Industrial Economics questions- Everything

Volvo- Scania case: case setup

"Quantifying the effects from horizontal mergers in European competition policy" by Ivaldi & Verboven, International Journal of Industrial Organization, 2005 Look at the Volvo-Scania Merger in three different ways: 1. Hypothetical market power test 2. Actual market power test 3. Comparative market power test • The proposed merger between Volvo and Scania was notified to the European Commission on 22 September 1999 • A Commission Decision on 15 March 2000 declared the merger incompatible with the Common Market and the functioning of the EEA Agreement

Analyse a homogeneous good Bertrand duopoly where marginal cost is constant c. Suppose h= probability that a third firm discovers superior product that makes the duopolists' product obsolete. What is the critical threshold for discount factor in this case?

(1 − h) = probability that duopolists still have a market in the next period. (p= pi= monopoly profit) 1/2p [1+ d(1-h) +d2(1-h)2 + ...] 1/2pm x 1 / [( 1-d(1-h)] >= p d.= 1/ 2(1-h) Note further that h < 1/2 is a necessary condition for collusion. d* > 1 if h > 1/2. in a homogenous good bertrand duopoly there is no effect on collusion; removes the slack from one market to another

How does this incentive constraint change if there is Cournot competition (firms choose quantities)?

(In the lecture we have analysed collusion with Bertrand competition) With Cournot competition there are two main differences. 1. the profits in the punishment path are not zero but equal to static Cournot profits. 2. we need to work out the optimal way to deviate from collusion. That is, how much to produce

leniency programmes

(form of competition policy) Leniency policy introduced in Europe in 2002 Full immunity from fines if 1. first cartel member to provide sufficient information to allow launch of investigation or 2. cartel member provides evidence that proves the cartel infringement (if firm did not coerce others to participate in cartel) Reduction of fines if cartel member provides evidence that has significant added value Heavy fines for other firms Unprecedented number of cartels has been detected and prosecuted due to leniency policy

Rochet-Tirole (2003) Monopoly Platform Benchmark-Private Monopoly Proposition 1

(i) A monopoly platform's total price, p = pB + pS, is given by the standard Lerner formula for elasticity equal to the sum of the two elasticities η = ηB + ηS: (ii) The price structure is given by the ratio of elasticities (and not inverse elasticities): • The price they are charging the buyers is not just related to the elasticity on the buyers' side but also related to the elasticity on the sellers' side

explicit v tacit collusion: experimental evidence

1. Communication leads to higher prices 2. Without communication only duopolies can maintain collusive prices 3. With communication collusive price is decreasing in the number of firms 4. Gains from communication U-shaped 5. Industries continue to collude successfully after communication is disabled (hysteresis effect) Fonseca and Normann also testing where allowed them to communicate a few periods and then disabled the communication prices.... Result able to sustain cartel pricing even when communication disables. Students playing games but encouraging that a lot matches empirical evidence. Also human behaviour more complex than standard models.

what are the 3 BASIC PRINCIPLES to define markets

1. Competitive Constraints • Demand-side substitution • Which products are seen as substitutes by the consumers? o E.g.: soft drink industry o Is Coke a substitute for Pepsi? Is Sprite a substitute for Pepsi? o If Coke and Pepsi are substitutes, the cross-price elasticity is positive ♣ Increase price for Coke, demand for Pepsi will increase o If they are compliments (e.g. tea and milk) the sign is negative 2. Supply-side substitution • Can suppliers switch to produce relevant product in short term without incurring significant costs? o E.g.: paper industry o Machines can produce different quality of paper easily without thigh switching costs o Quality of paper for an article and handouts are different, demands can be different o Higher demand for say art books, and a firm normally produces cheap paper, can switch production without incurring much costs o That restricts the main incumbents' (who are producing high quality paper) pricing actions o Can be seen like an entry but different, happens in short time without high costs 3. Potential competitors o Potential competitors, those firms that are not in the market but may enter the market o If there is a threat, that restricts the behaviour of the existing firms in the market

why may market shares not be a good indicator of market power (7 reasons)

1. Data problems o May not have good data that gives exact numbers for us to make firm conclusions 2. Low entry barriers o Irrespective of high market shares the firm may need to charge competitive prices if there are potential competitors o Vice versa, low market share but high entry barriers 3. Bidding markets o Buyers choose suppliers via auctions or tenders o A firm might have large market share but had to pass competition at the bidding stage 4. Successful innovation o High market share → innovation o If this is the case, it means they are investing in R&D, competing in terms of innovation o May not be abusing market power 5. Product differentiation o Even the products with low market share might have market power because the others are not close substitutes 6. Responsiveness of costumers o Firms with similar market share does not necessarily mean similar market power o It may result from customers' ability or willingness to switch to alternative suppliers o E.g. strict preference for apple, keep buying apple so they may abuse their market power o E.g. broadband agreement, when agree on 24-month deal, pay a penalty if end before, BT increase price but still can't change to virgin 7. Price responsiveness of competitors o The competitors may be producing at full capacity o When you want to increase your price they can't respond

factors to consider when assessing market power

1. Economic regulation: Prices or level of services may be controlled by government/industry regulator 2. Evidence about the behaviour and the financial performance of the firm 3. Competitive constraints: How strong are the competitive constraints? Is it possible to sustain high prices given these constraints? o Existing competitors- firms in the relevant market ♣ Strong enough- unlikely firm can abuse market power as other firms will just undercut if raise prices o Potential competitors- new entry, if entry barriers are low, high prices may not sustainable ♣ Strong enough- new competitor will undercut firm if they increase prices

examples of other evidence that can also help define markets

1. Evidence of substitution in the past 2. Quantitative tests o Estimates of elasticities and cross-price elasticities (to see which products are substitutes o Similarity of price movements over time ♣ Similarly over time: substitutes ♣ Moving differently: no relation o Causality analysis between price series 3. Views of customers and competitors 4. Consumer preferences, marketing studies - surveys 5. Barriers and costs associated with switching demand to potential substitutes o E.g. no cost of switching drinking to coke from Pepsi o However, may be some cost switching broadband provider 6. Different categories of customers & price discrimination

leniency programmes: experiment results on cartel formation

1. Fines and particularly leniency are effective in deterring cartel formation. 2. Leniency significantly increases cartel detection. 3. Leniency significantly reduces cartel re-forming. Encouraging results

leniency programmes: experiment results on prices:

1. Fines increase average price compared to no competition policy. Leniency gives almost the same average price than no competition policy. 2. Fines and leniency significantly increase cartel prices relative to no competition policy. Strange results; just having fines increases the average price. Both fines and leniency reduce no of cartels but increases average prices. Less cartels taking place but they are stronger. Thus the average prices are doing this. This is puzzling; where do they coordinate

examples of market concentration indices

1. Herfindhal-Hirschman Index (most common) • Sums the squares of market shares of firms in the market 2. Entropy index (Not used that much) 3. k-firm concentration ratio: Total market share of top 'k' firms Issue: how to determine k

Further issues with UPP: Measurement issues of UPP

1. Measuring gross margins ♣ In general the firms report their gross margins, especially if they want to merge ♣ With the troubling mergers, gross-margins are fairly high so the UPP test may not be sensitive to precise measurement ♣ Firms in general keep track of their margins and they need to share it with the agency 2. Measuring diversion ratios: customer surveys, estimation methods using data, etc. ♣ Diversion ratios are almost like an elasticity ♣ May need to use data to get this 3. Sensitivity analysis ♣ Gross margin of firm is this much, if diversion ratio is more than 10% likely to increase prices

Further issues with UPP

1. Non-Bertrand Behaviour: • Outside that case the definition of diversion ratio may not be correct o E.g. Don't know what will happen with Cournot competition 2. Measurement issues •How practical is it to measure pre-merger gross-margins and the diversion ratios between the two products? 1. Measuring gross margins • In general the firms report their gross margins, especially if they want to merge • With the troubling mergers, gross-margins are fairly high so the UPP test may not be sensitive to precise measurement • Firms in general keep track of their margins and they need to share it with the agency 2. Measuring diversion ratios: customer surveys, estimation methods using data, etc. • Diversion ratios are almost like an elasticity • May need to use data to get this 3. Sensitivity analysis • Gross margin of firm is this much, if diversion ratio is more than 10% likely to increase prices 3. Mixed test results: o What if UPP1 > 0 and UPP2 < 0? o Just finding one positive is enough to flag the merger o A positive test result for any product should be enough to trigger further scrutiny 4. Firms controlling multiple products prior to the merger: o If Firm A owns just Product 1 and Firm B owns substitute products n = 2, ..., N, the cannibalization term should be: N T 1 = D 1 n ( P ̄ n − C ̄ n ) n=2 o Should be careful if one product is substitute and the other is a complement to firm A's product o Additional sales may also generate intangible benefits (or costs) as well as direct and readily quantitative net receipts o E.g.: sales of spare parts, complements, markets with network effects, etc. o Not considering the other profits or benefits that the firms might have, e.g. after sale services ♣ Didn't increase prices but charging too much after sale

measuring market share issues (5 in total)

1. Production, sales and capacity o In general use the sales not total production but sometimes important to look at capacity 2. Sales values o If the product is homogenous then look at sales data by volume o If differentiated products look at values 3.Choice of exchange rates o They are volatile and change over time o Need to do sensitivity analysis after constructed it 4. Imports o International market: just look at the firms in international market o If the market is domestic market, and some goods that are imported, need to calculated share of each imported good separately o Shouldn't be taking total share of all imports 5. Internal production o Supplier may use some of its capacity or product ('captive capacity') ♣ E.g. some paper you produce you keep for yourself ♣ Keeping some production for yourself ♣ Then buy bad paper and sell your own good paper o It can divert its captive capacity/product to the open market if it's more profitable

factors contributing entry barriers

1. Sunk costs: E(π) <sunk costs → do not enter 2. Poor access to key inputs and distribution outlets o Essential facilities ♣ One firm has full use over essential facility, or not easy to get access, harder to enter o IPR: in the long-run it may stimulate competition in innovation ♣ Intellectual property rights (IPR) ♣ These make it harder but in the long run it may stimulate competition/innovation 3. Regulation o Licences may limit the number of firm operating in a market o Setting standards, e.g.: health and safety standards o If the regulation does not apply to all firms equally it may increase entry barriers ♣ Some firms may lobby to set high standards that not everyone can reach, keeping entry barrier high 4. Economies of scale o High economies of scale in the industry: may have to enter with high level of production o New entrant may not have enough distribution outlets/customers to sell product o High economies of scale make it more profitable for firms already in the market and harder for new firms to enter the market 5. Network effects o Users valuation of the network increases as more users join network o Not just an entry barrier, it just may contribute to make it harder to enter o E.g. telephone industry 6. Exclusionary behaviour: anti-competitive behaviour which harms potential or existing competition o Predatory response to entry ♣ Cutting the prices below marginal cost just to put off potential entrants o Discounts to foreclose market o Vertical restraints

Anti-Trust Issues Raised by 2SM Introduction- differences between one sided markets and 2SM

1. The individual prices charged on either side of the market do not track costs or demand on that side of the market 2. One cannot talk about the individual prices in isolation 3. Products in two-sided markets cannot come into existence and cannot remain in existence unless firms in those markets get "both sides on board" o E.g. if sellers don't accept credit cards, couldn't exist 4. Analysis of social welfare must account for the pricing level, the pricing structure, and the feasible alternatives for getting both sides on board

UPP in 2SM: Formal derivation Comparison of two-sided UPP and one-sided UPP:

1. There are two relevant formulas in two-sided markets, one for each side. 2. Each two-sided UPP formula has more terms than one-sided UPP one. o This is due to the fact that pricing decisions on one side also affect profits earned on the other side of the market. 3. The diversion ratios used in the two-sided UPP formulas also account for the feedbacks between the two sides of the market

Further content of Armstrong 2006

1. Uniform prices 2. Two-part tariffs 3. Competitive bottlenecks

give 4 extra aspects of the SSNIP test that are worth a closer look

1. What might make such a price rise unsustainable? 2. The starting point is "the narrowest group of products and geographical area" • Better to define smaller than too large • Too large: market share defined will be too small • Too large: may miss a case you should be flagging 3. One of the challenges in applying it is its hypothetical nature and gathering the information needed to put it in to practice 4. The basic idea behind it is that a market is a collection of products and geographical areas which can be profitably monopolized

factors that help collusion (6)

1. smaller number of firms 2. high entry barriers 3. high demand growth 4. regular orders 5. frequent orders 6. less innovation

some considerations to bear in mind when defining markets

1. temporal markets, time dimension e.g. peak v off peak/ seasonal variations o E.g. off peak and peak train may not be substitutes, people have to take peak to get to work 2. cellophone fallacy o SSNIP found cellophane and aluminium foil are in the same market o Problem: they increased the price which was already above the competitive level, when done further people will look for other products which aren't necessarily substitutes o Defined the market as a larger group of products, share of cellophane looked small 3. Secondary products, spare parts: printer &cartridge, razor &head of razor o Canon printer may be in competitive market but the cartridge may make it so that they aren't real substitutes 4. Chains of substitution: A-B-C-D-E o Products in A and B are substitutes for B as close to B o D may seem that it isn't in the same market as B, but because substitutable with C you end up putting them in the same market 5. Wholesale products o Although you may have two competing firms producing some retail products o Might be restricted by their wholesaler o The products at the retailer level may or may not be a good way of defining the market at the wholesale level o Cannot define the market by just looking at the retail level for the wholesale product o When you are analysing the market, good idea to consider market structure o In general: the effect of a 10% price increase at the wholesale level on downstream depends on two things ♣ How much the price increase will be passed to consumers at the downstream level ♣ Substitutability of the products at the downstream level o If you are trying to define market at the wholesale level, just because they are also selling at the retail level doesn't mean it's the whole market o You need to be aware what's happening downstream but the market is the wholesale market 6. Portfolios of products

what are the steps to carry out merger simulation

1.Model the industry and nature of the competition 2.Calibrate the model using pre-merger data

Lysine cartel case: Lysine fines

5 corporate fines U.S. corporate fines: $92.5 million EU: $97.9 million Canada: $11.5 million Lysine cartel U.S. fine was 7x previous highest fine 7 personal fines Prison sentences for 3 people, 99 months total In Europe only the company can be fined, in America it is a criminal offence and the CEO went to prison

Dominance

Abuse of dominant position OR European Court defines dominant market position as: 'a position of economic strength enjoyed by an undertaking which enables it to prevent effective competition being maintained on the relevant market by affording it the power to behave to an appreciable extent independently of its competitors, customers and ultimately of consumers.

Lysine cartel case: Lysine agreement

Allocated sales quotas across firms Used industry organization to meet and collect data Shared cost data Had a punishment scheme in place

upward pricing pressure (UPP)

Alternative to merger simulation Upward pricing pressure compares two opposing forces: 1. The loss of direct competition between the merging parties ⇒ upward pricing pressure 2. Marginal cost savings from the merger (cost efficiencies) ⇒ downward pricing pressure •Farrell & Shapiro (2010) created UPP to compare these two forces without predicting the full equilibrium (eqm) adjustment of the industry after the merger (easy to read article) •If the net effect of these two forces creates upward pricing pressure → the merger should be flagged to further scrutiny

collusion and multiple equilibria

Any π' such that 0 ≤ π' ≤ πm can be sustained in equilibrium if and only if δ ≥ δ* = ½ We focus on full collusion, πm, although there can be multiple solutions to the game

what is competition policy; in the EU? in the US?

