- Email: [email protected]
Does direct connect benefit travellers?
Journal Pre-proof Does direct connect benefit travelers? Jorge Padilla, Salvatore Piccolo
PII: DOI: Reference:
S0165-1765(20)30007-0 https://doi.org/10.1016/j.econlet.2020.108952 ECOLET 108952
To appear in:
Economics Letters
Received date : 23 October 2019 Revised date : 4 December 2019 Accepted date : 7 January 2020 Please cite this article as: J. Padilla and S. Piccolo, Does direct connect benefit travelers?. Economics Letters (2020), doi: https://doi.org/10.1016/j.econlet.2020.108952. This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
© 2020 Published by Elsevier B.V.
Journal Pre-proof
Does Direct Connect Bene…t Travelers? Salvatore Piccoloz
pro of
Jorge Padillay
December 4, 2019
Abstract
Direct connect is a business practice that, after a period of irrelevance, has become fashionable again in the travel industry. Airlines provide direct access to their
re-
sale systems to travel retailers, who can thus avoid dealing with traditional indirect distribution intermediaries, such as GDS aggregators. Although this practice is often advertised as pro-competitive, we show that it may actually harm consumers, especially in very competitive environments and when it does not involve all retailers in
lP
the market.
JEL Classification: L42, L50, L81.
Jo
urn a
Keywords: Agency Model, Direct Connect, Distribution Channels, Travel Industry.
For many helpful comments, we would like to thank Joe Harrington (the Editor) and one anonymous referee for useful comments and suggestions. We also thank Michele Bisceglia and Ra¤aele Fiocco, as well as seminar participants in Naples (CSEF), Compass Lexecon London, Catholic University of Milan. We acknowledge …nancial support from Amadeus. The views expressed in this paper are the sole responsibility of the authors and cannot be attributed to Compass Lexecon or its clients. y Compass Lexecon. z University of Bergamo, Compass Lexecon and CSEF.
1
Journal Pre-proof
1
Introduction
Jo
urn a
lP
re-
pro of
Direct connect refers to a business practice through which airlines used to distribute their content to travel agents in the past, prior to the emergence of the GDS1 aggregators such as Amadeus, Sabre, Travelport and Travelsky. After the advent of GDSs it then became largely irrelevant, but it is now growing again, as prominent airlines are seeking for sources of leverage against the GDSs. Notably, airlines are progressively providing direct access to their sale systems, which enables third party travel retailers (e.g., travel management companies) to search for availability, make and manage bookings on a one-to-one basis. This practice is often advertised as a pro-competitive tool (especially by American Airlines and Lufthansa)2 , limiting the market power of traditional distribution channels and promoting consumer welfare. The idea is that, thanks to this technology, passengers and retailers can avoid paying high booking fees and retail prices, often associated with the presence of platforms’(pro…t) margins. Yet, to the best of our knowledge, there is no formal economic model supporting this claim. Is the positive view of direct connect grounded on solid economic principles? If not, why? And, when should it be a concern for competition policy? We consider an agency model (see, e.g., Johnson, 2017) where a supplier (airline) distributes its product (content) both directly through its own distribution channel and indirectly through two retailers (travel agents). Absent direct connect, the intermediaries rely on the IT infrastructure provided by specialized platforms (GDSs) to buy the seller’s content on behalf of …nal consumers. By contrast, with direct connect, the supplier grants one retailer direct access to its content while the other continues to access the market through its exclusive platform (in Section 2.3 we discuss why this ‘partial’form of direct connect is a fairly good re‡ection of reality and what would happen when both retailers operate under direct connect). In both regimes, platforms charge the supplier a commission for each sale they process, and the supplier pro…ts by charging retailers access prices — i.e., the prices (net of commissions, surcharges and discounts) that the supplier requests to authorize a transaction in the indirect channel, whether it is carried out on a platform or through the direct connect technology. We characterize the equilibrium of the game with and without direct connect, and compare consumer surplus across the two regimes. We show that the e¤ect of direct connect on consumer surplus (with only one retailer being served directly by the supplier) is ambiguous and depends on the degree of competition between retailers within the indirect channel and across distribution channels — i.e., between the supplier in the direct channel and the retailers in the indirect channel. Speci…cally, when 1 2
Global Distribution System. See, e.g., https://lufthansa-city-center.com/leisure/en/distribution-platform
2
Journal Pre-proof
2
The model
lP
re-
pro of
the supplier grants a retailer direct access to its sale system, it has an incentive to reduce the demand for the content distributed by the other retailer in order to minimize the commissions paid to the platform dealing with that retailer. The retailer using the GDS platform is thus charged an access price larger than the retailer operating under direct connect. Hence, since access prices are passed on to travellers, the demand for the content distributed directly by the supplier and indirectly by the retailer operating under direct connect increases. This consolidates the supplier’s market power, who has an incentive to increase the price paid by travellers in the direct channel without being afraid of loosing business to the indirect channel (in particular to the retailer using the GDS platform). Yet, when competition across di¤erent distribution channels is strong — i.e., when many passengers regard di¤erent distribution channels as close substitutes — this leads retailers to increase their prices too, whereby harming consumers. By contrast, when competition is su¢ ciently weak, direct connect bene…ts consumers. In this case, the e¤ect described above is negligible because demand functions in the direct and indirect channels are nearly independent. Hence, the main e¤ect of direct connect is that of reducing double marginalization — i.e., the supplier saves the commissions it would have paid to the platform dealing with the retailer operating under direct connect. Obviously, when competition is neither too strong nor too weak, the two e¤ects balance out.
3
Jo
urn a
A monopolistic travel supplier (airline) denoted by S distributes its product (content) both through a direct sale system, and indirectly through an intermediated channel where two platforms (denoted by Pi , with i = A; B) are accessed by exclusive retailers (denoted by Ri , with i = A; B) competing to attract …nal consumers (see, e.g., Bisceglia et al., 2019, Boik and Corts, 2016, Gaudin, 2019, and Johansen and Vergé, 2016, among others).3 In contrast to the existing models, however, consider the possibility that, in addition to the direct distribution channel, the supplier may also grant direct access to its content to one of the retailers operating in the indirect channel: ‘direct connect’. Suppose, without loss of generality, that RB always deals with PB in order to access S’s content, while RA can either access the market by dealing directly with S or, alternatively, operating through its own exclusive platform PA when S gives up direct connect (and allows that platform to market its content). The assumption of exclusivity seems realistic in the travel industry because travel retailers usually access only one platform, and learning by doing economies together with switching costs may actually make competition for retailers harder. In Bisceglia et al. (2019) we discuss how competition between platforms a¤ects the equilibrium of the game.
3
Journal Pre-proof
pro of
Figure 1 below provides a graphical illustration of the di¤erent industry structures we consider.
Figure 1
lP
re-
S sets a price pd on its direct sale system, and contracts with platforms and retailers. Following the literature and business practice, contracts are linear and secret. The industry’s business structure is the agency model proposed by Johnson (2017). Speci…cally, a contract between S and Pi speci…es a commission (fee) fi paid by S to Pi for each unit distributed by Ri . In addition, S sets the access price i that it charges Ri for every unit sold through Pi or via direct connect. The timing of the game is as follows: t = 0 S decides whether RA operates via direct connect or via PA ;
urn a
t = 1 Platforms (only one or both depending on S’s decision in t = 0) o¤er commissions to the supplier; t = 2 The supplier accepts or refuses the o¤ers, and sets access prices; t = 3 The retailers and the supplier simultaneously post …nal prices and demand allocates across the two channels.
