyze competition between the two largest brands
in the market: Skippy and Jif. Together they constitute about 66% of sales in this category. The next
largest national brand has less than 10% of market
share in both categories in these stores. Private labels
have a very small market share in these stores.
Because store-level data do not provide any information on the share of the no purchase alternative,
12
These stores claimed to be EDLP. We verified these claims by comparing the variance of prices between these self-professed EDLP
chains and self-professed high-low chains in this market. As expected, the price variance in the EDLP chains is about half of that
of high-low chains. Furthermore, consistent with Bell and Lattin
(1998), EDLP chains had a lower average price than the high-low
chains.
we follow BGJ in using information from the household-level data to compute the share of no purchase. Assuming that the panel is representative of
the universe of consumers, BGJ take the share of no
purchase as the proportion of store visits by the panelists in the data that did not lead to a purchase in
the category.13
Estimation
Our estimation equations are the demand equations
in (3) and the pricing equations in (20). As discussed
earlier, details such as consumer promotions and advertising are not available in scanner data and are
unobserved by the researcher and therefore captured
by the demand error terms jt. However, manufacturers, retailers, and consumers have information on
these, and therefore they affect the manufacturers
and retailers choice of prices and the consumers decision to buy. Villas-Boas and Winer (1999) and BGJ
offer evidence that this is indeed true. We do the same
variation of the Hausman (1978) test that BGJ use and
find price is endogenous. We use lagged price as an
instrument in conducting this test.
Because features and displays are also strategic decisions made by the retailer, theoretically they should
be considered endogenous. Putsis and Dhar (1999)
found that features and displays are indeed endogenous. We test for endogeneity of features and display
in two ways. First we follow Villas-Boas and Winer
(1999) and look at the correlation between estimated
residuals (using both three-stage least square (3SLS)
and full-information maximum likelihood (FIML),
where only price was treated as endogenous and feature and display were treated as exogenous) and feature and display. The correlations were not significant
(p values .2), indicating that there would be no
endogeneity bias if feature and display were treated
as exogenous. One may note that in our demand
model, we allow for interaction effects between price
and feature and between price and display. This in13Because multiple purchases in a store visit are fairly common in
the yogurt category, in this category we also reweight store visits
by the number of purchases made during the visit in computing
the no purchase share. The results, however, are not very sensitive
to the reweighting.
255
SUDHIR
teraction variable was crucial in eliminating the endogeneity bias problem.14 The second test we do is a
variation of the Hausman test. We estimate two models with price treated as endogenous using 3SLS: (i)
with feature and display as endogenous and using
instruments (lagged feature and display) and (ii) with
feature and display as exogenous. If feature and display were indeed exogenous, then both models would
be consistent, but the first models estimates would
have a higher variance because it is an instrumental
variable method. However if feature and display were
endogenous, only the first model would be consistent.
If q denotes the vector of parameters to be estimated,
1 V
2)1(q1 q2),
then the test statistic (q1 q2)(V
where V is the estimated covariance matrix, is distributed as 2(#q) distribution, where #q is the number of parameters in the vector q. We find that the
null hypothesis of exogeneity cannot be rejected (p
.17).
This endogeneity of prices, however, implies that
we still need to use an instrumental variables estimation procedure. Furthermore, errors across the
supply and estimation equations are correlated due
to price. We therefore need to use a simultaneous
equation instrumental variables estimation procedure
such as FIML or 3SLS. The major advantage of FIML
estimation is that we do not need to provide instruments for prices, because the optimal instruments are
generated within the estimation procedure. To compute the optimal instruments in FIML, however, we
need to make the normal distribution assumption on
the error terms. That is, we assume that 1t, 2t, 1t,
2t are assumed multivariate normal. In 3SLS there is
no need to make the multivariate normal assumption,
but we need to provide instruments. We prefer FIML
in this paper, because formal tests for nonnested
models (Vuong 1989) are available using likelihood
estimates.
14We briefly explain the intuition for why modeling the interaction
can help in eliminating the endogeneity bias problem. Let y1 x11
x22 x1x23 be the true model. Suppose x1 is endogenous,
but x2 is not. Instrumenting x1 will eliminate the endogeneity bias,
but suppose we only fit the model y1 x11 x22 1, then 1
x1x23. In this case we will find that cov(1, x1) 0 and
cov(1, x2) 0. Therefore, it will seem that both variables are endogenous if we do not include the interaction term.
256
However, to check whether our results are robust

