Business Economics with a game theory focus Patrick
Business Economics with a game theory focus Patrick Mc. Nutt Follow @tuncnunc www. patrickmcnutt. com Abridged ©
Why the game theory focus? Real companies at the frontier of economic analysis…. . Understand management as ‘they are’ not as theory ‘assumed them’ to be Management can be ranked (by type) and are faced with indifference trade-offs => something must come ‘top of the menu’: the 3 rd variable or z. Trade off (x, y) to max z. Firms are conduits of information flows (vertical chain) Supply chain capacity constraints and technology-lag Reducing price does not necessarily lead to an increase in revenues (elasticity) Prices are primarily signals (observed behaviour) Companies understand the competitive threat as (recognised) interdependence (zero-sum and entropy) Predicting competitor reaction: price and entry strategies. Law of One Price: BIN price < END price
Lesson Plan for Business Economics Check Lecture Schedule on Blackboard Proposed Class Schedule & Topics Learning Plan is to follow Besanko’s Economics of Strategy 6 th Edition. . selected Chapters and cases/examples. Introduction and setting the scene using Mc. Nutt’s Decoding Strategy 2 nd Edition Chapters 1 to 9 as required Focus from behavioural analysis to understanding and application of extensive form and normal form games.
Relevance for Management What is game theory about? Relevance to MBA learning. Management as a player with type. Besanko Ch 3 & 4 and 5, Mc. Nutt Ch 1 Cost leadership and economics of capacity. Besanko Ch 2 and Mc. Nutt Ch 5 Dynamic price games, entry deterrence, market structure, oligopoly, signaling & Nash payoffs: . Besanko Ch 5, 6, 7 and 8 and Mc. Nutt Ch 6 -9. Patterns and Real Time case Analysis…go to Appendix in Decoding Strategy text.
What is game theory about? Visit www. patrickmcnutt. com Observed behaviour (inductive) in a game, G. Identify the players in the game and the player’s type. Finding the patterns in rival behaviour. Game => information on opponent type, recognised interdependence, action-reaction, belief systems. Payoff depends on what each player believes about the other. . Updating belief systems. What is a player’s true payoff? Independence v interdependence; one-shot v repeated play. Consumers’ preferences as technology in a game.
Decoding Strategy & Pattern Sequencing Complete knowledge on the type and complete information of the identity of a near rival: Actionyou -> Reactionnear-rival ->…. . -> Reactions……Nash. Replyyou…. . Strategy defined in terms of an equilibrium: how well either player does in a game depends on what each player believes the other player will do.
Example A: What is type? If you believe it to be true that Leo the Liar will never tell the truth, how do you respond to his helping hand as you cling for your life over the precipice of a cliff? Do you ignore his helping? Do you rely instead on the many apps on your smartphone, so tightly grasped in your other hand, trying to make contact with your best friend to come and rescue you? Define Strategy Cooperation arises in this instance if you and Leo as players in a game can infer from past behaviour that both of you are likely to be trustworthy. Leo may forgo the short term gain of keeping to type for the long term benefit of your friendship. He rescues you from the cliff. You, however, will use the experience in order to determine whether or not to believe Leo in the future.
Example B: Player’s belief system Your company’s strategy is s 1: delayed launch of a new innovative product for 2 years. Rumors do appear of an impending launch date. You do not deny such rumors. In the interim, an article appears a reputable trade journal reporting that a not dissimilar product is about to be launched by your competitor in the next few weeks. Define Strategy Do you stop and think about s 1? Do you reshape your strategy to s 2: launch the product as soon as possible?
Player Types I Baumol type: player in a Bertrand game who will reduce price if demand is elastic. CL type: in Cournot capacity game we have a cost-leader type, CL, with reserve capacity. Incumbent and entrant: In the geography the incumbent already exists in the geography and the entrant is intent on entering or presents a threat of entry (contestable market). Dominant incumbent is a player with at least 40% of the market share. Often linked with Stackelberg or ‘top-dog’ in Besanko.
Costs of not being a Player Agency costs can accrue across the shareholders (esp institutional & activist shareholders). . changing CEOs. Bounded rationality and opportunity costs with trade-offs First Mover Advantage (FMA) v Second Mover Advantage (SMA) Play to win v Play not to lose! Follower status ‘behind the curve’ Technology lag and failure to differentiate ‘fast enough’ to sustain a competitive advantage Near rival will try to minimise your gains by playing a minimax strategy Prisoners’ Dilemma
Prisoners’ Dilemma Bayesian persuasian problem Player type Co-operation v competition
Binary Preference Co-operate Reward No Reward Compete Punished Not Punished
Binary Preference Co-operate 1 st preference ? Compete 4 th preference ?
