Quiet Bubbles H. Hong D. Sraer July 30, 2011
Motivation: Loud versus Quiet Bubbles Credit bubble in AAA/AA tranches of subprime mortgage CDOs important in financial crisis (Coval et al. 09). Classic speculative episodes associates high prices, high price volatility and high turnover (Hong and Stein 07). South Sea bubble of 1720 loud, Carlos, Neal and Wandschneider (2006). Accounts of stock-market boom of late 1920s emphasize overtrading in anticipation of capital gains in 28-29. New record not set til April 1, 1968. Internet stocks during 1996-2000: (1) price volatility excess of 100% and (2) more than 20% of stock market turnover.
Figure 1: Monthly Prices and Share Turnover of Internet Stocks The figure plots the average monthly prices and turnover of internet stocks compared to the rest of the market.!"#$%&'()*'+,")-.%"'/-"'0)1%")%1'()*'2-)30)1%")%1'41-$5&6'78893:;;:'!!!!!!!!!
Motivation: Loud versus Quiet Bubbles However, the credit bubble was quiet: high price but low price volatility and low turnover. 1. ABX prices of CDO tranches, especially AA and AAA, not volatile until beginning of crisis. 2. Little turnover of these securities. (anecdotal evidence) 3. CDS prices for insurance against default of finance companies extremely cheap and not volatile. Still had speculation about home prices and short-sales constraints binding.
Figure 2: ABX Prices The figure plots the ABX 7-1 Prices for various credit tranches including AAA, AA, A, BBB, and BBB-.
Figure 3: CDS Prices of Basket of Finance Companies The figure plots the average CDS prices for a basket of large finance companies between December 2002 and December 2008. Financial firms CDS and share prices 1.2% Exhibit 1.27: Composite Time Series of Select Financial Firms' CDS and share prices 2.50 Average CDS Spread in Percent 1.0% 0.8% 0.6% 0.4% 0.2% 2.00 1.50 1.00 0.50 0.0% Dec 02 Apr 03 Aug 03 Dec 03 Apr 04 Aug 04 Dec 04 Apr 05 Aug 05 Dec 05 Apr 06 Aug 06 Dec 06 Apr 07 Aug 07 Dec 07 Apr 08 Aug 08 Dec 08 MarketCap Index - CDS SHARE-PRICE-ADJUSTED Source: Moody s KMV, FSA Calculations Firms included: Ambac, Aviva, Banco Santander, Barclays, Berkshire Hathaway, Bradford & Bingley, Citigroup, Deutsche Bank, Fortis, HBOS, Lehman Brothers, Merrill Lynch, Morgan Stanley, National Australia Bank, Royal Bank of Scotland and UBS. CDS series peaks at 6.54% in September 2008. 12
Figure 4: Monthly Share Turnover of Finance Stocks The figure plots the average monthly share turnover of finance stocks compared to the rest of the market.
Our Paper Role of payoff concavity in a model of speculative bubbles Static overpricing w/ disagreement and binding short-sales constraints (Miller 77, Cheng-Hong-Stein 02) Dynamic resale option (Harrison-Kreps 79, Scheinkman-Xiong 03, Hong-Scheinkman-Xiong 06) Main intuition: Pure resale option framework: investors buy an asset anticipating tomorrow s (1) disagreement and (2) binding short-sales constraints. With concave payoff, less scope for disagreement lower resale option. lower resale option lower turnover, volatility
Concave payoffs reduce the scope for disagreement. Belief about expected payoff! G+!! Equity! Belief about expected payoff! Credit! Scope for disagreement! G G-!! Scope for disagreement! "(G+!)! "(G)! "(G-!)! G-!! G G+!! Belief about G+!! G G-!! Belief about fundamental! fundamental!
Main Results 1. Credit bubble has smaller resale option than equity bubble. debt less disagreement sensitive than equity. 2. Deterioration in fundamental leads to (1) larger bubble (2) more volume (3) more volatility. Contrasts with models of adverse selection (Dang et al 10).
Deterioration in fundamentals louder and larger bubbles Belief about expected payoff! High fundamental! Belief about expected payoff! Low fundamental! Scope for "(G+!)! "(G)! disagreement! "(G-!)! "(G+!)! Scope for "(G)! disagreement! "(G-!)! G+!! G G-!! Belief about fundamental! G+!! G G-!! Belief about fundamental!
