1. Bruce Ian Carlin, Miguel Sousa Lobo, S. Viswanathan: Episodic Liquidity Crises: Cooperative and Predatory Trading (The Journal of Finance, 2007) Presented by Márta Bisztray for the Lecture: Financial Economics 18/03/2010

2. The problem Empirics: short-living, rare liquidity problems – puzzle Explanation of episodic illiquidity by the breakdown of cooperation among strategic investors Liquidity event for a trader  predatory trading vs. cooperation Looking for the determinants of cooperative equilibrium where apparent liquidity is provided Dynamic model of trading based on liquidity needs 1.One-period trading game 2.Multi-period model a.Fixed amount shock b.Stochastic amount shock 3.Multi-market extension 2

3. The model – setup Continuous time Players: – Oligopoly of n strategic traders – inside information about liquidity needs on the market, affect asset prices – Competitive fringe (long-term investors) – no inside information Risk free (zero yield) and risky asset, latter with S >0 supply, divided between strategic traders ( X t ) and long-term investors ( Z t ) S= X t + Z t 3

4. The model – asset price Trading influences price: P t = U t + γX t + λY t – U t : expected value of dividends, stochastic process – X t = Σ n i=1 X i t : aggregate amount of the asset held by strategic traders at time t (lower residual supply) – γ > 0 : permanent effect of trading (cumulative amount) Higher asymmetric information  higher γ – dX t = Y t dt and Y t = Σ n i=1 Y i t : change in asset holdings by strategic traders (sell faster  lower price) – λ : temporary effect of trading (current rate of trading) Trading volume, shares outstanding influence 4

5. The model – stage game Complete information 1. types: distressed or predator Liquidity event: a trader is distressed and has to liquidate a large block of an asset (Δx) within a short period T Observed by other strategic players  profit by predatory trade, have a trading target of zero at the end of the period 2. Trading schedule Y i t given type – determined at t=0 s.t.,, Open-loop Nash equilibrium: – for small t the first term dominates – race to the market (average trading target) – for larger t the second term dominates – magnitude of fading (reversing trade direction) when there are predators (individual trading target) 5

6. Stage game – Race to the market Benchmark: n=1, distressed  optimal to trade at a constant rate Y t = Δx/T (no short-term price pressure) n>1 traders, all distressed with Δx/n trading targets  optimum: - surplus loss – Racing at the beginning of the period  lowering prices even more – Lower permanent price impact ( γ ) compared to temporary one ( λ )  race is smaller 6 Walk down the demand curve

7. Stage game - Predation Distressed and predatory trader: Racing at the beginning of the period  lowering prices even more Fading the distressed trader at the end of the period – adverse price impact  buy back assets at a lower price  making profits 7

8. First best: n=1  Lower aggregate surplus if racing or predatory trading Expected total loss from racing: – Positive, higher if γ, T, n higher or λ lower Expected total loss from predation: Expected value of the predator: – Both positive, higher if γ, T higher or λ lower Higher γ/ λ ratio  more aggressive predation Stage game - Surplus 8

9. Dynamic game – setup Total surplus lower at predatory trading  room for Pareto- improvement: cooperation (no racing, no predatory trading) – market seems more liquid as large blocks can be moved for a better price, no volatility, no trading volume peaks Two strategic traders, many long-term investors Common discount factor ( δ ) and parameters U t, γ, λ Timing of each stage: 1.Nature: type of each strategic trader, distressed with independent probability p 1 and p 2 - has to liquidate Δx  traders have perfect information 2.If not distressed, decides if predate or cooperate Infinitely repeated game: trigger strategy 9

10. Dynamic game – fixed size liquidity event Grim trigger strategy works only for large enough δ – Cooperate forever > predate then deviate forever (i.e. each will race and predate if there is an opportunity) Threshold values: – Increasing in λ, T decreasing in γ – Cooperation is more likely if 1.higher permanent price impact (higher surplus loss) 2.Probabilities of distress are higher and more symmetric (incentive to share the market) 10

11. Dynamic game – fixed size liquidity event 11

12. Dynamic game – stochastic size liquidity event Trigger strategy, but allow to predate if (punishment only if predate when )  episodically increased volatility (risk management) – Optimal : highest feasible Higher relative benefits of cooperation, if higher δ, p i or lower p j Threshold value of relative benefits, below which no cooperation exists – Increasing in λ, decreasing in γ Usually stable liquidity, rare illiquidity but with high spikes – Assets with high level of asymmetric information (high γ ) – Assets which traded in a large volume (low λ ) – e.g. growth stock with diffuse ownership (vs. thinly traded AA bond) 12

13. Dynamic game – multiple markets Oligopoly can cooperate in n markets Liquidity shock occurs in the same direction in all markets (size is independent) Multimarket contact supports cooperation (punishment in all) Predation is more significant, as it affects all the markets C/K relative benefit from cooperation n number of markets r radius of cooperation region (like before) 13

14. Main results Episodic liquidity can be explained with the breakdown of cooperation when stakes are high Cooperation is more likely when – Need for liquidity over time (distress probability) is symmetric – Permanent price impact of trade is high, temporary price impact is low – but if predation, it’s more aggressive Asymmetric information  higher permanent price impact – Concentrated ownership, high insider ownership, high-growth firms  steady apparent liquidity, high illiquidity spikes – Diffuse ownership, mature firms  lower illiquidity spikes Higher average daily trading volume, higher number of outstanding shares  lower temporary impact of trade Cooperation across markets  less frequent episodic illiquidity but contagion No testing possible 14

15. Thank you for your attention! 15