Payoffs reflect preference order. Guaranteed a 2 but there is an elusive 3 What if? Strategy I: cooperate What if? Strategy II: compete. Then if I is the consensus……. . ? Strategy II Strategy I 2, 2 0, 3 Strategy II 3, 0 1, 1
Why the emphasis on behaviour (of players)? The Firm as a ‘nexus of contracts’ Vertical chains and agency costs Make-buy dilemma & incomplete contracting => embedded patterns of behaviour Type of management & Bounded rationality Shareholders-as-principals and management-as-agent. . The industry as a ‘market-as-a-game’ => players with a playbook
Maximising Market Share: Table 1. 1 p 9 Mc. Nutt Recognise zero sum constaint and entropy (redistribution within market shares) Market Shares (before): 40+30+20+10 Zero-sum (after): 30+40+20+10 Entropy (after): 30+35+25+10 Hypothesis: Iff {∆qi/∆Q} > 0 market exhibits non-price competition: Check {∆q. NOKIA/∆QSmartphones} < 0
Game Dimension: product and geography Framework T/3 Capacity and Credible Threat Camouflage Commitment
Total Cost £ Total Revenue Min Profit Constraint Output Sales driven beyond the point of max profit but within the minimum profit constraint Profit/Loss
Capacity and Cost leadership [CL] as a type (of player) Profitabiltiy v scale and (size and scope) Production as a Cost-volume constraint Understanding the economcis of productivity as exemplar for incentives Normalisation equation Sources of Cost Efficiency [next slide] Cost leadership 5 Steps Checklist. . Mc. Nutt p 78
Why? Capacity Constraints: Case A: Unexhausted economies of scale due to product differentiation Case B: Firm-as-a-player does not produce large enough output to reach MES Case C: Firm-as-a-player restraints production (deliberate intent). . Mc. Nutt’s dilemma as production drives demand…(Veblen monopoly type) Convergence of technology increases the firmspecific risk of Case C: Strategic Choice A or B or C?
Sources of cost efficiency Measure of the level of resources needed to create given level of value Capacity utilisation How much to produce given capital size? Other Economies of scale X-inefficiencies, location, timing, external environment, organisation discretionary policies How big should the scale of the operation be? Transaction costs Production-cost relationship Which are the vertical boundaries of the firm? Economies of scope What product varieties to produce? Learning and experience factors How long to produce for?
MES Point: Production - demand - production to attain cost leadership £ SAC 1 SAC 2 Lower per unit cost for more units sold SAC 3 LAC Av. Cost = marginal cost 0, 0 q 1 qt Current plan of plant closures to lower cost base not completed q 2 Q
Game type and signalling Decisions are interpreted as signals Observed patterns and Critical Time Line (CTLs). Go to Appendix in Mc. Nutt Recognition of market interdependence (zero-sum and entropy) Price as a signal v Baumol model of TR max Scale and size: cost leadership
Oligopoly, Games & T/3 Framework Study of strategic interactions: how firms adopt alternative strategies by taking into account rival behaviour Structured and logical method of considering strategic situations. It makes possible breaking down a competitive situation into its key elements and analysing the dynamics between the players. Key elements: • Players. Company or manager. • Strategies. • Payoffs Equilibrium. Every player plays her best strategy given the strategies of the other players. Objective. To explore oligopolistic industries from a game embedded strategy (GEMS) perspective. The use of T/3 framework, which considers 3 key dimensions (Type, Technology & Time), will allow players to better predict the likely strategic response of competitors when analysing rival competition.
Player Types II Extant incumbent: An incumbent that has survived a negative event such as a price war of a failed innovation or technology-lag. De novo entrant: An entrant intent on entering – the incumbents can observe plant building or product launch. Potential entrant: An entrant that presents a threat of entry into a game through signalling with noise or ‘moonshot’ or planned capacity building in another game [with economies of scope). Stackelberg type: A price leader in a Bertrand game moving first in the belief that others will follow or in the knowledge that other are disciplined (often linked to collusive behaviour).