Main results (continued) 1. Credit bubble has smaller resale option than equity bubble. debt less disagreement sensitive than equity. 2. Deterioration in fundamental leads to (1) larger bubble (2) more volume (3) more volatility. Contrasts with models of adverse selection (Dang et al 10). 3. Large credit mispricing requires either: more leverage (magnify disagreement) more average investor optimism. 4. Optimist bias makes credit (not equity) mispricing quiet. A rise in optimism (sentiment) makes credit bubbles (not equity ones) larger and quieter.
Sketch of the model: risky asset Three dates t = 0, 1, 2. Risk neutral agents. No discounting. Supply Q of risky credit w/ face value of D and date-2 payoff: ) m 2 = min (D, G 2 where G 2 = G + ɛ 2, and ɛ 2 Φ(.). Expected payoff with unbiased belief: π(g) = E[m 2 v] = Z D G (G + ɛ 2)φ(ɛ 2)dɛ 2 + D (1 Φ (D G)). Works more generally with any concave payoff function π().
Sketch of the model: agents beliefs Two groups of agents (A and B) w/ homogenous priors about fundamental. Ṽ 2 = G + b + ɛ 2, where b is aggregate bias At t=1, agents beliefs about fundamental becomes: { G + b + η A + ɛ 2 for group A agents G + b + η B + ɛ 2 for group B agents Where η A and η B are i.i.d. with normal C.D.F. Φ().
Leverage and trading costs Reduced form view of leverage: cost of borrowing. Agents endowed with 0 liquid wealth but large illiquid wealth W (pledgeable at date 2). Access to an imperfectly competitive credit market: banks charge > 0 interest rates for risk-free loans (parameter µ larger cheaper is leverage). Quadratic trading costs to have finite positions: Short-sales constraints c( n t ) = (n t n t 1 ) 2, 2γ Trading costs allow equilibrium to exist results similar in CARA/Gaussian framework.
Moments Construct a dynamic equilibrium and analyze following moments: 1. Ex ante mispricing: P 0 relative to no short-sales constraint / no aggregate bias (b=0) prices. 2. Price volatility between 0 and 1: ( 2 σ P = P 1 (η A, η B ) m) dφ(η A )dφ(η B ) η A,η B m = η A,η B P 1 (η A, η B )dφ(η A )dφ(η B ) is average date-1 price. 3. Share turnover between 0 and 1: T = T (η A, η B )dφ(η A )dφ(η B ) η A,η B with T (η A, η B ) = n A 1 (ηa, η B ) n A 0 (ηa, η B )
Moments Construct a dynamic equilibrium and analyze following moments: 1. Ex ante mispricing: P 0 relative to no short-sales constraint / no aggregate bias (b=0) prices. 2. Price volatility between 0 and 1: ( 2 σ P = P 1 (η A, η B ) m) dφ(η A )dφ(η B ) η A,η B m = η A,η B P 1 (η A, η B )dφ(η A )dφ(η B ) is average date-1 price. 3. Share turnover between 0 and 1: T = T (η A, η B )dφ(η A )dφ(η B ) η A,η B with T (η A, η B ) = n A 1 (η A, η B ) n A 0 (ηa, η B )
Moments Construct a dynamic equilibrium and analyze following moments: 1. Ex ante mispricing: P 0 relative to no short-sales constraint / no aggregate bias (b=0) prices. 2. Price volatility between 0 and 1: ( 2 σ P = P 1 (η A, η B ) m) dφ(η A )dφ(η B ) η A,η B m = η A,η B P 1 (η A, η B )dφ(η A )dφ(η B ) is average date-1 price. 3. Share turnover between 0 and 1: T = T (η A, η B )dφ(η A )dφ(η B ) η A,η B with T (η A, η B ) = n A 1 (η A, η B ) n A 0 (ηa, η B )
Date-1 equilibrium 1. Both groups are long (low leverage/high supply/small shocks): π(η A ) π(η B ) < 2Q µγ P 1 = µ π(ηa ) + π(η B ) 2 and T = µγ 2 π(η A ) π(η B ) 2. Group i sidelined (high leverage/low supply/large relative shock): π(η i ) π(η j ) 2Q µγ P 1 = µπ(η i ) Q γ and T = Q
Date-1 equilibrium 1. Both groups are long (low leverage/high supply/small shocks): π(η A ) π(η B ) < 2Q µγ P 1 = µ π(ηa ) + π(η B ) 2 and T = µγ 2 π(η A ) π(η B ) 2. Group i sidelined (high leverage/low supply/large relative shock): π(η i ) π(η j ) 2Q µγ P 1 = µπ(η i ) Q γ and T = Q
Date-0 equilibrium Agents select date-0 holdings anticipating date-1 equilibrium. Market clearing condition (n A 0 + nb 0 = 2Q) gives P 0. Symmetric equilibrium: n A 0 = nb 0 = Q. P 0 = Z 2 6 µπ(y) 2Q «Φ (x(y)) 4 γ {z } short-sales constraint Z + x(y) µπ(x)dφ(x) 7 5 dφ(y) Q γ {z } no short-sales 3 {z} supply
Equilibrium moments: bubble Bubble can be decomposed in two terms: bubble = Z Z x(y) µπ(y) µπ(x) 2Q γ «! dφ(x) dφ(y) {z } resale option P 0 is the price when b = 0 and no short-sales constraint + ˆP 0 P 0 {z } optimism ˆP 0 is the no-short-sales constraint price with aggregate bias b.
