Industrial Organization and High-Tech Markets, Dispense di Economia Industriale

Scarica Industrial Organization and High-Tech Markets e più Dispense in PDF di Economia Industriale solo su Docsity! CHAPTER 1: WHAT IS INDUSTRIAL ORGANIZATION? Industrial organization (IO) is concerned with the workings of markets and industries, in particular the way firms compete with each other. Whereas microeconomics typically focuses on the extreme cases of monopoly and perfect competition, industrial organization focuses primarily (but not only) on the intermediate case of oligopoly, which is an imperfectly competitive market, and studies the firms’ most important strategic interactions: ➢ Price and quantity competition, differentiation, product positioning, advertising, collusion, entry & exit strategies, R&D, type of contracts, data collection, business models of ecosystems… Market power is the ability to set prices above the production marginal cost. About this, we can say that the goal of industrial organization is to address the following four questions: 1) Do firms have market power? 2) How do firms obtain and maintain market power? → legal protection (patents), pre-emption (moving first to achieve a significant installed base), anti-competitive conduct (cartels) 3) What are the implications of market power? 4) Is there a role for public policy regarding market power? CHAPTER 2 and 3: RECAP OF BASIC MICROECONOMICS CONCEPTS Markets are made of buyers and sellers. In many cases, buyers are consumers and sellers are firms. CONSUMER PREFERENCES AND DEMAND Creating demand: move the masses to buy your product, service, or idea. If there is no demand, there is no business. Therefore, before deciding what to do, it's important to know the consumers’ tastes, and from this to derive their demand function, that is the willingness to pay for different amounts of a given product. The Consumer Surplus (CS) is the area below the demand curve (which represents the WTP) and above the price of the good. CS = WTP – p* In this graph, quantity is a function of price: q(p) such that A change in price leads to a movement along the demand curve; a change in other factors leads to a shift in the demand curve itself. PRICE ELASTICITY OF DEMAND → % variation of quantity demand divided by the % variation in price If Ԑ > 1, the demand is elastic; if Ԑ ˂ 1, the demand is inelastic. N.B.: elasticity and slope are not the same, and it’s independent of units Ԑ tends to be higher (in absolute value) for luxuries, and in the long run. Often, we use the concept of the inverse demand function p(q): price is a function of quantity demanded. COST FUNCTION C(q) → total cost of inputs the firm needs to pay in order to efficiently produce output q. • Fixed Cost (FC): does not depend on q • Variable Cost (VC): depends on q and would be zero if q = 0. • Total Cost (TC): FC + VC • Average or Unit Cost (AC): TC/q • Marginal or Incremental Cost (MC): cost of one additional unit. TC(q+1) – TC(q) or Given that the derivative of a constant (FC) is zero. Example: the weekly cost for leasing a machine which produces on T- shirt per hour is $20. This machine must be operated by one worker, whose hourly wage is paid as follows: ➢ $1 during weekdays (up to 40 hours) ➢ $2 on Saturdays (up to 8 hours) ➢ $3 on Sundays (up to 8 hours). Assuming that current output (q) is 40 T-shirts per week, we have that: • FC = $20; VC = 40 T-shirts * 1h * $1 = $40. • TC = 20+40 = $60; AC = 60/40 = $1.5 • MC = $2 → it’s the cost of producing one t-shirt more (from 40 to 41 the worker must work on Saturday) The market price of a T-shirt is $1.8. Should the firm operate on Saturday and produce more than 40? AC (q=40) = $1.5 at p = $1.8 → the firm makes profits because AC ˂ P, BUT MC(q=40) = $2 > $1.8 What is relevant in taking the decision is the comparison between price (marginal benefit) and marginal cost ➢ MC is the appropriate cost concept to decide How much to produce Suppose now that the market price goes down to p = $1.3 per T-shirt → no matter how much the production is, the firm loses money: p ˂ AC. The firm is better off not producing ➢ AC is the appropriate cost concept to decide Whether to produce This is a particular example because there isn't much flexibility in production. In general, the MC and AC functions would be continuous functions, as shown in the figure: P0 is the minimum of the AC function. ➢ p ˂ P0: a price-taking firm would prefer not to produce at all. ➢ p > P0: the optimal output level is given by the MC function. More generally, the MC (for p > minAC) gives the firm’s supply function. OPPORTUNITY COST → unavoidable benefit from not using a resource in the best alternative use. Example: suppose you own and use a building in Padua to run a small pizzeria. Food supplies are the only (accounting) costs. At the end of the year, these costs are $20000 and revenues $100000. ➢ Consequently, (accounting) profits are 100000 – 20000 = $80000 These profits, however, overstate economic profits: accounting costs don’t include opportunity costs, for example the cost of the time spent running the business. ➢ Suppose you could alternatively work for someone else for $30000: this is an implicit cost of running the pizzeria, which reduces accounting profits down to $50000 SUNK COST → cost of a specific asset with no alternative use (it has no impact on a future decision) Example: a company spends $50000 on a marketing study to see if a new product is willing to launch will succeed in the marketplace. The study concludes that the launch will not be profitable. ➢ This is a sunk cost: the company will stop developing the product and its investment is gone. ECONOMIES OF SCALE: the AC of production declines with the output produced - Natural monopolies: a big firm serving the whole market is more efficient than two or more firms ECONOMIES OF SCOPE: the cost of producing outputs q1 and q2 together is lower than of producing them separately. → C(q1, q2) < C(q1, 0) + C(0, q2) PERFECT COMPETITION (PC) It’s the extreme opposite of monopoly; both are rarely observed. We must remember these assumptions: 1. Atomicity: there are many-many suppliers (market for agricultural products) 2. Product homogeneity: all firms sell the same product 3. Perfect information: all agents (firms/consumers) know the price set by all firms (financial firms) 4. Free entry and equal access to all production technologies (textile industry/mature technology) As for the monopoly, firms’ objective is to maximize profits • If a firm charges a higher price, no one buys from it (homogeneity + full information). • If a firm undercuts, it serves all the market unless the other firms make the same The unique equilibrium condition of a PC market is P = MC: consequently, the output level (q) is the only one equilibrium, and this is efficient (all consumers with WTP>cost are served). Firms are price takers: ❖ if a firm charges a higher price, it sells nothing ❖ if it lowers its price, it sells an infinite amount (⇒ the demand is horizontal as if the price is given) Firms do not earn extra profits in the long run • If, given the technology freely available to everybody, firms make profits: new firms are attracted to the market (assumption 4) ⇒ entry occurs until extra-profits are driven down to zero CHAPTER 7: GAME THEORY The basic element is a game, which consists of: a set of players, rules and actions (who can do what when), and payoff functions (the utility each player gets as a result of each possible combination of strategies). Therefore, a game is a stylized model that represents such situation of strategic interdependence between people: each agent (individual, firm, player) recognizes that its payoff depends not only on their own actions but also on the action of the other agents playing in the game. Example 1: consider firms which compete in a market selling a homogeneous product. ➢ At which price each firm should set its product? ➢ May a firm decide its price without taking in consideration the rivals’ price? ➢ Each firm’s profit depends on its price only or on the rivals’ prices, too? Example 2: suppose you’re interested in buying a small statue on eBay. My offer depends upon (at least): ➢ Your personal evaluation of the sketch (abbozzo) ➢ How much do you expect the other bidders are going to offer ➢ The rule of the auction (first, second price…) All players want to maximize their own utility and nothing else (they don’t care about the others!): they have to assume that their strategy works good as well, given the fact that it depends on another player’s strategy. They have to be as more rational as possible. Different actors act in different ways: - Altruistic agents want to maximize utility of other players - Cooperative agents want to maximize group (social) utility - Byzantine agents want to minimize utility of other players Limitations: agent may not o foresee all consequences of their decisions (bounded rationality) o know all relevant information about the game structure and its current state (incomplete and imperfect information, in particular the adverse selection) NORMAL FORM GAMES We can describe finite strategic games using payoff matrices. Example: Two-player game where player 1 has actions T and B (Top and Bottom), and player 2 has actions L and R (Left and Right), with payoff matrix: If player 1 plays T and player 2 plays L, then player 1 gets payoff w1 and player 2 gets payoff w2, and so on. Payoff: gains or losses from implementing a strategy Prisoner’s Dilemma Two prisoners are interrogated separately and have the options to either cooperate (C) with their fellow prisoner and stay silent, or defect (D) and accuse the fellow prisoner of the crime. Possible outcomes: ▶ Both cooperate: no hard evidence against either of them, only short prison sentences for both. ▶ One cooperates, the other defects: the defecting prisoner is set free immediately, and the cooperating prisoner gets a very long prison sentence. ▶ Both confess: both get medium-length prison sentences. In equilibrium {D,D} both confess and get total utility of 2 (1+1); if they were cooperating, they would get total utility of 6 (3+3) → equilibrium {C,C} Prisoner dilemma ⇒ failure of cooperation in static environments Hawk and Dove (colomba e falco) In a fight for resources two players can behave either like a dove (D), yielding, or like a hawk (H), attacking. Possible outcomes: ▶ Both players behave like doves: both players share the benefit and get the highest total utility possible, that is 3+3=6 → {D,D} ▶ A hawk meets a dove: the hawk wins and gets the bigger part. ▶ Both players behave like hawks: the benefit gets lost completely because they will fight each other → {H,H} There is a coordination problem ⇒ multiplicity of equilibria {D,H} and {H,D} → in both cases the total utility is 1+4=5, but only one of the players wins The solution of a strategic game is a profile where all players strategies that are rational (“optimal”). Note that there different formal rules for predicting how a game will be played → solution concepts. Consequently, an agent should eliminate (or at least avoid) all obviously irrational strategies, which means that there is another strategy that is always better, no matter what the other players do. → Therefore, a solution concept is the iterative elimination of strictly dominated strategies Dominant str.: a player knows what’s the best thing to do independently on what the other player does Dominated str.: a player knows what’s the worst thing to do independently on what the other player does Example: ▶ Step 1 is to eliminate row C (strictly dominated by row D), because player 1’s utility would be higher if the plays D in both cases ▶ Step 2 is to eliminate column C (strictly dominated by col. D), because player 2’s utility would be higher if he plays D in both cases The solution is the combination of strategies {D,D}. Here we have a Nash Equilibrium (N.E.), the most famous solution concept in game theory and surely the most applied in economic models. Definition: a pair (vector) of strategies is a N.E. if no player can unilaterally change his/her strategy in a way that improves its payoff (both players have no incentive in changing their position). Formally, o Strategy profiles where no player benefits from playing a different strategy o Strategy profiles where every player’s strategy is a best response to the other players’ strategies o Nash equilibrium strategies are not eliminated EXTENSIVE FORM GAMES The normal form is a very useful tool in representing games where agents play simultaneously (or, in other words, when each player has to decide its strategy without knowing what the other player(s) decided). But often in practice simultaneous moves are not reasonable (for example when playing chess). Extensive games (with perfect information) reflect such situations by modelling games as game trees. The main idea is that players have several decision points where they can decide how to play. → Strategies: mappings from decision points in the game tree to actions to be played. We focus on finite games (finite horizon) and all ingredients of the game are common knowledge. CHAPTER 6: PRICE AND NON-PRICE DISCRIMINATION Price discrimination: practice of setting different prices for the same good. It requires the absence of resale: - Resale implies that individuals may exploit arbitrage opportunities, and for this reason the “law of one price” (=legge del prezzo unico) in a perfectly competitive market would not prevail. We will analyse the best discriminating strategies for a monopoly ⇒ Why and how to price discriminate When a monopoly sets a unique price for all consumers, there are many individuals that do not buy the good ➢ This happens despite their willingness to pay (WTP) may be higher than the cost of production ➢ From the firm’s point of view, there is an opportunity cost of not serving them! However, the firm does not find optimal to reduce its price because it is more profitable to sell at a higher price to a reduced number of consumers. Price discrimination allows the firm to serve more consumers. The reason why monopolists set different prices concerns the fact that each consumer has a different evaluation for every product. The monopolist has to make sure that, if an individual buys its products for €100, he/she couldn’t sell it again in an “after-market” at a higher price ⇒ NO resale Uniform pricing vs Personalized pricing The optimal output level qM is so that MR = MC The optimal price pM is given by the demand curve and the optimal output level qM At this price and output level, the seller makes a profit given by (pM − c)qM, ignoring the fixed costs. Things change if the firm can set different prices: in this case, the optimal strategy is to personalize them (first degree price discrimination). Each consumer pays his/her WTP, so that the firm can appropriate the areas A and B. → this strategy is difficult to implement because firms usually don’t have perfect information on consumers’ WTP, and therefore cannot charge prices accordingly. The “positive” aspect is that personalized pricing is getting real (OECD Report): data brokers, platforms and their client firms have access to a lot of characteristics about pretty much everyone of us. - We are all different so, potentially, one price per each of us - Full knowledge of all the correlates to our willingness to pay for the good How to discriminate? Suppose individual information is NOT available. Firms can do price discrimination in two ways: ▪ Focusing on some observable buyer characteristics (third degree price discrimination, also known as group pricing) geographic location, age, physical aspect, usage occasion (e.g., Mac/Android) ▪ Designing prices to let buyers self-select among different offerings Group pricing Two groups of consumers {1, 2} are identifiable. The monopolist engages in price discrimination by maximizing Π(p1, p2) = p1q(p1) + p2q(p2) − c[q(p1) + q(p2)], where c is the common MC of production. Profit maximization implies that MR1 = MR2 = MC. F.o.c.: Under discrimination by market segmentation, the seller charges a higher price to those consumers with lower price elasticity (which is the inverse of elasticity). Exercise: from uniform pricing to group pricing Consider a monopoly firm. P = 100 − Q is the inverse demand; constant marginal cost c = 20 for simplicity. π = TR(Q) − TC(Q) = pQ − cQ = (100 − Q)Q − 20Q → → QU = 40 (U = uniform) Substituting in the demand and profit functions, we get pU = 60 and πU = 60 · 40 − 20 · 40 = 1600 DWLU = deadweight loss: social inefficiency caused by the monopolist CSU = consumer surplus πU = monopolist’s profits MR(Q) = 100 – 2Q Consider now the example in which consumers can be divided into groups. The estimated demand functions are as follows (marginal cost does not change, c = 20): ➢ p1 = 120 − 2q1 for non-students; p2 = 80 − 2q2 for students The profits of the firm in this case are: π = TR1(q1) + TR2(q2) − TC(Q) = p1q1 + p2q2 − c(q1 + q2) = (120 − 2q1)q1 + (80 − 2q2)q2 − 20(q1 + q2) F.o.c.: → q1 G = 25 and q2 G = 15 (G = group) Substituting in the demand and profit functions, we get: p1 G = 70 and p2 G = 50 πG = 70 · 25 + 50 · 15 − 20 · (25 + 15) = 1750 Let’s compare and discuss the results Market power is sufficient for implementing group pricing; firms with lots of data and platforms usually also have substantial market power. No arbitrage: students cannot order for non-student friends (in the reality, the amount of arbitrage is rather limited). In our example, the recognition of different segments of demand is achieved through data but could be implemented in many other ways (e.g., presenting a student card...) Non-Linear Pricing In many industries consumers must decide not only whether to buy a given product, but also how much to buy of it. Thus, they pay a (flat) price f to access the service plus a price p for each unit consumed. ▶ Total price T = f + p × q. The per unit cost is T/q = f /q + p → here we have a fixed cost ▶ This is a form of price discrimination based on quantity: the more you buy, the less you pay each unit How does a monopolist set the two components p and f ? Consider the profit function Π(p, f): o For any p, the firm may also charge a fixed fee f (access charge-subscription fee). o It sets f as high as possible ⇒ at the maximum level compatible with the consumers’ participation constraint (making the consumers indifferent between buying or not): f = CS(p) Π = π(p) + f = π(p) + CS(p) NOTE: Π(p, f) is equal to W(p), which is the consumer Welfare (benessere del consumatore) • The firm extracts all surplus from the consumers • With non-linear tariffs, a profit maximizing firm is socially efficient (we don’t have the DWL) Assumptions: 1) consumers have the same demand D(p) ⇒ we can consider the ”representative” consumer or, equivalently, suppose the market is made of one individual 2) the firm (monopoly) has linear costs c(q) = c × q. The firms reduces P as much as possible to stimulate D(p) It then ”taxes” away CS(p) via the fixed fee f. The optimal/efficient tariff is as follows: p = c f = CS(p = c) = A + B + C A firm charging two-part tariffs (with identical consumers) maximizes social welfare (efficiency), but all the surplus goes to the benefit of the firm (trade-off between allocative efficiency and consumer welfare). Consequently, we don’t have deadweight loss. → similar to perfect price discrimination Summing up, the general rule to induce self-selection into high/low prices is to set the 1. price of the low-type version = low-type customers WTP 2. price of the high-type version such that high-type consum. are indifferent between the two versions ▶ Incentive constraint: don’t make the inexpensive version too attractive to those willing to pay more ▶ Participation constraint: cheap version must be sufficiently cheap that low types are willing to purchase ▶ Why it works: correlation between absolute valuation and cost (in terms of valuation) of restriction Three effects of versioning 1. higher price to HIGH type customers (+) 2. cannibalization (-) ▪ look at the first table: without versioning, normal readers (patients) buy at 50, but with versioning, they buy at 30... but inpatients buy at 90! 3. market expansion effect (+), as the firm wants to expand sales (this is not in our example…) The firm needs to balance these effects: this means that versioning is not always the most profitable strategy! Efficacy of versioning depends on individuals’ preferences and distribution (how many patients/inpatients). Bhargava & Choudhary (2008 Management Science) show that versioning is optimal when the optimal market share of the lower-quality version, offered alone, is greater than the optimal market share of the high-quality version, offered alone ⇒ when the market expansion effect (+) dominates cannibalization (-) Example: Spotify let users sort into groups and expands the market • Ad-based version: Free version for those who don’t mind ads (=nuisance) ⇒ Monetization via ads • Individual/Family plan for those who mind ads and premium services BUNDLING Bundling and tying arrangements are widespread in the modern commercial economy. Bundling is the practice of selling a number of separate products together in one bundle at a single price. ▶pure bundling: the seller only offers the individual products in a package - also called tying: whereby a seller requires that the purchaser of the “tying” product also must purchase a separate specified “tied” product or products (e.g., coffee machines and pods) ▶mixed bundling: the products are offered in a bundle and also separately, with the bundle sold at a discounted price. Example 1: TV on demand 3 different groups: movie/sport lovers, and medium group They are uniformly distributed ⇒ WTP heterogeneous How can a cable TV company identify the different types of consumers? → direct discrimination often unfeasible If the TV network charges €11 for each channel: - movie-lovers do not subscribe to sports channel and sport-lovers do not subscribe to movie channel - Profit will be 11 × 4 = 44 If it charges €22 for movie + sport channel package (bundling) - Every group subscribes to this package - Profit will be 22 × 3 = 66 (same as with direct discrimination) → 66 > 44 Example 2: the bundle allows the firm to reduce (eliminate) dispersion in the individual valuations. Best strategy without bundling: • charge Pa = Pb = Pc = 5 and sell two channels each. Π = (5 × 2) × 3 = 30 • alternatively charge Pa = Pb = Pc = 10 and get the same profit Π = 30 Best strategy with the bundle ABC: ➢ I1, I2 and I3 attach a value of 16 to the bundle ➢ The firms extracts all the surplus by setting PABC = 16(−ε) with ε → 0 and the profit is Π = 48 > 30 Profitability of product bundling by - keeping bundle price low: the firm can keep the demand high. - setting separate price high: the firm can extract rents from consumers who value only one product. A more general approach Consider a firm which sells n goods. For each consumer the valuations of the n products are independent and uniformly distributed over [0, 1]. Demand function for each product: q = 1 − p ▶ Without bundling, the firm offers n products separately. Optimal price: p = ½. Profit: Π = np = n/2 The demand for a bundle n = 2 clusters around the mean evaluation (in our case is 1). In general, the demand function for a bundle n > 2 clusters around n/2. As n goes to infinity, the bundle gets larger. By charging a price just below n/2, the firm is able to reap (=raccogliere) most of the consumers’ surplus. Another reason to bundle is leverage when a monopolist bundles a complementary good with its monopolized product (in order to avoid the monopolist’s profits to increase exponentially). Exercise 1 A publisher has to decide about a versioning strategy in the distribution of a novel. The market is made up of very passionate and less passionate readers of the author. A proportion β of readers are normal and the remaining 1 − β are passionate/impatient. Readers have mass 100 and there are no production costs. The willingness to pay for the two versions of the book of the two types of readers are: • Hardcover version: 90 (Impatient) vs. 60 (Normal) • Paperback version: 70 (Impatient) vs. 50 (Normal) 1. Determine prices and profits in case the editor was perfectly able to discriminate users. - How the profits vary with β? 2. Suppose now that the publisher does not observe the identity of the readers (it knows only that the population is divided between the two categories on the basis of the share β). - What is the best strategy in the absence of versioning? 3. Could the publisher do better? Identify profits under versioning and under which circumstances this strategy is the most profitable one (TIP: use a diagram with β on the x-axis and profits on the y- axis). 1. Price and profits with perfect discrimination If the firm only sells the Hardcover version, it charges pimp = 90 and pnor = 60. ➢ The profits are as follows: Π = 90 × 100(1 − β) + 60 × 100 × β = 9000 − 3000β Same thing for the Paperback version: the firm charges pimp = 70 and pnor = 50. ➢ The profits are as follows: Π = 70 × 100(1 − β) + 50 × 100 × β = 7000 - 2000β → Profits decrease in β ⇒ the smaller β, the smaller the number of normal readers, the higher the profits 2. No versioning, no discrimination. There are two possible strategies: - Set a high price and serve the inpatients only - Set a low price and serve all Formally, if 90 × 100 × (1 − β) > 60 × 100 → 9000 x (1 − β) > 6000 → 3000 > 9000β → β ˂ 1/3: only inpatients will buy the Hardcover version at p = 90 If 90 × 100 × (1 − β) ˂ 60 × 100 → 9000 x (1 − β) ˂ 6000 → 3000 ˂ 9000β → β > 1/3: everybody buys the Hardcover version at p = 60 Computing profits in the two cases, we can define a threshold value of β such that Blue line represents the profits for serving all. Black line is profit for serving only inpatients. 3. Versioning Let’s determine the optimal price for the two products such that the participation constraint (PC) and the incentive-compatibility constraint (IC) are satisfied. Formally, (PCI) pH ≤ 90 so that Inpatients buy Hardcover version (PCN) pP ≤ 50 so that Normals buy Paperback version (ICI) 90 – pH ≥ 70 – pP so that Inpatients do not find it optimal to buy Paperback version (ICN) 50 – pP ≥ 60 – pH so that Normals do not find it optimal to buy Hardcover version Consider pP = 50. In this case, ICN holds, and they are left with zero utility: 50 – 50 ≥ 60 – pH → pH ≥ 60 To induce Inpatients to buy Hardcover version, ICI must hold. Thus, 90 – pH ≥ 70 – 50 → pH ≤ 70 Note that with a price equal to pH = 70, also PCI and ICN hold. Therefore, • the optimal prices are pH = 70 and pP = 50 • profits are Πvers = 70 × 100(1 − β) + 50 × 100 × β = 7000 − 2000β Is versioning profitable? Which strategy do we have to follow? To identify under which circumstances versioning is the most profitable option, we must compare Πvers with a) the profit obtained when all consumers are served → Πall = 6000 for β > 1/3 and b) the profit obtained when only inpatients are served → ΠImp = 9000(1 − β) for β < 1/3. 4. Any better alternative? One alternative is to offer ES and ESN at the following prices E|S|N Al\io|i|1 ® pes = 11, thus A& C buy the bundle ES for a total of 22 B 1|5 ® pesn=15, thus D&B buy the bundle ESN for a total of 30 c 1 10]1 Therefore, this strategy is much better than the previous one as it D 1]9]5 generates a profit of 55 > 50. 