Two basic forms or types of model are used to analyse games: The Normal (Strategic) Form of a game • Summarises players, strategies and payoffs in a ‘payoff matrix’ • Particularly suitable for analysing static games (e. g. games with simultaneous moves). Making choices simultaneously The Extensive Form of a game • Summarises players, strategies and payoffs in a ‘game tree’ • Useful where the timing of players actions, and the information they will have when they must take these actions, is important (e. g. games with sequential moves)
Normal Form Simultaneous games Normal form game dimension with payoff matrices, wherein payoffs reflect preference order. Player type and camouflage Dominant strategy, Prisoners’ dilemma, Nash equilibrium.
Extensive Form Constructing an action-reaction sequence of moves in search for a pattern. Non-cooperative sequential (dynamic) games Extensive form game dimension with decision tree and backward induction Credible Threats Commitment strategies Signaling and Belief Systems
Competitor Reaction Binary reaction; Will Player B react? Yes or No? Non-binary reaction: Player B will react. Probability = x% If YES, decision may be parked Decision taking on conjecture of likely reaction If NO, decision proceeds on error Surprise No Surprise
The competitive threat! Traditional Analysis is focused on answering this question for Company X: what market are we in and how can we do better? Economics of strategy (T/3) asks: what market should we be in?
Describe (prices as signals) game dimension Players and type of players Prices interpreted as signals Understand (price) elasticity of demand cross-price elasticity Patterns of observed behaviour Leader-follower as knowledge Accommodation v entry deterrence Reaction, signalling and Nash equilibrium: ‘best you can do, given reaction of competitor’
Perfect market: perfect competition Defining a perfect market as follows: If ΔPi increases, then the firm’s output = 0 or rivals follow the price increase. In a perfect market price differences cannot persist across time Perfect competition = perfect market + near rivals So perfect market ≠> perfect competition but perfect competition => perfect market
Entry Deterrent Strategy & Barriers to entry Reputation of the incumbents Capacity building Entry function of the entrant De novo and entry at time period t Potential entrant - forces reaction at time period t from incumbent Coogan’s bluff strategy (classic poker strategy) and enter the game.
Game Strategy Nash premise: Action, Reaction and CV matrix Non-cooperative sequential (dynamic) games TR Test Mc. Nutt pp 48. . one-shot move Limit price [to avoid entry] and predatory pricing to force exit. Near rival plays Minimax, so I play Maximin [focus on my worst minimum payoff and try to maximise]. Segmentation strategy to obtain FMA Relevance of ‘chain-store’ tumbling price paradox Dark Strategy and 3 Mistakes in Mc. Nutt pp 117 -118
Game Dimension - Patterns Constructing an action-reaction sequence of moves in search for a pattern. Non-cooperative sequential (dynamic) games Normal form game dimension with payoff matrices, wherein payoffs reflect preference order. Dominant strategy, Prisoners’ dilemma, Nash equilibrium. Extensive form game dimension with decision tree and backward induction
Limit Pricing Model & Credible Threat Besanko pp 207 -211 and Mc. Nutt pp 85 -88 Outline the game dimension: dominant incumbents v camouflaged entrant type Define strategy set for incumbents: commitment and punishment Allow entry and define the equilibrium Extensive form preference - entry deterrent strategy v accommodation [next slide]
CLASS EXERCISE Strategy Profile - Fid the Nash Trap Observations and Intelligence In the decision tree narrative there is no other firm to compete with in this game – it is the incumbent v entrant. But if the entrant does not enter, fight and accommodate yield the same payoffs to both players Hypothesis 1 If the entrant does not enter, it does not matter what the incumbent chooses to do. Hypothesis 2 The incumbent will not lower prices if the entrant does not enter.