Jo
Production costs are linear and marginal costs are normalized to zero without loss of generality. Hence, S’s aggregate pro…t is S
(),(
A
IfA ) qA + (
B
f B ) qB + p d qd ;
with qi and qd denoting the quantities sold through platform i = A; B and the direct distribution system, respectively. The indicator function I takes value 0 if RA deals directly with S and 1 otherwise. Pi ’s pro…t is Pi ( ) , fi qi ; while Ri ’s pro…t is R i ( ) , i qi : 4
Journal Pre-proof
1
b
and qd , Dd (pd ; pA ; pB ) =
(1 + b) pi + b (p (1 b)(1 + 2b)
re-
qi , DA (pi ; p i ; pd ) =
pro of
Since contracts are secret, following Johansen and Vergé (2016) and Rey and Vergé (2017), the solution concept is Contract Equilibrium (Crémer and Riordan, 1987, and Horn and Wolinsky, 1988). This equilibrium concept has some of the features of a perfect Bayesian Nash equilibrium with passive beliefs: even if it receives an out-of-equilibrium contract, each retailer chooses its price assuming that the rival remains under the equilibrium contract. This is in line with the market-by-market bargaining restriction of Hart and Tirole (1990) and with the passive beliefs or pairwise-proofness assumption of McAfee and Schwartz (1994). In line with the evidence (see, e.g., Cazaubiel et al., 2018) we assume that consumers perceive the contents distributed through the direct and the indirect channel as imperfect substitutes.4 Hence, following Johansen and Vergé (2016), we consider the following demand functions
1
b
i
+ pd )
8i = A; B;
(1 + b) pd + b (pA + pB ) ; (1 b)(1 + 2b)
(1)
(2)
4
Jo
urn a
lP
where b re‡ects the degree of substitutability between products within and across distribution channels.5 We assume that b 2 0; b , with b 0:9 to guarantee that second-order conditions hold (see Appendix 2 in Bisceglia et al. 2019). Finally, notice that in the model described above it is assumed that platforms provide no useful role — i.e., they do not bring any e¢ ciency compared to direct connect. This is clearly a simpli…cation made only for the sake of analytical tractability. There are, indeed, many good reasons to believe that (other things being equal) GDSs may increase consumer welfare. For example, it is well known that GDSs have developed (over time) IT technologies that o¤er wider and better services compared to what airlines can do on their own: while Direct Connect is speci…c to a particular airline, and does not automatically aggregate content from multiple airlines, GDSs have over the years developed technologies whose goal is precisely to sort out the best travel plan for every passenger. Hence, in light of this, the negative welfare result on direct connect should be interpreted as a conservative result, that does not take into account the potential additional bene…ts that platforms may bring. Some consumers may in fact prefer to purchase through the indirect channel because intermediaries o¤er (un-modelled) additional services that M is unable or unwilling to supply in the direct channel. 5 To simplify exposition, we assumed that the degree of intra- and inter-channel competition is the same (b). Nevertheless, our results hold qualitatively when consumers perceive the contents distributed by the two retailers as less di¤erentiated than the content distributed directly by the supplier. In this case, it can be shown that direct connect harms consumers as long as inter- and intra-channel competition is su¢ ciently intense. Proofs and numerical simulations are available upon request.
5
Journal Pre-proof
2.1
Equilibrium analysis
pro of
Before solving the game, it is useful to observe that a multiproduct supplier, who fully internalizes the e¤ects of intra- and inter-channel competition, charges the same price for all products pM = 12 and, because demand functions feature preference for variety, it sells the same quantity of each product.6 In what follows, we will …rst develop the analysis for the case of direct connect, then we brie‡y review the logic of the model in the regime without direct connect. In Section 2.2 we then compare the retail prices and consumer surplus across the two regimes.
re-
Equilibrium with direct connect. Suppose that S forecloses PA and deals directly with dc RA . Consider an equilibrium in which: (i) Ri charges pdc i and S charges pd in the indirect dc and direct channel, respectively; (ii) S charges dc i to Ri ; (iii) S is charged fB by PB . We solve the game with a backward induction logic. Ri solves max D pi ; pdci ; pdc i) ; d (pi pi 0
whose …rst-order condition yields i
2
+
i.
max(
B
8i = A; B;
Clearly, pdc i ( i ) is increasing in
urn a
representing Ri ’s best response to prices pdci and pdc d . S solves
pd
b + b pdci + pdc d 2(1 + b)
1
lP
pdc i ( i) ,
dc fB )D(pdc B ; pA ; pd ) +
dc dc A D(pA ; pB ; pd )
(3)
i
and in the rivals’
dc + pd D(pd ; pdc A ; pB );
whose …rst-order condition is 1
6
|
b
dc (1 + b) pd + b pdc A + pB (1 b)(1 + 2b) {z
pd
Monopoly rule
(1
1+b + b)(1 + 2b) }
Jo
Indeed, the above demand system re‡ects the preferences of a representative consumer, whose utility function is X X X 1 X 2 U( ) , qj qj b qj qi pj qj + m; 2 j=A;B;d
j=A;B;d
i;j=A;B;d;;j6=i
j=A;B;d
which is strictly concave and thus exhibits preference for variety. For example, in the airline ticket industry, consumers may want to have the ability to book in di¤erent ways because they may not always be able to reach directly the airline website.
6
Journal Pre-proof
+ (1 |
b ( b)(1 + 2b) {z
fB +
B
Channel externality
A)
(4)
= 0:
}
pdc d (
A;
B ; fB )
,
pro of
Clearly, a higher pd increases S’s pro…t margin but lowers demand in the direct channel; moreover, it also increases S’s pro…t from the indirect channel because (other things being equal) consumers switch from the direct to the indirect channel, so that S collects higher fees from RB . The solution of (4) yields S’s best response to the access prices and the fee charged by PB — i.e., 1 b b b dc + pdc ( A + pB + 2 (1 + b) 2 (1 + b) 2 (1 + b)
A
+
B
fB ) ;
(5)
max ( A
0;
0
B
B
re-
which is increasing in the rivals’retail prices and in the access prices, and decreasing in the fee charged by PB (the higher this fee, the lower the incentive to increase pd since volumes diverted to RB are less pro…table). Moving back to stage 2, S solves dc dc fB )D(pdc B ( ); pA ( ); pd ( ))+
dc A D(pA (
dc dc d dc dc dc ); pdc B ( ); pd ( ))+pd ( )D (pd ( ); pA ( ); pB ( )):
1
lP
Let Ii be an indicator function taking value 1 for i = B and 0 for i = A. The …rst-order condition with respect to i is dc dc (1 + b) pdc i ( ) + b p i ( ) + pd ( ) (1 b)(1 + 2b) {z
b
|
(1
1+b ( b)(1 + 2b)
urn a
Monopoly rule
+ (1 |
b b)(1 + 2b)
i
I if {z
i
+ pdc d ( )
Channel externality
i
Ii fi )
@pdc i ( ) + @ i }
@pdc i ( ) = 0 8i = A; B: @ i }
(6)
Jo
S’s pro…t increases with i because (for a given demand) S collects higher revenues from Ri . However, since contracts are secret, a higher i only increases the retail price charged by Ri . Hence, other things being equal, Ri ’s demand drops, whereas R i ’s demand on the indirect channel and S’s demand on the direct channel increase. Solving (6), we obtain access prices in terms of fB — i.e., dc B
(2 + 3b)2 16 + 48b + 10b2 59b3 33b4 (fB ) = + fB ; 2 (1 + 2b) (4 + 3b) 2 (1 + 2b) (1 b) (4 + 3b) (4 + 5b) | {z } >0
7
(7)
Journal Pre-proof
and dc A
(2 + 3b)2 (fB ) = 2 (1 + 2b) (4 + 3b)
b (1 + b) (4 + 8b 3b2 ) fB : 2 (1 + 2b) (1 b) (4 + 3b) (4 + 5b) | {z }
(8)
>0
pdc d (
dc A (fB );
pro of
While dc B (fB ) is increasing in fB because a higher fee will be passed on by S to RB , dc A (fB ) is decreasing in fB because as RB becomes less e¢ cient due to a higher ‘marginal cost’S has an incentive to divert business towards RA . We can …nally move to stage 1. Let dc B (fB ); fB )
dc dc and pdc R ( R (fB )) , pR (fB ): PB solves
, pdc d (fB );
dc dc max fB D(pdc B (fB ); pA (fB ); pd (fB )); fB 0
|
b
dc dc (1 + b) pdc B ( ) + b pA ( ) + pd ( ) (1 b)(1 + 2b) {z
Monopoly rule
+fB (1 |
lP
1
re-
whose …rst-order condition is
fB
(1
1+b @pdc B (fB ) + b)(1 + 2b) @fB }
@pdc (fB ) b @pdc A (fB ) + d = 0: b)(1 + 2b) @fB @fB {z }
(9)
Strategic e¤ect <0
Jo
urn a
Three e¤ects shape PB ’s optimal fee. First, for a given demand, by increasing the commission charged to S, PB earns a higher pro…t. Second, since S reacts to such a higher fee by increasing the access price charged to RB , the demand on platform PB drops. Third, since fB a¤ects the access prices and S’s pro…t margin on RB ’s sale volume, there is also a strategic e¤ect: pdc A ( ) drops as fB increases because RA ’s access price decreases as explained before; dc pd ( ) also drops because RA charges a lower price as just explained and S’s pro…t margin on RB is lower, which means that S has a weaker incentive to increase the price in the direct
8
Journal Pre-proof
channel in order to pro…t on the indirect one.7 Solving (9) we have fBdc =
(4 + 5b) 4 (1 b)3 (1 + b) (1 + 2b) : 2 3 179b{z 122b4 + 85b5 + 87b6} |32 + 96b + 10b (>0) PB ’s mark up
pro of
dc Substituting fBdc into (7) and (8) we obtain the equilibrium access prices dc A and B , then dc dc using (4) and (5) we obtain the equilibrium retail prices pdc A ; pB and pd (see the Appendix). dc dc dc Proposition 1 With direct connect, access prices are such that dc B > A > 0 and B > fB . dc dc dc Retail prices are such that with pdc and pdc 0:75. B > max pA ; pd A > pd if and only if b The supplier and both retailers sell positive quantities in the direct and indirect channel, respectively — i.e., qAdc > qBdc > 0 and qddc > 0.
