Pricing, Quality-Setting, and Order of Plays
in an Online Information Market
Sang-Ho Lee*
Department of Economics, Chonnam National University, Korea
Abstract
This paper considers the issue of quality differentiation in an online information market in which one network provider and two content providers compete over both prices and con-tent qualities for the real-time network delivery of digital information products. We examine the order of plays in games of vertical product differentiation and show that sequential price choice of network providers reduces content differentiation among online information prod-ucts, so that intense price competition among content providers results. However, sequential price competition reduces not only social welfare but also consumers’ surplus.
Key words: sequential price competition; content differentiation; online information product JEL classification: L1; D4
1. Introduction
The online information market is characterized by the real-time network delivery of digital information products. Information production means collecting, editing, packaging, storing, displaying, and so on. In this online environment, different categories of digital content from differentiated information product markets are emerging, each with competing content providers. Examples of differentiated products are online sports, movies on demand, news on demand, and so on. For detailed economic analysis of business strategies such as selling advertising, content versioning, and bundling in the online information market, see Shapiro and Varian (1999) and Kahin and Varian (2000).
In the industrial economics literature, these phenomena are called product (quality) differentiation. Since the seminal work of Gabszewicz and Thisse (1979) and Shaked and Sutton (1982), it has been well known that the existence of quality differences relaxes price competition among competing firms and can lead to
Received December 12, 2001, revised April 4, 2002, accepted October 17, 2002.
*Correspondence to: Department of Economics, Chonnam National University, 300 Yongbong-dong
Bukgu, Kwangju, 500-757, Korea. Email: sangho@chonnam.ac.kr. I am indebted to the Managing Editor and two anonymous referees whose valuable suggestions and insights improved the paper. This study was financially supported by Chonnam National University in the Humanity and Social Science Program, 1999.
tive profits in equilibrium. Usually, quality differences are formalized in terms of a framework for preferences due to Mussa and Rosen (1978) in which individuals with identical preferences may choose different goods because their respective mar-ginal utilities of income differ. However, one can find a simplified version of quality choices in a duopoly model of vertical product differentiation in Tirole (1988), Choi and Shin (1992), and Wauthy (1996). They introduce an alternative utility function which captures the spirit of the earlier papers and yet leads to similar solutions. That is, firms in the same market want consumers to believe that their products are highly differentiated from competing products for the purpose of increasing their profits.
However, the market structure of online information products is somehow dif-ferent from the standard structure of the product difdif-ferentiation model. In the online distribution environment, the lack of a delivery network of proper quality for certain information products discourages the interests of consumers. This issue of “quality of service” gives rise to the so-called “last mile problem.” Put differently, the deliv-ery of on-demand multimedia content is bottlenecked by only gradual increases in capabilities of last mile facilities to the home provided by network providers such as local exchange carriers (LECs), cable TV suppliers (CATVs), or internet service providers (ISPs). These NPs (network providers) interact with CPs (content provid-ers) in such a way that information product demand is affected by the price of net-work delivery.
Therefore, one might ask whether the presence of this third party (NP) affects previous results in the standard product differentiation model. This question has not been addressed before; most researchers have paid attention to interactions within a market and have found that the decisive factor for product differentiation is the dis-tribution of consumers’ tastes or income in that market. That is, strategic interactions between markets in the context of product differentiation have not been investigated. NPs and CPs in online markets comprise a vertical complementarity structure since NPs form upstream product providers and CPs form downstream product pro-viders. In addition, NPs usually have a (local) monopolistic character due to the large costs of establishing their network facilities, while CPs are in a situation of competition. In this vertical relationship, NPs may not allow for maximal product differentiation among CPs. An NP wants to increase both its price level and its final market demand not only by reducing the prices of CPs but also by lessening the level of product differentiation. That is, lessened product differentiation has a similar strategic effect as reduced CPs’ prices.
In the vertical relationship literature, to my knowledge, this point has been not investigated. Waterman (1993) demonstrates that the economy of scale in product distribution present in the information product market strengthens the well-known “double-markup” effect of vertical integration theory. Economides (1999) shows that independent, vertically-related monopolists provide products of lower quality than a single, integrated monopolist. Oh and Chang (2000) examine the economic situation where one NP and one CP can compete or collude on price levels and qual-ity investment levels and show that the incentive of strategic vertical alliance de-pends on the relative market power. They also find that their result on quality levels
is consistent with that in Economides (1999).