Article 101 of the Treaty on the Functioning of the European Union prohibits agreements to fix prices, to limit production or to share markets or customers In US Sherman Antitrust Act 1890. In US, competition policy is called antitrust policy (early cartels like Standard Oil were called trusts). Rockerfeller made his fortune by gaining almost complete control of oil production in America, Standard Oil, broken up in 1911

Kaiser & Wright (2006) The Magazine

Assume profit of magazine: gross margin they make on the readers side, gross margin on advertising side and cost of producing content Where The magazine sets pi, ai and N_i^cto maximize its profits, given the choices of its rival Solving for unique demands, substituting these into profit for i = 1, 2, taking FOCs, give the equilibrium conditions for both magazines: Markup on readers side depends on how much benefit the readers are giving to advertisers (roe) Markup on advertising side depends on how much benefit advertisers are giving to readers (gamma) Then optimal number of content pages Estimate share equations, use the parameter estimates to solve for equilibrium price and cost margins

collusion at the extensive margin: example

Australian two local newspapers; you get advertising in one town and I get it in another. Letter; not focusing on your area of operation..if you continue to deviate..there will be a punishment A letter from a 'competing' newspaper "The Mannum advertising was again evident, which suggest your Waikerie operator, John Pick, is still not focussing on the traditional area of operations... If you continues to attack in Mannum, it may be we will have to look at expanding our operations into areas that we have not traditionally services" Waikerie exited Mannum after the letter.

leniency programmes: benefits and drawbacks

B: Leniency program can help break cartels ex post D: But can have negative ex ante effect: possibility of receiving amnesty increases the expected profits of forming cartel, cartel amnesty effect (Motta and Polo (2003) IJIO) D: Could leniency program make collusion easier? perhaps it makes cartels more profitable. When possibility that the heavy fines can be waived completely; increase expected profits from colluding

multimarket contact: model setup

Bernheim and Whinston (1990) Rand Multimarket contact = Firms compete against each other in many markets (geographic or product markets) Does multimarket contact make collusion more or less sustainable? ( now suppose there is competition in many markets... )

leniency programmes: experiment aims

Bigoni, Fridolfsson, Le Coq and Spagnolo (2012) Rand Laboratory experiment on leniency policy 1. Does leniency policy reduce number of cartels? 2.Importantly, what are the welfare effects?

Outline the previous and current contracts between online travel agents (OTAs) and hotels.

Booking.com and Expedia previously included wide MFN (most favoured nation) clause in their contracts with hotels which prevented hotels from offering cheaper room rates or better availability or conditions on competing OTAs. After investigations by competition authorities, Booking.com and Expedia changed their policies in 2015 and no longer require wide MFN. Narrow MFN remains, i.e. hotel cannot charge a lower price on their own website than what they offer through OTA. (The hotels can, however, offer lower prices directly to customers e.g. by telephone, direct email or loyalty groups.)

static collusion: effect on profits

By colluding firms charge monopoly price pm and earn ½πm > 0 Both firms agree to charge the monopoly price rather than MC, from tomorrow onwards. If they both keep their promise, they both get half the profit. Homogenous good; if one cheaper buy from cheapest, if equal assume that consumers split 50/50 between two.

collusion at the extensive margin

Byford and Gans (2014) IJIO Collusion at the extensive margin = cartel assigns different markets to different firms Collusion at the intensive margin = collusion in pricing within markets Intensive margin- colluding in price Extensive margin- I am monopoly in this market, you in the other. However this has to be done in a hidden way.

How do firms' incentives to collude differ depending on whether collusion is at the intensive or at the extensive margin?

Collusion at the extensive margin allows cartels to assign different markets to different firms whereas collusion at the intensive margin means firms can only form a cartel within a certain market. Firms will therefore have different incentives to deviate from collusion depending whether they are at the intensive or extensive margin. The incentives to collude will differ depending whether there is an extensive or intensive margin. The basic intuition for the extensive margin is illustrated by 2 firms operating in 2 identical markets. The firms are able to choose which markets they enter, so firm 1 will have the monopoly in market 1 and not enter market 2 and firm 2 will have the monopoly in market 2 and not enter market 1. A firm deviates by entering the rival's market. However, unlike at the intensive margin where firms cut prices, entry to a market has no element of surprise and is easier to observe. As soon as one firm notices its rival has entered they are able to adjust prices accordingly which will help to sustain collusion. If goods are differentiated, once a firm enters the rival market they will make Nash equilibrium profits. The equation below shows collusion at the extensive market. with differentiated goods, is sustainable if and only if: The left hand side shows each firm earns the discounted monopoly profit in their own separate markets. The right hand side shows what happens if a firm deviates. The first term shows the firm earns the monopoly profit in their own market but only the Nash equilibrium profit in their rival's market, because their entry was observed and the rival firm adjusted their prices accordingly. The second term shows that in the future the rival firm will also adjust their prices in the other market. Therefore, in both markets the firm is now only able to earn the discounted Nash equilibrium profit. The equation simplifies to: This expression is able to demonstrate that at the extensive margin collusion is easier to sustain when the industry is competitive. We can see this from this equation because as the Nash equilibrium profits tend to zero, the discount factor will also tend to zero. Competition will reduce the Nash equilibrium profits as it drives down prices. If Nash equilibrium profits are smaller, the gain from deviation is less, so the discount factor needed for the cartel will also be smaller. This illustrates that at the extensive market collusion is easily sustained when there is more competition. In contrast, the opposite is found at the intensive margin as a larger number of firms will hinder collusion. As stated above, the discount factor which facilitates collusion at the intensive margin with n number of firms is: This shows that when the number of firms increases, i.e. there is more competition, the discount factor needed to sustain a cartel will increase. The incentives to collude at the intensive margin weaken with competition whereas they become stronger at the extensive margin. The fact that the incentives to collude will be different depending whether collusion occurs at the intensive or extensive margin holds for another factor which influences collusion, barriers to entry. It is clear that low entry barriers make it harder to collude at the intensive margin for two reasons. Firstly, if a new entrant decides not to join the cartel they can easily undercut it. They can break the collusive equilibrium and set a price below the monopoly price to capture all of the market's profit. Secondly, if the new entrant does collude there will be more members in the cartel and this makes collusion harder to sustain, as shown above. However, we have already shown that competition facilitates collusion at the extensive margin and therefore the new entrant would make it easier to sustain the cartel. On the other hand, a factor such as growth facilitates collusion at both the intensive and extensive margin. An increasing growth rate of demand ensures there is more to lose from deviating. The equation below shows when collusion is sustainable at the intensive margin, with the industry growth rate of demand g, assuming there are 2 firms: The left hand side shows that if collusion is sustained the profits each firm earns are not only discounted but will grow at the rate g. The right hand side is unaffected by growth as once the cartel is broken the deviant firm will earn zero profits. As this makes clear, higher growth lowers the critical discount factor which makes it easier to collude. This is due to the fact the profits earned in the cartel are now increasing in the future and when a firm deviates they will lose all of this. This reasoning extends to the extensive margin, as shown below: This result is the same as the intensive margin case, the fact the industry is growing ensures collusion is more likely. As the growth rate of demand increases, so can the price and therefore profits if the firm remains in the cartel. No matter if collusion is at the intensive or extensive margin the firms have more to lose if there is growth in the industry, resulting in stronger incentives to remain within the cartel. One empirical example of an extensive margin occurred in Australia between two newspapers. Having noticed newspaper 1 had entered their market, newspaper 2 contacted the first to ensure they only advertised in their own towns. This enabled both of them to capture the whole of the market within their own town and earn the monopoly profit. The Australian courts found that this was an anti-competitive agreement.

factors affecting collusion: frequent orders

Collusion is more likely the greater the frequency with which firms interact Similar effect for frequency of adjusting prices for example, consider that now instead of weekly markets, it becomes daily. Firms interaction greater which makes collusion more likely. If meet everyone more frequently, collusion is easier as the punishment will be quick and swift... keeps the cartel on their toes. Punishment is far away until more tempting to deviate.

Is collusion more or less difficult to sustain under Cournot competition than under Bertrand competition?

Collusion is therefore sustainable if and only if (see pdf for the equations) d >= 1/2 Critical threshold is higher than under Bertrand competition. Both the gain and the loss from deviation are lower under Cournot competition and the loss effect is dominant. Loss is lower because in every period still make positive competitive profit rather than in Bertrand case where it drops after the first period Gain is lower too; in Bertrand by deviating, you can get all the customers by undercutting slightly. In Cournot, expand production but rival is still producing so cant be gaining all the profit thus under cournot (competing in q); the consequences are less severe and thus collusion more severe Therefore it is more difficult to collude under Cournot competition.

static collusion: issues

Collusion often fails because each firm has an incentive to cheat Cheating firm charges pm - ε and earns πm > ½πm In static game cheating is a dominant strategy therefore the issue with static collusion (one starts undercutting)/ if one undercuts by epsilon very little amount. Then they get more than half the profits which is what they get in colluding. If only one short game, the dominant strategy is to cheat. Repeated games may be more realistic

intro

Collusion with Bertrand competition allows firms to earn higher profits by setting prices above marginal cost, at a higher monopoly price. This allows the firms to split the monopoly profits; for example if there were two firms they would each get half the monopoly profit. This will unquestionably be larger than the zero profit they would make under perfect competition. There may be an incentive for a firm to deviate from collusion as they can gain the whole monopoly profit that period by setting a lower price, undercutting their rival and capturing the whole market. However, the other firm will punish the deviation in the following periods by also lowering their price. So no profit can be made in the periods after a firm chooses to deviate. Multimarket contact allows overall collusion to be facilitated as markets which are easier to collude in are able to help markets which are harder to collude in. Collusion at the extensive margin allows cartels to assign different markets to different firms whereas collusion at the intensive margin means firms can only form a cartel within a certain market. Firms will therefore have different incentives to deviate from collusion depending whether they are at the intensive or extensive margin.

UPP:Quantifying cannibalisation: Theoretical setup

Consider a merger with rival firms A and B with profits πA and πB •Suppose the merged firm now operates former firms A and B as Divisions A and B •Suppose for now that there are no efficiencies (as a result of merger) •Let's consider the simple case of single product static Bertrand competitors with differentiated products • Firm A produces Product 1 whose price is P1 and marginal cost is C1 • Firm B produces Product 2 whose price is P2 and marginal cost is C2 • Pre-merger values: P1, P2, C1, C2

summarise article 102

Don't reduce your market power in a way that limits production or reduces prices. according to european treaty firms shouldn't be doing any actions that harm the market

UPP in 2SM Formal derivation

Empirical example in the paper is for the daily newspaper market, so let's say that the firm sets P_1^A on the advertising market and P_1^r on the readership market Evaluating a merger in a two-sided market is more complex Suppose firm 1 increases its price on the advertising side P_1^A: Q_1^A quantity on advertising side decreases by one unit (less advertisers) This decrease of Q_1^A also decreases Q_1^R If the indirect network externality is positive Decreasing the number of advertisers decreases the number of readers Lost advertisers but also lost readers The additional profits of firm 2 are then the recaptured units of advertising times the advertising margin Firm 2 is able to capture the advertisers and readers that firm 1 lost In addition to the recaptured advertisers, firm 1 now also internalizes the recaptured readers in firm 2 times the margin on readers A similar reasoning applies to the effect of an increase in P_1^R On top of this, there are feedbacks between the two market sides • Another difference: need to consider two UPPs, advertising and readers side

explicit v tacit collusion: is the set of equilibria the same?

Evidence of firms communicating about prices is per se violation of competition law But the set of equilibria is the same with or without communication Does talking matter? COORDINATION can be difficult and communication can help but with and without communication, we have the same set of equilibria. How does coordination affect communication? Remember just talking about pricing is illegal... SO people have used Experiments; games that replicate the market; how does communication affect the occurrence of cartels. In any game there are multiple equilibra. Focus on where each gets half profit, but anything is possible and sustainable in the repeated game. Which one do they pick?

why do firms collude

Firms earn higher profits by coordinating their activities rather than acting independently

UPP in 2SM Formal derivation UPP on readers side:

First term and third term taking into account the externalities • First term: how many more advertisers firm 2 will get when the quantity of readers in firm 1 changes (times the gross margin) o How much its revenue will increase if it gets some extra advertisers because the number of readers in firm 1 has changed • Third: the cost increase that may happen because may lose readers and advertisers • Where: SEE CARDS)

leniency programmes: example

Harrington (2008) JIE allows for both cartel amnesty effect and deviator amnesty effect Deviator amnesty effect = deviating firm enters the leniency program and undercuts the collusive price simultaneously Gain from deviation higher with leniency With full amnesty, deviator amnesty effect is dominant and leniency

collusion at the extensive margin: differences from intensive margin collusion

How is this different from intensive and what does the model look like. Firm 1 takes market 1, Firm 2 takes market 2 (they assign markets). Deviation is that Firm 1 enters market 2. entry can be observed (rather than pricing) so the firm can adjust prices so they are not taken completely by surprise. The intensive margin; this works even when they are identical. The deviation doesn't have a completely surprise effect; the firm can respond, and if the market is competive they can gain not very much by entering. In practice.. In many industries this would be under great suspicion, so pricing would be a better way to do this.

repeated interaction: what prevents a price war

If the firms care enough about the future, they can sustain collusion because one-shot gain from undercutting the rival does not compensate for long term losses from entering a price war.

Lysine cartel case:

In 1996 Archer Daniels Midland pleaded guilty to price fixing in lysine (used in livestock feed) and citric acid (used in soft drinks and detergents) (Youtube video, the first cartel case Huge fines received, 7x the previous fines)

secret price cuts: theory

In customer markets (e.g. ready-mixed concrete and ocean shipping) prices are not public but negotiated Collusion is difficult to monitor when secret price cuts are possible Lack of transparency does not necessarily prevent collusion but makes it more difficult Up to now in the theoretical Bertrand model: firms set prices, pricing observable to other firms/ rivals eg online pricing thus they know when they break cartel pricing In reality; prices are usually negotiated, so final price can be secret and may not be something available to the public. Are they not selling because other firm has undercut or because demand is low? Secret pricing doesn't prevent collusion but does make it more difficult.

explicit v tacit collusion: experiment setup

In experiment by Fonseca and Normann (2012) repeated Bertrand price competition with inelastic demand Communication for one minute by instant messaging before price setting Unit demand, price upto 100, buy 1 unit. If price greater than 100 they don't buy. The best cartel price here would be 100. the most competitive price (Bertrand price= 0) In principle we expect a price between 0 and 100.

repeated interaction: model setup ( six points) ; strategy? to collude or not to collude?

In repeated game the firms can collude and charge monopoly price pm Time t = 1,2,... Firms follow trigger strategy 1. Charge pm in period 1 2. In period t (t > 1) charge pm if rival chose pm in period t-1 3. Otherwise choose p = c forever (punishment) here: the game is repeated infinitely. They start by charging monopoly price in first period. If rivals get at any time undercut, the firm goes to the Bertrand equilibrium of static game and charge this price forever. If one deviates from the promise there is a consequence for it in the future. TRIGGER STRATEGY- cartel forever collapses. Need to analyse which is better; is charging a monopoly price actually sustainable/ an equilibrium in the repeated game.... Need to compare the two i.e. colluding or not; what are profits if I keep the promise and collude or deviate. Stick to the cartel - get half profit today tomorrow and infinitely. If I undercut and deviate once, today I get more but i am punished in the future forever

Which of the models covered in the lecture is best suited for analysing this case?