lP
re-
In order to make positive pro…ts PB charges a positive commission to S, who will then pass on this commission to RB by increasing A . In turn, RB marks up consumers by setting a price larger than pM . Because contracts are linear and prices are strategic complements, S will also charge a positive access price to RA , which creates a mark-up also for consumers dc buying through direct connect. Obviously, in equilibrium, dc B because S diverts A < demand from RB to RA in order to minimize the commissions paid to PB . Finally, due to strategic complementarity, S will also charge a positive mark-up on the direct channel.8
urn a
Equilibrium without direct connect. The analysis of the regime without direct connect has already been developed in Bisceglia at al. (2019). Hence, for brevity, we will only state the result (details are in the Appendix). Proposition 2 (Bisceglia et al. 2019) Without direct connect there exists a unique symmetric equilibrium such that platforms charge a fee f n > 0 to S, who then sets n > f n to each retailer and charges pnd in the direct channel, retailers set the same price pn > n in the indirect channel.
7
Jo
The equilibrium features multiple marginalization: platforms charge S positive commissions, which will be passed on to the retailers via higher access prices, which in turn induce Formally, by (5) we have
@pdc d (fB ) @fB
= =
8
b 2 (1 + b)
@
dc A
(fB ) @ dc (fB ) + B @fB @fB
b 2 (1 + b) 3b2 < 0: 2 (1 b) (1 + 2b) (4 + 3b)
Because consumers feature taste for variety, S never shuts down PB .
9
1
Journal Pre-proof
a …nal price higher than pM . Due to strategic complementarity, S charges a very high price in the direct channel too.
2.2
Consumer welfare analysis
pro of
We now examine the e¤ects of direct connect on consumers. We start by comparing retail prices. Proposition 3 Direct connect always increases the price charged by S in the direct channel — i.e., pdc pnd with equality only at b = 0. As for the indirect channel, there exist two d thresholds b0 and b1 , with 0 < b0 < b1 < b, such that: pdc A
pn if and only if b 2 [0; b1 ], with equality at b = b1 ;
pdc B
pn if and only if b 2 [0; b0 ], with equality at b = b0 and b = 0.
urn a
lP
re-
dc With direct connect, S charges dc B > A in order to divert business from RB towards RA . This induces RB to pass on the higher access price to consumers. Hence, demand in the direct channel increases and, in turn, S charges a higher price in that market segment. However, a higher price in the direct channel also induces retailers to charge higher prices in the indirect channel when b is su¢ ciently large because of a relatively stronger e¤ect of pd on RA and RB ’s demand functions. Figure 2 shows the equilibrium prices as a function of b.
Jo
Panel a: direct channel
Panel b: indirect channel
Figure 2: Equilibrium prices
In panel a, the solid curve is pnd while the dashed one is pdc d ; in panel b, the black curve n cd cd is p , the red pA and the green pB . In line with the intuition provided above, the region of parameters in which direct connect increase prices in the indirect channel corresponds to the region of parameters in which the di¤erence pdc pnd is larger. d We can now study consumer surplus. 10
Journal Pre-proof
Proposition 4 There exits a threshold b 2 (0; b1 ) such that direct connect harms consumer surplus if and only if b b :
lP
re-
pro of
dc The intuition is straightforward. Because pdc d > pd , direct connect unambiguously harms consumers who buy directly from the supplier. However, in the indirect channel the e¤ect is ambiguous. Clearly, for b su¢ ciently large (e.g., b > b1 ) direct connect increases prices in the intermediated channel too, and hence it harms consumers in both channels. On the contrary, n dc n dc n for b ! 0 it can be seen that pdc A < p , pd ! pd and pB ! p . Hence, for b su¢ ciently small direct connect overall bene…ts consumers: it has a …rst-order bene…cial e¤ect in the indirect channel and a second-order negative e¤ect in the direct channel. When b takes intermediate values the two e¤ects balance out. Figure 3 provides a graphical illustration of the result
Figure 3: Consumer surplus
The solid (dashed) curve represents consumer surplus with (without) direct connect.
Discussion
Jo
2.3
urn a
Finally, it should be noticed that while direct connect harms consumers for a large enough b, it unambiguously bene…ts the supplier who (absent e¢ ciencies on the platforms’ side) always has an incentive to grant direct access to its sale system to at least to one retailer. By revealed preferences, in fact, the supplier saves on the commission it would have paid to the platform dealing with RA , while being able to control the retail prices through an appropriate choice of A and B .
Throughout the analysis we have assumed that S grants direct access to its distribution system to one retailer only. Clearly, if it could, S would like to foreclose both platforms: by doing so it would (other things being equal) save the commissions paid to the platforms, whereby increasing pro…ts. Notably, this would also wipe out the e¤ect of commissions on multiple marginalization, whereby leading to lower prices and higher consumer surplus. Hence, our model does predict that when direct connect occurs on a wide-market scale, it 11
Journal Pre-proof
9
urn a
lP
re-
pro of
unambiguously bene…ts consumers. However, at least in the short-medium run, it seems hard to imagine that airlines will be able to capture all retailers. As a matter of fact, not all travel agencies operate via direct connect (according to the CMA, direct connect at the moment only covers 2% of the market, although it is expected to grow). In fact, foreclosing all (or a large part of) the GDSs at once might be too costly for airlines (at least in the short-medium run). First, GDSs have developed (over time) IT technologies that o¤er wider and better services compared to what airlines can do on their own. As the CMA itself recognizes the supply of Direct Connects, and therefore the speci…c services these entail in terms of retailing, distribution (including aggregation) and ful…lment, is less standardized than the supply of GDS services and can vary signi…cantly between suppliers. The CMA also notices that in terms of aggregation, as a Direct Connect is speci…c to a particular airline, it does not automatically aggregate content from multiple airlines. Finally, in general, Direct Connects are not able to fully facilitate interlining and corporate reporting/compliance, as it is possible via a GDS.9 Second, by migrating to direct connect travel agents may incur high switching costs that must be compensated by airlines. GDS systems require training and travel agents must learn the ‘language’of the system in order to search for ‡ights and understand the results. Di¤erent systems may use di¤erent ‘languages’. Hence, switching to direct connect may not be costless. In addition, some GDSs also typically o¤er Technology Funding to travel agents in a form of cash or reimbursement. This funding is important for the travel agent to buy PCs, laptops, servers or any hardware essential to the business. Migrating to direct connect may require the provision of new hardware and software in addition to duplicating training and adoption costs. Third, as a matter of fact, when dealing with platforms, travel agents have direct access to the revenues generated by a range of ancillary services and products, to which they might not necessarily have access when dealing directly with airlines.10 When booking at travel agencies, travellers often also book some of these services in advance and travel agents typically charge them commissions for these services. Hence, giving up the revenues coming from these services might be costly for travel agents. In particular, in order to provide
Jo
An interesting quote from the from the very recent case US vs Sabre-Farelogix (2019) states the following: “Some traditional travel agencies also rely on other lines of their GDS’s business for IT products. For example, many travel management companies in Sabre’s network use mid- or back-o¢ ce software supplied by Sabre to perform monitoring or reporting for their corporate clients. To ensure consistent support and reporting for their travellers, corporations typically rely on only one travel management company. . . . ”(page 8). 10 For example, in addition to sharing revenues from traditional services, like hotel accommodations, car rentals and travel insurance, Amadeus typically shares with its clients revenues from other ancillary services like on-board sales of food and beverages; checking of baggage and excess baggage; assigned seats or better seats such as exit rows; priority check-in and screening; early boarding bene…ts; on-board entertainment systems; wireless internet access, etc.
