Table Multimarket Contact MMC for the firms present on Archos Price Transparency by MikeJenny


Table Multimarket Contact MMC for the firms present on Archos Price Transparency

More Info

               Mehdi Farajallah

               Thierry Pénard

   CREM, University of Rennes 1, Marsouin
  7 place Hoche 35 065 Rennes Cedex, France

              Preliminary Draft

ABSTRACT : Despite the existence of shopbots that enable online consumers to search and
compare prices, competition is far from being pure and perfect on the Internet. Indeed, online
markets are characterised by persistent price dispersion (often higher than on conventional
markets) and a high level of concentration due to significant barriers to entry. Moreover,
online market structure seems to be prone to tacit collusion: high level of price transparency
that facilitates monitoring of rivals’ behaviours and high price flexibility and reactivity that
enable to retaliate price undercutting. Other structural factors can also relax competition and
facilitate collusion. Multimarket contact (MMC) is one of these facilitating factors. The aim
of this paper is to empirically examine the impact of multimarket contacts on pricing
strategies on the Internet. For this purpose, we have selected a sample of 98 electronic
branded products (digital cameras, printers, DVD players) and collected all the offers for
these products listed on the French shopbot kelkoo over 3 months. Assuming that each
branded product constitutes a market, we find that the price of an product is higher when the
firms offering this product have many contacts in other markets (i.e. for other products).
Econometric results also show that multimarket contacts contribute to red uce price
dispersion. This paper confirms that the Internet retailers behave strategically and that online
competition is far from being perfect or frictionless.

        Key words : Multimarket contacts, electronic commerce, collusion, shopbots

1. Introduction:

         Has the Internet radically changed the nature of markets and competition? In the late
nineties, some economists used to describe the Internet as a “forthcoming” pure competitive
market, characterised by a perfect transparency that should lower price dow n to marginal costs.
However, such a “perfect world” has never happened on the Internet and is unlikely to happen in
the future. Despite the existence of shopbots that enable online consumers to search and compare
prices 1 , competition is far from being pure and perfect on the Internet. Indeed, online markets are
characterised by persistent price dispersion (often higher than on conventional markets, see
Brynjolfson and Smith (2000) or Baye, Morgan and Scholten (2002) on the US market, and
Pénard and Larribeau (2004) on the French market) and a high level of concentration due to
significant barriers to entry. Moreover, online market structures seem to be prone to tacit
collusion: high price transparency that facilitates monitoring of rivals’ behaviour and high price
flexibility and reactivity that enable rivals to retaliate with price undercutting. These two features
strengthen the ability of the internet retailers to sustain collusion: each retailer is deterred to
cheat from the collusive price knowing that it will be immediately detected and punished.
         Many empirical studies have attempted to measure the intensity of competition on the
Internet and to explain the pricing strategies of Internet retailers. Baye, Morgan and Scholten
(2002) collected 4 million prices on, an online price comparison site, during eight
months, representing thousand best-selling electronics products. They found that price dispersion
is a persistent phenomenon, whatever the age of products is. However, price dispersion te nds to
decrease with the popularity of products. They also showed that the way to measure price
dispersion matters. Indeed, the number of retailers has a positive impact on the range (the interval
between the highest and the lowest price) and a negative impact on the coefficient of variation
(the standard deviation divided by the mean price).
         Clay, Krishnan and Wolff (2001) studied price dispersion on the US online book market.
First, they observed that the price of a book tends to decrease when it is offe red by one of the

 If price dispersion is “the manifestation of ignorance in the market”(Stig ler, 1961), one should observe s imilar
prices for ho mogeneous goods on the Internet (since the Internet consumer can be fully informed without cost).