to the choice of estimation procedure, we also estimated the model using nonlinear 3SLS, using lagged
prices as instruments for prices. The results that we
report are consistent with both types of estimation
procedures. We, however, report only FIML-based estimates, because we use these to do our tests of model
selection.
Results
Model Selection: Choice of Functional Form. We
formally test whether the difference in log-likelihoods
significantly favors one model using the Vuongs
(1989) test of model selection for nonnested models.
The best-fitting multiplicative model in the yogurt
category had a log-likelihood of 343.09 for Store 1
and 423.9 for Store 2. For the peanut butter category,
the best-fitting multiplicative model had a log-likelihood of 320.14 for Store 1 and 421.39 for Store 2.
Clearly, the logit demand model performed much better in terms of fit compared with the multiplicative
demand model. This is especially true considering
that there were three extra parameters in the multiplicative model compared with the logit model. We
performed the Vuongs test for this case, and reject
the multiplicative model at p 0.001. It therefore appears that VSS implication of the logit model fits the
data better than VSC implication of the multiplicative
demand model.
However, an alternative explanation for the poorer
fit of the multiplicative model could be that it also fits
the demand equations rather poorly, compared with
the logit model. We therefore estimated only the demand models using both functional forms using
3SLS.15 The logit model had a higher sum of squared
errors than the multiplicative model, indicating that
the multiplicative model with a greater number of parameters fits the data better if we use purely the demand equations. Hence, the reduction in fit when we
include the pricing equations can be attributed to the
explanation that the supply-side implications of the
logit model (VSS) fit the data better than the supplyside implications of the multiplicative model (VSC).
15
Note that we cannot use FIML if we do not have the same number
of endogenous parameters and equations.
SUDHIR
Table 4 Model Likelihoods and the Vuong Test Statistics
Manufacturer
Interaction
Tacit collusion
Bertrand competition
Retailer Objective/ManufacturerRetailer Interaction

Vertical Nash
Vertical Nash
Constant margin
Vertical Nash
Vertical Nash
Constant margin
Likelihoods for Yogurt

(Vuong Test Statistic)
Likelihoods for Peanut Butter

(Vuong Test Statistic)
Store 1
Store 2
Store 1
Store 2
196.90
247.27
240.66
212.96
()
205.19
(2.259)
202.38
(2.028)
208.25
(2.643)
207.29
(2.529)
203.66
(2.040)
208.25
(2.643)
266.78
(6.558)
204.39
(2.147)
211.03
(2.949)
()
298.29
(2.101)
292.19
(1.972)
321.17
(2.529)
309.35
(2.318)
249.9
(2.030)
309.35
(2.318)
318.84
(2.489)
310.93
(2.347)
325.85
(2.608)
()
289.24
(2.734)
259.44
(1.700)
288.82
(2.722)
297.95
(2.969)
271.11
(2.164)
288.82
(2.722)
266.78
(2.005)
291.55
(2.798)
303.71
(3.114)
()
265.47
(2.132)
257.41
(1.961)
275.28
(2.322)
280.95
(2.426)
269.08
(2.204)
275.28
(2.322)
277.74
(2.368)
301.75
(2.772)
307.85
(3.821)
Note: The Vuong test statistic is with respect to the best-fitting model (with the highest log-likelihood).
The VSS implication that retail passthrough is less

than 100% appears to be intuitive for the two categories that we analyze, since yogurt or peanut butter
are not usually used as loss-leaders to drive store
traffic. This implication is particularly valid for this
data set, because our earlier analysis revealed that yogurt and peanut butter prices had a very limited impact on competitors sales. Hence, these categories are
unlikely to have been used to build store traffic.
Model Selection: Results for the Logit Model. The
likelihoods for the different models of the yogurt and
peanut butter markets for the logit functional form
are reported in Table 4. Based on the log-likelihoods
it seems obvious that, for both stores, the manufacturer Stackelberg model where manufacturers are involved in tacit collusion and the retailer maximizes
category profits fits best in both categories. The
Vuong statistics for the different models are computed with respect to this best-fitting model. Since the
Vuong statistic is distributed as a standard normal,
the critical value for rejecting a model at a p value of

0.05 is 1.645. Given that all of the Vuong statistics are
greater than 1.645, we can reject the other models in
favor of the best-fitting model.
Retailer Objective. The stores that we analyze were
part of large regional chains. Category management
was becoming an increasingly popular concept at the
beginning of the 1990s. In a survey of retailers, McLaughlin and Hawkes (1994) found that many large
retailers that had the technological infrastructure to
collect and analyze data had adopted the category
management concept. Hence the result that retailermaximized category profits has face validity. Our results also support the assumption widely made in the
theoretical literature that retailers maximize category
profits (Choi 1991, Lee and Staelin 1997).
Manufacturer-Retailer Interaction (VSI). The result that the manufacturer Stackelberg model best fits
the data is consistent with the institutional practice
257
SUDHIR
Table 5a Estimates for Best-Fitting Model in Yogurt Category

Store 1
Intercept (Dannon)
Intercept (Yoplait)
Feature
Display
FeaturePrice
DisplayPrice
Price
Marginal cost (Dannon) outside
good retail margin
Marginal cost (Yoplait) outside
good retail margin
Store 2
Parameter
SE
Parameter
SE
2.740
2.920
2.580
2.122
4.378
0.800
6.364
0.563
0.473
0.939
2.227
1.356
3.447
0.633
4.870
6.170
2.750
0.950
3.230
0.230
10.060
4.539
4.616
2.479
4.078
3.883
6.507
7.669
0.495
0.617
0.349
1.783
0.639
3.445
0.678
9.180
7.490
7.110
2.290
6.080
1.890
11.310
0.337
0.042
8.100
0.281
0.050
5.620
0.457
0.045
10.250
0.428
0.053
8.060
Table 5b Estimates for Best-Fitting Model in Peanut Butter Category