12
Journal Pre-proof
3
pro of
the right incentives to switch to direct connect airlines must compensate travel agents for these missed pro…t opportunities. This creates a cost for the airlines associated with direct connect. And, when this cost is su¢ ciently high, airlines might then endogenously decide to foreclose only some of the platforms. Hence, although in our context it would be socially desirable to have a market that fully operates under direct connect, in the short-medium run this seems rather unfeasible and in this case having only a partial form of direct connect does not necessarily bene…t consumers. This negative result would clearly be reinforced when taking into account the e¢ ciencies discussed before that GDSs bring in terms of product quality compared to direct connect.
Conclusion
Jo
urn a
lP
re-
The idea that direct connect unambiguously bene…ts consumers in the travel industry needs to be quali…ed. By providing some (but not all) retailers direct access to their sale systems, travel suppliers can actually consolidate their market power vis-à-vis traditional distribution channels, especially in competitive environments. Last, but not least, our model has important implications on the current debate on platform parity provisions — i.e., contractual agreements according to which airlines commit not to charge di¤erent prices for the same product distributed through di¤erent platforms. In Bisceglia et al. (2019) we have shown that, in a model without direct connect, platform parity bene…ts consumers as long as competition between and across distribution channels is su¢ ciently intense. Hence, the likelihood that direct connect harms consumers when the fall back option is an indirect channel with two platforms and price parity is very high. Therefore, in this case (under the assumption that regulations cannot impose airlines to drop direct connect) imposing a price parity agreement according to which airlines cannot charge di¤erent (access) prices to the retailer operating under direct connect and those operating through platforms certainly increases consumer welfare. The reason is that direct connect would still allow airlines to save on the commissions paid to the active platforms (whereby reducing multiple marginalization), but it would also prevent them from consolidating their market power by increasing the access price charged to the retailers accessing the market through platforms.
13
Journal Pre-proof
Appendix Proof of Proposition 1. Substituting for (3), (5), (8) and (7) into (9) and solving for fB dc dc dc yields fBdc . Substituting fBdc into (8) and (7), yields dc A and B . Showing that A < B and dc dc dc dc dc B > fB is immediate. Substituting fB , A and B into the system (3)-(5), solving for pA , pB and pd
pdc A
=p
M
b (1 + b) (24 + 78b + 24b2 141b3 128b4 + 65b5 + 87b6 ) ; (3b + 4) (1 + 2b) (32 + 96b + 10b2 179b3 122b4 + 85b5 + 87b6 )
pro of
M pdc + d = p
64 + 296b + 284b2 492b3 879b4 + 95b5 + 703b6 + 124b7 159b8 ; + 3 (4 + 3b) (1 + 2b) (32 + 96b + 10b2 179b3 122b4 + 85b5 + 87b6 )
M pdc + B = p
96 + 416b + 292b2 860b3 1087b4 + 471b5 + 967b6 4b7 255b8 : 2 (4 + 3b) (1 + 2b) (32 + 96b + 10b2 179b3 122b4 + 85b5 + 87b6 ) 0:75. Checking that qAdc > qBdc > 0
re-
dc dc dc and pdc with pdc A > pd if and only if b B > max pA ; pd and qddc > 0 is immediate.
Proof of Proposition 2. From Bisceglia et al. (2019) it follows that S sets pnd = pM +
b (1 + b) (8 9b2 ) 64 + 152b 28b2 171b3
pn = pM +
lP
while retailers charge the same price
48 + 56b 130b2 2(64 + 152b 28b2
54b4
;
59b3 + 81b4 : 171b3 54b4 )
urn a
n dc dc Proof of Proposition 3. The proof follows by comparing pdc A and pB with p , and pd with pnd (Figure 2).
Proof of Proposition 4. Consumer surplus under direct connect obtains by substituting the equilibrium prices into the utility function that generates the demand system (1)-(2) — i.e., CS dc =
X
qjdc
where
b
qddc =
1
b
X
qjdc qidc
i;j=A;B;d;;j6=i
Jo
j=A;B;d
1 X dc 2 (q ) 2 j=A;B;d j
dc dc (1 + b) pdc d + b p A + pB ; (1 b)(1 + 2b)
qidc =
1
b
X
dc pdc j qj + m;
j=A;B;d
dc dc (1 + b) pdc i + b p i + pd ; (1 b)(1 + 2b)
for every i = A; B: Similarly, CS n = 2q n + qdn
(1 + b) (q n )2
1 n 2 (q ) 2 d 14
2bq n qdn
2pn q n
pnd qbn + n;
Journal Pre-proof
where, qn =
1 b pn + bpnd ; (1 b)(1 + 2b)
qdn =
Comparing CS dc and CS n it follows that b
1
b (1 + b) pnd + 2bpn : (1 b)(1 + 2b)
0:5 (Figure 3).
pro of
References
[1] Bisceglia, M., Padilla, J., & Piccolo, S. (2019). When Prohibiting Platform Parity Agreements Harms Consumers. Available at SSRN 3454527. [2] Boik, A., & Corts, K. S. (2016). The e¤ects of platform most-favored-nation clauses on competition and entry. The Journal of Law and Economics, 59(1), 105-134.
re-
[3] Cazaubiel, A., Cure, M., Johansen, B. O., & Vergé, T. (2018). Substitution Between Online Distribution Channels: Evidence from the Oslo Hotel Market. University of Bergen, Department of Economics (No. 8/18). [4] Cremer, J., & Riordan, M. H. (1987). On governing multilateral transactions with bilateral contracts. The RAND Journal of Economics, 436-451.
lP
[5] Gaudin, G. (2019). Vertical relations, opportunism, and welfare. The RAND Journal of Economics, 50(2), 342-358. [6] Hart, O., & Tirole, J. (1990). Vertical integration and market foreclosure. Brookings papers on economic activity. Microeconomics, 205-286.
urn a
[7] Horn, H., & Wolinsky, A. (1988). Bilateral monopolies and incentives for merger. The RAND Journal of Economics, 408-419. [8] Johansen, B. O., & Vergé, T. (2016). Platform price parity clauses with direct sales. IDEI working. [9] Johnson, J. P. (2017). The agency model and MFN clauses. The Review of Economic Studies, 84(3), 1151-1185.
Jo
[10] McAfee, R. P., & Schwartz, M. (1994). Opportunism in multilateral vertical contracting: Nondiscrimination, exclusivity, and uniformity. The American Economic Review, 210230. [11] Rey, P., & Tirole, J. (2007). A primer on foreclosure. Handbook of industrial organization, 3, 2145-2220. [12] Rey, P., & Vergé, T. (2017). Secret contracting in multilateral relations. Working paper.
15
Journal Pre-proof Highlights
pro of relP
•
urn a
•
Direct connect is a practice according to which airlines provide direct access to their sale systems to travel agencies; Thus, travel agencies can avoid dealing with traditional indirect distribution intermediaries, such as GDS aggregators; Although this practice is often advertised as pro-competitive, we show that it may actually harm consumer in very competitive environments.