In terms of the models of the lecture notes, hotel is the manufacturer and OTAs are the retailers (although hotel can also sell directly to customers). Retail competition model fits this case well. (Double marginalisation may not be as issue in a potentially competitive industry and hotels do not make investments in OTAs.) DOUBLE MARGINALISATION: monopoly power in more than one level eg Bertrand competition with homogenous goods/ price comp but the problem is that here there is only one monopoly; unfair competition Unlike in the lecture notes, here OTAs have the bargaining power in setting the terms of the contract.

Do innovations hinder or facilitate collusion?

Innovations can make a firm more productive and grow faster thus give it a greater growth rate. Collusion can be more likely the greater the growth rate of the industry Higher probability of innovation hinders collusion. The gain from deviation does not change. The loss from deviation is lower.

diagram for monopoly features

L.H.S: linear demand curve, fixed marginal cost, then the monopoly price is given where MR=MC • Why need to regulate: issues on R.H.S • In a competitive model the whole triangle is consumer surplus • Total welfare: consumer + producer surplus • Monopoly model, DWL is a loss to society • Don't want firms to charge too high price or restrict output as loss to society

Level for detail in exam:

Level for detail in exam: o Assume two stage decision o Like example of buying tractor truck or rigid truck o Then once decide what to buy then decide on brnad o When you model it that way, rigid group are more substitutable with each other than tractor

factors affecting collusion: entry barriers

Low entry barriers hinder collusion If entrant does not join cartel and behaves aggressively, it can break collusive equilibrium. If entrant joins cartel, it is more difficult to sustain collusion with larger number of firms. If a cartel is profitable it will attract new entry.

multimarket contact: non-identical markets model

MARKETS not ALWAYS IDENTICAL. Eg look at differing growth rates pooling the incentive constraints in different markets can help remove slack from one market to charge Now suppose markets differ in growth rate g(i) for i = 1,2 Example g1 = 0, g2 = 0.5, δ = 0.4 First examine single-market case Collusion is not sustainable in market 1 since δ < 0.5 In market 2 collusion is sustainable since δ ≥ 1/(2(1+ 0.5)) = 0.33 Market 2 grows only. The discount factor is lower than critical threshold 0.4 so collusion occurs

Kaiser & Wright (2006) Empirical Specification: main identifying assumption

Main identifying assumption: • Cost factors are common across magazine published by a magazine's publisher and that other (demand-side) shocks specific to the magazine are not correlated with these factors

give an example of measuring market power

Measuring Market Power in the Ready-to-Eat Cereal Industry by Aviv Nevo, Econometrica 2001 • Ready-to-eat cereal industry: High concentration, High price-cost margins • Previous research: nearly collusive pricing behaviour • Nevo examines the price-cost margins and separates the margins into three sources: o due to product differentiation o due to multi product firm pricing o due to potential price collusion • Found that the first two effects explain the high price-cost margins and prices are consistent with non-collusive behaviour

Merger simulation

Merger simulation is a natural alternative to "market shares" approach

benefits of merger simulation

Merger simulation provides a precise, quantitative prediction of the unilateral effects of merger BUT prediction is only valid as long as o The model capture the essence of competition in the industry o If the post-merger interaction between the other firms does not change

The EU Merger Guidelines: what are the two ways mergers may deter competition

Mergers may deter effective competition in two ways: o Non-coordinated/unilateral effects: by eliminating important competitive constraints on firms → ↑ market power o Coordinated effects: by making anti-competitive coordination between the remaining firms more likely ♣ E.g. as a result of the merger the other firms respond by colluding

Does multimarket contact hinder or facilitate collusion?

Multimarket contact allows overall collusion to be facilitated as markets which are easier to collude in are able to help markets which are harder to collude in. If markets are identical multimarket contact will have no effect on collusion. Multimarket contact means that the same firms are competing with each other in many markets, for example geographic markets or product markets. There is more to gain from deviating because there are now two markets where the firm can undercut its rival to earn double the monopoly profit. However, this reasoning also holds for the loss from deviation as the profits will drop to zero in both markets. These losses exactly offset the gains from deviation meaning there will be no effect on collusion. The situation is represented by the equation below, collusion with two firms is sustainable if and only if: : monopoly profit : discount factor The left hand side shows the situation with collusion. Firm 1 and 2 get half the monopoly profit in both markets, with the future periods being discounted. The right hand side shows what would happen if one firm deviated. The deviating firm would earn the full monopoly profit in both markets in the first period and zero profits thereafter. Collusion is only sustainable if the left hand side outweighs the right hand side meaning the discount factor has to be greater than or equal to ½. This is the same critical discount factor that arises when there is just one market for the firms to collude in. Therefore, when markets are identical multimarket contact does not facilitate or hinder collusion. However, in reality markets are going to be non-identical and this can be shown to facilitate collusion. This is because one of the markets will be easier to sustain collusion in and this market is able to support the cartel in the second market. There are a number of reasons a cartel may function better in one market over another. These could include: a smaller amount of firms, higher barriers to entry or a higher growth rate of demand. If take the case where the two markets having a different number of firms we can demonstrate how multimarket contact facilitates cartels. The equation below outlines the general case for when collusion is sustainable with n firms: The left hand side shows the monopoly profit is divided among the number of firms each period there is a cartel, discounted in future periods. The right hand side shows the case where one firm deviates and gets the whole monopoly profit in the first period and zero profits thereafter. The discount factor is therefore equal to: If market 1 has 3 firms, market 2 has 2 firms and the discount factor is equal to 0.6, we can see that without multimarket contact on market 1 would fail to sustain collusion whilst market 2 would have a cartel. Market 1: Market 2: If we now allow the firms to have multimarket contact we are able to show that by adding the incentive constraints together we allow the firms to transfer the slack in market 2 to market 1. With multimarket contact collusion is sustainable in both markets if and only if: With the values stated above this simplifies to: This shows collusion is sustainable across both markets, not just the second, proving multimarket contact facilitates cartels. When you add the incentive constraints together the second market is able to pull the first market into collusion. Even though market 1 is very tempting to deviate from the firm now suffers the punishment in both markets. There are many empirical examples of multimarket contact. For example, airlines and mobile phones will unavoidably have firms competing in many markets. One case with air cargo carriers allowed a cartel to coordinate over a six year period. Although there were other factors which facilitated the collusion, such as frequent contact and high barriers to entry, multimarket contact across geographical areas would have made it easier to collude.

how to asses whether entry barriers exist

NOTE: THIS IS HARD TO DO! 1. Ask potential entrants and incumbents 2.Entering a neighbouring market is easier 3. Trade barriers make it harder 4. Easier in growing markets 5.Easier in markets with differentiated products 6. If brand image is important, advertising needed before entry

Kaiser & Wright (2006) Advertisers

Na advertisers, all place a single ad in one magazine πi, profit of a firm, that advertise in magazine i is a linear function of N_i^r (↑) and fee for placing ad, ai(↓). It also has a fixed affect θ_i^a and a mean-zero error term, ϵ_i^a Under the assumptions that the preference for a magazine is represented by a transportation cost, and it is ∼ U[0, 1]: SEE CARDS where : share of ads in magazine i End up with two equations, advertisers and reader sides Advertisement side: take into account the number of readers Effect of number of readers on advertisers As well as price of advertisement Readers side: take into account number of advertising pages Effect of advertising pages on readers As well as content pages and price

Two-Sided Single-Homing Basic Model

Now want to see what will happen with competing platforms Two platforms competing, competition will become even stronger 2 groups of agents: 1 and 2 2 platforms: A and B (1) Under the Hotelling Specification, where Agents in a groups are uniformly located along a unit interval 2 platforms are located at the 2 endpoints Platforms may be two newspapers, more conservative vs liberal view t1,t2 > 0 are the product differentiation (transport cost) parameters that describe the competitiveness Then the number of each group who join platform i: (2) Putting (2) together with (1) and using the fact that gives the following expression for market shares: (3) (4) n is the number of agents on each side on platform i On side 1: if you keep the prices on side 1 fixed, each extra side 2 agent will increase the number of side 1 agents by alpha1/t1 Keeping its group-2 price fixed expression above shows that an extra group-1 agent on a platform attracts a further (α2/t2) agents to that platform Suppose platform A and B offer respective price pairs (p_1^A,p_2^A) and (p_1^B,p_2^B) Given these prices, solving simultaneous equations in (3) and (4) gives the following market shares: Write the market share of the demand on side 1 only as the function of prices and parameters of the model No longer a function of people on the other side If α1, α2 > 0, demand by 2 groups is complementary, in the sense that a platform's market share for one group is decreasing in its price for the other group For the case of a symmetric equilibrium where each platform offers the same price pairs (p1, p2), FOCs for equilibrium prices are: In a Hotelling model without network externalities price would be p1 = f1 + t1 In a two-sided setting price is adjusted downward by (α2/t2)(α1 + p2 − f2) (α1 + p2 − f2): External benefit to the platform when an additional group-2 agent joins Platform attracts (α2/t2) extra group-2 agents when it has extra group-1 agent Profit: if there is another group 2 agent joining the platform it is going to cause a benefit of α1 to side 1 agents, pay price p2 and that agent cost the platform f2 The price as a group 1 agent is adjusted by the benefit I make to the platform, platform exploiting extra benefit making on the platform when deciding on price

why is availability given different treatment to prices?

OTA's could also favour and give more visibility to hotels that provide them extra; lectures; problem when verticle restraints can be welfare increasing but anti competitive; could be collusion between OTAs offering price competition anticomp effects of wide MFN; suffer price competition or even collusion if prices are too low at the OTA's this could be a barrier to entry ; if someone has to enter with a very low price, can only do so if their MC is lower than this which is less likely the lower the prices are here this couldnt be done as an entrant can offer lower prices since if hotel lowers price; it must lower the price on all its platforms this can block innovation of something that can be even lower TO summarise: barrier to entry: entrant thus needs to offer low price to enter wide MFN prevents entrant lower price and blocks innovation if there is too much free riding the service may no longer be provided AVAILABILTY given different treatment since it is not directly related to individual welfare; price is key here

what is the rationale for allowing hotels to offer cheaper prices in other platforms but not on own website?

OTA's provide an important service for consumers -however if you can actually book cheaper on the hotels website from using the OTA's services this would be free riding and the OTA's wouldnt be able to cover their costs they wouldnt be able to survive thus to ensure survival of the OTA's in the market, allows more customers to be reached OTA's could also favour and give more visibility to hotels that provide them extra;

What is the economic rationale for different treatment of hotel's own website and other OTAs?

OTAs provide an important pre-sale service: easy identification of available hotels and price and quality comparisons. Without narrow MFN (most favoued nation) the hotel could freeride on the service provided by the OTA. A customer can identify a suitable hotel via OTA but make a booking via hotel's own website at a cheaper price. As hotel does not pay a commission to OTA for this booking, OTA does not get a return for their investment in pre-sale service. Therefore OTA would underinvest in pre-sale service. Narrow MFN protects OTA from unfair competition and restores their incentives in providing the pre-sale service. This benefits the customers who value the service and also the hotel who will get more customers via OTA's search facility. Narrow MFN is therefore welfare increasing as it benefits all the parties. The main effect of broad MFN is the reduced competition between OTAs which can drive up their commissions and the hotel room rates. As the main effect is anticompetitive, Booking.com and Expedia had to remove their requirement of broad MFN. It is yet to be seen what effect this change will have on hotel room rates and consumer welfare.

Estimating Market Power in 2SM The Model-Supply Side

Oligopolistic Competition The profit function of each newspaper i is: Price times circulation (P depends on price of newspaper and price of other newspapers) on the readers side Price times quantity on the advertising sides, note that the quantity on advertising side is affected by advertising price and also the cover price, price for the reader because demand for advertisement is affected by number of readers Joint profit maximization Firms collude, so each firm i chooses the prices which maximize FOCs with respect to p_i^A and p_i^N give mark-ups for each side under each assumption

repeated interaction: gains and losses

One-shot gain from deviation = πm - ½πm = ½πm Long term loss from deviation = δ½πm + δ2½πm + ... - 0 = ½πmδ/(1-δ) Thus collusion is sustainable if and only if gain ≤ loss If firms don't care about the future....they may have an incentive to cheat. Cheating depends on the discount factor. If they care enough about the future... able to collude since they realise there is a consequence of undercutting

In what other platforms can similar issues arise?

Other cases, such as e-books and price comparison websites, are discussed in 'Most Favoured Nation Clauses' below and Akman (2015) 'A Competition Law Assessment of Platform Most-Favoured-Customer Clauses', CCP Working Paper 15-12. eg restaurant/ food delivery service eg Amazon; price parity across ebooks

Perfect competition ( 6 points)

Perfect Competition • Firms are price-takers so they decide on how much to produce o Price takers as many many firms o Can't decide the price, given the price they decide how much to produce • Then the maximization problem is given by: max pq − c(q) • Optimal output will be the one where p = MC • Marginal cost curve of a competitive firm is its supply curve • Upward bit of mc curve, gives you supply curve of the competitive firm • Competitive firm: price always equals MC

Lysine cartel case: Lysine market

Perfectly homogeneous product Share of global production of 4 largest manufacturers of lysine in early 1990s > 97% in U.S. > 95% world Purchases of four largest buyers < 30% Large infrequent purchases Cost of a new plant $150+ million (over 3 years to build)

market for cargo air services: Is the air cargo service provided likely to be a homogeneous product?