5. Sky entry in the market Recall that the price for the bundle of the three channels is pig = 12. » lf Sky enters by charging psy = 10, C is indifferent between Sky Sport and Rai DIG (0 net utility), thus it goes to Sky Sport, which earns 10 in response. E N È . alolili Sky can lower the price and attract also D. B 15 b Dpas 12 for RAldig but values it 15. Thus, net benefit o £ È n i » To attract it, Sky must offers the same net utility. ® Since Sky is worth 9 for D, it must offer at most a price of 6 » With pi, = 6 Sky can attract € & D with a price and obtain a profit of 12 dSEl LESSON 4: OLIGOPOLY A market is said to be an oligopoly when there are a few firms operating (duopoly if there are only two firms). What makes monopoly/perfect competition different from oligopoly? ● In monopoly, firms do not care about competitors. ● In oligopoly, firms know that their decisions do affect rivals’ strategies/decisions: each firm maximizes its own profits, but firm j’s actions affect firm is profits. BERTRAND COMPETITION – ASSUMPTIONS 1. product homogeneity 2. perfect information: each player’s payoff function is common knowledge 3. a single firm may eventually serve the whole market 4. firms compete choosing the price of their product: players receive payoffs that depend on the combination of actions chosen. Suppose there are two firms (A and B) with the same technology (constant MC), that produces identical (perfectly substitutable) goods and makes decisions simultaneously (without knowledge of the others’ choices). They play the following game: 1) firms choose the prices (PA; PB) 2) consumer chooses goods with the lowest price 3) firms supply the amount demanded D(p). Consumers, obviously, buy the cheapest product (considering the assumptions 1. & 2.) ▶ If PA = PB, then the market is equally divided between A and B The demand of A (and similarly B) is We look for the Nash Equilibrium of this game: - both firms do not have UNILATERAL incentive to deviate from their price strategy There are three possible situations (given that no firm will charge any P < MC): Case 3. Pj = Pi = MC (i, j = a, b) → this is the equilibrium! Both firms make zero profit, but they cannot do better. If A (B) raises its price, it does not sell anything; if A (B) cuts the price, it makes negative profits: no incentive to unilaterally deviate. Note the difference with Perfect Competition: strategic interaction. ➢ Firm i ’s optimal price is function of the price charged by firm j ➢ Reaction function: Pi*(Pj) is a function which gives the best price for each firm given any price set by the rival (also known as best response function) Reaction functions intersection (no incentive to deviate): (MC, MC) Both firms get zero profit, because P = MC. It seems that two firms are enough for obtaining a perfectly competitive outcome. But is this true? If not, why? ▶Remember, the model rests on the following strong assumptions: - A single firm may eventually serve the whole market (ass. 3) - Product homogeneity: firms may be imperfect substitutes... results may change (ass. 1) - Perfect information: firms may not know all prices available in the market (ass. 2) - Firms play only once: if repeated game, they may collude. - Elastic supply assumption: firms often face capacity constraints. To conclude, introducing even a small friction in the Bertrand framework is enough to drive the equilibrium away from MC. Capacity constraints A firm which doesn’t have a capacity constraint can always cover the entire market (ass. 3). Example: firms have capacity K. Suppose that at a certain price P > MC, if P = PA = PB both firms sell at capacity, i.e., K. → no firm has incentive to undercut since it cannot sell more than its capacity. There is an equilibrium characterized by P > MC and with positive profits for both firms. ▶ General result: if total industry capacity is sufficiently low in relation to market demand, then equilibrium prices are greater than MC → No Bertrand Paradox This analysis applies to the case when firms must decide in advance the maximum amount of production (plant dimension) and then compete in prices. Search cost – Diamond Paradox Assume that consumers must search to learn prices (there’s no perfect information). ▶ Sequential search - a consumer looks at a particular product, learns the price, then moves on to the next product, searching costs s. STACKELBERG COMPETITION N.B.: in the basic Cournot model the two firms compete simultaneously and set their outputs at the same time. What if one firm decides how much to produce before the other? → e.g., think about digital markets where price adjustments occur many times per day. Example: according to the Washington Post’s report, Amazon used a secret algorithm called ”Project Nessie” to see just how high it could raise prices while still staying lower than its competitors. So, to some extent, Amazon’s was moving first as a price leader: - If competitors raised their prices, Amazon would follow suit. - If competitors didn’t, the algorithm would automatically move Amazon’s price back to its original, lower price. In the Stackelberg model, the timing now changes as follows: 1) Firm 1 is the LEADER: it sets q1 before Firm 2 2) Firm 2 is the FOLLOWER: it observes q1 and decides its output q2. The game is solved by backward induction: ● Firm 2 decides about q1 considering what has been already produced by 1 ⇒ Firm 2 acts rationally. ● Firm 1 moves first and acts rationally ⇒ Firm 1 can anticipate 2’s choices: it chooses q1 to maximize its profits and considering what firm 2 will do in the second period. Formally, consider the standard Cournot model with linear demand and constant MC: P(Q) = a – b*(q1 + q2) C(q) = cq, with a > c We solve by backward induction starting from period 2. ▶ given q1, we determine firm 2’s optimal output. This is nothing else that firm 2’s reaction function: ▶ firm 1 acts rationally (knowing firm 2’s reaction function) when setting its output, i.e., f.o.c.: Substituting it into the reaction function of firm 2, q2(q1*), we get Thus, when Firm 1 and 2 choose sequentially: 1. The leader produces more than the follower and gets higher profits (advantage) 2. The leader gets more profits than in the case of simultaneous choice, while the follower produces less and gets less profits. 3. The total quantity is greater than in the standard Cournot game: ⇒ Market price is lower, and welfare is higher. Exercise 1 Two firms produce a homogeneous product and compete à la Bertrand in the final market; for marketing reasons firms prefer to set only discrete prices (we know that consumers are upset by .95 prices!), P = 0, 1...10. Firms have the same marginal cost of production, MC = 2; fixed costs are 0 for both rivals. Demand function is Q(P) = 10 − P. (Note: if one firm sells K units, the rival faces a residual demand Qr = 10 − K − P) 1. Assuming that, when firms charge the same price, they equally share market demand Q(P), define equilibrium prices, quantities, and firms’ profits. One equilibrium is the standard Bertrand Nash equilibrium: P1 = P2 = 2 → Q(P) = 10 - 2 = 8 → q1 = q2 = 8/2 = 4 ⇒ π = 0 But if P1 = P2 = 3 → Q = 10 - 3 = 7 → π = (P - MC)*Q/2 = (3 - 2)*7/2 = 3.5 > 0. ▶ in this case, both firms do NOT have incentive to deviate → N.E.: let’s see why ★ If P1 = P2 = 4 → Q = 10 - 4 = 6 → π = (4 - 2)*6/2 = 2*3 = 6 ➢ it cannot be an equilibrium: they both have (unilaterally) incentive to deviate. ★ If P1 = 4 and P2 = 3 (firm 2 deviates) → q1 = 6 and q2 = 7 → π1 = 6 and π2 = 7 ➢ firm 2 sets a lower price, getting the entire market and higher profits (π2 > π1) 2. Suppose now that firm 2 cannot produce more than 3 units of the good (capacity constraint → K2 = 3). Are the equilibria found in 1 still Nash equilibria of this second game? ● Consider the equilibrium pair (P1, P2) = (2,2). If charging P1 = P2 = 3, then Firm 1 still sells positive quantities: ❖ Q = 10 - K2 = 10 - 3 = 7 ⇒ residual demand faced by firm 1 (QR) = 7 − P1 = 7 - 3 = 4 ❖ firm 1 has incentive to deviate and makes positive profits: → π1 = (3 −2)*4 = 4 > π2 = (3 - 2)*3 = 3. Thus, (P1, P2) = (2,2) cannot be a N.E. of this game. ● Consider the equilibrium pair (P1, P2) = (3,3). If charging P1 = P2 = 4, both firms sell the same quantities: ❖ Q = 10 - K2 = 10 - 3 = 7 ⇒ residual demand faced by firm 1 (QR) = 7 − P1 = 7 - 4 = 3 ❖ q1 = q2 = 3: they do NOT have incentive to deviate and, in addition, both firms’ profits are higher: → π1 = π2 = (4 - 2)*3 = 6 > 4. Thus, (P1, P2) = (3,3) cannot be a N.E. of this game ⇒ (P1, P2) = (4,4) is the new N.E. 3. Suppose now that both firms cannot produce and sell more than 3 units (K1 = K2 = 3). ● Can (P1, P2) = (4,4) still be a Nash equilibrium of this third game? If charging P1 = 5 (considering that Firm 2 has still P2 = 4), Firm 1 still make positive profits: ❖ Q = 10 - K2 = 10 - 3 = 7 ⇒ residual demand faced by firm 1 (QR) = 7 − P1 = 7 - 5 = 2 ❖ π1 = (5 − 2)*2 = 6: Firm 1 has a weak incentive to deviate, where deviation leads exactly to the same profit (same for π2). Thus, (P1, P2) = (4,4) is not a N.E. anymore. ● Consider the equilibrium pair (P1, P2) = (5,5). If charging P1 = 6, Firm 1 still makes a profit. ❖ Q = 10 - K2 = 10 - 3 = 7 ⇒ residual demand faced by firm 1 (QR) = 7 − P1 = 7 - 6 = 1 ❖ π1 = (6 − 2)*1 = 4: Firm 1 does NOT have incentive to deviate, because it would have less profit than in the previous case. Thus, (P1, P2) = (5,5) is a N.E. Exercise 2 Consider a market characterized by the presence of a leader L and two follower firms, F1 and F2. The two followers observe the leader of production and compete simultaneously in quantity (Stackelberg). Market demand is p = 1 − Q, where Q represents the total amount of product (Q = qL + q1 + q2). The three firms use the same production technology characterized by a MC = c. 1. Determine the output of F1 and F2 given the amount of production decided by the leader. The game is solved by backward induction starting from the second period, when the followers move. Their profit maximization objective is: 2. Determine the market equilibrium (equilibrium price, quantities, and profits). Solving the system of equations, we obtain (1) → → i.e., a firms find it optimal to collude if it is patient enough. means that a firm has a high evaluation for the future stream of profits Therefore, it forgoes short-term high profit from deviation for long-term rewards. As mentioned before, δ ∈ (0, 1) is the discount factor, that is, today’s value of a euro gained tomorrow (fattore di sconto). It relates to ▶ individual’s (intertemporal) preferences opportunity cost of time ▶ interest rate: if I lend one euro today, I get (1 + r) euro tomorrow, where r > 0 Formally, Vt = Vt+1 ⇔ Vt+1 = (1 + r )Vt , which implies that (fattore di attualizzazione) Grim Trigger Strategy Assume for simplicity that pi = pj ⇒ Di = D(p)/n ▶ charge a monopoly price pm from now on if all collude and obtain πi C = πm/n in each period If a firm deviates today (setting a slightly lower price pm − ε): implement a punishment ▶ charge a Bertrand price p = c from tomorrow onward - this is the harshest punishment for a firm, as it leads to zero profit (Nash reversion) Does firm i have an incentive to collude against the possibility to get (almost) monopoly (=deviation) profits and be punished with Nash reversion? YES, if and only if Therefore, collusion is sustainable if and only if FACTORS FACILITATING COLLUSION Understanding them is crucial for antitrust authorities in order to intervene. Each firm compares the immediate gains from deviating (take advantage of high profits) with the reduction of profit suffered as a result of the reaction. A) Structural factors 1. Number of companies and concentration: ceteris paribus, the smaller the number of firms in an industry, the easier it is to collude (when firms are of similar size). To see why, recall that . The derivative with respect to n is Therefore, - Increasing the number of firms makes the critical discount factor higher - Since collusion is sustainable if and only if , then increasing the number of firms makes the collusive constraint more stringent → collusion is relatively less sustainable 2. Entry: the easier it is to enter the market (low barriers to entry), the more difficult to sustain collusion. To see why, we note that: - Profits of collusion “attract new entrants” - Entrants may not want to collude (because “asymmetric”), but in any case, the number of forms increases, and this makes collusion less stable 3. Cross-ownership, cross-shareholdings and other links between rivals: they increase the possibility knowing each other, which makes collusion more sustainable (ceteris paribus). Example: A’s management can be informed on rivals’ deviation from the collusive agreement. This can make the punishment stronger, thereby more likely to deter (=evitare) deviation. 4. Regularity and frequency of orders: companies receiving them are more likely collude (they want to divide equally the orders to operate efficiently). To see why, suppose the industry today receives an unexpected, large and unique order: - Di = k*D(p) with k > 1 (and current deviation profits Π = kπ(p)). The firms have a strong temptation to deviate (to enjoy the full benefit of the order). Collusion is sustainable if The higher k, the more stringent the participation constraint , the less stable the cartel. 5. Buyers’ concentration: the greater the degree of buyer concentration, the stronger their bargaining power (=potere contrattuale), the less likely sellers are able to collude. In a monopsony, the price is set by the buyer: a large buyer can induce the breakdown of the cartel agreement (similar to secret price cuts). 