three big e-bookstores (Amazon, Barnes or Noble). Second, they found that a more concentrated
market reduces price dispersion.
       Pénard and Larribeau (2004) analysed the French online CD market and examined the
relationship between CD’s popularity, price strategies and price dispersion. They found that
collusion between Internet retailers is more likely when the demand for a CD declines, meaning
that competition is lessened on the older and less popular CDs.
       Other structural factors than supply concentration and demand intensity can influence the
degree of competition and price dispersion. Among these factors, multimarket contacts may play
a major role. Since Edwards’ paper (1955), multimarket contacts have been analysed as a
facilitating device to sustain collusion. Edward asserted that firms who were operating in parallel
markets were more able to collude, because they could retaliate against a cheating firm on all
shared markets. This idea was reformulated and assessed through a game-theoretical lens by
Bernheim and Whinston (1990). They identified the conditions under which multimarket contacts
are likely to facilitate collusion. They proved that multimarket contacts have no impact if markets
and firms are identical, but if markets or firms differ in their characteristics, then multimarket
contacts can improve the conglomerate firms’ ability to sustain collusion, by pooling the
incentives to respect the price- fixing agreement.
       Several empirical studies attempted to measure the effects of multimarket contact in the
U.S. airlines industry (Evans and Kessides, 1994), in the U.S. cellular telephone market (Parker
and Roller, 1997, Busse, 2000) or in the cement industry (Jans and Rosenbaum, 1997). Evans and
Kessides (1991) obtain a statistically significant positive effect of inter-city routes contacts on the
airline fares. Parker and Roller (1997) study whether the duopolistic regulatory system prevailing
for the U.S. cellular telephone industry in the nineties led to collusive behavior. The ir empirical
analysis reveals that multimarket contact between mobile phone operators significantly increases
price mark-up. Based on a panel of 25 regional cement markets over 16 years, the study of Jans
and Rosenbaum (1997) shows that the mark-up obtained in a regional market is positively linked
to the extent of multimarket contact. More recently, Fu (2003) studies the impact of multimarket
contact on local daily newspaper competition. He attempts to determine whether multimarket
contact between two newspapers chains can facilitate agreement on exclusive areas for their
respective newspapers. His results indicates a negative relationship between multimarket contact
and circulation competition for 218 papers in the Midwest.

       The aim of our paper is to analyze the impact of multimarket contact on online markets. Is
competition less intense when Internet dealers face more multimarket contacts? To examine this
question, we assembled retailer- level panel data consisting of price observations from 98
electronic products (digital cameras, MP3 players and DVD players) sold on the Internet. Prices
were collected on the French leading shopbot for 3 months (September 2004-
December 2004). Assuming that each branded product constitutes a market in itself, an Internet
retailer offering several products can be in contact with another retailer regarding a few products
or markets. Consequently, for each product and each date, we measure the number of firm-pair
contacts on the other products/markets at the same date and estimate the effect of these contacts
on prices and price dispersion. Our econometric results are in line with previous studies
conducted on conventional markets. Multimarket contacts matter and have a significant impact
on strategic behaviors of Internet retailers.
       The remainder of the paper will be organized as follows. The next section presents the
game theoretical framework. Section 3 describes data and explains how multimarket contacts are
measured. Section 4 displays econometric results relying on panel data techniques.

2. Theoretical frame work

       First, we present a simplified version of the Bernheim- Whinston (1990) model to
understand how multimarket contacts can actually relax price competition. Consider two identical
markets A and B, except the number of firms (i.e. in supply concentration). N firms (N’ firms)
are competing on market A (on market B), with NN’. All firms are symmetric and compete à la
Bertrand. As competitive profits are driven to zero, firms have strong incentives to collude.
       Let us examine the conditions under which a self- enforcing price- fixing agreement is
feasible. Assume that firms agree upon trigger strategies à la Friedman (1971). These strategies
are designed to set the collusive price as long as nobody has deviated from this price. But as soon
as a firm deviates, then firms revert to the competitive price forever.
       Let pc denote the collusive price and  c the collusive industry profit. If firms equally
                                                   c                     c
share sales, then each collusive firm will receive    on the market A and    on the market B.
                                                   N                      N'
Finally, let  denote the discount factor (identical for all firms).