CLASS EXERCISE QUESTIONS A. Convert the decision tree into a normal form payoff matrix. B. Find the Nash equilibria C. Repeat A and B on the credible threat of entry from a spherical competitor [check pp 173 -175 in Mc. Nutt Decoding Strategy] D. Results in A+B+C written up as an Aide Memoire for management. E. Strategy Profile = Aide Memoire
Nash Equilibria & Prisoners’ Dilemma Define the Nash equilibria [next slide] Analyse the Payoff matrix Apply The Thief of Nature Handout Commitment and chat: one-shot and repeated play Punishment ‘grim’ strategy
Payoffs reflect preference order. Guaranteed a 2 but there is an elusive 3 What if? Strategy I: cooperate What if? Strategy II: compete: price war. Then if I is the consensus……credible? Strategy II Strategy I 2, 2 0, 3 Strategy II 3, 0 1, 1
In order for the NE of a game to be a compelling solution we assume: Players are rational The rules of the game are common knowledge Common knowledge of players’ rationality A source of ‘Common Beliefs’ Focal Points Pre-game communication Learning
Define a price war Check Real Time class Handouts Determine the Bertrand reaction function: Besanko Fig 5. 3 pp 190 and Mc. Nutt Fig 9. 4 p 143 Compute a Critical Time Line (CTL)from observed signals. . Examples of CTL in Mc. Nutt in the Appendix Find a price point of intersection Case Analysis of Sony v Microsoft at Mc. Nutt pp 141 -144 and also in Kaelo v 2. 0
PATTERN – 2000 -2006 PS 2 launched at $299 PS 2 at $199. 99 PS 2 at $179. 99 14 May 02 13 May 03 26 Oct 00 15 Nov 01 15 May 02 14 May 03 PS 2 at $149. 99 11 May 04 29 Mar 04 Xbox at $179 Xbox at $149 PS 2 at $129. 99 20 April 06 8 May 06 1 Nov 05 30 Oct 05 22 million Xbox shipped Microsoft Xbox launched at $299 Xbox at $199 100 million PS 2 shipped Announcement PS 3 production schedule to ship 6 million units by 31 Mar 07 at $499 22 Nov 05 6 Feb 06 27 April 06 Xbox at $179 Revised production schedule for Xbox 360 to 5 - 5. 5 million launched at units by 30 th June $399 2006
Strategic Pricing: Competitor Reaction & Folk Theorem READ the HBR articles Hypothesis: Bertrand Price Wars occur due to a mismatch in price signals. Mismatch can occur due to (i) declining volumes ∆qi/∆Q < 0; (ii) uncompetitive cost structure; (iii) decreasing productivity; (iv) management type (predator); (v) calling-my-bluff; (vi) bounded rational on player type.
The ‘signalling’ payoffs & assurance A & B Have common interest in coordinating strategies. Player A never choose ‘Bottom’ if rational, only ‘Top’, and Player B should play weakly dominant ‘Left’. B A Left Right Top 3, 3 1, 2 Bottom 2, 0 0, 0 Problem of coordination where players have different preferences but common interest in coordinating strategies. Classroom discussion on Folk Theorem Next slide for Assurance Game on coordination and trust: Payoff-dominant v riskdominant play.
Alliance/JV No Alliance/No JV 2, 2 Payoffdominant 0, 1 1, 0 1, 1 Risk-dominant
Minimax criteria. If you look at examples in the book Decoding Strategy pp 148 -151 we discuss this next slides for Near-rival v Apple but it can be applied also in any market-as-agame Strategy Simply, identify the near rival [reacting first] and set up the game tree assuming that near-rival plays minimax, that is, confining you to the least of the greatest market shares in the game - so then you play maximin, to maximise the least loss.
n Player B S 4 Player A: S 1 S 2 S 3 Column maximum Minimax strategy by B 95 60 30 S 5 S 6 S 7 5 70 35 50 55 30 40 90 10 95 70 55 90 Row Minimum Maximin strategy by A 5 55 10
Apple’s maximin = NR minimax Apple’s market shares Near Rival Strategy A Near Rival Strategy B Row min Apple Strategy 1 20 60 20 Apple Strategy 2 10 80 10 Column max 20 Minimax Smartphones games with entropy See next slide for camouflage deceiving play 80 20 Maximin
Apple’s ‘loading the dice’ strategy Apple’s market shares adapted from pp 152 in Mc. Nutt. . camouflage, deceiving mixed strategy Near Rival Strategy A B Apple Strategy 1 20 Apple Strategy 2 80 60 i. Phone 7 in 2016 nano i. Phone 2018 10
Visit Kaelo v 2. 0 and www. patrickmcnutt. com Check Examples of Critical Time Line in the Appendix of Mc. Nutt’s text Decoding Strategy. Play the PD game and investment game in Kaelo v 2. 0 as outlined in class Selfish gene [one-shot], dominant strategy to cheat. Schelling move, ‘loading the dice’. Near rival will play minimax – repeated play/learning and mixed strategy.
- Slides: 52