6. Evolution of demand - If the demand is auto-correlated (trend), a positive (or negative) trend in demand reduces (or increases) firms’ incentive to deviate. - If the demand is not auto-correlated (shock), an increase rises firms’ incentive to deviate. It is indeed subject to shocks, which is as if receiving an unexpected and isolated order. Formally, the stability of the demand facilitates the observability of rivals’ strategies, which makes collusion more likely to be stable. ▶ Suppose that, from tomorrow on, firms expect a demand increase/decline: θtD(p) for any t > 1 (and profits at time t are θtπ(m) If θ > 1 (continuous growth), the participation constraint is less stringent and collusion more stable; if θ < 1 (continuous decline), the participation constraint is more stringent and collusion less stable. 7. Symmetry: firms of similar size, with similar characteristics, make a cartel more sustainable. Similar firms may more easily find an agreement (they have common characteristics). Symmetry matters. Consider a duopoly: ▪ Firm 1 has a market share λ, while Firm 2 has a market share (1 − λ), with λ > ½ (not equal!) Note that collusion is sustainable if and only if both firms have an incentive to collude δ ≥ max{δ1−λ, δλ}. Since by assumption λ > 1/2, then δ1−λ > δλ. Therefore, the small firms faces a more stringent constraint and has more incentive to deviate: in case of deviation, it serves the entire market (for a period). The larger the asymmetry ⇒ the more divergent incentives ⇒ the more unstable the agreement 8. Multimarket contact: repeated interactions (across/within markets) make collusion more sustainable. When firms meet the same rivals in multiple markets, they learn and coordinate Firms asymmetries are reduced, transparency increases and there is possibility of retaliation. B) Market transparency 1. Demand fluctuations are observable: they make collusion more likely to be sustainable. The intuition is simple: firms’ ability to detect deviations makes a collusive agreement stronger - A firm receiving unexpectedly low demand understands whether it’s the result of a secret price cut of a member of the cartel (hence it may retaliate) or it’s the signal of an effective reduction in market demand. 2. Observability of rivals’ strategies: it makes collusion more likely to be sustainable. There are two ways to increase cartel affordability: price transparency and exchange of information. Exercise Consider a duopoly. Firms produce homogeneous goods and compete in quantities for an infinite number of periods. Market demand is p = 1 – Q, where Q is the total quantity. Firms have identical marginal costs, normalized to zero for simplicity. The common discount factor is δ < 1. 3. Explain why the collusive equilibrium described above is not sustainable in a one-shot game. Determine the optimal deviation of each firm from the collusive equilibrium. Let’s write the game in Normal Form, and use the above output We need to identify the payoffs in the asymmetric case. Best response functions (from point 1.): If Firm 1 produces the Collusive output q1* = 1/4, Firm 2’s reaction function is to set q2*(1/4) = 3/8 Total output is Q = 3/8 + 1/4 = 5/8. Firm’s profits are equal to The matrix payoff is completed as follows: Solving the game, (Collude, Collude) is not a N.E. There is always an incentive to deviate, and the unique N.E. is to play (Cournot, Cournot). Recall that cooperation can be sustainable if and only if the game is repeated a large number of times (this is a necessary but not sufficient condition). LESSON 6: VERTICAL RELATIONS Very often firms do not sell only to final consumers but also to other firms. Vertical relations → relation between two firms in sequence along the value chain. Examples: ▶ Cosmetic firms buy their inputs from chemical companies and sell the final product to retailers. ▶ Stellantis buys tires from Pirelli and then sell cars to downstream users through its resellers Upstream suppliers produce products. Downstream retailers intermediate products; their market power can decide the price that the upstream suppliers will set. They determine the terms of trade and the prices of products. Many variables of consumer demand are beyond sellers’ control (retail/sales price, advertising…). Retailers compete with each other: they care about wholesale price paid (=prezzo all’ingrosso). The number of customers/retailers is relatively small: less market power for suppliers. THE WHOLESALE MODEL In this classic model of trade, suppliers choose wholesale prices and retailers choose retail prices. Let’s consider the extreme case, where the upstream is a monopolist producer (M) and the downstream is a monopolist retailer (R). Assume the simplest production technology: MC = c. p is the price for consumers set by R, w the wholesale price set by M to R. D(p) is the market demand. BENCHMARK: VERTICAL INTEGRATION Let’s consider as a benchmark the case in which both M and R are within the same company (M&R). In this case, M&R sets price p to consumers, and does not have to buy input, i.e., w = 0. The problem is to set p so that . As in all monopoly problems, the price is the monopoly price p = pm(c) → of course, w is irrelevant. In the vertical chain when R buys from M, the price w(c) which R pays to M is > c (since M is a monopoly with respect to R). Since the monopoly price (w) is increasing in c, the final price of the retailer is higher than the price charged by a vertical integrated monopolist. Therefore, there are two margins: each monopolist makes a margin, which boost each firm’s price ▶ Double marginalization implies a loss of profits: the sum of R and M’s profit is lower than the vertical integrated profit → there is nothing worse than a chain of monopolies The final price for the consumers (the retail price) is larger than the monopolist has set before. Case 2: RPM In order to induce R’s investment, M sets RPM = 80 and w = 50 ▶ Is this vertical restraint successful in stimulating R’s investment? Each retailer must decide whether to Invest (I) or not (N). We have to consider 4 possible cases: ΠR(I,I) = (p − w)q − 500 = (80 − 50)*30 − 500 = 400 → Q = 140 – 80 = 60 ; q = 60/2 = 30 ΠR(I,N) = (p − w)q − 500 = (80 − 50)*30 − 500 = 400 → I invest, the other benefits (free riding) ΠR(N,I) = (p − w)q = (80 − 50)*30 = 900 → I don’t invest (free riding), the other does ΠR(N,N) = (p − w)q = (80 − 50)*10 = 300 → Q = 100 – 80 = 20 ; q = 20/2 = 10 Solving the game Main results: - at least one R invests - all firms are better-off than without RPM - ΠP = (50 − 20)*60 = 1800 (consumers too) Anticompetitive RPM Note that RPM and exclusive territories are clearly anticompetitive practices. ▶ Antitrust authorities have to evaluate whether the gain in efficiency (more investment) justifies the restriction on competition. Example: in 1999 Panasonic in US agreed to reimburse something like 16 million $ to consumers as a contribution for the higher expenses paid (up to 10% more) due to price fixing (Panasonic threatened to stop supplying its retailers who sold for less). How to avoid free riding without RPM? To deal with free riding without incurring in antitrust violations, a manufacturer: - May advertise on behalf of its distributors - May directly support retailers lowering their costs of investment MERGERS AND ACQUISITIONS (M&A) A radical strategy to solve the problem with Inter-firm vertical relations is via M&A. ADVANTAGES DISADVANTAGES Avoid double marginalization or free riding Direct cost of merger Full control of the demand in the downstream market Increase business complexity for management Reduce transaction costs More difficult to monitor the dealers’ efforts Reduce competition in the market and increase market power Antitrust concerns EXERCISE 1 Consider a monopoly U (upstream) which distributes its product through a retailer D (downstream); D buys the product at a unit price t and then sells it on the retail market. Firm U incurs MC = c and a fixed cost F. The retailer does not incur in additional cost (the only marginal cost it bears is the wholesale price). Market demand is Q(P) = 10 − P. 1.Determine the equilibrium retail and wholesale prices and firms’ profits. We solve the game by backward induction. In the second period, the profit of D (the retailer) is Πd = (p − t)*Q = (10 − Q − t)*Q where P = 10 – Q F.o.c.: 10 – 2Q – t = 0 → Q(t) = (10 – t)/2 t = 10 – 2Q We move now to the first stage, in which the upstream firm decides about the wholesale price. ; F.o.c. with respect to t: → t* = (10 + c)/2 Plugging t* into Q(t), we have Q* = (10 – c)/4 and p* = (30 + c)/4 ; 2. Can the upstream firm do better than this? How? Yes, the upstream firm can do better by just setting a contract different from the simple linear wholesale price: a two-part tariff. Denote T such a tariff, such that T = t × Q + f The wholesale price and fixed fee is set such that t* = c and In this way, the upstream firm (U) uses f to extract all the profits of the downstream firm. 3. Suppose now that two new firms enter the retail market; the entrants have the same technology of the incumbent distributor. The competition is on quantity (à la Cournot). Determine the equilibrium prices and profits of D and U. Explain and interpret your result. Let’s look at the problem of one distributor, say D1 (note that for symmetry, thing also apply to the other distributors): Π1 d = (p − t)Q = (10 − q1 − q2 − q3 − t)q1 F.o.c.: For symmetry, q1 = q2 = q3: 10 – 4q1 – t = 0 → q1(t) = q2(t) = q3(t) = (10 – t)/4 and Q(t) = 3(10 – t)/4 4. How would the result change if the entry process would continue indefinitely? Explain intuitively how the wholesale price and the upstream firm’s profits would change in this case. Moving to the upstream market, the profit function is F.o.c.: with respect to t and equalize it to 0, we have t* = (10 + c)/2 Replacing t* into q1*, q2* and q3*, we have Note that p* = (50 + 3c)/8, which is lower than when there is a downstream profit. Downstream competition reduces the market price as the double marginalization problem is less severe; the upstream firm gets larger because of the expanded output. EXERCISE 2 Firm D is a monopolistic retailer (it controls the network of supermarkets). Two firms, U1 and U2, produce a homogeneous good and need access to D’s network of supermarkets to distribute their products. U1 and U2 have the same MCU = 10 and sell their product to D at w1 and w2 respectively (they compete in prices). In addition to the wholesale price, D incurs MCD = 6. Market demand is equal to Q = 100 − P (P is the retail price). D buys from the producer charging the lowest price. (ASSUMPTION: in case w1 = w2, D buys half from U1 and half from U2). 1. Determine the equilibrium prices (wholesale and final) and profits in case of vertical separation. Due to the upstream competition, D buys from the upstream manufacturer that offers the cheapest price: thus, each manufacturer has an incentive to undercut the rival and the equilibrium wholesale price if w1* = w2* = 10 (= MCU) → their profits are equal to the Bertrand profits... Π1 U = Π2 U = 0 The profit of the downstream retailer is Πd = (P − 6 – w*)Q = (84 − Q)Q F.o.c.: 84 – 2Q = 0 → Q* = 42 ; P* = 100 – 42 = 58 ; Πd = (58 – 6 – 10)*42 = 42*42 = 1764 2. D is considering integrating upstream, buying one of the two Us. If you were the management of D, would you integrate (= how much D would be willing to pay to acquire control of U1 or U2)? If so, with which manufacturer? As w* = 10, which is the lowest possible wholesale price D can pay, D does not have any incentive to integrate! → the first markup is 0 and D could not do any better by vertically integrating backwards. 3. Go back to the first step. U1 and U2 are considering the possibility to agree on charging D the same price. What is this price? Can this be an equilibrium? First, we should find the wholesale price charged by the cartel (wc), which behaves like an upstream monopolist. Then, let’s solve the model backwards, starting from D that chooses Q to maximize Πd = (P − 6 − wc)Q = (94 − wc)Q F.o.c.: 94 – wc – 2Q = 0 → Q(wc) = (94 - wc)/2 In the first stage, the cartel maximize joint profits F.o.c.: 94 – 2wc + 10 = 0 → wc* = 52 ; Q* = (94 – 52)/2 = 21 ; P* = 100 – 21 = 79 Profit of the cartel: Πcartel = (52 − 10)*21 = 882 → since it’s split equally, then Π1 U = Π2 U = 441 D obtains Πd = (79 − 6 − 52)*21 = 21∗21 = 441 Note that wc = 52 cannot be an equilibrium: Both U1 and U2 have an incentive to undercut the rival, by just setting w = 52 - ϵ, with ϵ → 0 and serve the entire market. 