Store 1
Intercept (Skippy)
Intercept (Jif)
Feature
Display
FeaturePrice
DisplayPrice
Price
Marginal cost (Skippy) outside
good retail margin
Marginal cost (Jif) outside good
retail margin
Parameter
SE
Parameter
SE
10.414
10.523
3.470
14.400
4.942
12.544
10.539
1.538
1.528
4.303
2.327
4.985
2.284
1.326
6.770
6.890
0.810
6.190
0.990
5.490
7.950
9.504
9.616
8.807
5.023
8.066
2.944
9.049
2.162
2.142
2.376
2.970
2.221
2.895
1.831
4.400
4.490
3.710
1.690
3.630
1.020
4.940
0.718
0.054
13.220
0.798
0.082
9.760
0.714
0.056
12.870
0.771
0.086
9.020
in which manufacturers announce their wholesale

prices and the retailers choose their retail prices as a
response to these wholesale prices (Quelch and Farris
1983). This sequential approach has also been used
in theoretical models of manufacturer-retailer pricing
interactions (McGuire and Staelin 1983, Agrawal
1996).
Some theoretical research (e.g., Choi 1991) has put
forth the argument that leadership in the manufacturer-retailer interaction may be a result of power balance between manufacturers and retailers. Using this
argument, BGJ argue that vertical Nash is an appropriate assumption because increasingly there is balance of power between manufacturers and retailers.
Our result that manufacturer Stackelberg leadership
258
Store 2
may be a reflection of the fact that the balance of

power in these categories is in favor of manufacturers.
Given the dominance of the two leading brands in
these categories, our result also has substantial face
validity on the basis of the power argument. In an
interview by Manning et al. (1998), one executive says
it tellingly: We cant walk away from the kind of
shares that they have in some of the categories, so we
have to work with them and we cant afford not to
promote these products. 16
16
The author notes that Lee and Staelin (1997) show that Stackelberg
leadership leads to greater profitability (and therefore implies greater power) only in the case of demand models characterized by VSS.
The power balance argument the author makes holds only because
SUDHIR
Manufacturer-Manufacturer Interaction (HSI). A

cooperative outcome (relative to Bertrand competition) can be achieved in a noncooperative game by
using appropriate punishment strategies when firms
deviate from cooperative behavior. Such cooperative
behavior is easier to sustain under certain market
conditions than others. Besanko et al. (1996) provide
an excellent discussion on how such coordination can
be achieved in a noncooperative setting. A highly
concentrated market facilitates cooperative behavior.
This is because coordination is easier among a small
number of firms. Further, detection of deviations
from cooperative behavior is relatively easy, making
the threat of punishment strategies more credible,
thus enforcing cooperation. Because Dannon and
Yoplait have a high concentration in the yogurt market (82% in our data), it is not surprising that these
firms behave cooperatively relative to Bertrand competition. The same arguments hold for Skippy and
Jif, who collectively have over 66% market share in
the stores that we analyzed.
Demand and Cost Estimates for the Best-Fitting
Model. We report the results of our estimation for the
best-fitting models in the yogurt and peanut butter
category in Table 5a and 5b (see p. 258), respectively.
All of the coefficients have the expected signs, giving
the analysis substantial face validity.
As expected, features and display have a positive
effect on sales in both categories and stores. Price has
a negative effect on sales. In the final estimations we
did not estimate brand-specific price coefficients, because it turned out in the estimation that they were
not significantly different from each other when we
incorporate the price-feature and price-display interaction effects. It is well known that a decrease in price
when accompanied by a feature or display has a
greater impact on sales than their separate effects.
This was confirmed in that the interaction terms between price and display and between price and feature are negative and significant.
The demand intercept for Yoplait is marginally
he has inferred that the VSS logit model is more appropriate than
the VSC multiplicative model. We thank a reviewer for suggesting
that we highlight this.
higher than for Dannon. However, the estimate of

