Two-sided Markets. Özlem Bedre-Defolie* July 12, 2018 ESMT. Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

Transcription

1 Two-sided Markets Özlem Bedre-Defolie* ESMT July 12, 2018 Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

2 Introduction Definition:Two (multi)-sided markets are markets in which a platform coordinates and connects two or more distinct groups of users. Examples: Group1 Platform Group2 game players video game platforms game developers users operating systems app. developers eyeballs portals/newspapers/tv advertisers cardholders debit/credit card networks merchants buyers online market places sellers searchers online search engines advertisers shoppers shopping malls shops women heterosexual clubs men Chicken & egg problem: must get both sides on board while making money overall. Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

3 Externalities between the two sides Network (membership) externalities: One side benefits/loses from participation on the other side of the market. More cardholders hold cards of a network, say Visa, more merchants are willing to accept Visa cards. More merchants accept Visa cards, more consumers would like to hold Visa cards. More searchers/eye-balls a platform has, more advertisers would like to be present on that platform. More gamers/users a platform has, more game/software developers would like to join the platform. More advertising in a magazine/journal might make it less (or more) attractive for readers (open question). Usage externalities: One side benefits/loses from usage decision made by the other side of the market. When cardholders pay by card, merchants have to pay a commission to their bank. When searchers click on an advertised link, the advertiser of that link has to pay a fee to the search platform. Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

4 Externalities between the two sides- Ctd The two sides might not perfectly internalize those externalities. This might be due to price restrictions imposed by platforms, e.g., No-surcharge-rule prohibits merchants from surcharging expensive cards. Even when they are allowed, not all merchants surcharge, probably due to transaction costs of surcharging. Advertisers can pass the costs of advertising on the prices of the goods they sell, but not all viewers/searchers purchase a good after viewing/clicking on its advertisement. As a result, the volume of transactions between the user groups depend on the allocation of total transaction price between the two sides. Different from the standard theory of taxation where it does not matter whether the tax is on sellers or on buyers. A platform can use prices to balance the demand between the two sides. Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

5 Two-sided platforms vs vertical chains In general platforms are sellers of complementary goods/services which are consumed by different groups of users. Some platforms are suppliers of an input for sellers Key difference of platforms from standard vertical chains is that the platform (supplier) also interacts directly with final consumers (buyers). For instance, online market places provide infrastructure and search mechanism which allocates buyers search to sellers products Payment card platforms provide infrastructure which enables merchants to be paid by payment cards Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

6 Review The optimal pricing involves pricing below cost (subsidizing) the side which is more valuable to the other side and earning revenues from the side which generates less value to the other side. Examples: Cardholders get rewards, e.g., cash backs, miles, over their card usage, whereas merchants pay (high) commission fees for card transactions. Searchers enjoy free access to a wide source of information online, whereas search platforms make money from advertisers. Yellow page directories/some newspapers are offered for free to consumers and earn revenues from advertisers. Software platforms make their money mostly from users and software developers get access to platform services free. Video game platforms get most of their gross margin from licensing access to the software and hardware platforms to game developers; they sell the videogame console to consumers at close to or below manufacturing cost. Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

7 Model of a monopoly platform Two-sides of the market: i B, S N B, N S potential number of participants. Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

8 Setup: 1. No payment between end users U i = (b i a i ).N J + B i A i b i : per transaction benefit b i [b i, b i ] B i [B i, B i ] B i : fixed benefit N i = Pr(U i 0) c: platform s per transaction cost = 0(wlog). per transaction price Hence p i a i + Ai N J N i = Pr(b i + Bi N J pi ) D i (p i, N J ), i {B, S} Solving: { N B = n B (p B, p S ) N S = n S (p B, p S ) Define Volume (V ) NB.N S Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

9 Some redundancy in the pricing policy This modeling assumes (implicitly) that all potential users interact; N i = extensive margin = intensive margin on side i. If the platform changes a i while keeping p i constant that is changing A i accordingly to keep p i unchanged, the number of users would not change. On each side, the platform has two tools (a i, A i ) to achieve one objective N i. The platform is indifferent between any levels of (a i, A i ) as long as the average price is the same. When there is platform competition, this redundancy leads to multiplicity of equilibria with two-part tariffs. (Armstrong, 2006) Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