4. Suppose that U1 an U2 cannot produce more than 20 units: i) are wholesale prices found in 1. still an equilibrium? ii) (PREMIUM) verify that w1 = w2 = 14 (U1 and U2 are at capacity) is not a N.E. First, the previous solutions are not compatible with the constraint since q = 21 > 20 ⇒ w* = 10 cannot be an equilibrium: both firms may raise the price unilaterally and still face a positive residual demand, as the other firm cannot solve the entire market This model of product differentiation was proposed by Hotelling in 1928. Main assumptions: 1. Each firm is located at the extreme of a unit length segment (e.g., firm 1 in 0, firm 2 in 1); 2. Consumers are uniformly distributed along the segment (i.e., this is a simplifying assumption) ▶ in each point on the line there is always 1 consumer. ▶ each point of the segment represents one single consumer with a specific preference 3. Transportation costs equal to t ▶ This is a preference mismatch, i.e., the cost of not buying the preferred variety (the least distant) ▶ e.g., a consumer closer to 0 must travel (and spend travel cost t) to buy product from firm at 1 4. The utility of the consumer located at the point x ∈ [0, 1] who buys at a price p is v − tx – p Consumer demand Consider the utility of an individual located at x ∈ [0, 1]. If buying from ▶ firm 1, s/he obtains U1 = v − tx − p1 ; tx = travel cost from moving from x to 0. ▶ firm 2, s/he obtains U2 = v − t(1 − x) − p2 ; t(1 − x) = travel cost from moving from x to 1. Consumers in x ∈ [0, 1] The indifferent consumer, denoted by x*, is such that U1 = U2: v – tx* − p1 = v − t(1 – x*) − p2. This means that ▶ those to the left of x*, i.e., x < x*, buy from firm 1 ⇒ D1(p1, p2) ▶ those to the right of x*, i.e., x > x*, buy from firm 2 ⇒ D2(p2, p1) How to find the indifferent consumer? Solving the equation for x, Therefore, demand are given by How do firms set prices? Moving to the first stage of the game, both firms set prices to maximize profits simultaneously. max π1 = (p1 − c)D1(p1, p2) F.o.c.: max π2 = (p2 − c)D2(p2, p1) Reaction functions (symmetric): p1 = p2 = p = t + c → at equilibrium, π1 = π2 = π = t/2 Note that the transportation cost matters: this is a measure of preference mismatch. ▪ t = 0: it is costlessly for a consumer to choose the least preferred option ⇒ weak preferences: products are homogenous and there is Bertrand competition. Thus p = MC ▪ t > 0: it is costly for a consumer to choose the least preferred option ⇒ preferences for a product, thus there is differentiation and p > MC Two intuitions: the larger t, - the less intense the competition, as the consumers have “strong” preferences for one product and moving to the most distant one is very costly - the more differentiated the products, the higher the market power and the higher the profits PRODUCT POSITIONING The choice of where to position a product is a strategic one, in physical terms and/or in terms of ”features/characteristics” of the product. So far, we assumed firms’ location as given (at the extremes of the segment): products characteristics were given. What if firms may choose where to locate? Example 1: petrol stations must decide where to locate before competing (close or far apart?) Example 2: cosmetic firms must decide the characteristics of their products (the fragrance) The “distance” with respect to competitors must be defined with respect to their choices on the characteristics of their products (strategic interdependence). Note: while prices adjust simultaneously to market changes (short run variable), product characteristics are more difficult to change (long run variables) - Firms choose their location (characteristics) first and then compete in prices What equilibrium locations will firms choose in the first stage, knowing that once location has been chosen, they play the Hotelling game? The choice of location) comes from two opposite forces: Direct effect: business-stealing effect What happens to Firm 1’s profit when changing location for given prices by the two firms? Suppose Firm 2 is located at 2/3 of the segment. Where should Firm 1 locate itself? - The closer Firm 1 is located to firm 2, the greater its demand (it steals more customers to the rival): ⇒ for given prices, being closer to the rival implies higher profit (business stealing) Strategic effect What happens to Firm 1’s profit because a change in the location determines a change in price too? Once firms are located, price competition begins: the closer the firms, the more substitutable the products, and therefore the more intense the competition. Limit case: in the case location, no product differentiation ⇒ Bertrand competition - the closer the two firms are, the more intense the price competition, the lower the profits Summing up, changing location entails a trade-off between two opposite effects 1. Direct effect induces firms to locate close to each other (- product differentiation) 2. Strategic effect leads firms to locate far apart to soften the competition (+ product differentiation) Equilibrium product positioning is a balance between these two effects: ▶ If price competition is intense, firms differentiate (to survive); ▶ If price competition is not intense, firms do not differentiate (to steal customers to the rival) NOTE: product differentiation allows firms to gain market power to have some control on the price of their product (in order to escape the Bertrand Paradox). Nevertheless, what is relevant is not product differentiation per-se, BUT the fact that consumers treat products differently. In many cases: products are identical, but individuals treat them as they were different. A similar framework to that of product differentiation can occur when consumers 1. have IMPERFECT INFORMATION Suppose that consumers do not know in advance the price for a certain product charged by each shop in town: to know the price, they need to visit the shop before: this has a cost (search cost) - With positive search costs, the equilibrium price may be the monopolistic level even with perfectly homogeneous products! Let u be the individuals’ WTP for a product and s be the search cost of each visit: net of the search cost, u-s is the net WTP for the good. ▶ Each consumer chooses which shop to visit randomly ▶ If p ≤= u − s the consumer buys, otherwise s/he visits a new shop Assumption: once a shop has been visited and its price has been known, the consumer expects that the other shops charge the same price (similar result if consumers expect a lower price). Therefore, • if s = 0 : consumers visit all the shops, they become perfectly informed: Bertrand competition • if s > 0 : shops charge monopoly price, although products are not differentiated This is a very simple model, but it nicely shows that adding a small friction result might change. 2. incur in SWITCHING COSTS Often consumers must pay a cost to switch between suppliers. The effect of switching cost is very much the same as before: if s is now the cost of changing supplier, then the equilibrium may be again the monopoly price: up to a certain level, it pays for a firm to charge a high price (monopoly) without losing customers. The greater the value of search costs or switching costs, the greater the sellers’ market power. For this reason, very often firms tend to artificially create switching costs (frequent flyer programs, incompatible software, etc) and search cost (random prices) Note: switching and search costs may explain price dispersion: different prices for the same product. EXERCISE Two firms are active in the market for bottled water: A produces plain water and B produces carbonated water. Use the Hotelling line to represent the possible variety of products with a unit length segment (from natural water to the very carbonated water); consumers are uniformly distributed in [0, 1] and A and B are located at the ends of the segment. Let u = 10 be the gross utility that each consumer derives from buying a bottle of water; ceteris paribus each consumer buys from the firm closest to her preferences: each consumer incurs a ”travel cost” equal to 2, for each hypothetical unit of distance travelled to purchase the water. Both firms have MC = 1. Denote by pi the price per bottle by firm i (i = a, b). LESSON 8 – MARKET DYNAMICS: ENTRY, EXIT, MERGERS So far, we have considered the presence of an exogeneous number of firms. In the reality: - Active firms may exit the market (can fold up, discontinue a particular product, leave a specific geographic market segment…) - Entrant firms may enter the market (can be a brand-new firm or be an established firm that is diversifying into a new product/market). Why are entry and exit strategic decisions? Example: an incumbent firm may “react” aggressively to the entry of a new rival: the reaction may induce the entrants firm to exit the market. Strategic interdependence: entrant firms must take reactions into account. Example: when Italo entered the High-Speed Rail market, Trenitalia reduced prices and increased supply. ENTRY DETERRENCE Consider the following simple market: Firm 1 is the incumbent and Firm 2 is the entrant. Assume that in case of entry, there is a Cournot duopoly: ▶ Firm 1 announces its output level q1. ▶ Firm 2 observes q1 and decides whether to enter or not (if yes: it incurs a fixed cost of entry). Formally, 2 decides its output “à la Cournot”: using its reaction function q2(q1) ▶ Firm 1 knows 2’s reaction function. Given q1, firm 1 is able to anticipate firm 2 output level. ▶ The larger q1, the smaller q2 → by choosing q1, the incumbent affects q2 The incumbent finds it optimal to increase production to deter entry ▶ Firm 1 sets a quantity such that Firm 2 gets 0 profit It’s not always optimal for 1 to discourage entry by producing q1 = q1 D: If the cost of entry is low, q1D is very large, hence π1(q1 D) are very low. There are cases in which Firm 1’s optimal strategy is to accommodate 2’s entry (entry accommodation) → too costly to deter entry There are also cases in which the firm does not need to anything If the cost of entry is large, Firm 1 may ignore 2’s threat of entry (entry is blockaded) ACCOMMODATED ENTRY BLOCKADED ENTRY The Incumbent leaves room to the entrant the incumbent may ignore the threat of entry → quantity War is too costly CREDIBILITY AND COMMITMENT ▶ q1D represents 1’s announcement about its output level in case of entry. ▶ The ISSUE: is q1D credible? Commitment is the art of limiting your options for strategic advantage. The commitment must be known and believed by others. In case of entry, by definition the best thing that Firm 1 can do is to produce the Cournot output (it maximizes duopoly profit) q1 = q1(q2) < q1D - q1 = q1D is not CREDIBLE ⇒ Remember the concept of SPNE? - Firm 2 anticipates that in case of entry q1 = q1(q2), henceforth it enters! How to make q1 D a credible strategy? Suppose that, before entry, Firm 1 must choose its capacity: capacity costs are very large and sunk. In case Firm 1 later chooses to produce less than its capacity, it cannot avoid the cost ▶ Investment in capacity expansion may make the announcement q1=q1 D credible when capacity costs are high and sunk → It is costly for Firm 1 to produce less (opportunity cost) ALTERNATIVE PREEMTION STRATEGY The previous model involves deterrence by capacity expansion. However, firms may adopt other strategies than capacity expansion, and incumbents may deter entry through investments in service quality, marketing/brand recognition/advertising, cost reductions, product proliferation. Example: the ready-to-eat breakfast cereals market in the U.S. is characterized by 1. firms enjoying high profits, 2. fixed number of firms over time (no entry) 3. many varieties of cornflakes for sale Compared to other industries: - Costs of entry are not so large (essentially marketing costs) - Mature technology (little product/process innovation) Given these characteristics, entry should be rather easy (yet market concentration is persistent). 4 big incumbent firms; they have stable market shares BUT increasing number of products; large profits. 1. Why high profits do not attract new firms? 2. Why has the number of products increased while the number of firms has stayed the same? Answer: product proliferation as an entry deterrence mechanism PRODUCT PROLIFERATION IN HOTELLING Assume that cereals are defined by only one characteristic: sweetness. Individuals are located uniformly in (0;1) and they have mass 1. Firms choose which and how many varieties. Firm 1 moves first and anticipates firm 2’s decision. For simplicity, there is no price competition (that is, P1 = P2 = P price is given) • If Firm 1 produces only one variety of cereals, the best thing it can do is to produce an intermediate variety; Firm 2 at best can steal half of the market by entering with a variety just to the left/right of 1/2). Let F be the cost of creating a new variety of cereals and assume that F < 1/2P. If Firm 1 produces only one variety: The entrant (Firm 2) will enter the market with a variety very close to ½; in this case Firm 2 will cover 50% of the market and Π2 = Π1 = 1/2P – F. If Firm 1 produces two varieties at 1/4 e 3/4 (one bitter and one sweeter), the maximum market share the entrant can reach is ¼. Firm 2 will enter with an intermediate variety ½, and it will sell its product to the individuals located between 3/8 and 5/8. Hence its profits: Π2 = 1/4P – F. If F > 1/4P, Firm 2 does not enter, as in case of entry, it cannot cover its entry costs. In this case, Firm 1 remains monopoly and gets: . If F < 1/2 P Firm 1 is better off by adopting a proliferation strategy to deter entry: Profits with 2 varieties are larger than those with only one variety (and entry occurs). PREDATION (EXIT) If a firm fails to deter entry, it can still encourage/induce rivals exit → predatory pricing Example 1: in 1997 EasyJet enters the London-Amsterdam route. KLM (incumbent) reacted by matching EasyJet prices. August: EJ threatens to take KLM to court over what it alleges are ”unlawful attempts to exclude easyJet from the market”. Autumn: KLM put an end to its aggressive strategy. N.B.: There is little doubt that KLM’s tactics were aimed at forcing EasyJet exit. If the KLM strategy has lasted a little longer, EasyJet would have found itself forced to exit. Example 2: in April 1996 Spirit, the biggest UltraLowCost airline company in the US, enters the Detroit-Boston route with a price 70% lower than that of the incumbent (Northwest). Northwest reacted by further dropping its prices. Spirit started an antitrust lawsuit against Northwest. The US antitrust authority sided with the defendant and Northwest won a summary judgement. ▶ In the US courts tend to be sceptical about allegations of predatory pricing. Merger waves This is an empirical puzzle: Mergers seem to happen in rushes or “waves” rather than gradually over time. Often a wave is ignited by an exogenous event (market liberalization, development of a new production technology, radical change in market functioning). Example: aviation sector in the U.S.: after liberalization (late ’80s-’90s) merger wave /alliances (more than 500). But this is not enough to explain this strong regularity… Consider a Cournot model with 4 active firms ▶ Demand function Q = 150 − p ▶ MC = 30 and F = 120 Firms 3 and 4 are taking into consideration a merger. The same do Firms 1 and 2. • If firms 1 and 2 do not merge, the merger between 3 & 4 is not convenient. → Per firm profits decrease from 456 to 780/2 = 390. • if firms 1 and 2 do merge, the merger between 3 & 4 is convenient → the merged entity obtains 1480, which is 740 for each firm of the merger In this simple example, two firms benefit from the merger when the other two firms merge as well. This result is not true in general, but it offers an intuition for the empirical observation: ▶ outsiders generally benefit from a merger (unless the merger obtains large cost synergies, outsiders produce and sell more output) ⇒ this reduces their incentives to merge ▶ but if the market is not too competitive, this effect is weak ⇒ outsiders may take greater advantage from the merger by merging themselves Overall, a merger wave ▪ may be initiated by an exogenous event, as market deregulation or technical progress, etc ▪ it may initiate endogenously: following a merger, other firms merge as well EXERCISE Consider a monopolistic industry (only firm 1 is active). The demand is q = 100 − p where q is the total output produced by the firm. The cost function is c(q) = 2q (there are no fixed costs). 