manufacturing cost plus retailer margin from outside
good for Dannon is about 30% lower than that of
Yoplait. This implies that Dannon is the low-cost
leader.17 These results are similar to those of BGJ, even
though our estimates are on a very different data set.
However, the differences in the estimates of manufacturing cost plus retailer margin from the outside
good indicates that Store 1 gets a greater margin from
the outside good in the yogurt category.
In the case of the peanut butter category, the demand intercept for Skippy is marginally lower than
for Jif. The estimate of manufacturing cost plus retailer margin from outside good for Skippy is marginally lower than that of Jif. However, the differences in the estimates of manufacturing cost plus
retailer margin from the outside good indicate that
Store 2 gets a greater margin from the outside good
in the peanut butter category.
We report estimates of elasticities for the two categories in Table 6a and 6b. Across both stores, we
find that Yoplait has greater price elasticity than Dannon. Further, Yoplait has greater cross-elasticity
with respect to Dannons prices than vice versa.
This indicates a very strong market position for Dannon, and it is not surprising when we consider that
Dannon is usually regarded as synonymous with
yogurt.18 In the peanut butter category, across both
stores, Skippy has greater price elasticity than Jif.
Further, Skippy has greater cross-elasticity with respect to Jifs prices than vice versa. This indicates a
stronger market position for Jif than for Skippy.
Our results in the peanut butter market are consistent
17According to the teaching case YoplaitUSA in the marketing text

by Berkowitz et al. (1999, p. 607), some serious concerns for
YoplaitUSA in 1993 (the period of our analysis) were (i) Retail
Prices: Yoplaits prices for a six ounce cup was higher on some lines
than competitors eight ounce cups. For example, the prices on Yoplaits 4 pack are about 20% higher per cup than Dannons and
Krafts 6 pack. (ii) Low gross margins: Margins have declined, at
least partly because of high production and overhead costs. The
concern about the relatively high production costs and its impact
on prices give face validity to our estimates.
18 General Mills Inc.: Yoplait Custard Style Yogurt (A), Harvard
Business School Case 9-586-087, 1986.
259
SUDHIR
Table 6a Elasticity Estimates for Yogurt Category

Store 1
Table 7 CV Estimates for the Logit Model: Category-Profit-Maximizing Retailer and Constant-Margin Retailer
Store 2
Yogurt
Elasticity of Elasticity of Elasticity of Elasticity of

Dannon
Yoplait
Dannon
Yoplait
w.r.t. Dannon price
w.r.t. Yoplait price
4.293
0.286
0.640
5.420
4.601
0.537
1.289
6.462
Category profit maximizing retailer

1
2
1
2
Peanut Butter
Store 1
Store 2
Store 1
Store 2
0.122
0.154
0.232
0.252
0.126
0.149
0.262
0.232
0.151
0.399
0.297
0.454
0.147
0.316
0.204
0.693
Table 6b Elasticity Estimates for Peanut Butter Category

Store 1
Store 2
Elasticity of Elasticity of Elasticity of Elasticity of

Skippy
Jif
Skippy
Jif
w.r.t. Skippy price
w.r.t. Jif price
11.026
2.488
2.328
10.505
8.755
2.583
2.118
8.370
with the strength of these brands at the national level

(Deveney 1993).
Implications of the Identification Problem. We
have shown that the constant-margin retailer model
is econometrically identical to a model measuring the
degree of retail coordination. Hence, intuitively we
should expect that CVs estimated for the HSI might
be exaggerated under the assumption of a constantmargin retailer compared with CV estimates when
the retailer strategically maximized category profits
with the manufacturer as the Stackelberg leader. We
test this intuition by comparing the CV estimates for
the two models.19 The CV estimates for the two models in both categories and stores are reported in Table
7. CV estimates are consistently higher, exaggerating
the degree of cooperation among manufacturers for
the constant-margin retailer model. This is consistent
with the intuition we gain from the analytical results.
The empirical result further underscores the need to
properly model the VSI in inferring the HSI.
4. Conclusion
In this paper, we empirically inferred the VSI between manufacturers and retailers and the HSI be19
Since FIML does not converge for the CV models, we use 3SLS
for estimation.
260
tween manufacturers simultaneously. The approach

is particularly appealing because it can be used in
the common situation in which the wholesale price
information of all competitors is not available. We
showed analytically that the simplifying assumption
that a retailer charges a constant margin (used in earlier research) may misinterpret category management
behavior by the retailer to be cooperative behavior by
manufacturers. Consistent with the intuition from the
analytical result, we find that the estimated cooperation among manufacturers is exaggerated for the
constant-margin model, highlighting the need to simultaneously model and infer the VSI when analyzing the HSI.
Since packaged goods manufacturers spend a sizable share of their marketing-mix budget on trade
promotions (Bucklin and Gupta 1999), several decision support systems (DSSs) have been developed to
aid manufacturers in their promotion planning (e.g.,
Neslin et al. 1995, Tellis and Zufryden 1995, Midgley
et al. 1997, Silva-Risso et al. 1999). Our results are of
substantive import in guiding the assumptions that
go into such DSSs. For example,
1. DSSs usually assume Bertrand behavior among
manufacturers (e.g., Midgley et al. 1997). By explicitly
testing for Bertrand behavior as well as cooperative
behavior among manufacturers, we found that manufacturers in these markets tend to be more cooperative. A DSS that assumes Bertrand pricing behavior
would have led to more aggressive pricing than
would be warranted by the actual behavior of the
market participants.
2. Many DSSs assume manufacturer Stackelberg
behavior between manufacturers and retailers (e.g.,
SUDHIR
Tellis and Zufryden 1995). However, the theoretical