10 The Platform s Strategy π = A B.N B + A S.N S + (a B + a S c)n B.N S which can be transformed into π = (p B + p S c)n B (p B, p S )n S (p B, p S ) Price structure choice: V (p) = max (p B,p S ){n B (p B, p S ).n S (p B, p S )} s.t.p B + p S = p Price level choice max{(p c).v (p)} p Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

11 Equilibrium Price Level p c p = 1 η where η = dv dp /V p (elasticity) Intuition: Standard monopoly markup on the total price level. Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

12 Price Structure Solution The total differentiation of n i = D i (p i, n J ) gives dn i Di p i.dpi Di n J.dnJ = 0 dni dp i = Di p i Symmetrically, dn J DJ p J.dpJ DJ n i.dn i = 0 dnj dp i + Di n J.dnJ dp i = DJ n i. dni dp i where the last equality is obtained by setting dp j = 0 and dividing both sides by dp i. Solving the two equations together gives dn i dp i = Di / p i 1 Di. DJ n J n i Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

13 Price Structure Solution- Ctd By max (p B,p S ){n B (p B, p S ).n S (p B, p S )} st. p B + p S = p we get n B p B.nS + n B. ns p B ( nb p S.nS + n B. ns p S ) = 0 Plugging this into the equilibrium condition for price level gives 1 p c = n B p B n B + n S p B n S = n B p S n B + n S p S n S Replacing formulas nb, ns, ns, nb, and multiplying by p B p S p B p S 1 DB. DS we obtain: N S N B 1 DB N S. DS N B p c = D B p B D B + D B p B. DS N B D S = D S p S D S + D S. DB p S N S D B Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

14 (a) Heterogeneity in per-transaction benefit b i only (Assume no fixed benefits, no fixed costs: B i = 0)(Rochet & Tirole, 2003): 1 p c = D B p B D B = D S p S D S p B p c = DB p B. pb D B & ps p c = DS p S. ps D S So for i, j = A, B and i j we have p i (c p J ) p i = 1 η i where η i = Di p i. pi D i The opportunity cost of user i = c p J Intuition: Inverse elasticity rule with an adjusted marginal cost to account for cross-group externalities. One more user on side i costs c to the platform, but enables the platform earn p J from side j. This, therefore, reduces the effective cost of attracting one more user on side i. Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

15 Ramsey Prices The average user surplus of side i is + v i (p i p D i (t)dt ) = i D i (p i ) The Ramsey planner sets prices maximising the total user surplus subject to the platform s participation constraint: max pb,p S [v S (p S ) + v B (p B )]D B (p B )D S (p S ) s.t.p B + p S = c The Ramsey prices are therefore characterised by two equations: where η i = Di p i. pi D i. p B + p S = c p B η B v B = ps η S v S Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

16 Ramsey prices vs Privately optimal prices There are two types of distortions: Standard monopoly markup on the total price, p B + p S = p > c such that p c p = 1 η where η = ηb + η S. While allocating the total price between the two sides, the platform cares about the marginal user whereas the planner cares about the average user (Spence distortion, as named by Weyl, 2010): Platform s optimal: pb η B = ps η S Planner s optimal: pb η B v B = ps η S v S Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

17 The platform charges only per-transaction prices If it can meter transactions AND If the network is mature, that is, where the number of participants on both sides is nearly stable, and so network externalities are not relevant AND When there are no substantial fixed costs per participant. Examples: Mature payment card networks, search engines. Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

18 (b) Heterogeneity in membership benefit B i only (Assume:a S = a B = c = 0) (Armstrong, 2006) N i = D i (A i, N j ) = Pr(B i A i + b i N J 0) Opportunity cost of user i = b j.n j A i [ b J.N J ] A i = 1ˆη i ˆη i = D i A i. Ai N i Intuition: When the platform attracts one more user on side i, the platform can increase its fee to side J by b j, so the platform could earn b j.n J on side j by attracting one more user on side i. Note: The result is analog to the one in Rochet and Tirole (2003) for c = 0. p i ( b j ) = 1 η i = Di. pi p i η i p i N i Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