Equilibrium moments Mechanical link between turnover and volatility and mispricing: Turnover maximized when short-sales constraints are binding. Resale option maximized when short-sales constraints are binding. Price volatility also higher since average of beliefs lower volatility then volatility of max beliefs.
Comparative statics: credit riskiness Proposition 1: An increase in D leads to larger mispricing, larger turnover and larger volatility. Intuition: as D increases, credit becomes more disagreemeent sensitive. Larger resale option Larger mispricing Larger turnover, volatility. Thus, credit bubbles are quiet and small. In the pure resale option framework, loudness and prices go hand in hand.
Comparative statics: optimism Proposition 2: An increase in b leads to larger mispricing, lower turnover and lower volatility. Intuition: as b increases, credit becomes safer in the agents eyes. credit becomes less disagreemeent sensitive. lower resale option lower turnover, volatility. Lower resale option, but larger bubble from optimism. When optimism rises, credit bubbles quieter and larger. Optimism decouple turnover/volatility and price. Important: b leaves unchanged an equity bubble.
Comparative statics: fundamental Proposition 3: An decrease in G leads to larger mispricing, higher turnover and higher volatility. Intuition: as G decreases, credit becomes riskier and thus more disagreemeent sensitive. higher resale option larger bubble (but lower price) higher turnover, volatility. Thus, deterioration in fundamentals leads to more trading, more volatility, larger bubbles. Opposite to models of adverse selection that predict trading freeze and low prices. Can explain rise in ABX vol in the months preceding the crisis.
Extension: Interim Payoff and Dispersed Priors Agents have heterogenous priors: G + b + σ for group A, G + b σ for group B Agents receive interim payoff π(g + ɛ 1 ). (Interest payments) This t = 1 cash-flow occurs before belief shock. Two rationales for holding credit: (1) short term payments and (2) speculation on capital gains. Another mechanism that decouples pricing and volatility/turnover: Proposition 5: if leverage is cheap, in σ makes bubble quieter and larger.
Miller quietness Intuition in σ increases group A date-0 holdings (interest payments) up to the point where they hold all the supply (provided leverage is cheap enough). in σ makes it more likely that short-sales constraints bind at date 1 and group A agents want to hold on to their shares. Thus as σ increases, turnover becomes lower. Yet, large date-0 bubble because of (1) mispricing of interest payments (Miller) and (2) binding date 1 short-sales constraints.
Implications Our model offers a new take on the crisis. Simple extension of speculative bubbles to the assets that were at the heart of the crisis. Unified theory relating credit bubbles to Internet bubbles. Dispersion can lead to concentration of positions and quiet bubbles. Anecdotal evidence on AIG-FP as being key to rise of subprime mortgage CDO market. Credit bubbles are potentially harder to detect. Associated with lower volatility and turnover quiet bubbles. Also suggests a Taxonomy of Bubbles based on this loudness criterion.
Figure 5: Traditional and Non-Traditional Issuance of Asset-Backed Securities (Quarterly) The figure plots the issuance of traditional and non-traditional asset-backed securities by quarter.
Figure 6: Synthetic Mezzanine ABS CDO Issuance The figure plots the issuance of synthetic mezzanine ABS CDOs by year.
Conclusion Insight for a broader agenda to build a speculation-based asset pricing model (Hong-Sraer 2011b). Annual trading volume in excess of $50 trillion is speculative in nature. Risk-sharing models: CAPM beta prediction rejected Liquidity models: CAPM beta prediction not rejected per se. But idio-vol should be priced. Also rejected. Speculation models: High beta asset more sensitive to disagreement about market. Speculative beta equals low expected return.