literature does not provide clear guidance on whether
such an assumption is appropriate; it shows that a
change in assumption can have substantive implications for pricing behavior of the channel member.
Given the apparent shift in power to retailers, this
question becomes even more important to managers.
We found that for the categories and markets that we
analyzed, the manufacturer Stackelberg assumption
is appropriate.
3. DSSs have used different assumptions about retailer-pricing rules. Silva-Risso et al. (1999) assume
that retailers follow a simple passthrough rule and
manufacturers incorporate that assumption when deciding their trade deal. Tellis and Zufryden (1995) assume that the retailers objective is to maximize category profits. By testing both these specific
assumptions, we found that both retailers objective
is to maximize category profits.
4. DSSs also make assumptions about the functional form of demand. While functional forms of demand have been evaluated in terms of the fit of the
model to sales data, its ability to accommodate the
behavioral implications of the supply side have not
been evaluated. We found that the logit model performed better than the multiplicative model in its
ability to accommodate the strategic behavior of firms
for the categories that we analyze.
We now discuss some of the limitations in this paper and possible avenues for future research. First, it
would be useful to validate the methodology by
checking how well the model predicts actual wholesale prices by estimating the model on a data set with
wholesale prices. Second, the current estimation
method assumes that the estimation equations can be
derived in closed form. To allow for greater flexibility
in the functional forms and also to extend the analysis to a larger number of brands, we need to enable
estimation even when closed-form estimation equations cannot be obtained.
Clearly more research in other categories at other
retail stores is needed before we can develop a body
of evidence that will enable us to understand the impact of supply characteristics (for example, degree of
concentration in market, manufacturer-retailer power
balance, availability of private labels) and demand

characteristics of category (e.g., high inertia or variety
seeking, stockpiling, average price sensitivity, loss
leader) on the inference of strategic interactions. For
example, Cotterill and Putsis (2001) find that the Vertical Strategic Interactions are different across categories, while we found manufacturer Stackelberg behavior in both categories we analyze. Understanding
the determinants of these differences should be of interest in future work.
Putsis and Dhar (1998) have also done a cross-category analysis of how competitive behavior between
private labels and manufacturers changes as a function of the category and market characteristics. Sudhir
(2001) develops hypotheses combining arguments
from game-theoretic research and the ability-motivation paradigm (Boulding and Staelin 1995) about how
competitive behavior in different segments of the
auto market will differ, depending on the demandand-supply characteristics of these markets, and then
tests these hypotheses. Such an analysis will help
provide deeper insights into the determinants of horizontal and vertical strategic interactions and appropriateness of functional forms of demand.
The issue of retail competition was not found to
be particularly important in the yogurt and peanut
butter categories. Future research into categories
such as detergents and soda, which are known to
have an impact on store traffic, is needed to study
the role of strategic retail competition on retailerpricing behavior. Also, retail competition may be at
the basket level across a number of categories (Bell
and Lattin 1998).
We have limited ourselves to the inference of manufacturer-retailer interaction only when the retailer is
EDLP. However, accounting for unobserved heterogeneity among consumers in the demand model can
enable us to investigate how loyalty and switching
behavior, asymmetric responses to lower-price-tier
brands, etc., may cause high-low pricing behavior. We
are investigating how to use household-level data to
infer heterogeneity distributions and how to incorporate this information into the aggregate model so
that we can extend our analysis to the case of highlow retailers. Horsky and Nelson (1992) and Gold-
261
SUDHIR
berg (1995) provide good starting points for this

stream of research. Recently, there has been interest
in recovering heterogeneity from aggregate data (Berry et al. 1995, Kim 1995, Nevo 2001, Sudhir 2001,
Chintagunta 1999). In recent work, Villas-Boas and
Zhao (2000) address many of the issues in modeling
channel behavior with household-level data.
We have not modeled any kind of dynamics that
affect demand or supply. Forward buying by consumers, habit persistence, variety seeking, advertising wearout, etc., can have intertemporal demand
effects, which in turn can affect supply-side behavior. Forward buying by retailers is another supplyside intertemporal effect. While the use of efficient
consumer response (ECR) has significantly reduced
forward buying on the part of retailers, it is still a
significant component of how a retailer reacts to a
trade deal. This is an important issue for future research. For a more comprehensive set of methodological and substantive questions that await research
in this area, we direct the reader to Kadiyali et al.
(2001).
Acknowledgments
This paper is based on an essay from the authors
1998 Ph.D. dissertation at Cornell University. He
thanks the members of the dissertation committee, especially Vithala R. Rao and Vrinda Kadiyali, for their
extensive comments and suggestions. The author also
thanks Pradeep Chintagunta, Bill Putsis, and Robert
Shoemaker for their comments, as well as the seminar
participants at Chicago, Cornell, Dartmouth, Duke,
HKUST, INSEAD, NYU, Purdue, Rochester, Stanford,
SUNY Buffalo, Toronto, UCLA, and Yale. The author
has benefited greatly from the review process and
thanks the editor, the area editor, and the two anonymous reviewers for their input in revising this paper.
The usual disclaimer applies.
p1t
1
1 s1t p1t
s1t p2t
1 2 s0t
w1t
s0t
1t p1t w1t
p2t w1t
s1t
1t
1 s1t p1t
s1t p2t
2t p1t w1t
p2t w1t
s1t p1t
s1t p2t
s2t p1t
p2t w1t
p1t w1t
1t w 1t
s2t
2t
s2t p2t
.
p2t w1t
Substituting the share derivatives with respect to prices from