Jo
•
PII: DOI: Reference:
S0165-1765(20)30007-0 https://doi.org/10.1016/j.econlet.2020.108952 ECOLET 108952
To appear in:
Economics Letters
Received date : 23 October 2019 Revised date : 4 December 2019 Accepted date : 7 January 2020 Please cite this article as: J. Padilla and S. Piccolo, Does direct connect benefit travelers?. Economics Letters (2020), doi: https://doi.org/10.1016/j.econlet.2020.108952. This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
© 2020 Published by Elsevier B.V.
Journal Pre-proof
Does Direct Connect Bene…t Travelers? Salvatore Piccoloz
pro of
Jorge Padillay
December 4, 2019
Abstract
Direct connect is a business practice that, after a period of irrelevance, has become fashionable again in the travel industry. Airlines provide direct access to their
re-
sale systems to travel retailers, who can thus avoid dealing with traditional indirect distribution intermediaries, such as GDS aggregators. Although this practice is often advertised as pro-competitive, we show that it may actually harm consumers, especially in very competitive environments and when it does not involve all retailers in
lP
the market.
JEL Classification: L42, L50, L81.
Jo
urn a
Keywords: Agency Model, Direct Connect, Distribution Channels, Travel Industry.
For many helpful comments, we would like to thank Joe Harrington (the Editor) and one anonymous referee for useful comments and suggestions. We also thank Michele Bisceglia and Ra¤aele Fiocco, as well as seminar participants in Naples (CSEF), Compass Lexecon London, Catholic University of Milan. We acknowledge …nancial support from Amadeus. The views expressed in this paper are the sole responsibility of the authors and cannot be attributed to Compass Lexecon or its clients. y Compass Lexecon. z University of Bergamo, Compass Lexecon and CSEF.
1
Journal Pre-proof
1
Introduction
Jo
urn a
lP
re-
pro of
Direct connect refers to a business practice through which airlines used to distribute their content to travel agents in the past, prior to the emergence of the GDS1 aggregators such as Amadeus, Sabre, Travelport and Travelsky. After the advent of GDSs it then became largely irrelevant, but it is now growing again, as prominent airlines are seeking for sources of leverage against the GDSs. Notably, airlines are progressively providing direct access to their sale systems, which enables third party travel retailers (e.g., travel management companies) to search for availability, make and manage bookings on a one-to-one basis. This practice is often advertised as a pro-competitive tool (especially by American Airlines and Lufthansa)2 , limiting the market power of traditional distribution channels and promoting consumer welfare. The idea is that, thanks to this technology, passengers and retailers can avoid paying high booking fees and retail prices, often associated with the presence of platforms’(pro…t) margins. Yet, to the best of our knowledge, there is no formal economic model supporting this claim. Is the positive view of direct connect grounded on solid economic principles? If not, why? And, when should it be a concern for competition policy? We consider an agency model (see, e.g., Johnson, 2017) where a supplier (airline) distributes its product (content) both directly through its own distribution channel and indirectly through two retailers (travel agents). Absent direct connect, the intermediaries rely on the IT infrastructure provided by specialized platforms (GDSs) to buy the seller’s content on behalf of …nal consumers. By contrast, with direct connect, the supplier grants one retailer direct access to its content while the other continues to access the market through its exclusive platform (in Section 2.3 we discuss why this ‘partial’form of direct connect is a fairly good re‡ection of reality and what would happen when both retailers operate under direct connect). In both regimes, platforms charge the supplier a commission for each sale they process, and the supplier pro…ts by charging retailers access prices — i.e., the prices (net of commissions, surcharges and discounts) that the supplier requests to authorize a transaction in the indirect channel, whether it is carried out on a platform or through the direct connect technology. We characterize the equilibrium of the game with and without direct connect, and compare consumer surplus across the two regimes. We show that the e¤ect of direct connect on consumer surplus (with only one retailer being served directly by the supplier) is ambiguous and depends on the degree of competition between retailers within the indirect channel and across distribution channels — i.e., between the supplier in the direct channel and the retailers in the indirect channel. Speci…cally, when 1 2
Global Distribution System. See, e.g., https://lufthansa-city-center.com/leisure/en/distribution-platform
2
Journal Pre-proof
2
The model
lP
re-
pro of
the supplier grants a retailer direct access to its sale system, it has an incentive to reduce the demand for the content distributed by the other retailer in order to minimize the commissions paid to the platform dealing with that retailer. The retailer using the GDS platform is thus charged an access price larger than the retailer operating under direct connect. Hence, since access prices are passed on to travellers, the demand for the content distributed directly by the supplier and indirectly by the retailer operating under direct connect increases. This consolidates the supplier’s market power, who has an incentive to increase the price paid by travellers in the direct channel without being afraid of loosing business to the indirect channel (in particular to the retailer using the GDS platform). Yet, when competition across di¤erent distribution channels is strong — i.e., when many passengers regard di¤erent distribution channels as close substitutes — this leads retailers to increase their prices too, whereby harming consumers. By contrast, when competition is su¢ ciently weak, direct connect bene…ts consumers. In this case, the e¤ect described above is negligible because demand functions in the direct and indirect channels are nearly independent. Hence, the main e¤ect of direct connect is that of reducing double marginalization — i.e., the supplier saves the commissions it would have paid to the platform dealing with the retailer operating under direct connect. Obviously, when competition is neither too strong nor too weak, the two e¤ects balance out.
3
Jo
urn a
A monopolistic travel supplier (airline) denoted by S distributes its product (content) both through a direct sale system, and indirectly through an intermediated channel where two platforms (denoted by Pi , with i = A; B) are accessed by exclusive retailers (denoted by Ri , with i = A; B) competing to attract …nal consumers (see, e.g., Bisceglia et al., 2019, Boik and Corts, 2016, Gaudin, 2019, and Johansen and Vergé, 2016, among others).3 In contrast to the existing models, however, consider the possibility that, in addition to the direct distribution channel, the supplier may also grant direct access to its content to one of the retailers operating in the indirect channel: ‘direct connect’. Suppose, without loss of generality, that RB always deals with PB in order to access S’s content, while RA can either access the market by dealing directly with S or, alternatively, operating through its own exclusive platform PA when S gives up direct connect (and allows that platform to market its content). The assumption of exclusivity seems realistic in the travel industry because travel retailers usually access only one platform, and learning by doing economies together with switching costs may actually make competition for retailers harder. In Bisceglia et al. (2019) we discuss how competition between platforms a¤ects the equilibrium of the game.
3
Journal Pre-proof
pro of
Figure 1 below provides a graphical illustration of the di¤erent industry structures we consider.
Figure 1
lP
re-
S sets a price pd on its direct sale system, and contracts with platforms and retailers. Following the literature and business practice, contracts are linear and secret. The industry’s business structure is the agency model proposed by Johnson (2017). Speci…cally, a contract between S and Pi speci…es a commission (fee) fi paid by S to Pi for each unit distributed by Ri . In addition, S sets the access price i that it charges Ri for every unit sold through Pi or via direct connect. The timing of the game is as follows: t = 0 S decides whether RA operates via direct connect or via PA ;
urn a
t = 1 Platforms (only one or both depending on S’s decision in t = 0) o¤er commissions to the supplier; t = 2 The supplier accepts or refuses the o¤ers, and sets access prices; t = 3 The retailers and the supplier simultaneously post …nal prices and demand allocates across the two channels.