A collusion is sustainable on the market A if and only if
                               c
                   N (1   )
where the left hand side represents the expected discounted profits when the firm respects the
collusive agreement and the right hand side represents the expected profit if the firm cheats
(given that the cheater will be punished in the next periods). This condition can be rewritten as
                        N 1
                            A

where the expression on the right hand side is called the threshold factor (the discount factor
value above which collusion is sustainable or feasible).
For market B, this condition is
                        N '1
                            B

which is a less stringent condition, since NN’. Indeed, collusion is more likely to occur in more
concentrated markets. The threshold factor increases with the number of firms competing in the
          Now, consider that m firms (mN’) in market B are also present in market A. In this case,
both markets are characterized by m contacts (i.e. m multi- market contacts). What are the new
conditions of collusion sustainability?
          Consider that the m conglomerate firms decide to slightly reduce their market share on the
less collusive market A, accepting to receive  c with          . Given that collusive profit are
always equally shared on market B, the incentive condition (to respect the collusive agreement)
for a conglomerate firm becomes
              c         c
                                2 c
          N ' (1   ) (1   )
After rearrangement
               N ' (2   )  1
                     2N '
On market A, single- market firms have no incentives to cheat if
            (1  m ) c
                            c
          ( N  m)(1   )

                 N  1  m(1   )
                      N m
                                              N  1  m(1   ) N ' (2   )  1
Collusion is feasible on market A if   Max                   ,                
                                                   N m               2N '      
The best conditions for sustaining collusion are obtained whenever conglomerate firms accept to
                       2 N '( N  m)  1
receive only 2   Max               ,0  . The threshold factor is now equal to
                       ( N  m) N '  N
          ( N  1  m) N 'm                                 N 1 m
M                          (for   0 ). For   0 ,  M 
              ( N  m) N '                                    N m

           Proposition : The likelihood for collusion is higher with multimarket contacts, than
           without, on market A :  M   S

                    Moreover, increasing the number of multi-market                      contacts strengthens the
           incentives to collude (         0)

           Multimarket contacts enable firms to transfer collusion capacities from high collusive
markets to less collusive markets, by pooling incentive constraints. However, we can remark that
with multimarket contacts, collusion is less likely to exist on market B (the high collusive
market). The threshold factor has increased from  B to 
                                                                    S        M
                                                                                 . Multimarket contacts generate a
new threshold factor that is a weighted average of the two single market threshold factors:
S M S .
 B       A

           From this theoretical framework, we can formulate testable hypotheses: if firms are
attempting to relax competition, then multimarket contacts should increase prices on less
concentrated markets (by increasing the capacity to sustain price coordination), but could
decrease prices on more concentrated markets (by reducing the capacity to sustain price

    This condition is obtained when the individual threshold factors are equalized :
    N  1  m(1   ) N ' (2   )  1
         N m               2N '

       In the next section we will present our data and give some descriptive statistics.

3. The data

       The data were collected on the price comparison site kelkoo, over 12 weeks, between
September 12 and December 14 of 2004. Kelkoo is the European leader of price comparison site,
operating in nine European countries. In December 2004, it received about four million single
visitors per month. It has been recently bought by Yahoo.
       The data include price observations for a selection of digital cameras, DVD players and
MP3 players. These three classes of products are online best-sellers (see figure 1). Among the
numerous models of digital cameras, DVD players and MP3 players, we have selected 98
branded products with different characteristics, qualities and price levels. That enables us to have
a representative sample of the models sold by the Internet retailers. For example, for DVD
players we have both DivX compatible and DivX incompatible players. The entire list of the
products is available in the appendix.

Figure 1 : Best-seller electronic products on the French Internet market in Decembe r 2003

                         Digital camera                   5

                           DVD players                  4,7

                             USB KEY                4,1

                            MP3 player              4

                              Webcam          2,3

                                          0   1     2         3       4      5        6

                                                    % of all sold products

       For each of our 98 products, we collected all the offers listed on Kelkoo twice a week.
The data set contain 10 638 individual price observations from 73 different Internet retailers. We
also downloaded information on availability and delivery conditions. Availability indicates