Providing air cargo services is likely to be a homogeneous rather than a highly differentiated product. There are unlikely to be substantial quality differences across suppliers. Collusion is more sustainable when products are homogeneous. Coordination of pricing is easier for homogeneous goods. Also, in this type of market innovative new products are unlikely to be developed, hence there is unlikely to be a new entrant introducing a new variety which can break the collusive equilibrium. Many of the cartel cases investigated involve industries where the products are quite homogeneous. There types of industries are also often quite concentrated, i.e. contain small numbers of firms, which also aids collusion

repeated interaction: how can factors hinder/ facilitate collusion

Recall: Collusion sustainable for δ ≥ δ* Facilitating factor reduces critical threshold δ* thus the industry characteristics making it easier or more difficult to collude in some markets, do this by moving this critical threshold ( e.g. if it is moved down = harder to collude)

leniency programmes: experiment setup

Repeated differentiated-goods Bertrand duopoly game 1. Subjects can communicate about prices 2. Subjects choose price 3. Subjects can give a secret report to competition authority if cartel 4. Subjects learn competitors' prices and profits and own profits and can give a public report if cartel 5. Detection (cartel detected with fixed probability if no previous detection)

give 2 examples of recent cartels

Smart card chips producers Infineon, Philips, Samsung and Renesas (JV of Hitachi and Mitsubishi) in 2003-2005 (fine €138M in September 2014) Otis, Kone, Schindler and ThyssenKrupp in 1995-2004 in Belgium, Germany, Luxembourg and the Netherlands (fine €832M in Feb 2007). These firms provided lifts and elevators to buildings. Normally bid to get contract, so agreed who would get each before this. Had secret meetings with high up execeutives. in recent years fines for collusion have been increasing a lot

secret price cuts: strategy

Strategy In t = 1 choose p1 = pm If sell in period t, set pt+1 = pm If don't sell in period t, set pt+1 = c for T periods. After that pt+T+1 = pm (periodic price wars) What is the smallest T for which this is an equilibrium?smallest price war; since deviation can now sometimes triggered by low demand; so smallest T which can discipline the cartel Trigger strategy slighty different. If they sold in this period means that other firm didn't undercut. If didn't sell this period; either because rival undercut or because no demand. So set c and we have a price war for a certain number of periods.

secret price cuts: smallest T

Substitute (1) V+ = (1 - α)(πm /2 + δV+) + α(0 + δV-) in (3) V+ ≥ (1 - α)(πm + δV-) + α(δV-) (PV in collusive phase into the IC) (4) δ(V+ - V-)≥ πm/2 From (1) and (2) (PV during punishment phase) (5) V+ = (1 - α)πm /[2(1 - (1 - α)δ - αδT+1)] (6) V- = (1 - α)δTπm /[2(1 - (1 - α)δ - αδT+1)] Substitute (5) and (6) into (4) and simplify (7) 2 (1 - α)δ + (2α - 1)δT+1 > 1 For T = 0 (7) is not satisfied LHS of (7) is increasing in T if and only if α < ½ For α ≥ ½ collusion cannot be sustained because expected gains from future collusion are low For α < ½ collusion can be sustained. Choose smallest T such that (7) is satisfied. T=0 no punishment; this equation is still greate than 1, so punishment necessary for cartel to be stable The LHS increasing in T if and only if a is less than .5 since d less than 1, increasing t reduces this. If a is greater than .5; no way we can satisfy this inequality. The side just gets smaller The probability of no demand is quite high Very likely that there is no demand This implies that there will be price wars in equilibrium since they are triggered by low demand, not undercutting. So likely in recessions. Not triggered by anyone undercutting So slightly unfair punishment but needs to be in place otherwise nothing to deter anyone from deviating

collusion at the extensive margin: model setup

Suppose 2 firms and 2 identical markets. Firms choose which markets to enter. Cartel assigns monopoly in one market to each firm Firm can cheat by entering also the second market The other firm observes entry and can adjust price to entry/cheating (not surprise as price cutting when colluding at the intensive market)

multimarket contact: identical markets model

Suppose 2 identical firms in 2 identical markets Collusion sustainable if and only if (½πm + ½πm)/(1-δ) ≥ (πm + πm) in this case: Get half the profit in each market. if one firm decides to undercut best to do it in both markets and at the same time. Multi market contact in identical markets doesn't matter; more to gain but also more to lose ( profits drop in both markets) both effects exactly outweigh each other and there is no effect. therefore the equation simplifies to δ ≥ ½ Multimarket contact has no effect if markets are identical

secret price cuts: motivation

Suppose firms do not observe rivals' prices If firm has low demand, it is because 1. industry demand is low or 2.rival is undercutting Because now of uncertainty punishment forever cannot be optimal. Price war for T periods. Uncertainty about the reason why the firm is not selling a lot. For any cartel punishment is necessary to discipline the cartel The previous model punishment will not work, period of low demand Price war for finite number of periods.

factors affecting collusion: demand growth

Suppose g = growth rate of demand ½πm(1 + δ(1+g) + δ2(1+g)2 + ...) ≥ πm ½πm/(1- δ(1+g)) ≥ πm δ ≥ δ* =1/(2(1+g)) Collusion is more likely the greater the growth rate of the industry Compare fast growing industries and low growing: Low growth rate of demand, with the same incentive constraint. Makes the critical discount lower; becomes easier to collude. What is driving the gain and the loss? Higher growth; for the immediate gain and loss from future... why does growth make collusion easier; by undercutting the rivals, the firm would lose profits from the cartel level to zero. If g, the cartel level is increasing thus there is more to lose tomorrow. So by being patient all the firms can keep increasing their prices, which may make it easier for firms to collude.

collusion at the extensive margin: is collusion sustainable?

Suppose goods may be differentiated so that πN ≥ 0 where πN = Nash equilibrium profits Collusion is sustainable if and only if πm /(1- δ) ≥ (πm + πN) + δ2πN /(1- δ) LHS: discounted monopoly profits from 'own' market 1st term RHS: monopoly profits from 'own' market, πm, rival observes entry to the second market and responds in pricing, πN 2nd term RHS: rival also enters the second market, both markets are duopolies and cartel has broken down, 2πN IC simplifies to δ ≥ δ* = πN/(πm - πN) Note that δ* → 0 as πN → 0 Collusion at the extensive margin is sustainable for very small discount factors when the industry is very competitive. Therefore: This sort of model helps collusion. Assume differentiated goods; the Nash profits can be positive(otherwise profits would be zero) Now even if they compete they can make profits. Deviation; I decide to enter my rivals market. Now I enter the other firms market. they will adjust so I can get nash profits as now we are competing. Critical discount factor; the nash profits are small or close to zero. Even for very small discount factors, if the market is competive, they can collude. Now it can go all the way to zero and this can be pretty powerful.

factors affecting collusion: size of cartel

Suppose n firms πm/n(1-δ) ≥ πm δ ≥ δ* = 1 - 1/n It is easier to sustain collusion with few firms More firms would make it harder to collude ( looking at same basic incentive constraint each time. each firm will get i/n of the profit in the cartel. If only one undercuts by e they get all monopoly profit and nothing in future. The critical discount factor is increasing in n; the higher the no of firms the more difficult to sustain a cartel. The one off benefit from deviation is higher. The cartel profit is now split between more firms and thus is lower, so by undercutting they can get same withfull monopoly profit. Now the gain is worth more and the loss is lower. Then in becomes more tempting to deviate... more difficult when there are more firms. (In practice n=7 on average) Cartel with many firms.. cooordination is harder eg communication.

factors affecting collusion: regular orders

Suppose unusually large order in one period with profit λπm while later profits remain at πm Collusion sustainable if and only if ½πm(λ + δ + δ2 + ...) ≥ λπm ½πm[λ + δ(1+ δ + δ2 + ...) ] ≥ λπm δ ≥ δ* = λ / (1+λ) > ½ given λ > 1 Collusion is more likely when there are regular orders. in irregular growth periods, some times the growth is high, then flat in future etc.. In one period there is unusually large profit then becomes normal. The firms can enjoy a bit higher profits. Then the normal monopoly profits or zero. This shows that the critical discount factor is greater than 0.5. Collusion more difficult since critical discount higher. INTUITION (IN TERMS OF GAINS N LOSS): the loss is the same, the difference is that today there is more to be gained, bigger pie, more tempting. Irregular orders make it easier.

Market Definition in Printed Media Industry: An Econometric Illustration Table Results

Table 1 First column: • Shows results without taking into account advertising effect on readers demand • Almost inelastic • Second column: • Does take into account advertising effect on readers demand • Elastic Table 2 • Without taking into account advertising effect on readers demand almost looks zero • Much bigger when taking into account advertising • Market definition analysis: If the consumer substitutes for other products, that product is also in the market • Looking at the results, without taking into account advertisements, they're not substituting those products • Elasticity very low • Cross-price elasticity zero or almost zero • When not taking into account 2SM you would define the market as that newspaper o Market for La Vie is La Vie o Because super inelastic and cross-price elasticity is zero • Shows the importance of taking into account the two-sidedness • Thinking about a merger analysis, defined the market of two magazines as two different markets, wouldn't find an effect on market power as say they are different markets • There would be no 'issue' of competition • The excersise shows the importance of considering 2SM

Volvo- Scania case: market assessment from tables

Table showing market shares per country in 1998 • The countries with the red squares: ones they find the merger will likely harm the market •Sweden: merger 91% verses second biggest at 6% • Norway: 70% vs 12% • Ireland: 49% vs • Denmark, second biggest only have 18% whilst merger have 59% • Market shares giving pretty good prediction of what will happen when they merge

The EU Merger Guidelines Market shares and HHI indices: when are competition concerns unlikely

The commission is unlikely to identify competition concerns: o If the post-merger HHI is below 1000 in the market o with a post-merger HHI between 1000 and 2000 and a delta below 250, o with a post-merger HHI above 2000 and a delta below 150

repeated interaction: condition to sustain collusion

The firms can sustain collusion if and only if ½πm/(1- δ) ≥ πm Equation simplifies to δ ≥ δ* = ½ thus the firm can only sustain collusion if discount factor is great enough.. Future punishment is enough to restrain defection from cartel. Compare the paths of profit under the two scenarios. stick to cartel; half profit forever. Deviate; full profit today, but 0 from tomorrow onwards. Look at the gain from deviation. The loss starts tomorrow and drops all the way down to zero. The firm can cheat only once whereas the loss or consequence is long term. (one off gains v long term gain)

Volvo-Scania Merger: Actual Market Power test METHOD

The previous test is conservative in that it underestimates the profitability of joint price increases in two respects 1. It considers a percentage price increase that is the same for all products of the merging partners 2. It ignores responses by competitors. In practice, competitors may respond to a price increase by also raising their prices • An alternative approach is the actual market power test, which imposes more specific assumptions about firm behaviour after the merger • It is assumed that firms continue to set prices non-cooperatively after the merger. • To implement the actual market power test, they numerically compute the post-merger Nash equilibrium, with a modified ownership structure, and compare this to the pre-merger Nash equilibrium • The test calculates what the new price they will be charging is and what its effect will be on consumers given their demand curve

market for cargo air services: How frequent are interactions between air cargo suppliers and customers (i.e. companies buying air cargo services) likely to be?

There is a large number of customers relative to suppliers of the air cargo services, and with a large number of flights per day there are likely to be frequent interactions between buyers and sellers in this market. More frequent interactions facilitate collusion because any deviation will be punished more quickly. quite frequent; frequency helps collusion as punishment comes quickly

conclusion

To conclude, multimarket contact facilitates collusion unless markets are identical, which is unlikely to ever occur in reality. As markets are non-identical there will always be one where collusion is easier to sustain. Although collusion may not be sustained in the second market, when there is multimarket contact the slack from this market can be transferred to the second market which facilitates a cartel. It is shown that there are different incentives to collude depending whether the firm is at the intensive or extensive margin. Factors such as competition and barriers to entry have opposing effects. They hinder collusion at the intensive margin because there is less to gain from collusion because the cartel is only operating in one market. However they make collusion easier at the extensive margin because the separate markets mean there is both more to gain from being in the cartel and more to lose from deviation. On the other hand, there are also factors which have the same incentive at either margin such as growth. An industry with a growing rate of demand means there are more profits to be gained by sustaining collusion. This holds whether a cartel is operating within a single market or firms hold their own monopolies in separate markets.

leniency programmes: experiment treatments

Treatments: 1. No competition policy (no fines) 2. Fines 3. Leniency (and fines)

Two-sided Markets

Two-sided Markets • If we compare the two pricing formulas for monopoly platform and two-sided single homing, we see that in the competitive case platform puts more emphasis on the external benefit • The reason is simple: o When monopoly platform ↑ its price, the agent disappears from the market o When duopoly platform ↑ its price, group-1 agent joins the rival platform and it becomes harder to attract group-2 agents

UPP in 2SM Formal derivation UPP on advertising side:

UPP on advertising side: SEE CARDS FOR FORMULA • Don't need to remember formula • First and third term are usual ones, UPP without network externalities • The second term: diversion ratio from firm 1 and 2 that occurs as a result of change of the advertising quantity on the readers side (timsed the margin) o The number of people who would switch to read newspaper2 because there are less advertisers on newspaper 2 o The gain that the second firm would have on the readers side as a result of change of price on the advertising side in the first firm o This is firm 1 to 2 from advertising side to readers side • Last term: considering the effect of having less readers as a result of a change in the number of advertisers that may have negative effects in terms of cost on the first firm o Lost readers for firm 1 means cost increased, lose economies of scale o This is in the same firm, the effect of change of quantity on advertising side on the readers in the same firm • Where: see cards • Note that the last diversion ratio shows the effect of price increase of one side on the other at the same platform

secret price cuts: incentive constraint

V+ = present discounted value from date t in collusive phase (1) V+ = (1 - α)(πm /2 + δV+) + α(0 + δV-) V- = present discounted value in punishment phase (2) V- = 0 + ... + 0 + δTV+ Incentive constraint (3) V+ ≥ (1 - α)(πm + δV-) + α(δV-) Last period both firms charged monopoly price so on collusive path; from tomorrow onwards; still on the collusive path. With prob alpha there is no demand. So they get 0 today and punishment starting tomorrow. Incentive constraint; if that is more than deviating from the cartel ( 1-a) there is demand and undercutting firm gets all of it, but from tomorrow onwards they get punished from tomorrow onwards. If subsititute; the gain from deviation vs. the loss The gain; temptation; increase profits from half to full Punshment; the discounted profits reduced from V+ to V- since moved from collusion to punishment path.

Appreciability

Whether a firm has appreciable effect on competition

multimarket contact: non identical model, is it sustainable?

With multimarket contact collusion sustainable in both markets if and only if ½πm/(1-δ) + ½πm/(1-δ(1+g2)) ≥ (πm + πm) Equation simplifies to 1/(1-δ) + 1 /(1-δ(1+g2)) ≥ 4 ⇔ 4.17 ≥ 4 With multimarket contact collusion sustainable in both markets Multimarket contact allows the firms to transfer the slack in incentive constraint in market 2 to market 1 Empirical evidence for airlines (Evans and Kessides (1994) QJE) and US mobile phones (Parker and Röller (1997) Rand). Airlines and mobile phone examples regulator in America gave only two contracts per region, if same players and had multimarket contact- prices higher... multimarket contact will increase prices.

secret price cuts: model setup

With probability α no demand, with probability (1 - α) demand is D(p), i.i.d.( every period a new one is drawn) Firms do not observe state of demand or rivals' prices

market for cargo air services: Is the market characterised by multimarket contact?

Yes. Each airline will be supplying cargo services on a number of flight routes. Many markets, many countries) Multimarket contact facilitates collusion if markets differ. Incentive constraints are pooled and any slack in IC in one market can be transferred to another market.

Two-Sided Markets: platform dating agency

[platform] [side 1]<===externalities===> [side 2] • Platform: dating agency • Two sides, e.g. men and women who want to join • Externalites, network externalities • Platform is exploiting the network externalities between two sides when it is deciding on its pricing

industrial economics

analyses when firms have market power, how they maximise profits, and new market forms. How should firms play differently than in standard markets. They can also end up abusing market power- comp policy needed to restrict this behaviour that can be harmful to consumers

diagram for perfect competition

but there is not two diagrams upward part of MC= supply curve

Quantifying cannibalisation: what is the diversion ratio

e.g. D12 or D21 The second term is the diversion ratio from Product 1 to Product 2 at pre-merger prices, D12 o Shows how much the quantity of product sought for firm 2 changes when the quantity of product sought for firm 1 changes o Increase price, loose some customers, what percentage of these customers are now buying from firm 2 o Hypothetical post-merger price increase for product 1 would reduce the demand for good 1 by some number of units. A fraction of the former product 1 purchases would turn to product 2. *The incremental product 2 unit sales divided by the 'lost' product 1 units is the diversion ratio*

ways of forming a cartel

explicit; actually meet and agree Mutual understanding; cut throat comp will harm everyones profits. May be possible to form a cartel without explicitly communicating

market for cargo air services: How easy is it likely to be for new carriers to enter this market? ie are barriers to entry likely to be high or low?

high entry barriers; helps collusion as an entrant would break the cartel There might be quite high barriers to entry if a new provided needed to enter on a large scale in order to compete effectively — e.g. a new entrant might need to establish a frequent service on a number of routes. Access to take-off and landing slots at airports might also constrain entry. On the other hand, it might be fairly straightforward for an existing supplier of air passenger services to expand into the air cargo services market. High entry barriers facilitate collusion. New entrant might undercut and destabilize the cartel. Furthermore, collusion is more sustainable with smaller number of firms in the market

static model of cartel behaviour: model setup ( 5 points and summary)

in the static model (firms compete once) Homogenous good duopoly Firms set prices simultaneously (Bertrand) Marginal cost is constant c (no capacity constraints) In static Bertrand model competition drives p = c and firms earn zero profit summary: there are 2 firms, they compete so that price goes down to marginal cost. Firms make zero profit. Everybody has same constant MC. If anyone charging higher than cost, another firm could undercut them slightly. Since they are producing a homogenous good, the lower price firm would get all the customers, by undercutting. Then, obviously the other firm would undercut and they keep undercutting each other until they hit the floor which is the cost. Thus this static model doesn't look good for firms.