In this paper, we consider games of vertical product differentiation in an online information market in which a monopolistic NP and duopolistic CPs compete over both content qualities and prices for the real-time network delivery of digital infor-mation products. We pay special attention to the order of plays in the games of ver-tical differentiation. This is because, in general, the online business environment is largely determined by the NP who has bottleneck facility and market power. [For a theoretical discussion on the bottleneck facility and access charges, see Laffont and Tirole (2000) and Shy (2001).]
In particular, in a price competition situation, we compare two cases where the NP and CPs compete in price simultaneously or sequentially. The former describes a Nash competition model where the NP and CPs have symmetric information con-cerning the others’ reaction to its own price decision, and thus they choose their op-timal prices simultaneously and independently. The latter is a Stackelberg game with sequential price competition in which the NP with market power can accurately an-ticipate CPs’ responses to its pricing decision and commit to its price level, whereas CPs do not have such information and market power. If the NP can commit in ad-vance to its pricing strategy successfully, then the natural outcome will be Stackel-berg leadership solution.
We then show that sequential price choice of the NP induces less content dif-ferentiation between online information products, so that intense price competition between CPs is the outcome of the game. In other words, in sequential price equilib-rium, less product differentiation has a similar strategic effect as reducing CPs’ prices. We also show that sequential price competition reduces not only social wel-fare but also consumers’ surplus.
In the industrial economics literature, several papers consider the order of play in a game model and compare the results of Cournot and Stackelberg. Dowrick (1986) considers the choice of roles of leader and follower in a duopoly model and indicates that a firm’s reaction functions are key determinants in choosing roles. Boyer and Moreaux (1987) analyze a differentiated model and find that price setting in a Stackelberg leadership situation is more profitable if the goods are complements. As a market equilibrium, Anderson and Engers (1992) examine a fixed m-firm model, and Economides (1993) investigates a free-entry model. In a vertically dif-ferentiated duopoly model, Lee (1996) considers sequential price competition and examines the leadership solutions when firms cover the full market. Aoki and Prusa (1997) consider the timing of investment and show that a game with sequential qual-ity choice induces two firms to make smaller qualqual-ity investment than they would in a game with simultaneous quality choice.
The organization of this paper is as follows. In Section 2, we construct a model of vertical product differentiation in an online information market. In Section 3, we analyze the market equilibrium by considering the order of play. Two cases are con-sidered, one a simultaneous pricing game and the other a sequential pricing game. In Section 4, we compare the results between two compelling equilibria and show that price leadership of a NP induces less content differentiation between CPs’ online
information products, so that the price competition between two CPs becomes more aggressive. The final section provides conclusions.
2. A Model of the Online Information Market
Consider a simplified configuration of the online (electronic) information product market. There are three firms in this market: one NP and two CPs. At one end of the network for the information product, two competing CPs produce differ-ent information contdiffer-ent and store them on a server. At the other end, consumers with varying valuations of the information product are connected to the CPs via the de-livery network (e.g., a telephone line, a television cable, or a mobile phone system) operated by a dominant NP.
Each CP’s information products have an average load size of s bits. Con-i sumers are connected to the delivery network provided by the NP via an access channel of bandwidth capacity b (bits per second), which is assumed to be fixed on average. Then, the delivery time (i.e., the duration starting from the moment the con-sumer orders a product to the time she finishes acquiring it) is determined by the ratio between capacity and load. That is, the delivery time is the transmission delay of the network, which is expressed as wi si b.
The NP charges a usage-based price P per unit time 0 w , and each CP in the i content market sells its information goods at per-period price Pi, i = 1, 2. So, each consumer should pay a network usage charge wiP0 to NP and a fixed charge P i to CPi when she orders firm i’s product. We also assume that there are no production costs for analytic simplicity.