1. Calculate the price and the quantity produced by the monopoly. Consider the monopolist, who maximizes ΠM = p × q − c(q) = (100 − q)q − 2q. From the f.o.c. and the standard maximization problem (not reported!), pM = 51, qM = 49. → In total, ΠM = 2401. 2. Firm 2 is considering whether or not to enter the market; in case of entry, competition occurs à la Cournot. Firm 2 has a total cost c2(q) = 10q and in case of entry it incurs a fixed cost of entry equal to F. Determine what Firm 2 will decide to do. In case of entry, there is Cournot competition, with the following maximization problem: • Π1 = (100 − q1 − q2 − 2)q1 and Π2 = (100 − q1 − q2 − 10)q2 – F F.o.c.: At the equilibrium, Entry by Firm 2 occurs if F < 6724/9. Else, Firm 2 does not enter and there is only a monopolist. 3. In order to deter rival’s entry, Firm 1 announces that if Firm 2 will enter it will increase its output. Determine the level of output to achieve deterrence. The output that achieves deterrence is q1 D and this is defined in a way that, in case of entry, Firm 2’s profit are driven down to 0. Recall that the profit of Firm 2 is Π2 = (100 − q1 − q2 − 10)q2 − F Recall that if entry occurs, the best response of Firm 2 is To identify the deterrence quantity q1, let us plug q2(q1) into the profit function of Firm 2, such that = The deterrence strategy pursued by Firm 1 is to set q1 D such that Π2(q1) = 0. Therefore 4. Suppose that F = 0 (no cost of entry). Is the announcement credible? If F = 0, then q1 D = 90, which implies that the price is just equal to p = 10. In case of deterrence, Firm 1 does not face any competition, but the quantity produced is again equal to 90. Thus, profits are equal to Π1 D = (10 − 2) × 90 = 720. The threat to produce q1 D = 90 must be credible. ▶ In the current scenario one can observe that Π1 D < Π1 accommodation as 720 < 1248. ⇒ Entry therefore will occur, and the monopolist will accommodate the entrant VERTICAL MERGERS (between firms in different or complementary stages of the value chain) CONGLOMERATE MERGERS (between firms that are industry-unrelated) LESSON 9 - ADVERTISING Many markets are characterized by intense advertising activity: • In the US there are firms that spend more than 20% of the value of their sales (Procter & Gamble, McDonald’s, Kelloggs, Warner-Lambert, …) → large advertising-to-revenue ratio • Investments in ads in many cases are larger than investments in R&D What are the implications of advertising on competition? Goods/products can be distinguished into two categories: ● Search good: one whose features and characteristics the consumer may know before purchase (e.g., the features of a PC can be determined by simple inspection) ● Experience good: one whose features can be ascertained only upon consumption (e.g., food, wine, software programs, beauty products in general) N.B.: this distinction is not entirely correct, because it’s often difficult to classify goods as experience or search ones. Moreover, advertising may be of both types together. Advertising expenditures can be classified into: ● Informative: describes the product’s existence, its characteristics and selling price ● Persuasive: aimed at changing consumers’ preferences (“our wine tastes better”) Which form of advertising is the most important? Which one is more efficient for society? - search goods⇒ informative advertising - experience goods⇒ persuasive advertising An important index is the advertising-to-revenue ratio (a/R): according to estimations, it’s three times greater for experience goods than for search goods. ADVERTISING AS A SIGNALING DEVICE Often advertising campaigns carry very little information other than the simple existence of the product, but they cost a lot to producers (e.g., ad campaigns for Chiara Ferragni’s water) Can these campaigns be considered informative? → YES! At least indirectly.... Spencer’s signaling game The players consist of the (potential) employees and the employer: the latter doesn’t know the candidates’ ability, (S)he only observes their CVs (Information asymmetry) How can a candidate inform the employer? 1. Via a signal about her/his ability level (e.g., acquiring COSTLY education credentials) 2.The informational value of the education comes from the fact that it is difficult for low-ability employees to obtain education credentials (e.g., earn a degree with honors), therefore the employer believes education to be positively correlated with having greater ability ⇒ only HIGH-quality employees obtain credentials (CREDIBILITY) thus, by observing the credentials the employer is able to distinguish low-ability workers from high-ability ones. NOTE: usually case 1 and 2 coexist Summing up: as markets become more competitive • each firm’s margin decreases (a/R ↓) • each firm captures a lower share of the demand-increasing effect of advertisement (a/R ↓) • each firm induces a shift of consumers demand toward its product (a/R ↑) The empirical evidence of the effect of concentration on advertisement intensity is ambiguous as well. ADVERTISING AND THE NATURE (INTENSITY) OF PRICE COMPETITION The issue: How does advertising change the nature of price competition among firms? Again, there are two opposite stories... 1) Especially when advertisement is informative and is aimed at informing people about product characteristics - it increases differentiation among products that would otherwise be seen as similar - it softens price competition (think to pure Bertrand oligopoly vs Hotelling) 2) If advertisement increases consumers’ knowledge about different prices charged by firms, - it reduces consumers’ ”search costs”: it is easier to make price comparisons and to shop at the lowest price - it increases price competition (especially if products are homogeneous) Both case 1 and 2 have been empirically supported: the impact of advertisement on the nature of competition is ambiguous. Exercise Trutel is a leader in mobile TLC. Market demand is: Q(p,a) = 10 + 2a − p, where total demand depends on the price p and the amount of advertisement a, measured in hours of TV commercials (note: advertising stimulates demand). TC(Q) = Q2; the total cost of advertisement is equal to A = a2 1. Determine the equilibrium price and amount of advertisement. Then, specify the amount of commercials the firm is willing to buy. 2. What is the A/R at the equilibrium? 3. Calculate the price elasticity of demand and the demand elasticity to advertisement and show that in equilibrium the Dorfman-Steiner condition holds. LESSON 10 - RESEARCH AND DEVELOPMENT Technical progress is a key factor for economic growth: firms continually introduce new products and new production processes, and the results of R&D effort. Main issues: - Why does the amount of investment in R&D vary significantly from sector to sector? - Is there a relationship between R&D and industrial structure? - Are today’s R&D leaders likely to remain leaders in the future? - What is the impact of R&D on market structure? - What can policymakers do to favor investments in R&D? PRODUCT VS PROCESS INNOVATION Process innovation: aimed at creating more efficient production technologies (that allow reducing the cost of production) Product innovation: aimed at developing new products Q1:Which market structure induces firms to invest more in R&D (and to innovate)? Schumpeterian view Large incumbents have more incentive to invest in R&D • Incumbents are better positioned to innovate than smaller firms. •Smaller firms often lack incumbents’ (knowledge of the market, experience, access to funding, distribution network) → yet, incumbents (monopolists) often fail to innovate Some examples of incumbents’ failure: • Polaroid: leader in the production of cameras during the ’80s, failed to innovate in digital technology and went bankrupt in 2001 • Kodak: very slow in transitioning to the digital photography market — 1990s • Major TLC companies and VOIP in the early 2000s • Sony (producer of walkman – analog technology) in the portable MP3 player market in 2002-2004 Arrow view Small/entrant firms have a greater incentive to innovate in competitive markets (where firms are small in size). Several path-breaking innovations are realized by small start-ups! Examples: • Google: L. Page and S. Brin invented the search engine based on PageRank in 1998 • 3D Systems: invented 3D printing in 1986 (a process to fabricate physical objects using inputs from Computer-Aided Designs) Process innovation and competition The competitive firm gets zero profits BEFORE innovation, but it serves the whole market if it innovates. Monopolistic firm makes money even before the innovation: this is called the ”Replacement effect” (M “replaces” itself): The firm with market power has less incentive to innovate because of the profits that it already obtains before the investment. Common perception⇒ competition stimulates innovation THE TWO VIEWS The two views are not necessarily alternatives to one another ❖ In the graphical analysis we have implicitly assumed that the innovator charges the lowest price once innovated and it gets the whole market... that is, it is like saying that the firm becomes a monopoly! If innovation were immediately imitated (the market is perfectly competitive also after the innovation): the firm does not have the incentive to innovate ❖ No threat of entry was considered: if the current monopolist may lose profits if it doesn’t innovate while a potential entrant does, then the first may have a stronger incentive to invest Schumpeter was not so wrong… These topics are extremely debated today. How to protect innovation? Which protection if innovation is sequential? It is crucial to distinguish between: • ability to innovate: large/dominant firms (they have resources) • incentive to innovate: small/competitive firms (they have more to gain in relative terms) An important aspect to take into account is the dynamics of innovation: • perfect competition implies an efficient use of resources, BUT.... in static terms! This may no longer be true when the dynamic aspects are taken into account. Dynamic aspects There are two companies: an incumbent/monopolist (M) which is already active in the market, and a potential rival (R). A laboratory has discovered and patented an invention (=innovation) but it is not able to exploit it. It decides to put it on sale (= to license the use of the innovation). The model is dynamic. We have two stages/scenarios: pre-innovation (monopoly) and post-innovation. R enters only if it buys the innovation. - Who is willing to pay more for the innovation? First scenario M is gaining πM. If it buys the innovation it continues to obtain πM (gross of the payment for the innovation). If the rival buys the innovation, R enters the market, competition occurs and the two firms obtain πD ; firm M keeps staying on the market even if it does not innovate/buy the innovation Incentives to invest in R&D (=WTP for the innovation): 1. M is willing to pay the patent up to πM − πD 2. R is willing to pay the patent up to πD (duopolistic profits) ➔ M has a higher WTP (and a higher incentive than R) if πM − πD > πD ⇒ πM > 2πD : this condition is usually verified competition destroys profits - unless R produces a highly differentiated product. The monopolist has greater incentives to invest in R&D than the rival: M has more to lose by not winning the race for the innovation than what the rival has to gain by buying the innovation (in order to enter). 