The EU Merger Guidelines Market shares and HHI indices: SIX special circumstances that may invalidate the HHI as a proxy for change in competitive conditions

o A merger involves a potential entrant, or a recent entrant with a small market share ♣ It's a potential and cannot really use HHI ♣ New entrant: very little market share as its not really established ♣ HHI give misleading results o One or more merging parties are important innovators in ways not reflected in market shares ♣ Even though HHI conclude happen, innovators can rapidly increase market share with merge R&D activities ♣ In the end they may have too much market power o There are significant cross-shareholdings among the market participants ♣ Only interested in each others profits ♣ By just looking at market share doesn't give any more information than pre merger case ♣ Before, already acting to increase their profits o One of the merging firms is a maverick firm with a high likelihood of disrupting coordinated conduct o Indications of past or ongoing coordination, or facilitating practices, are present ♣ Even if you say the merger won't be harmful, they are coordinating already ♣ Not giving new information ♣ Say the market share too high when they merge, but they are already acting together o One of the merging parties has a pre-merger market share of 50% or more ♣ No point using HHI ♣If you calculate just this one it will already be too high

tacit coordination conditions

o Each firm should be able to monitor the other (transparency) o Firms should have incentives to co-ordinate over time (sustainability) o Reactions of competitors or costumers should not jeopardize the results of common policy

Further issues with UPP: Firms controlling multiple products prior to the merger

o If Firm A owns just Product 1 and Firm B owns substitute products n = 2, ..., N, the cannibalization term should be: N T 1 = D 1 n ( P ̄ n − C ̄ n ) n=2 o Should be careful if one product is substitute and the other is a complement to firm A's product o Additional sales may also generate intangible benefits (or costs) as well as direct and readily quantitative net receipts o E.g.: sales of spare parts, complements, markets with network effects, etc. o Not considering the other profits or benefits that the firms might have, e.g. after sale services ♣ Didn't increase prices but charging too much after sale

The EU Merger Guidelines: SIX Factors which may induce non- coordinated effects in mergers

o Merging firms have large market shares ♣ New firm will have an even larger market share and maybe high market power o Merging firms are close competitors ♣ Prices might increase as a result o Customers have limited possibilities of switching supplier ♣ E.g. only 3 firms, the 2 firms are charging high prices but not able to move to the other one o Competitors are unlikely to increase supply if prices increase ♣ If they have capacity constraints ♣ After the firms merge, the competitors cannot decrease prices to get their customers because they can't increase their quantity any more ♣ As a result competitor will also increase price because then can make more money with same output o Merged entity able to hinder expansion by competitors ♣ Competitor can't expand or compete properly o Merger eliminates an important competitive force

Anti-Trust Issues Raised by 2SM Predation • U.S v. Microsoft Example

o Microsoft gave away a software tool to make it easier for Internet Service Providers (ISPs) and corporate IT departments to customize Internet Explorer o Netscape sold a similar kit for $1, 995 o Microsoft had a two-sided justification: it give away the tool to increase ISP demand (one side of the market) and provide other features to increase the demand for Windows (which included Internet Explorer) among users

conditions that can increase a buyer's bargaining strength

o The buyer is well-informed about alternative sources of supply and could switch easily o The buyer could commence production of the item itself or sponsor new entry by another supplier o The buyer is an important outlet for the seller o The buyer can intensify competition among suppliers through establishing a procurement auction or purchasing through a competitive tender

SSNIP Test for Two-Sided Markets: Should then the hypothetical monopolist be thought of as raising:

o The price level keeping fixed the price structure? E.g. keeping same ratio of say 1:2 o The price level adjusting optimally the price structure? o First one of the two prices keeping the other fixed and then the other price keeping the first fixed? o First one of the two prices and then the other price, each time adjusting optimally the price structure?

outline the pricing regulations for hotels

rate parity for different OTA's Hotels: 1. can now offer different prices for different OTAs (wide MFN; cant offer any better to anyone else= lowest price) 2. cant offer lower prices on their own website (narrow MFN) 3. can offer lower prices in "offline channels"

Price elasticity of demand

shows the change in unit output given a change in price

Lerner index

the margin = the inverse of elasticity • Markup and elasticity relationship, the more elastic a product is, the less able you are to price above the cost • If it is more elastic, inverse elasticity will be a smaller number, less able to price over the cost • If your product is highly elastic, change price a little, change in quality will be high, makes you less able to charge prices that are high • If it is less elastic then the inverse will be larger and can charge more above the MC to obtain it, divide the PED FOC's by price

why is it important to define the market correctly?

to assess whether it is behaving correctly. see what the market is defined in is it a set of products or is it a geographical location? e.g. is it in the US or UK

collusion; types?

when firms get together and agree to increase the price or restrict quantity or share markets in order to increase profits Collusion can be explicit or tacit Explicit collusion results from secret agreements Collusion can result from tacit agreements where no direct communication but mutual understanding

factors affecting collusion: demand growth and entry

with just demand growth, assumed that the number of firms remains fixed However, growing markets can attract entry For example, Vasconcelos (2008) JEMS finds (in a Cournot setup) that when growth is fast, collusion can be sustained even with entry when growth is slower, incumbents may deviate just before the optimal entry date to delay entry

Volvo-Scania Merger: Econometric Model-Consumers and Demand and the probability

δj =xjβ−αpj +ξj • pj is the price of product j and xj is a vector of observed characteristics of product j • ζig and (1 − σ)ij have extreme value distribution • Mean utility for outside good is normalized to 0, δ0 = 0 • Each potential consumer i chooses product j that max his utility. • The probability, sj that a potential consumer chooses product j: sj = exp(δ /(1 − σ)) D1−σ jg D 1+ D1−σ Dg = exp[δk/(1−σ)] k ∈Gg • At the aggregate level, the choice probability sj coincides with the market share of product j • Don't really need the algebra • Only thing you need to know: this mean utility comes up as the part of the demand curve o If you find alpha as positive: i.e. negative alpha then price increases demand decreases • Only thing you need to remember: this bit appears in the demand curve

Estimating Market Power in 2SM Estimation-Readers' demand

• (skipped) technical detail • The vector of product characteristics includes dummies for the issue of the weekly magazines that are sold together with the newspaper (both a dummy for the day of the week of issue and a general dummy for their presence), dummies for the presence of a website, for games with prizes played on the newspaper and for changes of editors, the number of local sections for each newspaper and monthly dummies to capture seasonality in the data. • Market shares and prices are endogenous • Instruments used: Sum of characteristics of other firms. ex: sum of number of games, sum of general information supplements, etc. • They are going to be correlated with p and within market shares through the condition for profit maximization but are exogenous to the single firm decision

UPP in 2SM: Differences between one-sided UPP and two-sided UPP

• 1) There are two relevant formulas in two-sided markets, one for each side • 2) Each two-sided UPP formula has more terms than a one-sided UPP one • This is because the pricing decisions on one side also affect priftis earned on the other side of the market • 3) The diversion ratios used in the two-sided UPP formulas also account for the feedbacks between the two sides of the market

3 factors determining the structure of prices offered to the 2 groups

• 1. Relative size of cross-group externalities o If a member of group one exerts a large positive externality on each member of group 2, then group 1 will be targeted more aggressively o E.g. a nightclub with the two sides men and women o If women exert more of an externality, man benefits more from interacting with women the nightclub may let woman in for free o This attracts the women and in turn this attracts the men who then get charged o E.g. magazine industry, advertisers benefit more from reaching the readers and that's why they pay above marginal cost • 2. Fixed fees or per-transaction charges o If an agent pays only for successful transaction, then she will not care how well the platform performs on the other side E.g. only pay to the magazine if get business, won't care how the magazine is doing on the readers side o Cross-group externalities are weaker with per transaction charges • 3. Single-Homing vs. Multi-homing o Single-homing: only using one platform, only buying 1 platform, e.g. only buying guardian o Multi-homing e.g. buying guardian and independent

what three features determine market competitiveness

• 1. The degree of intra-market rivalry o In the case of Dell, any competitors to Dell or Toshiba just following prices Dell setting o Really competing or agreement, not explicit, to follow prices o Real competition: won't be able to set prices that high • 2. The extent of buyer (or supplier) power o Few buyers that are strong, can organise and act strongly against the seller they can restrict the seller's actions o They can restrict the seller to act anti-competitively • 3. The state of entry o Potential entrants can mean other firms in market cannot really abuse market power or act anti-competitively

why may a firm with market power harm competition

• A firm with the market power may harm the competition by: o Weakening existing competition o Raising entry barriers o Slowing innovation

UPP: Merger efficiencies

• A merger can lower the costs • If it ↓ mc, then it can reverse upward pricing pressure measured by T1 • T1 is a virtual mc so it is comparable with mc efficiencies • Assume that product 1 is credited with default efficiency equal to E1C1 and product 2 is with E2C2 *Ei should not be thought as narrowly an efficiency parameter* *It can give a threshold in terms of upward pricing pressure, above which mergers are subjected to further scrutiny* *The efficiencies give the pressure to decrease the prices* *T1 is the upward pricing pressure and E1 is the downward pricing pressure* • Need to look at the difference o If the difference is positive, the prices are likely to increase and shouldn't let the merger happen

"Upward Pricing Pressure in Two-Sided Markets", Working Paper by Pauline Affeldt, Lapo Filistrucchi, Tobias Klein Original Formulation

• A merger changes the first order conditions in two ways. 1. Upward pressure on prices due to the loss of competition between the merging parties products o Not competing anymore, so can increase prices as won't lose customers because in the end we are same firm 2. Downward pressure on prices caused by merger-related efficiencies (marginal cost decreases) o Push costs down so pressure to decrease prices • The difference between these two effects is UPP UPP1 = D12(P2 -C2) - E1C1

Kaiser & Wright (2006) Introduction

• A model of competition in magazine market • Take into account the two-sidedness of the industry • Use data on German magazines • Adapt 2-sided single-homing model of Armstrong (2006)

Rochet-Tirole (2003) Monopoly Platform Benchmark-Private Monopoly

• A private monopoly chooses prices so as to maximize total profit: • Assuming that DB and DS are log concave, and so is π, the FOCs: • In particular: (DB)'DS = DB (DS)' ⇒ The volume impact of a small (absolute) variation of prices has to be the same on both sides • Introduce the elasticities of quasi-demands: • then the private monopoly prices can be characterized by a two-sided market formula which is similar to Lerner's formula: • The total price that they are receiving minus the cost is given by the price over the elasticity

Volvo-Scania Merger: Comparative Market Power test METHOD

• A relevant question that is currently ignored in the merger decision process is what will happen next. • For example, if claims of cost-savings in the form of returns to scale are valid, it seems reasonable to expect that other firms with similar scale may propose further mergers. • The merger between Volvo and Scania would trigger an additional regional merger, between Renault and Iveco, who have stronger positions in the southern parts of Europe. • This regional merger scenario is compared with an alternative scenario of two pan-European mergers. • This scenario assumes that when the Commission blocks the VolvoScania merger, Volvo would seek another partner, Renault, and subsequently Scania and Iveco will merge

Estimating Market Power in 2SM The Model-Newspaper Demand Demand function

• Almost exactly the same as before, demand for readers SEE CARDS • Beta: effect of observed charac, alpha: effect of price on demand • Didn't include number of advertising pages because insignficant • Where: SEE CARDS sN : market share of newspaper i at time t I sN 0t : market share of outside good I xN it : observed characteristics I N it : unobserved characteristics I pN it : cover price I sN itjg : share of newspaper i in group g at time t.

UPP in 2SM Assessing Hypothetical Merger between Dutch Daily Newspaper Publishers

• Apply UPP to a hypothetical merger in the Dutch daily newspaper market. • Focus on whether conclusions change when the correct two-sided UPP formulas, rather than the one-sided ones, are used to assess the merger • Only look at the UPP • Table 3: All negative in the first half, not suggesting a price increase with one-sided UPP, mostly negative below too • Table 4: negative values mostly change to positive • When you don't take into account 2SM you may have conclusions that are not correct and let merger happen although you shouldn't have let it happen

Estimating Market Power in 2SM Estimation-Readers' demand Table results

• As expected, effect of cover price is negative o Increase price demand will increase o Statisitally signficant • Generalist magazine, demand higher, day magazine higher • Woman magazine demand little higher • Games demand higher • Website, demand is lower o Some people will read on website • This is for readers demand

•Rochet-Tirole (2003) Monopoly Platform Benchmark

• Assume that o End-users incur no fixed cost o Platform pricing is linear • Good representation of credit card market: • When you buy a good from a merchant, pay your issuer (bank of customer) price of the good plus a little bit o Interest or credit card fee • The issuer pays the merchants bank the price of the good minus the interchange fee • The acquirer (merchants bank) pays back to the merchant the price of the good minus the fee • Economic value is created by "transactions" between pairs of end-users : buyers (B) and sellers (S) • Buyers are heterogeneous in their gross surpluses associated with transaction, bB • Sellers are heterogeneous in their gross surpluses associated with transaction, bS • Transactions take place in platform which has a marginal cost of c ≥ 0 • Ex: payment cards-a "transaction" takes place if and only if the buyer pays by card instead of using another payment instrument (cash) • Benefits bB and bS correspond to differences in utility of buyers and sellers when they pay by card rather than cash • In the absence of fixed usage costs and fixed fees, the buyers' (sellers') demand depends only on the price pB (respectively, pS) charged by the monopoly platform • There are network externalities in that the surplus of a buyer with gross per transaction surplus bB, (bB - pB)NS, depends on the number of sellers NS o Affected by the number of sellers positively • The buyers' "quasi-demand function": • If buyers benefit greater than the price paying that gives the demand and also related to number sellers, only get benefit if enough sellers on platform • The sellers' "quasi-demand function": • Assuming that bB and bS are independent, the proportion (volume) of transactions is equal to: DB(pB) × DS(pS) o If DB(pB)=0.5 50% of the market is entering the platform o DS(pS)= 0.5 50% of the sellers are entering the platform o The probability they interact is 0.5 x 0.5 • Quasi demand: not really demand because they can't demand it without sellers on the platform • Transaction only takes place if both on the platform