19 Suppose now that the platform can observe transactions, usage benefits are known and c > 0. There will be no change in profit. Platform can eliminate cross-group externalities by setting a i = b i. Then optimal membership fees will be: A i [ (b B + b S c)n J ] A i = 1 [( dni da i )/ Ni A i ] = 1ˆη i Remarks: (1) No cross-group externality, yet it is still a two-sided market. (2) Externalities depend on whether platform absorbs them. Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

20 The platform could charge only membership fees If it cannot meter the transactions. Examples: Traditional media platforms charge consumers a fee per access to the content and advertisers a fee per ad, heterosexual clubs charge men and women for entry to the club. Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

21 2. Payments between end-users Examples: Payment card networks, video game platforms, online market places Seller and buyer can bargain over the price. Suppose b i is drawn from F i (b i ) after the end-user becomes a member Assume monotone hazard rate, i.e., quasi-concavity of demands) f i 1 F i is increasing (for the Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

22 Coasian bargaining: trade iff b B + b S a B + a S = a OPTIMUM: a = c Intuition:There is no private information about marginal willingness to consume. So you do not want to distort transactions to capture some rent. (1) create efficient trade (2) then back to pure membership model, choose price structure (p B, p S ) Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

23 If there is price setting by seller or asymmetric information efficient bargaining (Myerson-Satterthwaite, 1983). OPTIMUM: a < c. Intuition: You subsidize the monopoly to reduce its price below the monopoly price (i.e., to compensate for the monopoly mark-up, you reduce price to the seller) since lowering a increases trade. (you do not care about extracting). Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

24 Neutrality of price structure Remark: Regardless of the fact that bargaining is efficient or inefficient, the price structure is neutral. To see this let a = a B + a S, b (b B, b S ), x(b, a): probability of trade, t i (b, a B, a S ): transfer between B and S that neutralizes the allocation of a between B and S. The expected utility of side i is The platform s profit is where X = E[x(b, a)]. β i (a) E[(b i a i )x(b, a) + t i (b, a B, a S )] π = i=b,s Ai N i + (a c)xn B N S Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

25 Neutrality of price structure- Ctd The platform s profit can be rewritten as where π = [p B + p S + v(a)]n B n S p i = Ai N J + a i.x E[b i x(b, a) + t i (b, a B, a S )] v(a) E[(b B + b S c)x(b, a)] Therefore, the platform s optimization problem decomposes into (i) the choice of prices (p B, p S ) (as above). (ii) an ancillary problem of finding the per-transaction total access charge a that maximizes the average social surplus from potential transactions. Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

26 3. What is a two-sided market? (a) Usage prices (a B, a S ): Definition: Market is one-sided if volume V depends only on level a = a B + a S, not on its structure. Otherwise the market is 2-sided. If market is one-sided, business and public policy attention to price structure is misguided. Examples of charges in one-sided markets VAT Telecom charges when caller and receiver side contract Buyer/seller bargaining, even under asymmetric information Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

27 (b) More generally, apply definition to per transaction prices (p B, p S ) For a market to be two-sided, the Coase theorem must not apply (necessary condition). But failure of Coase theorem does not suffice: allocation of a is irrelevant in bargaining games of asymmetric information. Factors conducive to two-sidedness: transaction costs (telecom, websites, card/cash payments). platform-imposed constraints on end-user bargaining (NSR, itune s $0.99 per song). transaction-insensitive end user costs (fixed membership fee/cost): If we allow also fixed access charges, distribution of total fixed price between B and S matters unless B and S do bargain ex-ante (ex-post bargaining does not take into account ex-ante cost). Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

28 Private vs Social Incentives Do profit maximizing platforms want to induce the efficient volume of trade? A monopoly platform upwardly distorts the total level of the per-transaction prices due to its market power and, Most notably, distorts the structure (allocation) of the total price between the two sides due to the platform s inability to price discriminate across heterogeneous users on each side (Rochet and Tirole, 2003, 2006). Platform competition would correct the market power distortion on the total price level. Identifying the direction of the price structure distortion is not straightforward, and so it is unclear whether platform competition would reduce it (Rochet and Tirole, 2006; Weyl, 2010). Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