Jo
Production costs are linear and marginal costs are normalized to zero without loss of generality. Hence, S’s aggregate pro…t is S
(),(
A
IfA ) qA + (
B
f B ) qB + p d qd ;
with qi and qd denoting the quantities sold through platform i = A; B and the direct distribution system, respectively. The indicator function I takes value 0 if RA deals directly with S and 1 otherwise. Pi ’s pro…t is Pi ( ) , fi qi ; while Ri ’s pro…t is R i ( ) , i qi : 4
Journal Pre-proof
1
b
and qd , Dd (pd ; pA ; pB ) =
(1 + b) pi + b (p (1 b)(1 + 2b)
re-
qi , DA (pi ; p i ; pd ) =
pro of
Since contracts are secret, following Johansen and Vergé (2016) and Rey and Vergé (2017), the solution concept is Contract Equilibrium (Crémer and Riordan, 1987, and Horn and Wolinsky, 1988). This equilibrium concept has some of the features of a perfect Bayesian Nash equilibrium with passive beliefs: even if it receives an out-of-equilibrium contract, each retailer chooses its price assuming that the rival remains under the equilibrium contract. This is in line with the market-by-market bargaining restriction of Hart and Tirole (1990) and with the passive beliefs or pairwise-proofness assumption of McAfee and Schwartz (1994). In line with the evidence (see, e.g., Cazaubiel et al., 2018) we assume that consumers perceive the contents distributed through the direct and the indirect channel as imperfect substitutes.4 Hence, following Johansen and Vergé (2016), we consider the following demand functions
1
b
i
+ pd )
8i = A; B;
(1 + b) pd + b (pA + pB ) ; (1 b)(1 + 2b)
(1)
(2)
4
Jo
urn a
lP
where b re‡ects the degree of substitutability between products within and across distribution channels.5 We assume that b 2 0; b , with b 0:9 to guarantee that second-order conditions hold (see Appendix 2 in Bisceglia et al. 2019). Finally, notice that in the model described above it is assumed that platforms provide no useful role — i.e., they do not bring any e¢ ciency compared to direct connect. This is clearly a simpli…cation made only for the sake of analytical tractability. There are, indeed, many good reasons to believe that (other things being equal) GDSs may increase consumer welfare. For example, it is well known that GDSs have developed (over time) IT technologies that o¤er wider and better services compared to what airlines can do on their own: while Direct Connect is speci…c to a particular airline, and does not automatically aggregate content from multiple airlines, GDSs have over the years developed technologies whose goal is precisely to sort out the best travel plan for every passenger. Hence, in light of this, the negative welfare result on direct connect should be interpreted as a conservative result, that does not take into account the potential additional bene…ts that platforms may bring. Some consumers may in fact prefer to purchase through the indirect channel because intermediaries o¤er (un-modelled) additional services that M is unable or unwilling to supply in the direct channel. 5 To simplify exposition, we assumed that the degree of intra- and inter-channel competition is the same (b). Nevertheless, our results hold qualitatively when consumers perceive the contents distributed by the two retailers as less di¤erentiated than the content distributed directly by the supplier. In this case, it can be shown that direct connect harms consumers as long as inter- and intra-channel competition is su¢ ciently intense. Proofs and numerical simulations are available upon request.
5
Journal Pre-proof
2.1
Equilibrium analysis
pro of
Before solving the game, it is useful to observe that a multiproduct supplier, who fully internalizes the e¤ects of intra- and inter-channel competition, charges the same price for all products pM = 12 and, because demand functions feature preference for variety, it sells the same quantity of each product.6 In what follows, we will …rst develop the analysis for the case of direct connect, then we brie‡y review the logic of the model in the regime without direct connect. In Section 2.2 we then compare the retail prices and consumer surplus across the two regimes.
re-
Equilibrium with direct connect. Suppose that S forecloses PA and deals directly with dc RA . Consider an equilibrium in which: (i) Ri charges pdc i and S charges pd in the indirect dc and direct channel, respectively; (ii) S charges dc i to Ri ; (iii) S is charged fB by PB . We solve the game with a backward induction logic. Ri solves max D pi ; pdci ; pdc i) ; d (pi pi 0
whose …rst-order condition yields i
2
+
i.
max(
B
8i = A; B;
Clearly, pdc i ( i ) is increasing in
urn a
representing Ri ’s best response to prices pdci and pdc d . S solves
pd
b + b pdci + pdc d 2(1 + b)
1
lP
pdc i ( i) ,
dc fB )D(pdc B ; pA ; pd ) +
dc dc A D(pA ; pB ; pd )
(3)
i
and in the rivals’
dc + pd D(pd ; pdc A ; pB );
whose …rst-order condition is 1
6
|
b
dc (1 + b) pd + b pdc A + pB (1 b)(1 + 2b) {z
pd
Monopoly rule
(1
1+b + b)(1 + 2b) }
Jo
Indeed, the above demand system re‡ects the preferences of a representative consumer, whose utility function is X X X 1 X 2 U( ) , qj qj b qj qi pj qj + m; 2 j=A;B;d
j=A;B;d
i;j=A;B;d;;j6=i
j=A;B;d
which is strictly concave and thus exhibits preference for variety. For example, in the airline ticket industry, consumers may want to have the ability to book in di¤erent ways because they may not always be able to reach directly the airline website.
6
Journal Pre-proof
+ (1 |
b ( b)(1 + 2b) {z
fB +
B
Channel externality
A)
(4)
= 0:
}
pdc d (
A;
B ; fB )
,
pro of
Clearly, a higher pd increases S’s pro…t margin but lowers demand in the direct channel; moreover, it also increases S’s pro…t from the indirect channel because (other things being equal) consumers switch from the direct to the indirect channel, so that S collects higher fees from RB . The solution of (4) yields S’s best response to the access prices and the fee charged by PB — i.e., 1 b b b dc + pdc ( A + pB + 2 (1 + b) 2 (1 + b) 2 (1 + b)
A
+
B
fB ) ;
(5)
max ( A
0;
0
B
B
re-
which is increasing in the rivals’retail prices and in the access prices, and decreasing in the fee charged by PB (the higher this fee, the lower the incentive to increase pd since volumes diverted to RB are less pro…table). Moving back to stage 2, S solves dc dc fB )D(pdc B ( ); pA ( ); pd ( ))+
dc A D(pA (
dc dc d dc dc dc ); pdc B ( ); pd ( ))+pd ( )D (pd ( ); pA ( ); pB ( )):
1
lP
Let Ii be an indicator function taking value 1 for i = B and 0 for i = A. The …rst-order condition with respect to i is dc dc (1 + b) pdc i ( ) + b p i ( ) + pd ( ) (1 b)(1 + 2b) {z
b
|
(1
1+b ( b)(1 + 2b)
urn a
Monopoly rule
+ (1 |
b b)(1 + 2b)
i
I if {z
i
+ pdc d ( )
Channel externality
i
Ii fi )
@pdc i ( ) + @ i }
@pdc i ( ) = 0 8i = A; B: @ i }
(6)
Jo
S’s pro…t increases with i because (for a given demand) S collects higher revenues from Ri . However, since contracts are secret, a higher i only increases the retail price charged by Ri . Hence, other things being equal, Ri ’s demand drops, whereas R i ’s demand on the indirect channel and S’s demand on the direct channel increase. Solving (6), we obtain access prices in terms of fB — i.e., dc B
(2 + 3b)2 16 + 48b + 10b2 59b3 33b4 (fB ) = + fB ; 2 (1 + 2b) (4 + 3b) 2 (1 + 2b) (1 b) (4 + 3b) (4 + 5b) | {z } >0
7
(7)
Journal Pre-proof
and dc A
(2 + 3b)2 (fB ) = 2 (1 + 2b) (4 + 3b)
b (1 + b) (4 + 8b 3b2 ) fB : 2 (1 + 2b) (1 b) (4 + 3b) (4 + 5b) | {z }
(8)
>0
pdc d (
dc A (fB );
pro of
While dc B (fB ) is increasing in fB because a higher fee will be passed on by S to RB , dc A (fB ) is decreasing in fB because as RB becomes less e¢ cient due to a higher ‘marginal cost’S has an incentive to divert business towards RA . We can …nally move to stage 1. Let dc B (fB ); fB )
dc dc and pdc R ( R (fB )) , pR (fB ): PB solves
, pdc d (fB );
dc dc max fB D(pdc B (fB ); pA (fB ); pd (fB )); fB 0
|
b
dc dc (1 + b) pdc B ( ) + b pA ( ) + pd ( ) (1 b)(1 + 2b) {z
Monopoly rule
+fB (1 |
lP
1
re-
whose …rst-order condition is
fB
(1
1+b @pdc B (fB ) + b)(1 + 2b) @fB }
@pdc (fB ) b @pdc A (fB ) + d = 0: b)(1 + 2b) @fB @fB {z }
(9)
Strategic e¤ect <0
Jo
urn a
Three e¤ects shape PB ’s optimal fee. First, for a given demand, by increasing the commission charged to S, PB earns a higher pro…t. Second, since S reacts to such a higher fee by increasing the access price charged to RB , the demand on platform PB drops. Third, since fB a¤ects the access prices and S’s pro…t margin on RB ’s sale volume, there is also a strategic e¤ect: pdc A ( ) drops as fB increases because RA ’s access price decreases as explained before; dc pd ( ) also drops because RA charges a lower price as just explained and S’s pro…t margin on RB is lower, which means that S has a weaker incentive to increase the price in the direct
8
Journal Pre-proof
channel in order to pro…t on the indirect one.7 Solving (9) we have fBdc =
(4 + 5b) 4 (1 b)3 (1 + b) (1 + 2b) : 2 3 179b{z 122b4 + 85b5 + 87b6} |32 + 96b + 10b (>0) PB ’s mark up
pro of
dc Substituting fBdc into (7) and (8) we obtain the equilibrium access prices dc A and B , then dc dc using (4) and (5) we obtain the equilibrium retail prices pdc A ; pB and pd (see the Appendix). dc dc dc Proposition 1 With direct connect, access prices are such that dc B > A > 0 and B > fB . dc dc dc Retail prices are such that with pdc and pdc 0:75. B > max pA ; pd A > pd if and only if b The supplier and both retailers sell positive quantities in the direct and indirect channel, respectively — i.e., qAdc > qBdc > 0 and qddc > 0.