Price Wars in 2SM The Model

• Assume that firms play a non-cooperative two-stage game in which: o 1st stage: simultaneously chose their optimal locations on the political line o 2nd stage: simultaneously chose both prices and advertising rates • We assume four newspapers, the Guardian; the Independent; The Times; and the Daily Telegraph are differentiated on the political unit-line from left to right according to common consensus o Same news but different political angles o Not covering this article: because of two-sidedness firms choosing middle, don't want to look like they have strong views o Locate in the middle, more readers so they can increase advertising revenue o Important in this industry to consider position on political line

Volvo-Scania Merger: Econometric Model-Pricing

• Assume: Each firm f produces a set Ff of products • Its profits are given by the sum of its operating profits for each product minus fixed costs K o Price - mc times the quantity o Mc assumed to constant over quantity πf =(pj−cj)qj−K j ∈Ff • Maximise profit and • Multi product Nash equilibrium is given by a system of J FOCs: (pk−ck)∂qk +qj k∈F • Similar to simple monopolist case • Use these equations to get mark-up expression • This is equal to zero • qk comes from the demand curve we are estimating • So then we know the derivative • We don't know the mc we assume it has a specific form • Also estimating marginal cost given this relationship • In the end have a demand a supply curve, estimate them together cj =exp(wjγ+ωj) mc for each product j is constant and depends on a vector wj of observed characteristics of product j, and unobserved characteristics, ωj

Anti-Trust Issues Raised by 2SM Barriers to Entry

• Barriers to entry is discussed both in market definition and market power • In market definition it is considered to see whether it is possible for other competitors to come into the market • In market power analysis, it is considered to determine whether the incumbent can exclude competitors and thus maintain higher prices • In this case, just because of the structure of the industry cannot conclude there are entry barriers • As long as the entrant has sufficient funds to enter the industry they can still enter • In these markets, even with network externalities there can be success of entries • For a two-sided market to exist or be successful the platform needs to 'get both sides on baord' • Therefore, a platform having a large market share on one side is not a barrier to entry • Positive feedback effects do not prevent entry • The relevant question to ask is whether the incumbent has special advantages that couldn't be replicated by entrants • There are many examples of successive entries to two-sided markets

Two-Sided Markets Intro

• Benefit enjoyed by a member of one group depends upon how well the platform does in attracting customers from the other group • Ex: a heterosexual dating agency or nightclub can do well only if it succeeds in attracting business from both men and women • Interdependence of the two sides to enter the platform ⇒ Two-sided network externalities o E.g. players would prefer to have a games console which has many games and developers would prefer to have a games console which has many players • Pricing of the platform: Internalize the two-sided network externalities o E.g. sometimes readers of magazines paying below marginal cost and magazines make a loss but advertisers are profit makers

Anti-Trust Issues Raised by 2SM Predation

• Business strategies and their effects on consumers must be evaluated with respect to both sides of the market has important implications for the analysis of predation • It may be privately and socially optimal for prices on one side of the market to be below any possible measure of cost on this side

results from cereal industry example

• C3: shows market share of top 3 firms, 75%: really high • However high market shares may not induce high market power but the conditions of the industry • If you are going to SSNIP to define the market for a merger for one of the segments • You start with the smallest bundle, just two of the factors if the 10% increase is not profitable, you add another etc. • Use table of elasticities • The first number shows the % change in quantity in cornflakes when the price of cornflakes increases • The second number shows the % change in quantity in cornflakes when the price of frosted flakes increased • The number is high: they are substitutes

Collective Dominance

• Collective dominance: 2 or more legally independent firms adopt a common policy on the market o The links may be structural o They adopt a common policy ♣ E.g. all produce at medium quality which is cheaper o tacit co-ordination

ways to measure market shares (5)

• Data provided by firms • Data provided by trade associations • Gathering info by asking customers • Market research reports • Sales data by value or by volume

repeated interaction: monopoly price equilibrium 4 points

• Discounted payoff from colluding ½πm(1 + δ + δ2 + ...) = ½πm/(1- δ) where δ = discount factor, 0 < δ < 1 • By undercutting firm 1 can earn πm in one period • But from next period on the rival will charge p = c according to trigger strategy • If firm 1 deviates, its discounted profits are πm + 0 I can't do anything better than go equally low; in the future my profits will be zero. THUS is collusion better than deviating since the Discounted payoff from colluding forever vs. cheating today.

why define markets?

• Do firms have large market shares? • Do they enjoying a high market power and set prices above competitive level? • Is a merger going to be harmful for the relevant market? • To answer these questions, first tool we look is the market shares which cannot be obtained without identifying the market • On the other hand, although a firm have large market share it may not be enjoying its market power because its customers may switch to other suppliers • So, the real issue is not what the market is, but how competitive it is (or how a merger affects competition)

how to use entry barriers to asses potential competition

• Entry barriers are important to assess potential competition: • Potential competition → ↓ market power • A firm with large market share + low entry barriers → unlikely to have market power • A firm with large market share + effective entry barriers → likely to have market power

in the example of the cereal industry, what was the strategy with which the market was defined

• Estimate demand function which is a function of observed product characteristics, heterogeneous consumer preferences and unknown parameters • Use the estimated demand system to compute PCM implied by: o single product firm → product differentiation o few firms with many products each → multi-product firm effect o multi-brand monopolist producing all brands → collusion • Compare the PCM predicted by them and the PCM observed

Volvo-Scania Merger: Econometric Model-estimation

• Estimate equations sj =exp(δ /(1 − σ)) D1−σ jg D 1+ D1−σ ggg (pk−ck)∂qk +qj k∈F • Using δj =xjβ−αpj +ξj cj =exp(wjγ+ωj) Put observable characterics and estimate first equation, get alpha and beta hat • Can see what will happen to demand when price increases

Estimating Market Power in 2SM Estimation-Advertising demand

• Explanatory variables for advertising demand: newspapers prints, the readers average age, the share of male readers and the difference between the own real advertising price and the average advertising price across all daily newspapers not included in our sample • Newspaper circulation and ad price are endogenous • Instruments used: Sum of characteristics of other firms: sum of readers' average age, sum of proportion of male readers, number of free copies distributed by each newspaper TABLE: • Price increases, demand for advertising increases • Number of circulation increases, advertising demand increases • This captures the two sided effect • Number of prints is important for advertisers demand • The more readers there are for the newspaper, the more likely you will advertise as can reach more people • (Advertisers don't like male readers or old readers)

Hotelling Model: Derivation of Demand

• Find the demand by finding the marginal consumer who will be indifferent between buying from the first shop and the second shop o The two utilities will be equal • If prices are not too high, there exist a consumer with location x˜ who is indifferent buying from shop 1 or shop 2 o Anyone on side 1 of x will go to shop 1 and other side will go to shop 2 o This will be the demand p1 + t~x = p2 + t(1 - ~x) ) => ~x(p1, p2) = (p1 -p2 + t)/2t D1(p1, p2) = N~x D2(p1,p2) = N(1 - ~x) • where N is the total number of consumers • If price difference between 2 shops exceed t: D1(p1, p2) = N if p1 ≤ ξ - t D1(p1, p2) = N(ξ - p1)/t if p1 > ξ t, market is "NOT COVERED" • Can think about how far from the shop you are as how much you prefer that product

Monopoly

• Firms choose price to maximize their profits o As only one firm • So the maximization problem is given by: max π = pD(p) − cD(p) • Optimal price is the one that sets MR=MC

Anti-Trust Issues Raised by 2SM Market Power

• For market power analysis we also need to take both sides into account • Does the firm under consideration price above marginal costs?-to answer this question we need to check whether the total price is above marginal costs o If you considered separately, could say abusing market power o But could be due to structure of market, e.g. network externalities • As in the magazine industry example, one side can be loss leader while the other is the source of profits o If don't consider both sides could say one is abusing market power and the other is predatory pricing, but this isn't the case • Need to check if the total price is significantly above marginal costs

Estimating Market Power in 2SM, Case of Newspapers conjectures and situations

• Formulate two alternative conjectures: o Oligopolistic competition o Collusion in each of the two sides of the market. • This leads to four possible situations: o 1- The firms compete on both markets, o 2- The firms jointly maximize profits on both markets, o 3- The firms jointly maximize profits on the readers' market but compete on the advertising market o 4- The firms compete on the readers' market but jointly maximize profits on the advertising market • Estimate the model and calculated the mark-ups from the four different cases

Kaiser & Wright (2006) Data

• German magazine data from 1972-Q1 to 2003-Q4 • Data selection for model • Magazine groups used: Health, Entertaining, Photo, DIY, Food, Gardening, Monthly-high-priced women's, Weekly women's, PC

Two-Sided Market vs. Super Market

• Hagiu (2007), Merchant or Two-Sided Platform? • The differences between intermediaries and two-sided platforms • "merchant model": buying from sellers and re-selling to buyers o E.g. supermarkets • "2-sided platform model": enabling affiliated sellers to sell directly to affiliated buyers SEE DIAGRAM • Amazon is now a case in between, matches you with sellers but also sells own products • Pure merchants take the possession of sellers' good and take full control over their sale to consumers o Example: Walmart • Pure 2SP, leave that control entirely to sellers and simply determine buyer and seller affiliation with a common market place o Example: eBay o Control over strategic variables (pricing, advertising, distribution, etc.) The seller chooses the price not eBay The seller distributes the item not eBay o Sharing of economic risk (who bares the risk?) Seller can't sell that is their problem not eBay's o Ownership of buyers (which one is important? sellers' brands or platform's brand) May be going to Waitrose because prefer Waitrose, but go to amazon to buy a book

Hotelling Model ( 11 points)

• Hotelling 1929 • Model of differentiated products • "linear city" of length 1 • Consumers are distributed uniformly along the city • 2 shops located at the end of the city, both sell the same physical good. • Shop 1 is located at x = 0 and shop 2 is located at x = 1 • Consumers incur transportation cost of t per unit of length • Consumer have unit demand, they consume zero or none • p1 and p2 are the prices of good sold by shop 1 and 2 respectively. • "generalized price" of going to shop 1 for a consumer with coordinate x is: p1 + tx o for shop 2: p2 + t(1 − x) • s¯: surplus enjoyed by each consumer o Doesn't change by shop as same amount of surplus from same good • Utility of a consumer located at x is: o s¯− p1 − tx if she buys the good from shop 1 o s¯− p2 − t(1 − x) if she buys the good from shop 2

•Rochet-Tirole (2003)

• How the price allocation between the two sides of the market is affected by o Platform governance (for-profit vs. not-for-profit) o End users' cost of multi-homing o platform differentiation o platforms' ability to use volume-based pricing o The presence of same-side externalities o platform compatibility • It also investigates how privately optimal pricing structures compare with socially optimal ones

reasons why buyer power may not always benefit the final consumer

• If only some of buyers are powerful, the supplier may harm the downstream competition by charging high prices to weaker buyers • It may be weakened as a result of agreement or behaviour under investigation • If the buyer power has also market power as a seller in the downstream market, it may not pass on lower prices to the final consumer • Conduct by a dominant buyer may harm competition

Volvo-Scania Merger: Comparative Market Power test RESULTS

• If there will be two regional mergers: • Consumer surplus always decrease • Same with two pan-European mergers • Change in total welfare • Decreasing both two regional mergers and two pan-European mergers • Whatever analysis you use, all the results show you shouldn't let the merger happen because it won't be good for society/total welfare

Two-Sided Single-Homing solve FOCs

• If we solve FOCs: • is the unique equilibrium. • Thus the platform will target one group more aggressively than the other if that group is: o on the more competitive side and/or o causes larger benefit to the other group • Targeting more aggressively, charging smaller prices • Being on the more competitive side means t is smaller o T captures the competitiveness • In terms of elasticities: • These reflect the fact that now there is a competitor, for each agent we are losing on one side we are more likely to lose agents on the other side too • The competition is much stronger • (don't need to remember all the formulas, just need the intuition)

Volvo-Scania Merger: Hypothetical Market Power test RESULTS

• If you increase the price by 5% after the merger, the profits are increasing o In denamrk, finaland, Ireland, Norway, Sweden, those with large market share • Price increase by 10%, in these countries this is still profitable • Looking at 25% increase, unprofitable for all except Sweden (market share was 91%) • If we let them merge they can easily increase price by 5 or 10%

SSNIP Test for Two-Sided Markets: what can you distinguish?

• In a two-sided market one can distinguish: o The price level (roughly, the sum of the two prices) o The price structure (roughly, the ratio of the two prices)

Estimating Market Power in 2SM, Case of Newspapers: Journal of Applied Econometrics by Elena Argentesi & Lapo Filistrucchi Introduction; what does optimal publisher behaviour depend on?

• In general, the publishers' optimal behaviour would depend on four different elasticities: o The elasticity of readers' demand with respect to cover price o The elasticity of readers' demand with respect to the quantity of advertising Capturing network effect o The elasticity of advertising demand with respect to advertising prices o The elasticity of advertising demand with respect to newspaper circulation Capturing network effect

UPP test: Net pricing presssure

• In its pure form the test checks whether the merger causes "net" upward pricing pressure • If the merger creates net UPP for product 1, basic economic theory unambiguously predicts that the price of Product 1 will rise, holding fixed other prices and products • With two single-product, Bertrand price-setting firms, the merger creates net upward pricing pressure for Product 1 if: D12(P ̄2 − C2) > E1C ̄1 (1) • Note that C2 is written as the post merger cost • Need to use the cost we know • Crediting Product 2 with its default efficiencies implies that C2 = (1 − E2) in which case inequality (1) becomes D12(P ̄2 − C ̄2(1 − E2)) > E1C ̄1 (2) • When the efficiency increases, the cost decreases, so gross margin increases • In inequality (2), greater default efficiencies for Product 2 cause more upward pricing pressure for Product 1 • It has the unattractive property that it seemingly could flag a merger for further scrutiny because of credited efficiencies • Keep the pre-merger gross margin • Simpler formula for upward pricing pressure for Product 1: D12(P ̄2 − C ̄2) > E1C ̄1 (3) Using inequality (3), we define net UPP on product 1 as: UPP1 = D12(P ̄2 − C ̄2) − E1C ̄1 ( in the eons the line is meant to be above the number and the number is a subscript) • A merger causes upward pricing pressure for Product 1 if UPP > 0 o Likely that prices will then increase

Estimating Market Power in 2SM, Case of Newspapers: Journal of Applied Econometrics by Elena Argentesi & Lapo Filistrucchi Introduction

• In order to measure market power, it is necessary to compute price-cost margins taking into account the two-sided nature of this market o Analysing the observed mark-up o If found mark-up of 50% with collusion, 30% for competition from our model and observed is 52%, can say probably colluding o Checking if observed mark-up is in confidence interval of collusion or competition

Kaiser & Wright (2006) Empirical Specification: Instruments

• Instruments used: (l = log) • I avr. cover price of a publisher's other magazines • I avr. ad rate of a publisher's other magazines • I avr. content pages per copy of a publisher's other magazines • I avr. circulation of a publisher's other magazines • I avr. advertisers share of a publisher's other magazines • I ln(number of magazine titles published by own publisher in a given year) • I ln(number of pages printed by the own publisher) • (skipped)

"Price Wars in Two-Sided Markets: The Case of UK Quality Newspapers", NET Institute Working Paper by Stefan Behringer & Lapo Filistrucchi Price Wars in 2SM Introduction

• Investigate the price war in the UK weekly quality broadsheet newspaper industry in the 1990s which was portrayed as a case of presumed predatory pricing • Predation might in fact employ a wide set of strategic instruments, e.g.- product differentiation. • One can harm a competitor in the newspaper industry by the political location of the newspaper

The EU Merger Guidelines Market shares and HHI indices: problems with the contradiction avoiding approach based on concentration

• It may not work well with the markets with differentiated products since defining the market may be problematic. • It does not give information on post-merger dynamics • Example: The merger proposed in 2007 between Whole Foods and Wild Oats o Chains of grocery stores specialized in natural and organic food o FTC defined the market as "premium natural/organic supermarkets" o With that market definition they find that merger will increase concentration too much o The customer can also buy organic groceries from traditional supermarkets o So market definition inquiry do not directly answer the question if the merger is anti-competitive

Armstrong SW maximizing outcome

• Let νi(ui) be the aggregate consumer surplus of group i = 1, 2 • νi(.) satisfies the envelope condition: • Then welfare is: • Total welfare is given by the profit of the platform, the consumer surplus that the reader's are getting and the consumer surplus the advertisers are getting • Welfare maximizing outcome is: • prices will be given by: • Cost of the platform, minus the adjustment that captures the benefit that side 1 has on side 2 o Alpha 2 is how much side 2 benefits from interacting with side 1 o When join platform creating surplus for myself but also for the other side because they benefit from interaction o The platform will adjust the price given that I am exerting positive externality on the other side o Because the readers give benefit to advertisers, and the alpha will be positive, the price could be below marginal cost o If the readers don't like seeing adverts the alpha 1 will be negative and the price will be more than marginal cost o Platform offsets loss on one side with gain on the other side

What does the Commissioner's notice say about market definition?