29 3-party The card payment network card industry 4-party card network Payment Network Payment Network Issuer p- a Acquirer p+ f F p - m p+ f F M p - m Consumer good Merchant Consumer Merchant (a) 3-party (closed) card network (b) 4-party (open) card network p: price of the good a: interchange p: price fee of (IF) the good a: interchange fee (IF) (E.g., AMEX) (E.g., Visa) f: usage fee (per-transaction) m: merchant f: usage fee (per-transaction) fee (per-transaction) m: merchant fee (per-tran F: membership fee (fixed) F: membership fee (fixed) p : price of good, F : Card membership fee, f : card transaction fee, M : merchant membership fee, m : merchant fee per transaction, a : interchange fee (IF). IF determines the price structure; how the total transaction price is allocated between the two sides. good 04/11/14 Pricing Payment Cards, 04/11/14 TSE Pricing 2 Payment Cards, TSE Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

30 Policy makers concerns and interventions High IFs (so high merchant fees) inflate the cost of card acceptance by merchants without improving efficiency. Cap regulations on IFs in Australia, Canada, Chile, Denmark, Mexico, Singapore, Switzerland, and the US (2011). Mostly based on issuers costs. MIFs harm competition between acquiring banks, inflate merchant fees and so final consumer prices. (The UK OFT s MasterCard case, the EC MasterCard (2007) and Visa (2002, 2010) cases). No-surcharge-rules: Payment networks prohibit merchants from surcharging their payment cards in favour of other networks cards possibly distorting competition. In 2010 Visa and MasterCard reached a settlement with the US DOJ to stop using NSRs. AMEX refused the DOJ s rule and lost the US law suit. In Australia, Czech Republic, Denmark, Ireland, Netherlands, New Zealand, UK NSR is banned, but in Australia and UK merchants surcharges are subject to cap regulation based on merchants costs of card acceptance. Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

31 In the payment card industry Private vs Social Incentives The comparison depends on quantitative considerations, for example, cost and preference attributes (Wright, 2004; Schmalensee, 2002). The total of end user fees is too high if the platform has market power (Rochet and Tirole, 2003, 2006; Bedre-Defolie, 2013). The platform s pricing might distort the price structure by inducing too high IFs and merchant fees due to 1. Merchant internalisation (ex-ante: Rochet and Tirole, 2002, 2011; Wright, 2013, ex-post: Bourguignon, Gomes, and Tirole, 2014.) 2. Asymmetric choice between consumers and merchants (Bedre-Defolie and Calvano, 2013) 3. Network competition when consumers single-home (adopt one type of card) more than merchants (Rochet and Tirole, 2003; Guthrie and Wright, 2007) Platform competition will correct the first distortion, but it is unclear how it affects the price structure distortion. Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

32 Merchant internalisation Ex-ante: Card acceptance is a way to increase quality of the merchant services, so increase store demand and/or steal business from rivals and so to internalise (at least partially) consumer surplus from card transactions, v B = E[b B f b B f ]. Ex-post: Once consumers are at the shop, merchants do not want to miss sales at a point-of-sale by declining cards (Bourguignon et al. 2014). Must-take cards: When merchant internalisation holds, merchants accept cards even if the merchant fee is above their transaction benefit: m > b S (Rochet and Tirole, 2002, 2011; Bourguignon et al. 2014). When merchants are heterogenous, merchant (ex-ante) internalisation makes the merchant demand for card acceptance less elastic to merchant fee, and so raises the network s optimal IF (Wright, 2013) The social planner sets a lower IF than the network since it counts consumers card usage surplus, v B, only once. Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

33 Asymmetric choice between consumers and merchants Bedre-Defolie and Calvano, 2013: Idea: When one side decides both on membership and usage, there are two distinct margins, extensive and intensive margins, and so two-part tariffs become meaningful tools for a platform. Technical twist:to capture the fact that only some proportion (not all) of members transact introduce uncertainty on usage benefit from card transactions: buyers learn their transaction benefit only after membership decisions are made. Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