Equation (4), we get
p1t
s1t
p1t
s2t
p2t
1
.
w1t
(1 s1t s2t ) w1t
(1 s1t s2t ) w1t
(A1)
Similarly,
p2t
s1t
p1t
s2t
p2t
.
w1t
(1 s1t s2t ) w1t
(1 s1t s2t ) w1t
(A2)
Solving (A1) and (A2) for p1t /w1t and p2t /w1t, we have
p1t /w1t 1 s1t and p2t /w1t s1t. By symmetry, p2t /w2t
1 s2t and p1t /w2t s2t.
These reactions are summarized in Equation (12) of the paper.
Reactions for Brand-Managing Retailer

Taking the derivatives of the retail prices in Equation (10) of the
paper,
p1t
1
s1t p1t
s1t p2t
1 2
1t
,
w1t
1t (1 s1t ) 2
p1t w1t
p2t w1t
p2t
1
s2t p1t
s2t p2t
2
2t
.
w1t
2t (1 s2t ) 2
p1t w1t
p2t w1t
Substituting the share derivatives with respect to prices from
Equation (5), we get
p1t
s1t
p1t
2t s1t s2t p2t
1
,
w1t
(1 s1t ) w1t
1t (1 s1t ) 2 w1t
(A3)
p2t
1t s1t s2t p1t
s2t
p2t
.
w1t
2t (1 s2t ) 2 w1t
(1 s2t ) w1t
(A4)
Solving (A3) and (A4) for p1t /w1t and p2t /w1t, we have
p1t
(1 s1t ) 2 (1 s2t )
w1t
(1 s1t )(1 s2t ) (s1t s2t ) 2
and
p2t
1t
(1 s1t ) 2 (s1t s2t )
.
w1t
2t (1 s1t )(1 s2t ) (s1t s2t ) 2
By symmetry,
Appendix
p2t
(1 s2t ) 2 (1 s1t )
w2t
(1 s1t )(1 s2t ) (s1t s2t ) 2
Reactions for Category-Managing Retailer
p1t
2t
(1 s2t ) 2 (s1t s2t )
.
w2t
1t (1 s1t )(1 s2t ) (s1t s2t ) 2
Taking the derivatives of the retail prices in Equation (8) of the

paper,
262
and
These reactions are summarized in Equation (13) of the paper.
SUDHIR
References
Agrawal, D. 1996. Effect on brand loyalty on advertising and trade
promotions: A game theoretic analysis with empirical evidence. Marketing Sci. 15(1) 86108.
Allenby, G. M. 1989. A unified approach to identifying, estimating
and testing demand structures with aggregate scanner data.
Marketing Sci. 8 265280.
, P. Rossi. 1991. There is no aggregation bias: Why aggregate
macro logit models work. J. Bus. Econom. Statist. 9(1) 115.
Armstrong, M. K. 1991. Retailer response to trade promotion: An
incremental analysis of forward buying and retail promotion.
Unpublished doctoral dissertation, University of Texas at Dallas, Dallas, TX.
Bell, D., J. Lattin. 1998. Shopping behavior and consumer preference
for store price format: Why large basket shoppers prefer
EDLP. Marketing Sci. 17(1) 6688.
Berkowitz, E. N., R. A. Kerin, S. W. Hartley, W. Rudelius. 1999. Marketing, 6th ed. Irwin, McGraw Hill, New York.
Berry, S., J. Levinsohn, A. Pakes. 1995. Automobile prices in marketing equilibrium. Econometrica 63(4) 841890.
Besanko, D., D. Dranove, M. Shanley. 1996. The Economics of Strategy.
John Wiley and Sons, New York.
Besanko, D., S. Gupta, D. Jain. 1998. Logit demand estimation under
competitive pricing behavior: An equilibrium framework.
Management Sci. 44(1) 15331547.
Boulding, W., R. Staelin. 1995. Identifying generalizable effects of
strategic actions on firm performance: The case of demandside returns to R&D spending. Marketing Sci. 14(3) G222G236.
Bresnahan, T. 1989. Industries and market power. R. Schmalensee,
R. Willig, eds., Handbook of Industrial Organization. North Holland, Amsterdam, The Netherlands.
Bucklin, R. E., S. Gupta. 1999. Commercial use of UPC scanner data:
Industry and academic perspectives. Marketing Sci. 18(3) 247
273.
, J. Lattin. 1992. A model of product category competition
among grocery retailers. J. Retailing 68(3) 271294.
Carlton, D. W., J. M. Perloff. 1994. Modern Industrial Organization.
Addison Wesley, Reading, MA.
Chevalier, M., R. C. Curhan. 1976. Retail promotions as a function
of trade promotions. Sloan Management Rev. (Fall) 1932.
Chintagunta, P. 1999. Investigating category profit maximizing behavior at a retail chain. Working Paper, University of Chicago,
Chicago, IL.
Choi, S. C. 1991. Price competition in a channel structure with a
common retailer. Marketing Sci. 10(Fall) 271296.
Christen, M., S. Gupta, J. C. Porter, R. Staelin, D. R. Wittink. 1997.
Using market-level data to understand promotional effects in
a nonlinear model. J. Marketing Res. 34(3) 322334.
Cotterill, R. W., W. P. Putsis. 2000. Testing theory: Do assumptions
of models of vertical strategic interaction of national and store
brands meet the market test? J. Retailing
,
. 2001. Do models of vertical strategic interaction for na-
tional and store brands meet the market test? J. Retailing 17