whether or not the product is in stock (if not, it indicates the number of days which it takes the
retailer to be supplied with this product). Delivery time is the number of days to ship the product
to the customer. Table 1 displays the statistics on price and delivery conditions. On average,
products are available in a period of less than 10 days and are delivered in under 7 days. Shipping
fees amount to 5.1 euros on average and the mean price of a product is 349.22 €. 87% of our
electronic products are in the interval [100€ - 500€]. During the three- month period, we observed
a downwards trend for the prices, (from 473.69 € in September to 465.83 € in December for
digital cameras, from 255.06 € to 198.19€ for DVD players and from 256.19€ to 248.39€ for
MP3 players).
       We also collected information on the different versions offered by a retailer for a given
branded product. The difference between two versions can lie in the presence of additional
accessories (a memory card, a bag, a battery) or additional services. Some retailers can offer up
to 17 versions of the same branded product. On average, each retailer proposes 1.7 versions. The
price used in our econometric analysis is the price of the minimal or basic version (without
additional services and accessories).
       To measure price dispersion, we consider two standard statistics in the literature: the
coefficient of variation and the range. The coefficient of variation, defined as the standard
deviation divided by the mean price, and the range, defined as the difference between the highest
and the lowest price listed at, divided by the lowest price, are both scale independent
but are affected by the number of firms present on the market 3 . The average range is equal to 40
% of the minimum price and the average coefficient of var iation amounts to 13.6% (these
statistics are rather conform to those found by previous studies).
       Concerning supply concentration, we observe that for a branded product, 13.4 offers are
listed on average, with a minimum of 2 and a maximum of 37. The numbe r of competing retailers
is rather important, compared to the study of Ruppert, Gatti and Kattuman (2003) which found on
average 5.6 firms, considering a sample of 30 products. This means that French electronic
commerce has become a dynamic activity in recent years and has been progressively catching up
with the US where electronic commerce is much more mature (Baye, Morgan and Scholten
(2002) found 17.3 firms on average advertise their prices on from a sample of 1 000
electronic products).

                                               Table 1: Summary statistics

     Vari able                            Definiti on                    Mi n      Max       Mean         SD

     Price                         Price of the basic product            34.99   1799.95    349.224     248.225

     Fees                                Shipping fees                     0       69.95      5.139      5.805

     Full Price                  Price including shipping fees           36.81    1869.9    354.362      248.65

     Availability                      Availability time                   0        30        10.94      11.904

     Deli very                           Delivery time                    1.5       15        6.876      5.261

     Versioning            Nu mber of d ifferentiated versions of the
                                product offered by the seller              1        17        1.706      1.522

     MMC                    The sum of mu ltimarket contacts on a
                                           market                          1       3281      696.19     797.017

     Firms                   Nu mber of firms present on a market          2        37       13.462      9.199

     AMMC                   Average mult imarket contact per pair of
                                          competitors                      1        27        5.87       2.452

     Market coverage       Percentage of markets covered by a firm       0.01      0.555      0.216      0.138

     CV                             Coefficient of variation               0       0.977      0.136      0.103

                         Difference between the lowest and the
     Range               highest price on a market                         0       4.986      0.408      0.365

        The measure of multimarket contacts requires to define clearly what a market is. In this
paper, we assume that each branded product constitutes a market in itself. Thus two Internet
dealers experience multimarket contacts (MMC) whenever they offer sever al identical products
(when they have overlapping offers). Following this definition, we construct an index of
multimarket contacts for each market and each date like in Evans and Kessides (1995) and Jans
and Rosenbaum (1997). This index corresponds to the sum of all the contacts over the 97
remaining markets, for all pairs of retailers present on this market,.

  As price dispersion is only mean ingful, when two or mo re firms are quoting prices, we have excluded all the cases
(product-period) with only one retailer – leading to 1 462 observations.

       Let i= 1,…, I where i denotes a retailer, j=1,…, J where j denotes a market and t =1,…,T
where t denote a date. Let Dtij be equal to 1 if retailer i operates in market j at date t and 0
otherwise. Then we construct an IxI symmetric matrix at each date

            at11 .......... at1I 
                                  
                                  

       At = 
                                        where                    atkl =      D j1
                                                                                              tkj   D tlj
                                  
            atI 1 .......... atII 
                          ... 