Next, we consider a simple example of a vertical differentiation model for a consumer’s selection behavior. We assume that the quality of network service per-ceived by consumers is only determined by the content size s of the information i product. Then, a consumer’s utility with type θ is described by
0
)
(θ θs P wP
U i i i , where θ denotes the consumer’s preference for the infor-mation product. There is a continuum of consumers, and it is assumed that θ is uniformly distributed over the interval [0,1]. According to the interpretation of Ti-role (1988), θ represents the inverse of the marginal rate of substitution between income and content size. Thus, wealthier consumers with a higher θ have a higher marginal utility of information contents. Each consumer has unit demand (i.e., she consumes at most one unit of the good) and does one of three things: buy from CP1,
buy from CP2, or not buy at all. [That is, we restrict our attention to the uncovered
market case, which Choi and Shin (1992) explored. On relevant issues to this point, see Tirole (1988, pp.96-97) and Wauthy (1996).] Finally, all relevant parameters are assumed to be public and common knowledge.
3. Two-Stage Game Analysis
We consider a situation where one NP and two CPs play a two-stage game, where CPs choose the content size level simultaneously and then the NP and CPs
compete in prices. Specifically, in the first stage of quality choice, given the NP’s bandwidth capacity b, each CPi chooses s from the interval [0,i s] simultaneously and independently. That is, the NP does not play in the first stage, though it does play in the second stage for price choice.
In the second stage for price competition, we consider two cases of price com-petition between the NP and CPs. The first case is in Section 3.1, which examines a game played by setting prices simultaneously and independently. That is, in the first case, each player is symmetrically informed of how the opponent will react to his pricing decisions and maximizes respective profit with this information simultane-ously to yield a Nash equilibrium solution.
The second case appears in Section 3.2, where the NP acts like a dominant player who picks his price while the two CPs play a Nash game after observing the price of NP. That is, in the second case, the NP with market power is endowed with information of the others’ pricing decisions before making its own pricing decision, whereas CPs are not. However, CPs are symmetric in that they compete simultane-ously in price. Hence, the NP is a price leader and the two CPs are price followers. This asymmetric structure will emerge as a Stackelberg equilibrium solution.
3.1 Simultaneous Price Competition (1) Price Choice
Let us first examine a simultaneous price competition between one NP and two CPs. Notice that in the second stage of price choice, qualities are exogenous and thus we assume s1! s2!0. That is, CP1 is the top quality firm and CP2 is the
bot-tom quality firm. This is so because s 1 s2 cannot be an equilibrium since it yields Bertrand competition, i.e., zero profits in the price game. (It can be shown that
0
2 1z s z
s at equilibrium in the first stage, which will be discussed in the following analysis.)
In solving the game, consider the demand faced by each firm. If there are sev-eral qualities offered in the market, the consumers choose among these qualities as well as choosing whether to buy at all. Let a consumer with index θ be indifferent 1 between the two information contents, i.e., θ1s1P1w1P0 θ1s2P2w2P0. Then, all consumers with θ tθ1 (i.e., θ
>
θ1,1@
) always prefer quality type 1 to quality type 2 if they purchase at all. Similarly, consumers with index greater than2
θ and less than θ (i.e., 1 θ
>
θ2,θ1@
) will prefer to buy from firm 2 than not to buy at all. That is, a consumer with index θ is indifferent between buying from 2 firm 2 and not buying at all, so that θ2s2P2w2P0 0. However, any consumer with index less than θ (i.e., 2 θ>
0 θ, 2@
) will not buy any product in this market configuration.Thus, if the two CPs do not cover the market (i.e., there is any consumer who does not buy the product at all), the demand function for each CP’s product is given as follows:
) ( ) ( 1 ) , ( 0 1 2 1 2 1 P s P b P P s s D , (1) 2 2 2 1 2 1 2(P,s) (P P ) (s s ) P s D , (2)
where )P (P0,P1,P2 , )s (s1,s2 , and s i bwi. [In order to have a complete equilibrium solution, it should be shown that D1(P,s)D2(P,s) 1
0
2 2
0 bP s t
P at equilibrium. Below, we show that two CPs do not cover the market in equilibrium.] Notice that the demand function for firm 2 is not directly associated with P . 0
Since costs are zero, the profit function for a CPi is given by ) , ( ) , (P s PiDi P s i
3 . Taking s as given, the following reaction functions for i CPs are obtained from the first-order conditions:
b s s P b bP P 2 ( 20)( 1 2) 1 , (3) 1 2 1 2 P 2ss P . (4)
Then, solving (3) and (4) simultaneously, we have: ) 4 ( ) )( ( 2 2 1 2 1 0 1 1 s bb sP ss s P . (5)
It is noteworthy that the prices of the two CPs are strategic complements, while the prices of the NP and CP are strategic substitutes, i.e., wPi wPj !0 ,
0
0!
w
wPi P , i, j 1,2, and j zi. This implies that the profit-maximizing NP has an interest in raising P and reducing 0 P if possible. This is so because NP i can increase its profit when the NP can re-balance between its price and the CPs’ prices so that the demand size does not change. However, the NP can not do this under the simultaneous price competition situation.