83109.
,
, R. Dhar. 2000. Assessing the competitive interaction between private labels and national brands. J. Bus. 73(1) 109137.
Deveney, K. 1993. Peanut butter makers feel the crunch. Wall Street
Journal. September 16. Page B6.
Goldberg, P. 1995. Imperfect competition in international markets:
The case of the US automobile industry. Econometrica 63(4)
891951.
Guadagni, P. M., J. D. C. Little. 1983. A logit model of brand choice
calibrated on scanner data. Marketing Sci. 2 203238.
Gupta, S., P. Chintagunta, A. Kaul, D. Wittink. 1996. Do household
scanner data provide representative inferences from brand
choices: A comparison with store data. J. Marketing Res. 33(4)
383399.
Hausman, J. 1978. Specification tests in econometrics. Econometrica
46 12511272.
Horsky, D., P. Nelson. 1992. New brand positioning and pricing in
an oligopolistic market. Marketing Sci. 11(2) 133153.
Kadiyali, V. 1996. Entry, its deterrence and its accommodation: A
study of the U.S. photographic film industry. Rand J. Econom.
27(3) 452478.
, P. Chintagunta, N. Vilcassim. 2000. Power in manufacturerretailer interactions: An empirical investigation of pricing in a
local market. Marketing Sci. 19(2) 127148.
, K. Sudhir, V. R. Rao. Structural analysis of competitive behavior: New empirical organization methods in marketing. Internat. J. Res. Marketing 18 161186.
, N. Vilcassim, P. Chintagunta. 1996. Empirical analysis of intertemporal competitive product line pricing decisions: Lead
follow or move together? J. Bus. 69(4) 459487.
,
,
. 1999. Product line extensions and competitive
market interactions: An empirical analysis. J. Econometrics 89
March/April, 339364.
Kim, Byung-Do. 1995. Incorporating hetereogeneity with store-level
aggregate data. Marketing Letters 6 159169.
Lee, E., R. Staelin. 1997. Vertical strategic interaction: Implications
for channel pricing strategy. Marketing Sci. 16(3) 185207.
Manning, K. C., W. O. Bearden, R. L. Rose. 1998. Development of a
theory of retailer response to manufacturers everyday low cost
programs. J. Retailing 74(1) 107138.
McGuire, T., R. Staelin. 1983. An industry equilibrium analysis of
downstream vertical integration. Marketing Sci. 2(Spring) 161192.
McLaughlin, E. W., G. F. Hawkes. 1994. Category management: Current status and future outlook. Working paper, Food Industry
Management Program, Cornell University, Ithaca, NY.
Midgley, D. F., R. E. Marks, L. G. Cooper. 1997. Breeding competitive
strategies. Management Sci. 43(March) 257275.
Neslin, S. A., S. G. Powell, L. S. Stone. 1995. The effect of retailer
and consumer response on optimal manufacturer advertising
and trade promotion strategies. Management Sci. 41(5) 749766.
Nevo, A. 1998. Measuring market power in the ready to eat cereal
industry. Econometrica 69(2) 307342.
263
SUDHIR
Nichols, D. 1987. Supermarket promotions: Whos running the store.