       The term atkl measures the number of markets in which retailer k and retailer l are in
contact at date t. Diagonal terms measure the number of markets in which each retailer operates
at date t. Table 2 gives an example of the matrix of multimarket contacts on the Archos Gimini 400
(a MP3 player). On September the 22nd, six firms were offering the Archos Gimini 400 Mp3
player: Orysimage, Dabs, Magma, Maisoneo, Rue du commerce, Azimut.
       If the market j contains Ntj retailers at the date t, then we have Ntj(Ntj-1)/2 pairs of retailers
(15=6*5/2 pairs of retailers exist on the Archos Gimini 400 market). The number of multimarket
contacts on the market j is the sum of contacts existing between these Ntj(Ntj-1)/2 pairs.
Let MMCtj be the total contact points between all firms operating in the market j at the date t.
This measure is defined as
                                                  J      J
                                       MMCtj=         a         tkl   D tkj D tlj
                                                 k 1 l  k 1

We also define AMMCtj as the average multimarket contacts per firm-pairs in market j.
                                                                        J   J
                                           Ntj (Ntj - 1)/2
                                                                    a
                                                                   k 1 l  k 1
                                                                                    tkl   D tkj D tlj

   Table 2: Matrix of Multimarket Contact on the Archos Gimini 400 market ( September 22 2004 )
                   Azimut            Dabs         Mag ma    Maisoneo    Orysimage      Rue Du
   Azimut             2               2              2          1            1            2
   Dabs               2               19             4          3            8            5
   Mag ma             2               4             23          5            2            3
   Maisoneo           1               3              5          8            2            1
   Orysimage          1               8              2          2           11            2
   Rue Du             2               5              3          1            2           12

       In the case of the Archos Gimini 400 market, MMC is equal to 43 and AMMC to 2.8.
Table 1 shows that the number of multimarket contacts amounts to 696 on average and the
multimarket contacts per firm-pairs to 5.8.
       Finally, we construct a variable, called market coverage, that is the percentage of products
offered by a firm at each date.
       Table 3 presents statistics per category of products. We notice that price dispersion is
higher for DVD players and digital camera than for MP3 players. There also are more
multimarket contacts and versioning in digital camera markets than in the two other categories of

                          Table 3: Summary statistics by category of products

Product            CV        Range       MMC      AMMC     Firms    Versioning   Availability   Deli very
Mp3 pl ayers     10.49%     27.62%      261.561    6.136   8.863      1.234        10.020        6.561
DVD pl ayers     14.73%     43.19%      124.387    5.550   6.351      1.358         9.754        6.593
Digital camera   15.61%     52.88%     1149.848    5.859   18.686     2.089        11.889        6.213

       If we examine the evolution of multimarket contacts over the three-month period, we
notice a downwards trend (see figure 2 and figure 3). The total number of contacts has decreased
from 965 to 595 and the number of contacts per firms-pair from 6.8 to 5.2. We can also observe a
strong negative shock on early October (between the fifth and sixth price collecting). Several
products in our sample were removed from the price comparison site during the first week of
October and consequently affected the number of multimarket contacts negatively. But we have
no real explanation for this shock (perhaps, it is due to a technical problem on the Kelkoo

       In figure 4 and figure 5, we see how multimarket contacts have a positive correlation with
the number of firms offering the product. When the number of firms raises from 8 to 16, the total
multimarket contacts increases from 182.1 to 1041.7 and the average contacts from 4.6 to 6.2.

                  Figure 2: The evolution of multimarket contacts over time






                          1   2   3 4   5   6 7   8   9 10 11 12 13 14 15 16 17 18 19 20 21

                 Figure 3: The evolution of multimarket contacts per firm-pair


             A 5
             C 3

                      1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

   Figure 4: The relationship between the numbe r of competitors and the sum of MMC



             M   600


                             8        11     12      13      14      15      16
                                           Average number of firms

 Figure 5: The relationship between the numbe r of competitors and the average numbe r of
                                      MMC per pair



             M 4
             M 3


                         8       11        12       13      14       15      16

4. Econometric analysis

Next, we will empirically evaluate the impact of multimarket contacts on pricing decisions and
price dispersion. As we have panel data, we can adopt either a fixed effect approach or a random
effect approach. But we have opted for the first approach, which is more relevant for our

unbalanced panel (most of the products have not been continually offered during the three- month
period). We also used Hausman tests to compare the two approaches and these tests were
favourable to a fixed effect specification: the fixed effects have been put on our 98

        What explains pricing behaviors of Internet retailers ?