• Market definition is a tool to identify and define the boundaries of competition between firms • Relevant product market: all products and/or services which are regarded as interchangeable or substitutable by the consumer, because of product' characteristics, prices and intended use. • Relevant geographic market: geographic area where the conditions of competition are homogenous and which can be distinguished from neighbouring areas because of different competition conditions. • It is often practical to define the relevant product market first and only then to define the relevant geographic market

Market Power

• Market power is defined as Office of Fair Trading (OFT): o The ability to profitably sustain prices above competitive levels o OR the ability to restrict output or quality below competitive levels ♣ Producing less goods to charge higher prices, abusing position ♣ Able to do that: you have market power • Market power can be possessed by a single firm or group of firms

Volvo-Scania Merger: Econometric Model-Consumers and Demand SETUP

• N potential consumers • They may either buy one of J products (heavy trucks), j = 1, ..., J or otherwise choose the outside good 0 • The nested logit model: G groups of products, and one additional group for the outside good • Products within the same group are closer substitutes than products from different groups

Estimating Market Power in 2SM, Case of Newspapers: Journal of Applied Econometrics by Elena Argentesi & Lapo Filistrucchi Background

• Newspaper cover prices used to be regulated by a governmental agency, and it was only on January 1st 1988 that prices were liberalized • Price liberalization has no impact on price competition: prices have had a quite stable pattern and price increases by the different newspapers have always been quite simultaneous. • The question is whether the observed pattern in newspaper prices is consistent with profit-maximizing behaviour or coordinated practice • Four price curves on top of each other • They were increasing prices at the same time and charging same prices • These are cover prices, for readers • After price liberalisation nothing changed, increased price together • Raised question over whether they are colluding

Volvo- Scania case: market definiton

• No SSNIP test • Truck market has three categories: light duty trucks (6t), medium duty trucks (5t-16t), heavy duty trucks(> 16t) • Heavy trucks has two segments: rigid trucks and tractor trucks • Although rigid trucks and tractor trucks are not perfectly substitutable, the relevant market is defined as "heavy trucks" • No mention on second hand trucks • Relevant geographic market is defined to be national markets o Market conditions not homogenous across Europe but is across different countries, define geographic market as national o Substantial price and mark-up differences across countries o Model and technical configuration differences across countries o Sales and after sales services o Large variations in the market shares across countries

Kaiser & Wright (2006) Readers

• Nr readers, all read only one magazine SEE CARDS FOR EQUATION • Gamma: captures the two sides of the network effect o Positive: readers get positive externality from the present of different prices o Negative: readers get negative externality • The transportation cost: SEE CARDS FOR EQUATION • Assuming x ∼ U[0, 1] gives the Hotelling Specification of readers demand: SEE CARDS FOR EQUATION •where: SEE CARDS FOR EQUATIONS • share of readers of magazine i • (skipped)

how to use market shares to assess market power

• Office of Fair Trading starts by defining the market and examining market shares over time. E.g.: if a firm has persistently high market share → higher probability of market power • Relative shares are also important o E.g. one firm has 50% all others have 5%, likely this firm has market power • Pattern of market shares over time might be more informative: o Changing a lot → innovation o Rapid growth → low barriers to expansion

Armstrong Profit maximizing outcome

• Profit maximizing prices: • Price has an extra term - Adjustment is related to the elasticity of that side of the platform • So the price is equal to cost plus a downward adjustment by external benefit to group 2 (α2n2) plus an upward adjustment by a factor related to the elasticity of group participants o Elasticity is given the number of agents on the other side • Define the groups' elasticities of demand for a given level of participation by the other group as follows: • Elasticity of side one given the number of agents on side 2 • using the profit maximising prices • Then profit maximizing prices satisfy: • So it is possible to have p1 < f1: if the group's elasticity of demand is high and/or external benefit by group 2 is large • Can think of the whole thing as the net cost, the cost is adjusted by the extra benefit, from the interaction with the other side • So now the prices can be negative, might be less than the mc • If the group's elasticity of demand is high or if the external benefit to the group 2 is large • If that term is large enough, can pay a price that is less than mc

Market Definition in Printed Media Industry: An Econometric Illustration

• Provide an example showing the bias that can arise if the advertising side is not accounted for in the estimation of readers demand • Dataset on French magazines • A demand system based on a nested logit model of readers' demand • The nested logit model assumes that there are two levels in the choice of consumers: o A consumer first decides among the available magazines o Then she has to decide between buying a one-year subscription and buying the magazine each week at the newsstand • Data shows that the magazines could be broadly grouped in three classes according to the proportion of subscriptions with respect to newsstand sales • Black mostly by subscription • White mostly from newsstand

Estimating Market Power in 2SM The Model-Newspaper Demand

• Readers demand is assumed to be independent of advertising. o Tried and found it insignificant o Plausible approximation of reality, ads on newspapers can be skipped more easily • Nested-Logit model: o Readers first choose to buy a national newspaper or not o If they decide to buy, they decide which one to buy • Market size: population of Italy > 14 years-old

Two-Sided Markets Intro •Rochet-Tirole (2003) •Armstrong (2006)

• Rochet-Tirole (2003) is more suitable for credit cards while Armstrong (2006) fits better to dating agencies, newspapers, shopping malls

Anti-Trust Issues Raised by 2SM Market Definition

• SSNIP must pay attention to both sides • It should consider firms that currently supply both sets of customers. • The main issue is: whether the two-sided firms offer sufficiently differentiated products that they should not be considered in the same market. o Example: one of the dating clubs just serves to Christians. o Do credit cards compete with debit cards and checks and cash? • Price effects on one side of the market cannot be considered without taking into account the other side • In a merger analysis for example a price increase on one side may lead to a price decrease on the other, total price should be considered instead • For the merger guidelines price test, whether the firms under consideration, if they merged today and priced as a monopolist, could raise the total price by 10 percent or more o No reason to look at one side only

Price Wars in 2SM Price Wars in 1990s

• Sep 1993 NIN cut the price for The Times from 45p to 30p o Undercutting the Guardian (45p) Independent (45p) Daily Telegraph (48p) • The Independent estimated that at the current level of circulation this price cut cost £50,000 per day • Because of its substantial financial difficulties, the Independent decided to raise its price from 45p to 50p on the 12 Oct 1993 • The Telegraph also decided to drop its price from 48p to 30p Aug 1994 • June 1994 The Times decreased is price again from 30p to 20p • By this time the issue has received strong political attention • The Independent reacted in August 1994, reducing price from 50p to 30p o This was in order to stop the decline of its circulation that decreased by 20% since The Times had first reduced its price • Subsequently there was a period of increase in cover prices as the costs of news printing were rising for all firms • The Times decided to increase its prices from 20p to 25p on the 3rd July 1995 • At the same date The Telegraph also increased from 30p to 35p. • The Independent followed on the 17 July and increased its price to 35p. • Another wave of price increases was initiated by The Times and The Telegraph on the 20 November 1995 who raised their prices to 30p and 40p respectively. • The Independent followed on the 22 January 1996 • All these price changes raised questions over what was going on, price war to undercut copmetitors or predatory pricing?

Price Wars in 2SM Findings

• Simulation show that asymmetric production costs may explain much of the observed changes in prices, circulation and advertising volumes o The effect of differentiated products was important • An alternative explanation of the observed market changes is the breakdown of a collusive agreement on cover prices which was upset when Rupert Murdoch took over the Times and changed old habits • An eventual alternative and one that also takes seriously the role of product placement on the political line is that the observed changes in market structure result from a (expected) positive shock on the demand side for advertising. o Positive shock which allowed them to charge lower prices to readers o Structure of the industry, prices go down as a result of positive demand shock on advertising side o That's how price cuts explained • This shock would lead to an adjustment process that finally implies lower equilibrium prices on the reader's side as the new optimal mix of newspaper finance has more of its revenue resulting from advertisers than from readers

Estimating Market Power in 2SM The Model-Advertising Demand

• Single-homing may seem unrealistic for advertising side, however in each day of the week, 80% of advertisers put message in only one of four newspapers o 80% of advertisiers are single-homing • Simple logit specification for advertising demand o Should I advertise or not where I sA it : ad share of newspaper i at time t I sA 0t : market share of outside good I xA it : observed characteristics I A it : unobserved characteristics I pA it : advertising price I yN it : total circulation. • L.H.S: demand • Beta a: effect on advertising time, alpha is effect of price on demand • Also include circulation rate, number of people buying the newspaper • Total potential market size: market of daily publications

what is the SSNIP test

• Small but Significant Non-Transitory Increase in Price • Also known as the hypothetical monopolist test it is a method to define markets when the relevant data is available

how do you conduct the SSNIP test (5 steps)

• Start with the narrowest group of products • Suppose that the relevant product is produced by a monopolist • What would happen if the monopolist increase the price of product by 5-10 %? • If the price rise is not profitable, then the closest substitute is added to the monopolized bundle and the procedure is repeated • The procedure stops when it is profitable to increase the price of the bundle of product in hand

Quantifying cannibalisation: what is the tax on profits

• Suppose after the merger, HQ want the divisions to maximize total profits • Of course, pre-merger prices fail to do it o Pre-merger prices won't be maximising total profit as they weren't internalising price changes on each other products o Profit maximinsg prices will be different • So, the headquarters put an internal tax to each divisions output • Tax=incremental profitability of the business cannibalized • So the tax on product 1 equals: T1 ≡ || ∂ π(b)/ ∂X1 || T 1 ≡ (∂πB/∂X2) x ||∂X2 ∂X1 || • Tax on product 1: derivative of profit of firm B w.r.t changed quantity of firm A • Profit of firm A depends on quantity of firm B • Then use the chain rule • The first term on the R.H.S is the absolute gross margin of firm B, • The second term is the diversion ratio from Product 1 to Product 2 at pre-merger prices, D12 o Shows how much the quantity of product sought for firm 2 changes when the quantity of product sought for firm 1 changes o Increase price, loose some customers, what percentage of these customers are now buying from firm 2 o Hypothetical post-merger price increase for product 1 would reduce the demand for good 1 by some number of units. A fraction of the former product 1 purchases would turn to product 2. The incremental product 2 unit sales divided by the 'lost' product 1 units is the diversion ratio • Thus we can write: T1= D12 (P2-C2) T2 = D21 (P1-C1) • Since the products are substitutes T1 and T2 are positive • T1 and T2 can be thought of factors shifting the MC • Thus T1 is viewed as a measure of initial impetus for product 1's prices to rise as a result of loss of competition

Volvo- Scania case: market assessment

• The Commission used traditional market power proxies • The investigation is limited to 5 countries where the creation of a dominant position is found o Sweden, Norway, Finland and Ireland, or where this is found to be likely, Denmark • The merging firms joint market share have remained stable, and showed no tendency to decline • Points out the large difference between the joint market share of the merging parties and the market share of the largest remaining competitor in most of the five countries o The remaining competitor's share is too low • The Commission supplements its market share analysis with qualitative factors o Brand loyalty and little customer purchasing power o Likelihood of entry: huge entry costs

drawbacks of UPP (4)

• The UPP methodology takes the pre-merger structure of demand as a given • It assumes that prices in the industry are determined by a Bertrand equilibrium • The UPP framework does not allow for possible entry of new competitors when postulating a post-merger price increase • It does not allow for possible product repositioning by existing competitors in response to higher post-merger prices

Market Definition in Printed Media Industry: An Econometric Framework

• The aim of an empirical model for market definition is to provide estimates of demand parameters and in particular of price sensitivity parameters which are crucial to determine the substitution pattern between titles. • An empirical model of printed media industry should include three main issues: o Two-sidedness o Product differentiation (both vertical and horizontal) o Definition of total market size (outside good) • The potential market on the readers' side is defined as the total population above the age of 14 • On the advertising market, the largest possible market size would include all the media o I.e., TV, radio, internet, press. o Narrower definitions might be more appropriate given the limited substitutability between different media services • They chose all printed media • They propose to estimate a system of logit demands • (don't need to memorise all equations) • The two demands should be connected by an inter-market network externality • For readers: SEE CARDS • Demand on readers side depends on some newspaper characteristics, price and number of advertisements • For advertisers: SEE CARDS • Demand on advertiser's side depends on newspaper characteristics, price for advertisers and number of readers • The parameters γr and γa capture the link between the two markets

SSNIP Test for Two-Sided Markets: Proposed extension

• The author's proposed extension is relevant for Media markets • The test should consider an increase in price level while adjusting the price structure optimally o Increase price of one side and other side adjust optimally o Then increase the price on the other side and then the deamdn and price of other side adjust optimally o Each time let it adjust optimamly given price structure • In other words, you need to take into account all the dynamics of the market • Credit card market: increase the price level, prices of both sides then let the price structure adjust optimally o Media is sequentially not increasing prices for both sides at same time • Answer depends on type of market considering

Market Definition in Printed Media Industry: Case Review

• The definition of relevant markets in printed media industries should take into account the distinctive features of competition in this industry: o There is competition between newspapers belonging to the same product group o Another important dimension in printed media is its geographic and spatial location E.g. Bristol and south gloustershire news may overlap o The boundaries of the relevant market depend also on the competitive constraint provided by other, non-newspaper media, namely other printed media, television, or internet • Ex: In the Recoletos / Unedisa case, product categories are defined according to frequency, content, and quality of the publication • Ex: In the decision Regional Independent Media Ltd and Gannett UK Ltd/ Johnston Press plc/ Guardian Media Group, the competition assessment focuses on the local area analysis, as the titles involved compete primarily for a local readership

Market Definition in Printed Media Industry: An Econometric Illustration Demand model