In this simultaneous price competition, the NP has the profit function ) ( ) , ( 0 1 1 2 2 0 P s P wD w D
3 . Then, we get the following price reaction function of the NP in the simultaneous game:
1 1 1 0 b(s2s P)
P . (6)
It is also noteworthy that the NP price is directly related to the CP1 price, which
is a strategic complement of the CP2 price.
By integrating (4), (5), and (6), we have the following Nash equilibrium solu-tion in the second price competisolu-tion stage:
1 2 1 2 2 2 1 1 1 2 1 0 b(2s6s s ), P s 3s , and P s (s6s s ) P .
so that total market size is 1/2 in equilibrium. (It completes the solution in the un-covered market configuration.) Note that the NP price depends on the sizes of dif-ferentiated information contents in equilibrium. In particular, the NP price level has a negative (positive) relationship with respect to the quality level of CP1 (CP2), i.e.,
0
1
0 w !
wP s and wP0 ws2!0. This implies that the NP can set a higher price under less content differentiation.
Anticipating the simultaneous price competition equilibrium, the profit func-tions can be expressed in terms of s :
1 2 1 2 2 2 1 1 1 2 2 1 0 (2s36ss ) , s 9s , and s (36s s s ) 3 3 3 . (7) (2) Quality Choice
Next, we consider the choice of quality levels by firms in the interval [0,s]. We develop the analysis of firm i’s best response to s , j j zi, and a symmetric analy-sis prevails for firm j. Intuitively, given s , choosing j si cannot be a best re-sj sponse to firm i since it yields Bertrand competition, i.e., zero profits in the price game. Then, the best response to s differs according to whether we consider a i response si!sj or si sj. In the former region, firm i is the high quality firm and this allows her to sell to high θ consumers, in the latter region firm i is the low quality firm and sells to low θ consumers.
First, best responses in the domain si!sj are easily determined. Since firm i is the high quality firm, we can consider this firm i to be CP1. Then, profits of the
top quality firm 3 in (7) are increasing in its quality, so that the best response for 1 firm i against any si!sj is to set si , i.e., to choose the best available quality. s On the other hand, best responses in the domain si sj are also easily deter-mined. Given s , firm i is the low quality firm; we can consider this firm i to be j CP2. The best response of the bottom quality firm can be obtained from its profits
function, denoted 3 . The first-order condition for the maximization of 2 3 in (7) 2 yields s 2 s1 2.
In sum, at equilibrium, the top quality firm chooses s while the bottom qual-ity firm chooses s/2. However, it is not determinate which firm has top or bottom quality since the two CPs are symmetric in the first stage.
Without predetermined asymmetry between firms, each firm prefers top quality firm if the other will comply. Thus, we can anticipate that there is always a conflict over the choice of quality; this is a kind of preemption game. If one of the firms en-ters first, that firm would choose high quality and other firm low quality, so the equilibrium would be unique. Otherwise, the equilibrium would not be unique. [On this point, see Dowrick (1986), Tirole (1988), and Lee (1996).] Nevertheless, we can see that any subgame perfect equilibrium at which the two firms enjoy positive prof-its entails product differentiation in which one firm chooses the best available qual-ity s and the other firm chooses half of the best quality s/2.
In the case of s !1 s2, as assumed for the price stage, we have sN s
1 and
2
2 s
differentia-tion does not hold in the online informadifferentia-tion product market. Furthermore, we can get that P1N s6 and 24
2 s
PN , so that PN PN
2
1 4 and P0N 5b12. It is
noteworthy that content size difference relaxes price competition between competing content providers, so that they command positive profits in equilibrium. In particular, from (7) we have 31N s18, 144
2 s
N
3 , and 30N 25s144. However, as stated above, because the profit level of the NP has a positive relationship with re-spect to the quality level of CP2, i.e., w30 ws2!0, the NP might get higher profits
under less content differentiation.