        To explain individual prices, we introduce different variables: some of them are firm-
specific (delivery, availability, versioning, market coverage) and others are market-specific
(multimarket contacts, number of firms).
        We can expect a retailer to set a higher price when the number of multimarket contacts
increases (collusion effect), and to lower prices when market concentration declines (competition
effect). Moreover, a firm is likely to reduce the price of its basic offer when it intensively uses
versioning strategy. Such a strategy aims to discriminate customers with a menu of differentiated
offers (discrimination effect).
        Delivery conditions should also affect the full price of a product; a retailer can set a higher
price when it offers a short delivery time.
        Table 4 shows that multimarket contacts have the predicted effect: the larger the number
of multimarket contacts per firm-pairs in a market, the higher the price set by a retailer. This
structural factor seems to facilitate tacit collusion.
        Supply concentration has an unexpected effect: competition is more intense when the
market becomes more concentrated, but as the number of firms is also an indirect measure of the
multimarket contacts density4 , a high concentration could reduce the ability to collude and
negatively affect prices and mark- up.
        Market coverage has a negative effect. Retailers tend to be more aggressive when they are
present in many markets (when they are multi-product retailers – not specialized in a category of
product). We have also introduced the crossed variable Market coverage*MMC. We obtain a
negative effect ; this means that the total number of multimarket contacts has a positive effect on
price when a retailer has a market coverage below 44.8% (0.022/0.049) and a negative effect for
firms having a market coverage above 44.8%. Therefore, individual pricing decisions are

 The coefficient of correlat ion between these two variables is high. It is the reason why we have no t included these
two variables together in the regressions.

differently affected by multimarket contacts, depending on the nature of the Internet retailers
(number of offered products )
        For the other explanatory variables, versioning has the expected negative effect on prices.
By offering different variants, the retailer is seeking to discriminate among customers. Thus the
basic version is sold at a lower price to make it affordable for low reservation value customers
(the other versions being designed for customers with a greater willingness-to-pay).
        For availability and delivery time, the impact is significantly positive on prices. A
possible explanation for this counterintuitive result can be related to the studied period (the end of
the year). As the demand tends to be high before Christmas, we can imagine that some Internet
retailers use a high price to ration demand and avoid strong shortage: when it takes longer for an
item to become available, then the retailers react by raising prices to lower demand.

        What explains price dispe rsion on the Internet ?

        For a price dispersion estimation, we can only use market- level explanatory variables (the
mean versioning, the mean delivery and avaibility time, the number of firms and multimarket
        In table 5, we observe that the range tends to increase with the number of firms and the
total number of multimarket contacts. But these explanatory variables have no influence on the
coefficient of variation.
        In addition, the average MMC per firm-pair has a negative impact on price dispersion (for
both measures of price dispersion). We can interpret it as follows: inc reasing the intensity or
density of contacts makes the retailers more interdependent and can encourage them to set similar
or parallel prices.
        Delivery conditions have no significant influence. Finally, versioning contributes to the
increase of price dispersion. The latter result correlates with the theory of discrimination:
discrimination policy tends to increase price dispersion on the basic offer.

Table 4: The impact of multimarket contacts on full price (including shipping fees)

                               Model 1      Model 2       Model 3       Model 4

        MMC                       -           0.013         0.013         0.022
                                  -         (8.22)***     (7.93)***    (11.47)***

        AMMC                    1.037        0.496          1.131         0.787
                              (2.99)***      (1.42)       (3.17)***     (2.20)**

        Market Coverage           -             -          -40.654       -11.264
                                  -             -         (8.71)***       (1.95)

        Coverage* MMC             -             -             -           -0.049
                                  -             -             -         (8.54)***

        Firms                   0.738           -             -             -
                              (4.09)***         -             -             -

        Versioning              -13.359      -13.257       -12.368       -12.075
                              (31.22)***   (31.05)***    (28.27)***    (27.61)***

        Deli very                2.382        2.378         2.066         2.015
                              (20.15)***   (20.17)***    (16.83)***    (16.44)***

        Availability            0.170         0.174         0.181         0.181
                              (3.36)***     (3.45)***     (3.61)***     (3.61)***

        Constant                342.848      346.690       352.651       348.125
                              (95.03)***   (131.60)***   (129.99)***   (126.36)***

        Observations            10635        10635         10635         10635
        Number of products
        (fixed effects)           98            98           98            98
        R-s quared (within)      0.14          0.15         0.15          0.16
        Rho                   .94284451     .9428889     .94314841     .94356324
         Note: value of t-student in parentheses
         * Significant at 10%; ** Significant at 5%; ***significant at 1%