Incentive November.
Putsis, Jr., W. P., R. Dhar. 1998. The many faces of competition. Marketing Lett. 9(3) 269284.
Putsis, Jr., W. P., R. Dhar. 1999. Category expenditure, promotion
and competitive market interactions: Can private labels expand the pie? Working paper, London Business School, London, UK.
Quelch, J. A., P. W. Farris. 1983. Cases in Advertising and Promotion
Management. Business Publications Inc., Plano, TX.
Roy, A., D. M. Hanssens, J. S. Raju. 1994. Competitive pricing by a
price leader. Management Sci. 40(7) 809823.
Silva-Risso, J. M., R. E. Bucklin, D. G. Morrison. 1999. A decision
support system for planning manufacturers sales promotion
calendars. Marketing Sci. 18(3) 274300.
Slade, M. E. 1995. Product rivalry and mutliple strategic weapons:
An analysis of price and advertising competition. J. Econom.
Management Strategy 4(3) 445476.
Sudhir, K. Competitive pricing behavior in the auto market: A structural analysis. Marketing Sci. 20(1) 4260.
Tellis, G. J., F. S. Zufryden. 1995. Tackling the retailer decision maze:

Which brands to discount, how much, when and why? Marketing Sci. 14(3) 271289.
Tirole, J. 1988. The Theory of Industrial Organization. The MIT Press,
Cambridge, MA.
Tyagi, R. K. 1999. A characterization of retailer response to manufacturer trade deals. J. Marketing Res. 36(10) 510516.
Vilcassim, N., V. Kadiyali, P. Chintagunta. 1999. Investigating dynamic multifirm market interactions in price and advertising:
A conjectural variations approach. Management Sci. 45(4) 499
518.
Villas-Boas, J. M., R. S. Winer. 1999. Endogeneity in brand choice
models. Management Sci. 45(10) 13241338.
, Y. Zhao. 2000. The ketchup marketplace: Retailer, manufacturers and individual consumers. Working Paper, University
of California, Berkeley, CA.
Vuong, Q. H. 1989. Likelihood ratio tests for model selection and
non-nested hypothesis. Econometrica 57 307333.
Walters, R., S. MacKenzie. 1988. A structural equations analysis of
the impact of price promotions on store performance. J. Marketing Res. 25(1) 5163.
This paper was received October 7, 1998, and was with the author 19 months for 4 revisions; processed by Scott Neslin.
264

Structural Analysis of Manufacturer Pricing in The Presence of A Strategic Retailer

Uploaded by

Copyright:

Available Formats

Structural Analysis of Manufacturer Pricing in The Presence of A Strategic Retailer

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Structural Analysis of Manufacturer Pricing in The Presence of A Strategic Retailer

Uploaded by

Copyright:

Available Formats

Structural Analysis of Manufacturer

Pricing in the Presence of a

Consumer goods manufacturers usually sell their brands to

and infer VSI simultaneously to accurately estimate the HSI

STRUCTURAL ANALYSIS OF MANUFACTURER PRICING IN THE PRESENCE OF A STRATEGIC RETAILER

It is important to note that the derivation methods used to solve

MARKETING SCIENCE/Vol. 20, No. 3, Summer 2001

Table 1 Positioning Against Related Research

Category profit, brand

Logit (VSS), Multiplicative (VSC)

Kadiyali et al. (2000)

Sudhir (2001; this paper)

tion implies that wholesale prices can be obtained by

brand profit maximization, and (3) constant margins.

MARKETING SCIENCE/Vol. 20, No. 3, Summer 2001

the HSI between manufacturers and the VSI between

We note that Kadiyali et al. (2000) estimate conjectural variations

MARKETING SCIENCE/Vol. 20, No. 3, Summer 2001

tail prices in terms of a deterministic demand model.4

passthrough greater than 100% in some categories is that retailers

A Schematic Model of the Market

MARKETING SCIENCE/Vol. 20, No. 3, Summer 2001

be a serious shortcoming when analyzing categories

MARKETING SCIENCE/Vol. 20, No. 3, Summer 2001

the right sign, other store prices were insignificant in

where f it, d it, and p it are the features, displays, and

Therefore, when both brands reduce prices, the total

The first derivatives of market share with respect

where p1t and p2t are the retail prices of Products 1

MARKETING SCIENCE/Vol. 20, No. 3, Summer 2001

on the demand and profits from the other brand. The

ing the first-order conditions give us the pricing

For brand-profit-maximizing case, the first-order

Substituting the price derivatives from (4) and solv

MARKETING SCIENCE/Vol. 20, No. 3, Summer 2001

For the constant retailer margin model, the pricing

In Pricing Equations (8), (10), and (11) the first term

For the manufacturer Stackelberg model, with the

As we stated earlier, the logit demand model leads to vertical

2t (1 s1t )(1 s2t ) (s1t s2t ) 2

For the vertical Nash model, both the retailer and

brand-specific unobservable marginal cost at time t.

For the constant margin model, the retailer reaction

for tacit collusion and

where w it is the wholesale price for brand i that the

for Bertrand competition. The operator denotes element-by-element multiplication of a matrix.

where the term

is the vector of margins that manufacturers choose

MARKETING SCIENCE/Vol. 20, No. 3, Summer 2001

Table 2 Retail Margins

Constant margin retailer

The actual estimation equations for the five models

Wholesale price (wt )

Table 3 Retailer Reactions

(1 s2t ) 2 (s1t s2t )

Constant margin retailer

MARKETING SCIENCE/Vol. 20, No. 3, Summer 2001

2t (1 s1t )(1 s2t ) (s1t s2t ) 2