3.2 Sequential Price Competition (1) Price Choice
Now, let us examine the second case of a Stackelberg price-setting game. Con-trary to the first case of simultaneous price competition, the NP has information and market power and thus can commit to its price level in the sequential price competi-tion situacompeti-tion. That is, the NP is the price leader and the CPs are price followers in the second stage. As shown in (5), the NP has an incentive to increase its price, and thus decrease the price levels of the CPs. In addition, the NP has an interest in less product differentiation. To show this, we find a Stackelberg equilibrium solution below. Again, we assume that s1! s2!0.
Taking the CPs’ reaction functions in (3) and (4) as given, the NP chooses that P0 which maximizes its profit. In particular, the NP directly considers the reaction of
firm 1 in (5) when maximizing its profit, 30(P,s) P0(w1D1w2D2). Using the first-order condition under the constraint in (5) yields P 0 b2. Notice that the op-timal price of the NP does not directly depend on the degree of differentiation be-tween information content firms in the first stage equilibrium when the NP has mar-ket power in deciding the prices. This is so because equilibrium quality levels s i are directly dependent on b, i.e., si si(b), at equilibrium and thus the NP can al-ways find the optimal price as a function of b only.
In particular, we have the following Stackelberg equilibrium solution in the second stage: ) 4 ( 2 ) ( and , 4 ) ( , 2 1 2 2 1 2 2 2 1 2 1 1 1 0 b P s ss ss P s ss ss P .
The demand sizes of CP1 and CP2 are D1= 7/24 and D2= 7/48 respectively, so
that total market size is 7/16 in equilibrium. (It completes the solution in the uncov-ered market configuration.) Then, the profit functions can be expressed in terms of s:
2 2 1 2 1 2 1 2 2 2 1 2 1 2 1 1 2 1 2 1 1 0 (4 ) , and 4(4( )) ) ( , ) 4 ( 4 ) 2 ( s s s s s s s s s s s s s s s s 3 3 3 . (8)
Still the profit level of the NP has a positive relationship with respect to the quality level of CP2, i.e., w30 ws2!0, and the NP can get higher profit under less
content differentiation. However, the NP is a second mover when the CP chooses its content size. That is, the NP cannot fully internalize the effect of content differentia-tion so that it cannot obtain the maximized profit level in the vertical structure.
(2) Quality Choice
Analogous to simultaneous case, we can decide the choice of quality levels of each firm at the first stage by considering the best responses in an interval [0,s]. Again, given s , choosing j si cannot be a best response since it yields Ber-sj trand competition.
The best responses of firm i, which is the top quality firm in the domain j
i s
s ! , are easily determined. The profit of the top quality firm in a differentiated market is 3 in (8), and it is increasing in its quality, so that the best response for 1 firm i against any si!sj is si , i.e., choosing the best available quality. s
On the other hand, the best responses of firm i, which is the bottom quality firm in the domain si sj, are determined by the first-order condition for profit maxi-mization. Since the profit of the bottom quality firm in a differentiated market is
2
3 in (8), the first-order condition for the maximization of 3 yields: 2 0 ) ( 2 ) 2 )( 4 ( s1s2 s1 s2 s2 s1s2
,
which simplifies to s 2 4s1 7.In sum, at equilibrium, the top quality firm chooses s while the bottom qual-ity firm chooses 4s/7. By symmetry between the two CPs, however, it is again in-determinate which firm is the top or bottom quality firm. Nevertheless, we can see that any subgame perfect equilibrium at which the two firms enjoy positive profits entails product differentiation in which one firm chooses the best available quality
s and the other firm chooses a quality level 4/7 of the best quality, or 4s/7. In the case of s !1 s2, we have sA s
1 and s2A 4s 7 in equilibrium. Again,
the maximum principle of information content differentiation does not hold in se-quential competition equilibrium. But the degree of product differentiation is less than that in simultaneous competition equilibrium. Furthermore, we get P1A s8 and P2A s 28, so that 7 2
2
1A PA
P and P0A b2. Since the degree of content differences is less in sequential competition equilibrium, the two CPs compete more intensely in prices, so that their prices and profits levels decrease. In particular, from (8) we have 31A 7s192, 192
2A s
3 , and 30A 3s16. However, the price and profit levels of the NP increase under sequential price competition equilibrium. In sum, less product differentiation drives the strategic effect of reducing down-stream prices.