3. Suppose now that the quality of the new drink is so high that nobody wants to buy the old drink. Given ρ, find who buys the drink. So far we considered gradual innovation. The situation we are now considering is equivalent to that of drastic innovation: the firm which obtains the innovation gets the entire market. WTP(F) = πM − 0 = 4, WTP(P) = ρ(πM − πM) + (1 − ρ)(πM − 0) = 4(1 − ρ) Since ρ < 1⇒ WTP(F) > WTP(P) → Freshnet buys the patent. PUBLIC POLICY: HOW TO STIMULATE INVESTMENTS IN R&D? Direct intervention: ● Public universities and research centers ● Subsidies to companies that invest in R&D (i.e tax deductions) Indirect intervention: ● Patents ● Policies in support of R&D agreements Patents & Efficiency The main issue: How to reward innovators? Ex-post reward: once innovated, the inventor is granted monopoly rights for its invention Patent: a set of exclusive rights granted to an inventor for a limited period in exchange for the public disclosure of the invention. There is an intricate equilibrium: - PATENT: it grants monopoly rights to the patent holder for a certain period of time in order to promote dynamic efficiency (incentive to invest) - MONOPOLY is inefficient from a static point of view (allocative inefficiency; resources are not allocated efficiently) A market is dynamically efficient when agents have the correct/optimal incentive to innovate/conduct R&D⇒ focus on value generation A market is socially efficient when social welfare is maximized ⇒ focus on value allocation THERE IS A TENSION BETWEEN STATIC AND DYNAMIC EFFICIENCY! An optimal patent system must balance the benefits from greater incentives to invest in R&D and the costs associated with an increase in market power. The law imposes requirements of ”novelty” and ”non-obviousness” of the production process to produce innovation. • Patent length: determined by the law (up to 20 years) • Patent breadth: determined case by case by the Patent Office (indirectly determ. by law) The second one is a very subtle (=sottile) concept. Ex.: does a protein produced with recombinant DNA infringe a patent of the same protein but produced synthetically? → two different technologies to produce the same product - Technological territory claimed and protected by the patent - the area within which competitors cannot offer rival innovations without infringing the patent - Minimum size of improvements that another inventor has to make in order to obtain an independent (non-infringing) patent → defines the boundaries of patent protection A narrow patent (small breadth): does not provide strong protection to the innovator (an imitation may not infringe the patent). A broad patent (large breadth): effectively protects from competitors imitating the patented innovation. NOTE: the boundaries of patent protection are very much uncertain/unclear legal battles! Effect of reducing breadth: ★ Production efficiency: the innovator suffers the loss A but gains C (almost identical) ★ Allocative efficiency: social welfare increases by B+C (larger than the loss in monopoly) → this is a first-order welfare effect Effect of reducing length: ★ The monopolist loss from shorter protection is proportional to what society has to gain → this is a negligible effect Optimal patent system: patents should last many years but should be rather weak SOCIALLY INEFFICIENT PATENT RACES The firm obtaining the innovation is protected with a monopolistic position ➔ firms compete head-to-head in order to obtain the innovation This might result in a wasteful duplication of efforts⇒ over-provision of R&D ➔ this is the case of the so-called Patent races A simple model • n = number of firms competing in a perfectly competitive market (zero profits) • each firm invests c in R&D; nc = total (social) R&D cost of the investment • p(n) = prob. that the innovation will be produced - the larger the number of firms investing in R&D, the higher the probability p that the innovation will be produced p′(n) > 0 (but at a decreasing rate p′′(n) < 0) • p(n)/n = probability for a firm to innovate (i.e., win the race) • v = value of the innovation: the monopoly rent enjoyed by the firm that obtains and patents the innovation The individual firm expected profit: How many firms will enter a race for innovation? • Firms will enter the patent race as long as there are positive expected profits: up to n°, such that Eπ(n°) = 0⇒ p(n°)v = n°c The individual firm expected profit: What is the optimal number of firms (= optimal amount of investments)? • The socially optimal investment in R&D maximizes the difference between the expected social benefit p(n)v and the corresponding total social cost of R&D activities (nc). Excessive investment wastes the benefit that society derives from innovations. Number of firms participating in the race > number of firms that maximize welfare! Another purpose of patents: information disclosure In order to obtain the protection the innovator must provide a clear and precise description of the innovation → this ignites knowledge externalities, stimulates follow-on innovations and technological progress Notice that this may create the wrong incentives. If protection is not sufficiently strong: the innovator may prefer not to apply for patent protection – keep the innovation secret and benefit future developments. Costs and benefits of a patent system In traditional non-network industries, the WTP decreases with the number of units sold: this is the law of demand and is traditionally considered to hold for almost all goods → negatively sloped With network effects, as more goods are sold, the willingness to pay for the last unit may be higher. As a result, the fundamental law of demand is violated: - For network goods, some portions of the demand curve can slope upwards! Market Equilibrium Suppose that the network good is produced at a constant marginal cost c. Assume c = 1, a = 1, ν = ¼. • Focus on the Pareto superior equilibrium n2(p) The monopoly chooses the network size to maximize π = (a + νn(1 − n) − c)n subject to n ≤ 1 Solution: In a competitive situation pc = c. All consumers are affiliated nc = 1 and π = 0 • A monopolist charges a price that is too high compared to the competitive benchmark (a classical). COMPETING NETWORKS So far, we have considered the case with only one service/technology. The focus is now on competition between firms in a market with network effects o How do these markets evolve? o When there will be standardization on one technology? o When different technologies coexist? Properties 1) Path-dependence: outcome depends on the way in which adoptions build-up (i.e., on the path the process takes) 2) Inflexibility, or lock-in: the left-behind good would need to bridge a widening gap if it is chosen by adopters at all 3) Non-predictability: the process locks into monopoly of one of the 2 goods, but which good is not predictable in 4) Potential inefficiency: the good that ”takes the market” needs not be the one with the longer-term higher payoff Lesson: The competition between incompatible network goods is likely to lead, in the long run, to market dominance by a single good. The dominant good cannot be predicted beforehand and might not be the best available option. Example: Uber is the leading firm in the ride-sharing market. It’s a platform exhibiting large and multiple network externalities • Market capitalization: $70 bn (2017), after the loss of $2.8 bn in 2016 • >30 million clients monthly The company does not own a single taxi; Uber prices aggressively to beat the competitors and win the market. It also invests heavily in self-driving cars • Investors believe that this strategy can pay in the long run →They finance the operating losses and the investments With strong network effects, • Competition between incompatible network goods is likely to lead to a ’winner-takes-all’ situation →Firms compete for the market →Standards war situation • Compatibility between competing network goods (strategic decision of firms) →Reduces competition for the market →Increases competition in the market Tipping markets The presence of network externalities generates positive feedback. Examples of monopolizing technologies: - VHS vs Betamax - Wintel vs Apple - Facebook vs Myspace The winner takes all markets: the strong one gets stronger, the weak one gets weaker → tippy markets In 1975 Sony launches Betamax, and in the next year JVC launches VHS (videocassette). The main difference is that Sony operates “alone”, while JVC signs an agreement with Matsushita. Only two more big companies signed an agreement with Sony (Sanyo and Toshiba); in the meantime, the VHS alliance became stronger → VHS wins. Same thing for Facebook and Google. LESSON 12 – PLATFORMS WHY PLATFORMS? A platform is an entity that bring together economic agents ad actively manages network effects between them. To be a platform, it must respect two requirements: 1. Facilitate the interaction between users who are linked by some form of network effects. 2. Manage network effects in an active way. • Lower transaction costs to interact (search, matching, trust, reputation, etc.); • New tools to manage network effects more actively (recommender and rating systems, payment systems, data analytics, transaction monitoring, etc.) Multi-sides platforms Including additional sides on a platform is an endogenous decision. • Facebook, Google Search, and Amazon started by managing within-group network effects (they still do to a large extent); they have gradually included the management of cross-group network effects into their business model and monetization strategy. Direct network effects The presence of positive direct network effects gives rise to an attraction loop. The higher the activity level of the group, the more attractive it becomes for each group member to increase her activity level... feeding back into the group’s overall activity level. The case of Amazon - Electronic retailer selling books (not a platform as such) - Rating and review systems. User experience depends positively on number of users (within-group network effects); - Marketplace. Independent sellers are added as another group of users (cross- group network effects) The case of Waze Waze is an application for smartphones with GPS support. Its users can obtain real-time travel times, route details, updates and location-dependent information. - Waze relies on its users providing complementary information about accidents, traffic jams, speed and police traps, updates on roads, landmark, house members, etc. Positive direct network effects: more users → more information → app becomes more useful Attraction loop: more users → app becomes more useful and attracts more users Piggybacking: ‘borrow’ user base from another network • Airbnb exploited Craigslist’s large network by offering an improved experience for finding short- term rentals. - Craigslist: a website for classified advertisements with sections devoted to jobs, housing, for sale, items wanted, services, etc. Airbnb used a “Publish on Craigslist” button to make it easy for its hosts to publish their Airbnb listings on Craigslist. The price structure matters A market is two-sided if the platform can affect the volume of transactions by charging more to one side of the market and reducing the price paid by the other side by an equal amount. Idea: the price structure matters, not only the total price Taking the total price as constant, how is it divided between the two sides? Two types of prices - Usage fees: price paid for using the platform - Membership or subscription fees: price paid for joining the platform How should the platform charge the two different sides? Consider a monopoly platform with 2 distinct groups of users: buyers (b) and sellers (s). The platform sets subscription prices p1 for buyers and p2 for sellers. A basic model: take two standard inverse demand function and add the cross-network externalities Yi = number of users joining the platform on side i. We need to find the demand functions: y1(p1, p2) = y1 = 1 + θ21y2 − p1 and y2(p1, p2): = 1 + θ12y1 − p2 Solve for y1 and y2 the system of the two inverse demand functions: Standard negative relationship price/quantity: dy1/dp1 < 0 Cross market effect: dy1/dp2 < 0 → by reducing p2, y2 increases and this induces an increase in y1. This effect depends on θ21. What is the problem of the platform? Assume marginal cost equal to zero: How does the platform owner optimally set the price on side 1 for any price charged on the other side? → The two correspondences are negatively sloped Three possible cases: a) (almost) symmetric externalities: 0 < θ12, θ21 < 1 b) Asymmetric externalities: θ12 < 1, θ21 > 1 c) Asymmetric externalities: θ12 > 1, θ21 < 1 Assume θ21 + θ12 < 2. The optimal prices are The price is lower on the side where the effect of the externality is larger. In some cases, prices may be below cost (negative in this model where mc = 0) or null (free newspapers). The platform provider sets prices in order to find the correct balance between the externalities: Typically, this strategy leads to very asymmetric prices. Optimal to subsidize (even below cost) the side of the market which generates greater externalities - The side paying a low/below cost price is said Loss Leader side - The side paying a high price is said Profit Making side Asymmetric pricing NON-PRICING STRATEGIES How many sides to get on board? Sometimes, not a choice (e.g., eBay buyers and sellers are the two obvious sides) Sometimes, a platform can choose the n° of sides to attract and their identity (e.g., mobile platforms) • Google runs the Android platform as a three-sided platform connecting users, app developers and mobile phone manufacturers. • Apple iOS: two-sided model - users and application developers - while producing its own hardware. It can choose various features to influence access to the platform and interactions among users ➢ Functionalities to reduce search costs ⇒ Airbnb and Match.com (dating services) provide search functionality based on desirable characteristics ➢ Functionalities to reduce transaction costs ⇒ eBay users can settle transactions using PayPal It can regulate users’ behaviour on the platform o Apple places tight restrictions on third-party developers for its iOS two-sided platform o Google is much more liberal with respect to developers for its three-sided Android platform. ADVANTAGES OF MORE SIDES • Larger indirect network effects and larger scale. Example: Windows & Apple in the PC industry - 3-sided platform: users, app developers, hardware manufacturers - Apple 2-sided model: users, app developers - Larger ecosystem for Windows • Diversified sources of revenues. Example: LinkedIn - 3-sided platform: individual users, recruiters, advertisers - Revenues from 3 sides (2011: 20% - 50% - 30%) - Attempts to attract two more sides (corporate users and application developers) DISADVANTAGES OF MORE SIDES • Complexity and conflicts of interests between sides. Example: proposed extension of LinkedIn - Individual users may not welcome the presence of corporate users • Easier to solve the initial chicken-and-egg problem by starting with fewer sides and partially vertically integrating into some of the ‘missing’ sides. Example: video game market - Video game console manufacturers operate their own development studios to produce 1st- party games exclusive to their respective consoles Summing up • Network effects arise when consumers value positively the number of other users of a product or the number of compatible products. • Multiple equilibria can exist, depending on consumers’ expectations about the size of the network. A critical mass of users should be reached for the network to take off. • One side of the market pays generally a lower price than the other side to join or use a multi-sided platform. • The platform must set a very attractive price on one side of its market to attract a sufficiently high number of users and make the platform valuable to the other side. • The group of users that benefits from the lowest price are the one that is “needed more” (i.e., which generates the strongest indirect network effects) Results: • reviews matter most when there is limited information available on the product page. • reviews have a greater impact on purchases when the product’s characteristics are likely to appeal only to a subset of customers, measured by the variance in ratings. • individual reviews with a high-star rating are particularly effective when the average rating of the product is relatively low. Cabral and Xu (2021) study reputation and incentives to engage in price-gouging during the pandemic. Distinction between entrant and incumbent: the latter has more to lose than the former. Example: focus on 3M face masks and Purell hand sanitizers and their sales on Amazon. Seller reputation does act as a limit on how much sellers price- gouge. Higher-reputation sellers are less likely to take advantage of shortages of supply. Reviews and ratings contribute to the generation of positive externalities (attraction effect) ➢ Rating → more buyers → more information → positive network effects. BIASED REVIEWS Reviews can be noisy: when reviews are supposed to capture a quality dimension, but reviewers react by commenting based on their tastes (that may not be strongly correlated with others’ taste) They also tend to over-estimate positive experience (herding behaviour) and can be faked. Moreover, unsatisfied buyers tend not to leave negative comments; buyers can be afraid of retaliation (when they are also rated). This reduces the strength of network effects (repulsion effect) Rating → less trustworthy platform → few buyers → negative network effects. Mayzlin, Dover and Chevalier (2014). Consider hotels that want to boost their feedback at the expenses of the rivals. As it is hard to identify directly fake reviews, they exploit two main features - Tripadvisor.com lets anyone leave feedback, Expedia.com only buyers ⇒ for a given seller we should see more fake reviews on Tripadvisor - if competition gets fiercer, then the marginal gains from manipulating reviews are higher Results: • hotel neighbours of hotels with a high incentive to fake have more negative reviews on TripAdvisor relative to Expedia ⇒ competition matters • hotels with a high incentive to fake have more positive reviews on TripAdvisor relative to Expedia ⇒ verification matters He et al. (2022): A wide array of products purchase fake reviews, including those with many reviews and high average ratings. The authors study the market for fake reviews on social media platforms. Given the potential reputation costs, why does Amazon allow this? In the short run, platforms may benefit from allowing fake positive reviews if these reviews increase revenue by generating sales or allowing for higher prices. - high-quality sellers can use fake reviews to manipulate beliefs and increase market information (Dellarocas et al. 2006) - fake reviews could be an efficient way for sellers to solve the “cold start” problem and establish a good reputation How much is a fake review worth it? → Buying fake reviews on Facebook is associated with a significant but short-term increase in average rating and number of reviews. This also leads to an increase in the number of sales (and prominent position in search results). How to solve the problem? → Marketplaces measure quality and supplement feedback with governance ⇒ Certification (Badges). This can be done in different ways: external certification (e.g., Trustpilot) or internal data-driven certification of quality Does it work? What are the long-run effects of a stricter certification system on a platform? → Focus on entry by sellers, quality and prices by entrants and incumbents Example: eBay switched from Powerseller to the eTRS (e-Bay Top Rated Seller) badge in Sept 2009 - Certification requirements more stringent - eTRS = Powerseller + other stringent requirements. Powerseller badge became obsolete. The policy change caused a significant decrease in the share and number of badged sellers: i.e., from approx. 10% to approx 4% (with a gradual re-adjustment over time). Sellers’ effort (measure of quality) changed ⇒ effort increases at the margin ⇒ average effort by entrants increases (distribution of entrant quality exhibits fatter tails within each market) ⇒ effort increases by entrants more in those categories affected more by the policy shock Incumbents did not change much their effort: only those incumbents that ’lost’ the badge after the policy and regained it after some months increased their effort after the policy shock Prices changed as well: sellers who lost their badge experienced a slight decrease in prices, badged sellers experienced a larger increase in prices than unbadged sellers. Are buyers afraid of retaliation? → On some platforms (e.g., eBay, Airbnb), both buyers and sellers are rated ⇒ buyers might not leave negative feedback because they are afraid the seller will retaliate How can retaliation be detected? → Consider ”feedback pairs”: where f can be negative or neutral → If (−,−) is a lot more likely to happen when the seller leaves feedback right after the buyer, then (possible) retaliation! LESSON 14 – BIASED INTERMEDIATION RECOMMENDATION SYSTEMS AND RANKING Recommendations are part of a platform’s information-push strategy, based on characteristics and observed behaviour. Example: Spotify’s recommender system provides made-for-you suggestions based on user history, characteristics of what has been played, and preferences of ”similar” users. Positive aspects: • potential to reduce search costs • product discovery • positive spillover of frequent shoppers to occasional shoppers (the former make larger contributions to the functioning of the recommender system than the latter) Negative aspect: possibility of search bias for all platforms (such as Google, Amazon, YouTube…) Older internet e-commerce websites were price comparison sites: this gives you a ranked list of different products by their price (maybe with shipping costs). • Now, websites rank things based on various algorithms (e.g.: Past popularity, Seller quality, how much they think you will like the product, whether or not the seller paid them) Factual observation that links on top get more clicks than links at the bottom of the page. Can we conclude that a top-ranking leads to more clicks? → ”correlation” or ”causation”? Do the top links get more clicks because they are on top or because they are more relevant for users? RANKINGS MATTER Experimental evidence from Expedia: Ursu exploits a random variation in the ranking of hotels at the online travel agent. More click-through rate and almost no difference on conversions depending on the position. Congruent payoffs: What does this imply? → the presence of a bias induces the vertically integrated seller to offer more utility to consumers! As u1 > u2 ⇒ Recommendation to the ”better firm” Conflict payoffs: What does it imply? → the presence of a bias induces the vertically integrated seller to offer less utility to consumers! As u2 > u1 ⇒ Recommendation to the ”worse firm” When the magnitude of the bias b increases, the firm vertically integrated faces more captive consumers, demand becomes less price-elastic and the price increases. What about the other firm? → demand becomes more price-elastic and needs to compete more aggressively to attract consumers What are the economic consequences of bias? If there is CONFLICT ➢ an increase in bias results in more consumers being directed to a seller offering a worse utility ➢ lower utility and preference mismatch (=consumers preferring seller 2 but sent to seller 1) ⇒ consumers worse-off in aggregate. If there is CONGRUENCE ➢ seller 1 offers more utility and more consumers are sent to this firm ➢ although more consumers are sent to the better firm, there is a preference mismatch as these consumers would have preferred, other things equal, to go to seller 2. ⇒ effect on consumers unclear. However, consumers are better-off when - the mismatching is not very harmful (transportation costs are very low) - investment costs in quality for seller 1 are not very expensive FROM THEORY TO REALITY: GOOGLE SHOPPING In 2017, Google was fined 2.42bn euros by the European Commission after it ruled the company had abused its power by promoting its own shopping comparison service at the top of search results. Search engine market What factors can explain the concentration in the search engine market? 1. Economies of scale o High fixed costs of infrastructures → Google has huge computing capabilities (datacentres) o Economies of scale in the design of algorithms 2. Indirect network effects o Larger number of users ⇒ search engine more attractive to advertisers o Large amount of content ⇒ search engine more attractive to users 3. Direct network effects (data-driven) By analysing the users’ data, a search engine can improve the quality of its algorithm and provide more relevant search results - query enhancements - results are dynamically adjusted: global re-ranking based on learning relevance from user clicks engine ⇒ the larger the benefit for users. Hence: the larger the number of users and queries, the better the quality of a search 4. Switching Costs: costs of changing provider. The larger the SC, the higher the market power of the provider, and the less competitive the market. o For users: “competition is just one click away”, but personalization of search results: since 2004 Google personalizes search results based on content the user has read or on his/her history of interaction with Web pages o For advertisers: “advertisers follow the eyeballs” - no incentive to advertise on small search engines - Only a fraction of small to medium-sized advertisers advertise on Bing in addition to Google Two PhD students in computer science at Stanford (Sergey Brin and Larry Page) developed an algorithm to retrieve information in the web. → PageRank = measure of how many important sites link to a given site (patented in 1998) Important improvement over existing information retrieval algorithms (that were not adapted to retrieve info in a huge database as the WWW). Brin and Page tried to sell their algorithm to Yahoo for $1 million, but Yahoo declined! → Hence: how to monetize? OpenText in 1996: first to attempt to commercialize search by offering paid search listings → negative reaction to seeing paid ad placements and the idea failed to take off GoTo.com in 1998 (changed name to Overture in 2001) • Model: auction search results • sells its listings by auctioning off search terms to the highest bidder • Highest bidder gets the highest slot, etc. • First price auction: highest bidder pays his bid Google business model Advertisers choose: - the keywords - the ad - the maximum bid Ads are shown based on a matching between query + keywords. Google in 2000 ➢ ranking of ads based on expected revenue not only on bids (as in Overture) but also on bid amount and expected click-through rate (CTR) . ➢ Ad price: cost per click (CPC) RANKING OF ADS Every time anyone types a search query, an auction is run. The (maximum) bid is chosen by the advertiser when setting his offer (maximum payment he is willing to pay for a click). • Google considers the relevance and usefulness of an ad ranked by bid, measured by a quality score: - QS: based on historical CTR (click-through rate) - Ads quality: increase when ads include additional information, such as the phone number, or in relation to the quality of the landing page QS: a way to approximate the “expected clickthrough rate”: ads with higher quality are those expected to generate mode clicks ⇒ high clicks per impression This system makes sense: ads ranked on expected revenues • Ads that are likely to generate larger revenues for Google are placed on the top • These are the ads with the larger price per impression = (price per click) x (clicks per impression) = Bidders may be ranked first also if they have a low bid. Second-price auction: each bidder pays a price determined by the bidder below him (to avoid strategic behaviour). Note: Ads with high click-through rate, high relevance and high-quality webpage are useful to searchers.