34 Bedre-Defolie and Calvano (2013)- Ctd Member buyers transact if and only if b B a B. Let now D B denote usage demand of a member: D B = Pr(b B a B ) As sellers decide only on membership there is one margin (extensive margin), so wlog assume A S = B S = 0: Merchant demand is D S = Pr(b S a S ). At the membership stage, buyers average usage surplus from transactions is v B = E[b B a B b B a B ] A buyer becomes a member if and only if B B A B v B D B D S. The extensive margin on the buyer side is N B = Pr(B B A B v B D B D S ) The intensive margin on the buyer side is N B D B Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

35 Bedre-Defolie and Calvano (2013): Private vs Social Incentives Using two-part tariffs on the buyer side, the platform can capture the average buyer surplus via fixed fee, so sets marginal price at its marginal cost: a B = c B and puts a monopoly markup on its fixed fee. On the seller side the platform cares only about the marginal seller. A Ramsey planner would care about the average seller and the average buyer, so the privately optimal transaction fees are biased against sellers. In a payment card network, this implies that the privately set interchange fee is higher than the socially optimal one. Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

36 Platform Competition The optimal pricing decisions depend whether end users join one platform (single-homing) or both platforms (multi-homing). The competitive price on one market depends on the extent of multi-homing on the other market. Example: If Visa reduces the fee paid by merchants, merchants become more tempted to refuse more costly AMEX cards as long as a large fraction of AMEX customers also owns a Visa card. Multi-homing on one-side intensifies price competition on the other side since platforms use low prices in an attempt to steer end users on the latter side towards an exclusive relationship (Rochet and Tirole, 2003; Guthrie and Wright, 2007; and Armstrong, 2006). Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

37 Platform Competition- Ctd In general, when adoption costs are high, end users prefer to single-home. End users prefer to multi-home when they expect that other side single-homes and network effects are strong. Examples: Merchants mostly accept cards of more than one payment platform, e.g., Visa, MasterCard, AMEX. In Europe consumers mostly hold one type of card brand, in the US consumers mostly hold more than one card brand. Advertisers want to be present on several media platforms. Gamers generally buy one game console. Video game developers may port their games to several game platforms. Consumers mostly rely on one operating system. Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

38 Platform Competition- Ctd When both sides single-home, platforms should put more emphasis on network effects when they are competing: With competition, losing one user is more costly since a new user on rivals platform would attract more users on the other side (due to network effects) (Armstrong, 2006). When one side single-homes and the other side multi-homes, platforms have monopoly power over providing access to their single-homing customers for the multi-homing side (competitive bottlenecks). Networks compete for the single-homing side intensively and set monopoly prices to the multi-homing side. Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

39 Non-linear pricing by platforms Considering two-part tariffs, Rochet and Tirole (2006) note that there is redundancy in the pricing tools of the monopoly platform. insulating tariffs, (Weyl, 2010): The monopoly platform can implement the desired participation levels on both sides by appropriately choosing its tariffs, and so is indifferent between these multiple ways of implementing the same outcome. This indifference leads to multiplicity of equilibria when platforms are competing (Armstrong, 2006). If only an exogenous fraction of participants transact with each other, there is unique equilibrium with two-part tariffs and competing platforms (Reisinger, 2011) Indeed, whenever one side makes two distinct decisions (membership and usage), two-part tariffs to that side become meaningful since then the average tariff affects the extensive margin and transaction fee affects the intensive margin (Bedre-Defolie and Calvano, 2013). Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40

40 Open questions When do platforms prefer to restrict the trade between the two sides and make the market one-sided? Is a vertical restraint on one side of the market is good or bad for the total welfare? What would be the impact of a regulatory intervention, e.g., banning particular price or non-price restraint imposed by the platform, capping end-user fees, like interchange fee regulation on payment card networks? Estimating end user demands on both sides and analyzing the real impact of a price regulation on end user fees, by using structural estimation methodologies. Özlem Bedre-Defolie* (ESMT) Two-sided Markets July 12, / 40