4. Discussion
We now examine the economic consequences to welfare and consumers’ sur-plus in each equilibrium. The welfare function can be defined by the sum of produc-ers’ surplus (the NP’s profits plus the CPs’ profits) and consumproduc-ers’ surplus. Since
costs are zero, we have the following welfare function in an online information market configuration: ) ( ) ( ) (s W1 s1 W2 s2 W
³
³
1 2 1 2 1 1 θθ θ θsdθ θs dθ =s1D1(22D1)s2D2(22D1D2). (9)Substituting the equilibrium solutions into (9) yields W1N 5s/18 and 144
/ 7
2 s
WN in simultaneous price equilibrium, and 287 /1152
1 s WA and 1152 / 61 2 s
WA in sequential price equilibrium. Thus, total social welfare in si-multaneous price equilibrium is greater than that in sequential price equilibrium, i.e.,
1152 / 348 1152 / 376s W s
WN ! A . Therefore, the sequential price competition by the NP with market power reduces the social welfare.
Furthermore, by definition, the consumers’ surplus in each market can be ob-tained as CSi Wi3iPowixi. Then, we have CS1N 96s/1152 and CS2N
1152 /
8s in simultaneous equilibrium and CS1A 77s/1152 and CSA
2 1152
/
7s in sequential equilibrium. Therefore, consumers’ surplus also decreases under the presence of a dominant position of the NP. We then conclude that if the NP can commit to its price level, the equilibrium is inefficient from the viewpoint of social welfare. This result is similar to that of Boyer and Moreaux (1987) who com-pare Nash and Stackelberg equilibria in a differentiated products model. They find that both consumer surplus and welfare are larger in the simultaneous price equilib-rium than in the sequential price equilibequilib-rium.
Finally, we compare our results in sequential price competition and that in Choi and Shin (1992). They considered a duopoly model of vertical differentiation where the firms do not cover the market and found that maximal product differentiation does not hold. This is so because consumers’ tastes are sufficiently diverse so that some consumers do not buy from either firm. Put differently, there exists an outside option. This result is the exactly same as our result for CPs.
The difference between our model and their model is the presence of the net-work access provider (NP) as a strategic agent. Our analysis answers the question raised in the introduction: Does the presence of this third agent affect previous re-sults in the vertical differentiation model? We find that it depends on the economic relationship between the third agent and competing firms. Specifically, we have shown that the level of product differentiation depends on the ability of the NP to set prices in the online information market.
Intuitively, we immediately note that the presence of the NP affects consumers’ utility by U(θ) θsiPiwiP0 (θ P0 b)siPi . In other words, the presence of the NP affects the marginal willingness of consumers to pay for quality exactly by the price per bandwidth capacity. Then, if the NP cannot commit to its price, both the degree of content differentiation and the price levels of CPs depend on the stra-tegic effect of complementary prices. In particular, CPs can raise their prices through
more differentiation in a simultaneous competition equilibrium. Therefore, we can conclude that if the NP commits to its price before the CPs, its presence does not affect quality choice, whereas if they all set prices simultaneously, it does so and leads to more differentiation.
5. Concluding Remarks
This paper examines games of vertical product differentiation in an online in-formation market, in which the NP and CPs compete over prices for the real-time network delivery of digital information products. We have shown that the degree of information content differentiation depends on the competition environment in the online information market. More information content differentiation arises when the NP acts as a follower in the price choice game. The NP, however, can get more profit when it acts as a leader. Thus, if the NP who establishes a bottleneck facility in real-time network delivery has market power in its favor and incentives for acting as a leader, less content differentiation occurs. This lessened product differentiation has a strategic effect similar to reducing the CPs’ prices. As a result, the CPs choose their prices more competitively in the sequential competition equilibrium, so that the prices and profits of the CPs decrease, while the price and profit of the NP increases. However, price leadership reduces not only social welfare but also consumers’ sur-plus in this online information market.