• The demand model is specified as: • where • xj: is a set of exogenous variables to measure the specific reputation of each magazine by means of a dummy variable for each magazine, a dummy to signal the distribution mode, the number of issues per year and a time trend • pj: is the unit price or the subscription price • sj|m: the market share of magazine j in the group of magazines sold under subscription or unit purchase • Estimate the model by two sets of instrument • The second set of instruments includes a variable measuring the advertising revenue

Anti-Trust Issues Raised by 2SM Barriers to Entry: example of successive entries to 2SM U.S v. Microsoft Example

• The government argued that the stock of applications that had been written for Microsoft Windows was a barrier to entry that prevented other firms from competing in the market for operating systems for Intel-compatible computers • Application developers write to Windows because a large number of end-users, and end-users use Windows because it has a large number of applications • The district court agreed with the government • The fact that Microsoft has attracted a large number of applications is no more or less remarkable than the fact that it has attracted a larger number of users • Indeed, from society's standpoint, it is a good thing • By themselves positive feedback effects do not prevent entry o Payment cards: Diners Club (1950), American Express (1958), Visa (1966), MasterCard (1966), and Discover (1986) o Video games in the US: Magnavox (1972), Atari (1975), Coleco (1976), Fairchild (1976), Mattel (1979), Nintendo (1985), Sega (1989), Sony (1995), and Microsoft (2001). • The relevant question to ask concerning barriers to entry into a two-sided market is whether the incumbent has special advantages that could not be replicated by entrants

The EU Merger Guidelines: Market shares and HHI indices

• The guidelines specify that large market shares (50% or above) may indicate market power, flag that merger • Certain mergers are not likely to harm effective competition, in particular when the combined market shares of the merging firms do not pass 25% • HHI and the change in HHI from pre-merger to post-merger (delta) are the first indicators of the change competitiveness of the market after the merger ( recall HHI = Sum of the squares of the market shares of the firm in the industry N HHI = sn2 n=1)

how to use buyer power in the assessment of market power

• The strength of buyers and the structure of the buyers' side of the market may constrain the market power of a seller. • Size is not sufficient for buyer power; buyer power requires the buyer to have choice. • Buyer power is most commonly found in industries where buyers and suppliers negotiate

Network externalities recap

• There are network effects if one agent's adoption of good o Benefits other adopters (total effect) o Increases others' incentives to adopt (marginal effect) • The benefit to consumers is higher the more other consumers use it Products that consumers get utility out of them, the more that other consumers adopt the good The more others buy, the more it increases their incentive to do so too 2 sources the benefit can arise from direct; whole point is for product to connect with other; allows consumers to connect with others, the more people using the more people you can connect with 2. indirect; benefit arises from the availability of direct compliment product eg more windows users, more software available The more consumers of the same product, the more utility

"Competition in two-sided markets", RAND Journal of Economics Armstrong Monopoly Platform

• There are two groups of agents 1 and 2. Utilities are given by: o Readers and advertisers are the agents o Utility of agent 1 reader is related to how many advertisements there are minus the price paid Assuming adverts are useful o Utility of the advertisers is related to the benefit from reaching the readers (n1) minus the price • Where p1 and p2 are the prices charged to the two groups, n1 and n2 are the number of agents from each group on the platform. • α1: measures the benefit of a group-1 agent enjoyed from interacting with each group-2 agent on the platform. • α2: measures the benefit of a group-2 agent enjoyed from interacting with each group-1 agent on the platform • Numbers of agents who participates as function of utilities are: for some increasing functions φ1 and φ2 • Platform incurs a per-agent cost f1 (e.g. printing/delivery cost) for serving group 1 and f2 for serving group 2 • The platform's profit: • Number of people on side 1 times the price they are paying minus cost plus the number of people on side 2 times the price minus the cost • If we consider the platform offer utilities {u1, u2} rather than the prices {p1, p2}

Market Definition in Printed Media Industry: Results

• These results are just illustrative of the bias that can arise in the estimation of elasticities if the feedback effect between the two sides of the market is neglected. • It would be better to specify a full econometric model. • Going from the first set of estimates to the second set when advertising is used as an instrument could modify the policy conclusion. • If demands for magazines were inelastic and if magazines were not substitutes, the relevant market could shrink to the magazine itself. • In terms of the antitrust policy, a merger in this industry would not be investigated

Estimating Market Power in 2SM Estimation-Mark-ups TABLE RESULTS

• They considered 4 different scenarios: o Competition on readers side and competition on advertising side o Competition on readers side and collusion on advertising side o Collusion on readers side and competition on advertising side o Collusion on readers side and collusion on advertising side • First assume the readers side is competing and look at mark up on advertising side o Total mark up: competition in advertising side is a dominant strategy compared to collusion o If you fix readers side the total mark up is higher (0.430 vs 0.425) • If you fix readers side to collusion o Totol mark up higher with competition (0.797 vs 0.792) • Then what they conclude is that compeitiotn on advertising side is a dominant strategy as total mark up is higher • This decreased the analysis to deciding between 1st and 3rd outcome • In this case check which confidence interval this observed mark up would fall o The two 95% confidence intervals are [0.252, 0.608] for competition-competition and [0.428, 1.166] for competition-collusion o The observed markup is 0.943 • Statistical evidence in favour of the model competition-collusion is strong • So, the hypothesis that there exists some collusion is not rejected by the data, whereas the one that firms are competing in both markets is • Consistent with the observed pattern of cover prices • The advertising market seems to be much more competitive • Concluded that the newspapers were deciding the prices on the readers' side collectively whereas competing for advertisers • Support: advertising prices are highly non-linear • Advertising prices are subject to a lot of discounting and are not easily observable by competitors, so collusion is difficult to sustain • Why the estimated total mark-ups are lower when the newspapers collude in the advertising market than the case where they compete: • Consider the case of compeititon in the readers' market and compare the total makreupunder the two alternative hypotheses of compet/collusion on the advertising market, for a given level of prices and quantitites • Advertising prices would be higher if firms were colluding in the advertising market than if they were competiting • Advertising quantity would therefore tend to be lower • In order to keep advertising quantity unaltered, the firm can increase the readership by lowering the cover price, which implies a lower cover price markup • If the lower cover price makrup more than offsets the higher advertising price markup (w.r.t compettion) then the total makrup under collusion in advertising is lower than the total markup under competition (in advertising)

Volvo-Scania Merger: Econometric Model-Consumers and Demand notation

• This is a random utility model • The utility to consumer i from purchasing product j is given by: uij =δj +ζig +(1−σ)εij j: is the mean valuation for product j, common to all consumers ζig and ij , are random variables reflecting individual i's deviation from the mean valuation ζig is consumer i's utility, common to all products belonging to group g ij is consumer i's utility, specific to product j • σ ∈ [0, 1] σ = 1 ⇒ perfect correlation σ = 0 ⇒ no correlation • looking at how substitutable 1, perfectly substitutable • Assumption is the consumer buys the good that gives the maximum utility • Assume a specific distributional from for the unobservable terms we can get a closed form solution

Kaiser & Wright (2006) Results for 2-sided Single-Homing

• This shows the gamma • How much readers value adverts, how do they benefit from seeing the advertisements • Coefficient is positive and significant o Readers like seeing adverts on the magazine • Like the content as well: positive and sig • Advertisers value reader: positive and sig • Given the estimation results, they computed the price-cost mark-ups • The mark up on the readers side, per magazine, is negative (-2 something) • If you look at advertisement side, plus 10,000 something users • Advertisers benefit more from interacting with readers, hence magazine using this and making money form advertising side while subsidising readers • By charging a small price to readers, below mc, can attract more and more readers which is beneficial to advertisers as prefer having more readers • Hence charging readers less and interacting more readers, interact with more advertisers and charge higher prices • There are no predatory prices, need to look at both sides of the market • Moreover, if we multiply the estimated mark-up with the number of readers (and advertisers for the advertising side) we can obtain the contributions of each side to the profits of the magazine. • Readers' average contribution is: −2, 100,830 euros (se = 4, 530, 730) • Advertisers' average contribution is: 6, 911,360 euros (se = 3, 715, 350) • Magazines obtain no direct contribution from the readers' side but compensate their losses by gains on the advertising market

Volvo-Scania Merger: Hypothetical Market Power test METHOD

• This test computes the profitability of unilateral and nontrivial price increases by the merging firms. • They consider price increases by 5%, 10% and 25% o Then see if that is profitable or not • The test is related to the previously discussed SSNIP-test, only the focus is different o The SSNIP-test asks how many firms are needed to make a given price increase profitable, for the purpose of defining the relevant antitrust market. o The hypothetical market power test asks whether the two merging firms can profitably raise prices by alternative amounts o The purpose is here to examine the potential of increased unilateral market power • Since it relates to the SSNIP test, it has the advantage of being familiar to antitrust practitioners • It does not require specific assumptions about firm behaviour after the merger

Market Definition in Printed Media Industry: CEPR Discussion Paper by Elena Argentesi and Marc Ivaldi Introduction

• Three important ingredients in the definition of relevant markets in press industry o Two-sidedness o Product differentiation o Correct estimation of potential market size is required • The paper mostly focus on demand substitution • For SSNIP test to be successful, you need to estimate the correct econometric model • Thus, the paper also gives examples of econometric models for print media industry • The authors also illustrate what would happen if we do not take into account the two-sidedness of the market

Single-homing vs. Multi-homing: 3 cases that may arise

• Three possible cases my arise: 1. Both groups single-home 2. One group single-homes, the other multi-homes 3. Both groups multi-home • Case 2 is called COMPETITIVE BOTTLENECKS • The platform has monopoly power over the multi-homing side since the only way to reach the single-homing agents is via that platform o E.g. I am the advertiser and I am advertising in all newspapers, guardian, independent, the sun etc. o But young, uni educated groups only reading one, e.g. guardian o Can only reach this consumer base through the guardian so the guardian can charge whatever they wants, o The only way to reach is advertising in guardian so they are a monopoly in the sense that they are the only ones who have young, uni educated

The EU Merger Guidelines Market shares and HHI indices: how to avoid contradiction in results of HHI and market share indicators

• To avoid the contradiction, the Commission says the HHI indicators point in only one direction, to give the message a soft safe harbour meaning • Below certain levels, the Commission is unlikely to identify competition concerns in a merger • The Guidelines do not say that above these levels, there are likely to be competition concerns (the HHI indicators take a neutral stance in this respect) • In this way, the message of the HHI indicators is made consistent with the message of the 25% market share threshold

Kaiser & Wright (2006) Empirical Specification

• To remove fixed effects, data is first differenced • The equations are estimated jointly by GMM • All explanatory variables in estimation equations are potentially endogenous •

SSNIP Test for Two-Sided Markets: features of two sided markets

• Two-sided markets have different features amongst themselves • In credit card industry for example the transaction between the two sides is observable while it is not in most media industries. • In media, relevant market for one side may not be relevant for the other o TV or billboards are other relevant markets for advertisers while they are not relevant for readers • In credit card industry a card is either in the relevant market on both sides or not. • Thus, it is very hard to find a generalized extension of SSNIP

Merger simulation versus. UPP: Epstein and Rubinfield 2010

• UPP does not need to assume specific functional forms while Merger Simulation depends heavily on the choice of functional forms • Merger simulation gives better results if the data in hand is big enough, but UPP does not necessitate this • Both UPP and the more general merger simulation approach rely on the computation of a post merger Bertrand equilibrium with differentiated products and merger-specific efficiencies • UPP asks whether in this framework there would be a price increase, without attempting to quantify the magnitude • Merger simulation goes one step further and estimates the magnitude of a price increase • The two techniques, when consistently calibrated, yield the same qualitative conclusion • Neither UPP analysis nor merger simulation is well suited to account for entry or product repositioning as additional competitive constraints

Volvo-Scania Merger: Actual Market Power test RESULTS

• Under the assumption there is no cost efficiency o Still have the same costs • Consumer surplus decreases everywhere • Total welfare decreases everywhere • All confidence intervals are negative • 5% cost efficiency: • In Austria it may increase but confidence increase shows it may also decrease • Same with France, Germany and Luxembourg • Isn't one that is always positive values • With or without cost efficiencies the consumer surplus and total welfare will decrease or stay the same • In most of the countries it will decrease • Merger simulation that they should not merge • It gives you a number what the total welfare will be, given the specifications of the model

Anti-Trust Issues Raised by 2SM Introduction Example:

• Uniformdating.com (ud) and justchristiandating.com (jcd) want to merge • Suppose ud charges both men and women £15 while the jcd charges men £20 and women £10 • Suppose ud has a 40% market share while jcd has a 10% market share • How would you define the market? o All online dating clubs? o Should you include the ones which are not online as well? o Should you count also night clubs? o If magazine: still include billboards? As advertisers use these too even though one-sided • Suppose you define the market, how would you analyse the merger? o Which price increase should you look at? o You should do the simulation considering both sides, thus estimating two demand functions simultaneously for both sides • Suppose competition authority found that the merged site will charge men £20 and women £15, o So the average price charged in both clubs raise from £15 to £17.5 o Should the merger be banned? o Network externalities means hard to look at the issue at just the pricing level o Although the price has increased, when look at merger side, will have more options, benefit is higher than price increase o Not straightforward to flag merger just because price increased

SSNIP Test for Two-Sided Markets: NETInstitute Working paper, Lapo Filistrucchi

• Usual SSNIP cannot be applied to two-sided markets (2SM) • A firm in a 2SM sells two products or services to two distinct group of consumers • There is a link between the demand on the two-sides of the markets thus the price increase on one side will affect the other side as well • So, profits from both sides should be considered to check the effect of a price increase

summarise article 101

• You cannot fix the price or fix the quantity/reduce the quantity or share the market • Cannot do anything that affects competition • But, if there are certain circumstances (doing things that are better for the market) then you can do it

what is the Herfindhal-Hirschman Index:

• where sn is the market share of firm n, and N is the number of the firms in the market • Sums the squares of market shares of firms in the market • Given the number can conclude that the market is concentrated or not

Volvo-Scania Merger: Econometric Model-Consumers and Demand Diagram

•By modelling the market in this way imposing that you are more likely to substitute tractor truck model with a tractor truck than a rigid truck • Think of car example, more likely to switch between peugots than a peugot to a bmw

Volvo-Scania Merger: market power test results (3 tests)

•Hypothetical market power test o Similar to SSNIP o If increases prices by 5%, 10% etc. is it still profitable for them • Actual market test: o Similar to merger simulation o Estimate model, then look at what will happen when maximise profits jointly to the social welfare/consumer surplus • Comparative market power test: o Allow the merge, and as a reaction two others merge? o Look at whether they will merge with others if initial merger not allowed o Always looking at what will happen to social welfare •Analysis the same as the conclusion just from market shares/power

drawbacks of merger simulation

•Merger simulation tries to fit a structural model to pre-merger data and use the model to predict post-merger prices and/or quantities •⇒ It risks misspecification by omitting immediate and concrete aspects of firms' objectives o If mis-specify demand function, result may not be the same • It can be demanding in terms of data requirement • It generates results sensitive to functional form specifications • It is opaque to non-specialists o As an economist, hard time to explain to a lawyer or a representative

European Commission guidelines on merger analysis

•The new Regulation requires intervention in relation to mergers which would significantly impede effective competition, in particular as a result of the creation or strengthening of a dominant position in the markets concerned •E.g. then being able to cutting the quantity or increasing the price