lP
re-
In order to make positive pro…ts PB charges a positive commission to S, who will then pass on this commission to RB by increasing A . In turn, RB marks up consumers by setting a price larger than pM . Because contracts are linear and prices are strategic complements, S will also charge a positive access price to RA , which creates a mark-up also for consumers dc buying through direct connect. Obviously, in equilibrium, dc B because S diverts A < demand from RB to RA in order to minimize the commissions paid to PB . Finally, due to strategic complementarity, S will also charge a positive mark-up on the direct channel.8
urn a
Equilibrium without direct connect. The analysis of the regime without direct connect has already been developed in Bisceglia at al. (2019). Hence, for brevity, we will only state the result (details are in the Appendix). Proposition 2 (Bisceglia et al. 2019) Without direct connect there exists a unique symmetric equilibrium such that platforms charge a fee f n > 0 to S, who then sets n > f n to each retailer and charges pnd in the direct channel, retailers set the same price pn > n in the indirect channel.
7
Jo
The equilibrium features multiple marginalization: platforms charge S positive commissions, which will be passed on to the retailers via higher access prices, which in turn induce Formally, by (5) we have
@pdc d (fB ) @fB
= =
8
b 2 (1 + b)
@
dc A
(fB ) @ dc (fB ) + B @fB @fB
b 2 (1 + b) 3b2 < 0: 2 (1 b) (1 + 2b) (4 + 3b)
Because consumers feature taste for variety, S never shuts down PB .
9
1
Journal Pre-proof
a …nal price higher than pM . Due to strategic complementarity, S charges a very high price in the direct channel too.
2.2
Consumer welfare analysis
pro of
We now examine the e¤ects of direct connect on consumers. We start by comparing retail prices. Proposition 3 Direct connect always increases the price charged by S in the direct channel — i.e., pdc pnd with equality only at b = 0. As for the indirect channel, there exist two d thresholds b0 and b1 , with 0 < b0 < b1 < b, such that: pdc A
pn if and only if b 2 [0; b1 ], with equality at b = b1 ;
pdc B
pn if and only if b 2 [0; b0 ], with equality at b = b0 and b = 0.
urn a
lP
re-
dc With direct connect, S charges dc B > A in order to divert business from RB towards RA . This induces RB to pass on the higher access price to consumers. Hence, demand in the direct channel increases and, in turn, S charges a higher price in that market segment. However, a higher price in the direct channel also induces retailers to charge higher prices in the indirect channel when b is su¢ ciently large because of a relatively stronger e¤ect of pd on RA and RB ’s demand functions. Figure 2 shows the equilibrium prices as a function of b.
Jo
Panel a: direct channel
Panel b: indirect channel
Figure 2: Equilibrium prices
In panel a, the solid curve is pnd while the dashed one is pdc d ; in panel b, the black curve n cd cd is p , the red pA and the green pB . In line with the intuition provided above, the region of parameters in which direct connect increase prices in the indirect channel corresponds to the region of parameters in which the di¤erence pdc pnd is larger. d We can now study consumer surplus. 10
Journal Pre-proof
Proposition 4 There exits a threshold b 2 (0; b1 ) such that direct connect harms consumer surplus if and only if b b :
lP
re-
pro of
dc The intuition is straightforward. Because pdc d > pd , direct connect unambiguously harms consumers who buy directly from the supplier. However, in the indirect channel the e¤ect is ambiguous. Clearly, for b su¢ ciently large (e.g., b > b1 ) direct connect increases prices in the intermediated channel too, and hence it harms consumers in both channels. On the contrary, n dc n dc n for b ! 0 it can be seen that pdc A < p , pd ! pd and pB ! p . Hence, for b su¢ ciently small direct connect overall bene…ts consumers: it has a …rst-order bene…cial e¤ect in the indirect channel and a second-order negative e¤ect in the direct channel. When b takes intermediate values the two e¤ects balance out. Figure 3 provides a graphical illustration of the result
Figure 3: Consumer surplus
The solid (dashed) curve represents consumer surplus with (without) direct connect.
Discussion
Jo
2.3
urn a
Finally, it should be noticed that while direct connect harms consumers for a large enough b, it unambiguously bene…ts the supplier who (absent e¢ ciencies on the platforms’ side) always has an incentive to grant direct access to its sale system to at least to one retailer. By revealed preferences, in fact, the supplier saves on the commission it would have paid to the platform dealing with RA , while being able to control the retail prices through an appropriate choice of A and B .
Throughout the analysis we have assumed that S grants direct access to its distribution system to one retailer only. Clearly, if it could, S would like to foreclose both platforms: by doing so it would (other things being equal) save the commissions paid to the platforms, whereby increasing pro…ts. Notably, this would also wipe out the e¤ect of commissions on multiple marginalization, whereby leading to lower prices and higher consumer surplus. Hence, our model does predict that when direct connect occurs on a wide-market scale, it 11
Journal Pre-proof
9
urn a
lP
re-
pro of
unambiguously bene…ts consumers. However, at least in the short-medium run, it seems hard to imagine that airlines will be able to capture all retailers. As a matter of fact, not all travel agencies operate via direct connect (according to the CMA, direct connect at the moment only covers 2% of the market, although it is expected to grow). In fact, foreclosing all (or a large part of) the GDSs at once might be too costly for airlines (at least in the short-medium run). First, GDSs have developed (over time) IT technologies that o¤er wider and better services compared to what airlines can do on their own. As the CMA itself recognizes the supply of Direct Connects, and therefore the speci…c services these entail in terms of retailing, distribution (including aggregation) and ful…lment, is less standardized than the supply of GDS services and can vary signi…cantly between suppliers. The CMA also notices that in terms of aggregation, as a Direct Connect is speci…c to a particular airline, it does not automatically aggregate content from multiple airlines. Finally, in general, Direct Connects are not able to fully facilitate interlining and corporate reporting/compliance, as it is possible via a GDS.9 Second, by migrating to direct connect travel agents may incur high switching costs that must be compensated by airlines. GDS systems require training and travel agents must learn the ‘language’of the system in order to search for ‡ights and understand the results. Di¤erent systems may use di¤erent ‘languages’. Hence, switching to direct connect may not be costless. In addition, some GDSs also typically o¤er Technology Funding to travel agents in a form of cash or reimbursement. This funding is important for the travel agent to buy PCs, laptops, servers or any hardware essential to the business. Migrating to direct connect may require the provision of new hardware and software in addition to duplicating training and adoption costs. Third, as a matter of fact, when dealing with platforms, travel agents have direct access to the revenues generated by a range of ancillary services and products, to which they might not necessarily have access when dealing directly with airlines.10 When booking at travel agencies, travellers often also book some of these services in advance and travel agents typically charge them commissions for these services. Hence, giving up the revenues coming from these services might be costly for travel agents. In particular, in order to provide
Jo
An interesting quote from the from the very recent case US vs Sabre-Farelogix (2019) states the following: “Some traditional travel agencies also rely on other lines of their GDS’s business for IT products. For example, many travel management companies in Sabre’s network use mid- or back-o¢ ce software supplied by Sabre to perform monitoring or reporting for their corporate clients. To ensure consistent support and reporting for their travellers, corporations typically rely on only one travel management company. . . . ”(page 8). 10 For example, in addition to sharing revenues from traditional services, like hotel accommodations, car rentals and travel insurance, Amadeus typically shares with its clients revenues from other ancillary services like on-board sales of food and beverages; checking of baggage and excess baggage; assigned seats or better seats such as exit rows; priority check-in and screening; early boarding bene…ts; on-board entertainment systems; wireless internet access, etc.