Our results demonstrate that further research is necessary before one can be sufficiently confident to design appropriate antitrust laws for online information markets. For example, for more general parameterization of θ, we can obtain the relationship between the pricing behavior of the NP and the distribution of consum-ers’ willingness to pay for quality or market coverage. This treatment is in line with Wauthy (1996). In addition, the robustness of our results can be examined by incor-porating more general forms of quality costs and capacity investment. On this point, see Oh and Chang (2000). Furthermore, as an interesting future research, it is worthwhile to extend our policy-relevant results to the model of competing NPs and investigate the capacity investment in network architecture or/and compatibility in-centives of two NPs in implementing interconnection. For instance, Mackie-Mason et al. (1996) explore the potential impact of architectural differences in the distribu-tion network on the types of the content provided. Foros and Hansen (2001) analyze economic issues of network competition and compatibility strategies.
Finally, it is noteworthy that we do not consider the vertical integration and foreclosure problems, where the NP integrates with one of the CPs. This is so be-cause producers and distributors in the information product market maintain separate ownership and business activities due to different requirements for business strate-gies, different expertise, and antitrust laws. For example, the NP may choose to spe-cialize in high-quality services and thus attract customers preferring these services. However, we can also observe that vertical integration in online information product markets occurs frequently. Therefore, it is also worthwhile to consider alliances be-tween the NPs and CPs as a source of product differentiation. Extending our model
in these directions is left for future research.
References
Anderson, S. P. and M. Engers, (1992), “Stackelberg versus Cournot Oligopoly Equilibrium,” International Journal of Industrial Organization, 10, 127-135. Aoki, R. and T. J. Prusa, (1997), “Sequential versus Simultaneous Choice with
En-dogenous Quality,” International Journal of Industrial Organization, 15, 103-121.
Boyer, M. and M. Moreaux, (1987), “On Stackelberg Equilibria with Differentiated Products: The Critical Role of the Strategy Space,” Journal of Industrial Eco-nomics, 36, 217-230.
Choi, C. J. and H. S. Shin, (1992), “A Comment on a Model of Vertical Product Dif-ferentiation,” Journal of Industrial Economics, 40, 229-231.
Dowrick, S., (1986), “Von Stackelberg and Cournot Duopoly: Choosing Roles,” Rand Journal of Economics, 17, 251-260.
Economides, N., (1993), “Quantity Leadership and Social Inefficiency,” Interna-tional Journal of Industrial Organization, 11, 219-237.
Economides, N., (1999), “Quality Choice and Vertical Integration,” International Journal of Industrial Organization, 17, 903-914.
Foros, O. and B. Hansen, (2001), “Competition and Compatibility among Internet Service Providers,” Information Economics and Policy, 13, 411-25.
Gabszewicz, J. and J. F. Thisse, (1979), “Price Competition, Quality and Income Disparities,” Journal of Economic Theory, 20, 340-359.
Kahin, B. and H. R. Varian, (2000), Internet Publishing and Beyond: The Economics of Digital Information and Intellectual Property, Cambridge: The MIT Press. Laffont, J.-J. and J. Tirole, (2000), Competition in Telecommunications, Cambridge:
The MIT Press.
Lee, S. H., (1996), “Choosing Roles in a Model of Quality Differentiation,” Seoul Journal of Economics, 9, 191-201.
Mackie-Mason, J. K., S. Shenker, and H. R. Varian, (1996), “Service Architecture and Content Provision—The Network Provider as Editor,” Telecommunications Policy, 20, 203-217.
Mussa, M. and S. Rosen, (1978), “Monopoly and Product Quality,” Journal of Eco-nomic Theory, 18, 301-317.
Oh, C. J. and S. G. Chang, (2000), “Incentives for Strategic Vertical Alliances in Online Information Product Markets,” Information Economics and Policy, 12, 155-180.
Shaked, A. and J. Sutton, (1982), “Relaxing Price Competition through Product Dif-ferentiation,” Review of Economic Studies, 49, 3-13.
Sapiro, C. and H. R. Varian, (1999), Information Rules: A Strategic Guide to the Network Economy, Boston: Harvard Business School Press.
Shy, O., (2001), The Economics of Network Industries, New York: The Cambridge University Press.
Tirole, J., (1988), The Theory of Industrial Organization, Cambridge: The MIT Press.
Waterman, D., (1993), “A Model of Vertical Integration and Economies of Scale in Information Product Distribution,” Journal of Media Economics, 16, 23-35. Wauthy, X., (1996), “Quality Choice in Models of Vertical Differentiation,” Journal