Table 5: The impact of multimarket contacts on price dispersion

                       Model 1     Model 2      Model 3           Model 4
Dependent Variable            Range                         CV

MMC                       -         0.00010        -              3.93e-06
                          -        (3.58)***       -               (0.54)

AMMC                  -.0051402     -0.0073    -.0025857         -.0026638
                        (-1.70)*   (-2.36)**   (-3.45)***        (-3.54)***

Firms                 0.022054         -       .0007797              -
                      (8.06)***        -         (1.15)              -

Versioning            0.0467203     0.03123    .0086673          .0081346
                       (2.37)**      (1.56)     (1.77)*           (1.67)*

Deli very             -0.0015534   -0.00086    -.0009405         -.0009167
                        (-0.53)     (-0.29)      (-1.29)           (-1.26)

Availability          -0.0030933    -0.0041    .0000642          .0000284
                        (-1.45)     (-1.90)*    (0.903)            (0.05)

Constant              0.2504818    0.4236948    .1384309          .1445028
                      (4.79 )***   (8.95)***   (10.67)***        (12.54)***

Observations            1462         1462        1462              1462
Number of products
(fixed effects)           98           98          98                98
R-s quared (within)     0.0517       0.015       0.0115            0.0107
Rho                   .5329497     .5538786    .66024379         .65746412
Note: value of t-student in parentheses
* Significant at 10%; ** Significant at 5%; ***significant at 1%

5. Conclusion

       In this paper, we have examined the impact of the multimarket contacts on strategic
behaviours of Internet retailers. We have found that the nature of online competition is imperfect,
driven by strategic considerations, including tacit collusion, discrimination policy, rationing
policy, …
       This study is rather original by its emphasis on multimarket contact, a facilitating device
whose effects have largely been analysed on conventional markets, but never on the Internet
markets, to our knowledge. This study can also be considered original due to the nature of the
data collected. Firstly, we used French data, enabling international comparisons with other
studies on the online markets (mostly based on US data). Secondly, we collected information on a
number of different versions offered for the same branded product. Versioning is one of the main
elements of online commerce. The Internet enables retailers to offer a higher variety of products
and provides many ways to mass-customize offers (see Varian and Shapiro, 1998, Brynjolfson,
Smith and Hu, 2003). Our future research will focus on the determinants of versioning strategy
(by examining how our Internet retailers attempt to differentiate their offers).


Baye, M. R., Morgan and J. and Scholten, P, (2001) “Price Dispersion in the Small and in the
  Large: Evidence from an Internet Price Comparison Site”, Working Paper.
Bernheim B. D., and M. D Whinston (1990) “ Multimarket contact and collusive Behavior ”
  Rand journal of Economics, 20. Pp.344-365.
Brynjolfsson E., Smith M.D. (2000) “Frictionless Commerce? a Comparison of Internet and
  Conventional Retailers”, Management Science 46, pp.563-585.
Brynjolfsson E., Smith M.D. and Hu Y. (2003) “Consumer Surplus in the Digital Economy:
  Estimating the Value of Increased Product Variety at Online Booksellers”, Working Paper.
Busse M. (2000) “ Multimarket Contact and Price Coordination in the Cellular Telephony
  Industry ”, Journal of Economics and Management Strategy 9, N°3, Fall, pp.287-320.
Clay K., Krishnan R., Wolff E. (2001) “ Prices and Price Dispersion on the Web : Evidence from
   the Online Book Industry”, The Journal of Industrial Economics, Vol. XLIX, N°4, pp.521-
Edwards, C.D. (1955), Conglomerate Bigness as a Source of Power. In N.B.O.E.R.C. Report
  (Ed.), Business Concentration and Price Policy: pp.331-352. Princeton, NJ: Princeton
  University Press.
Evans, W. M. and Kessides, I. N., (1994) “Living by the “Golden Rule”: Multimarket Contact in
  the U.S. Airline Industry”. Quarterly Journal of Economics, 109: pp.341-366.
Friedman, J.W. (1971) “A Non-Cooperative Equilibrium for Supergames” Review of Economic
   Studies 28, 1-12.
Jans, I and Rosenbaum, D, (1997) “Multimarket contact and pricing: evidence from the
   U.S.cement industry”, International Journal of Industrial Organization, 15(3), pp.391-412
Larribeau, S., Pénard, T. (2004) “ Strategies and Price Competition on the Internet : Evidence on
   French Data ”, Working Paper,
Parker, P. M., and Roller, L. H., (1997) “Collusive Conduct in Duopolies: Multimarket Contact
   and Cross-ownership in the Mobile Telephone Industry,” Rand Journal of Economics, 28,
Rupert, J., Gatti, J. and Kattuman, P. (2003) “Online price dispersion within and between seven
  European countries.” In Baye, M.R. (ed.): Advances in applied microeconomics: vol.12:
  organizing the new industrial economy. London: JAI Press, pp.107-141.
Shapiro C. and Varian H. [1998] Information Rules: “A Strategic Guide to the Network
  Economy”, Harvard Business School Press.
Stigler, George J. (1961), “The Economics of Information,” Journal of Political Economy 69,
Varian, Hal R. (1980) “A Model of Sales”, American Economic Review, 70, pp.651–659.