12
Journal Pre-proof
3
pro of
the right incentives to switch to direct connect airlines must compensate travel agents for these missed pro…t opportunities. This creates a cost for the airlines associated with direct connect. And, when this cost is su¢ ciently high, airlines might then endogenously decide to foreclose only some of the platforms. Hence, although in our context it would be socially desirable to have a market that fully operates under direct connect, in the short-medium run this seems rather unfeasible and in this case having only a partial form of direct connect does not necessarily bene…t consumers. This negative result would clearly be reinforced when taking into account the e¢ ciencies discussed before that GDSs bring in terms of product quality compared to direct connect.
Conclusion
Jo
urn a
lP
re-
The idea that direct connect unambiguously bene…ts consumers in the travel industry needs to be quali…ed. By providing some (but not all) retailers direct access to their sale systems, travel suppliers can actually consolidate their market power vis-à-vis traditional distribution channels, especially in competitive environments. Last, but not least, our model has important implications on the current debate on platform parity provisions — i.e., contractual agreements according to which airlines commit not to charge di¤erent prices for the same product distributed through di¤erent platforms. In Bisceglia et al. (2019) we have shown that, in a model without direct connect, platform parity bene…ts consumers as long as competition between and across distribution channels is su¢ ciently intense. Hence, the likelihood that direct connect harms consumers when the fall back option is an indirect channel with two platforms and price parity is very high. Therefore, in this case (under the assumption that regulations cannot impose airlines to drop direct connect) imposing a price parity agreement according to which airlines cannot charge di¤erent (access) prices to the retailer operating under direct connect and those operating through platforms certainly increases consumer welfare. The reason is that direct connect would still allow airlines to save on the commissions paid to the active platforms (whereby reducing multiple marginalization), but it would also prevent them from consolidating their market power by increasing the access price charged to the retailers accessing the market through platforms.
13
Journal Pre-proof
Appendix Proof of Proposition 1. Substituting for (3), (5), (8) and (7) into (9) and solving for fB dc dc dc yields fBdc . Substituting fBdc into (8) and (7), yields dc A and B . Showing that A < B and dc dc dc dc dc B > fB is immediate. Substituting fB , A and B into the system (3)-(5), solving for pA , pB and pd
pdc A
=p
M
b (1 + b) (24 + 78b + 24b2 141b3 128b4 + 65b5 + 87b6 ) ; (3b + 4) (1 + 2b) (32 + 96b + 10b2 179b3 122b4 + 85b5 + 87b6 )
pro of
M pdc + d = p
64 + 296b + 284b2 492b3 879b4 + 95b5 + 703b6 + 124b7 159b8 ; + 3 (4 + 3b) (1 + 2b) (32 + 96b + 10b2 179b3 122b4 + 85b5 + 87b6 )
M pdc + B = p
96 + 416b + 292b2 860b3 1087b4 + 471b5 + 967b6 4b7 255b8 : 2 (4 + 3b) (1 + 2b) (32 + 96b + 10b2 179b3 122b4 + 85b5 + 87b6 ) 0:75. Checking that qAdc > qBdc > 0
re-
dc dc dc and pdc with pdc A > pd if and only if b B > max pA ; pd and qddc > 0 is immediate.
Proof of Proposition 2. From Bisceglia et al. (2019) it follows that S sets pnd = pM +
b (1 + b) (8 9b2 ) 64 + 152b 28b2 171b3
pn = pM +
lP
while retailers charge the same price
48 + 56b 130b2 2(64 + 152b 28b2
54b4
;
59b3 + 81b4 : 171b3 54b4 )
urn a
n dc dc Proof of Proposition 3. The proof follows by comparing pdc A and pB with p , and pd with pnd (Figure 2).
Proof of Proposition 4. Consumer surplus under direct connect obtains by substituting the equilibrium prices into the utility function that generates the demand system (1)-(2) — i.e., CS dc =
X
qjdc
where
b
qddc =
1
b
X
qjdc qidc
i;j=A;B;d;;j6=i
Jo
j=A;B;d
1 X dc 2 (q ) 2 j=A;B;d j
dc dc (1 + b) pdc d + b p A + pB ; (1 b)(1 + 2b)
qidc =
1
b
X
dc pdc j qj + m;
j=A;B;d
dc dc (1 + b) pdc i + b p i + pd ; (1 b)(1 + 2b)
for every i = A; B: Similarly, CS n = 2q n + qdn
(1 + b) (q n )2
1 n 2 (q ) 2 d 14
2bq n qdn
2pn q n
pnd qbn + n;
Journal Pre-proof
where, qn =
1 b pn + bpnd ; (1 b)(1 + 2b)
qdn =
Comparing CS dc and CS n it follows that b
1
b (1 + b) pnd + 2bpn : (1 b)(1 + 2b)
0:5 (Figure 3).
pro of
References
[1] Bisceglia, M., Padilla, J., & Piccolo, S. (2019). When Prohibiting Platform Parity Agreements Harms Consumers. Available at SSRN 3454527. [2] Boik, A., & Corts, K. S. (2016). The e¤ects of platform most-favored-nation clauses on competition and entry. The Journal of Law and Economics, 59(1), 105-134.
re-
[3] Cazaubiel, A., Cure, M., Johansen, B. O., & Vergé, T. (2018). Substitution Between Online Distribution Channels: Evidence from the Oslo Hotel Market. University of Bergen, Department of Economics (No. 8/18). [4] Cremer, J., & Riordan, M. H. (1987). On governing multilateral transactions with bilateral contracts. The RAND Journal of Economics, 436-451.
lP
[5] Gaudin, G. (2019). Vertical relations, opportunism, and welfare. The RAND Journal of Economics, 50(2), 342-358. [6] Hart, O., & Tirole, J. (1990). Vertical integration and market foreclosure. Brookings papers on economic activity. Microeconomics, 205-286.
urn a
[7] Horn, H., & Wolinsky, A. (1988). Bilateral monopolies and incentives for merger. The RAND Journal of Economics, 408-419. [8] Johansen, B. O., & Vergé, T. (2016). Platform price parity clauses with direct sales. IDEI working. [9] Johnson, J. P. (2017). The agency model and MFN clauses. The Review of Economic Studies, 84(3), 1151-1185.
Jo
[10] McAfee, R. P., & Schwartz, M. (1994). Opportunism in multilateral vertical contracting: Nondiscrimination, exclusivity, and uniformity. The American Economic Review, 210230. [11] Rey, P., & Tirole, J. (2007). A primer on foreclosure. Handbook of industrial organization, 3, 2145-2220. [12] Rey, P., & Vergé, T. (2017). Secret contracting in multilateral relations. Working paper.
15
Journal Pre-proof Highlights
pro of relP
•
urn a
•
Direct connect is a practice according to which airlines provide direct access to their sale systems to travel agencies; Thus, travel agencies can avoid dealing with traditional indirect distribution intermediaries, such as GDS aggregators; Although this practice is often advertised as pro-competitive, we show that it may actually harm consumer in very competitive environments.
Jo
•