                                 Table 6: Products List
     Product Name                                         Product Name
 1   Aiptek 256 M o                               50      M SI M ega Stick 256 M o
 2   A-M ax Napa DAV 314                          51      MUSTEK PL207
 3   Apple iPod (20 Go)                           52      Neosafa 512M o
 4   Apple Ipod 4Go                               53      Neoxeo D V301
 5   Archos Gmini 120                             54      nextbase SDV17
 6   Archos Gmini 400                             55      Nikon 3100
 7   Astry 1800                                   56      Nikon 3700
 8   Biostek XE200                                57      NIKON D70
 9   Canon 400                                    58      Olympus C310
10   Canon A60                                    59      Optio S4
11   Canon A80                                    60      Packard Bell 128 M o
12   Canon A85                                    61      PANASONIC DVDS35
13   Canon A95                                    62      PanasonicFZ10EG
14   Canon G5                                     63      Peekton PK6004
15   Casio EXZ40                                  64      Philips DVD737
16   Creative 1.5 Go                              65      Philips DVP720
17   Creative 4 Go                                66      Philips HDD060 1.5 Go
18   Daewoo 730                                   67      PHILIPS SA230
19   Denon 2900                                   68      Pioneer DV 575
20   Digital Dream Espion Xtra                    69      Pionner PDVLC20
21   Dual 650                                     70      RIO Nitrus 1.5 Go
22   Fuji Q1                                      71      Samsung 935
23   Fuji S5000                                   72      Samsung YP-T5V
24   Fuji S7000                                   73      Shinco1720
25   H&B DX3220                                   74      Sigmatek X300
26   H&B JK 20                                    75      Sigmatek PVR800
27   HARM AN DVD 21                               76      SIGM ATEK S-650 (128 M o)
28   Höher T 058                                  77      Sony DVP-LS785
29   HP 945                                       78      Sony DVPNS330
30   IIsonic II2316                               79      SONY NWM S70
31   Iriver H320                                  80      Sony P52
32   Iriver IHP-120                               81      Sony P72
33   Ism i-Bead 256 M o                           82      Sony P92
34   ISM M F-220                                  83      Sony PQ2
35   JVC XV-N316                                  84      Sony W1
36   Kiss DP1000                                  85      SonyF717
37   Kiss DP1504                                  86      STOREX M obiKey 256 Mo
38   Kodak CX6200                                 87      Thomson DTH233E
39   KodakDX6490                                  88      Thomson PDP 2448
40   Kyocera M 410R                               89      Thomson PDP 2860
41   LG 4820                                      90      Thomson PDP2335
42   LG DVD6183                                   91      Toshiba SD-433E
43   LIVE M USIC 128                              92      Toshiba SDP1400
44   Logitech 130                                 93      TwinM OS Red Rock 256 M o
45   M EGA STICK 256M                             94      Waitec Vision P7

46   M EMUP II 256 M o   95   Xen EM P200 256 Mo
47   M inolta A2         96   Yamaha PDV700
48   M inolta Z1         97   Yamaha S540
49   MPIO DMK 128        98   ZicPlay M egaKey 256


To top