The Race for Priority

The Race for Priority Martin Oehmke London School of Economics FTG Summer School 2017

Outline of Lecture In this lecture, I will discuss financing choices of financial institutions in the presence of a race for priority. In a race for priority, some claimholders seek to gain priority over other claimholders. This can be self-defeating and lead to inefficient outcomes. Often, but not always, the underlying friction is related to non-commitment to financing choices. We will cover two instances of the race for priority: Priority via maturity: The maturity rat race (Brunnermeier and Oehmke, JF 2013) Priority via the bankruptcy code: The special treatment of derivatives (Bolton and Oehmke, JF 2015) At the end, I will briefly mention some other applications/papers.

Is There Too Much Maturity Mismatch? Households have long-term saving needs Firms have long-term borrowing needs Why is borrowing so short-term? particularly for financial intermediaries Rationale for beneficial maturity mismatch: Diamond and Dybvig (1983) Calomiris and Kahn (1991), Diamond and Rajan (2001) But, there may be excessive maturity mismatch in the financial system due to a maturity rat race

The Maturity Rat Race A borrower raises financing for a long-term investment from multiple creditors at different maturities Negative externality can cause excessively short-term financing: shorter maturity claims dilute value of longer maturity claims depending on type of interim information received at rollover dates Externality: general mechanism (can arise for any borrower) but is particularly relevant for financial institutions Successively unravels all long-term financing: A Maturity Rat Race

Model Setup: Long-term Project Long-term project: investment at t = 0: $1 payoff at t = T : θ F ( ) on [0, θ] Over time, more information is learned: s t observed at t = 1,..., T 1 S t is sufficient statistic for all signals up to t: θ F ( S t ) S t orders F ( ) according to FOSD Premature liquidation is costly: early liquidation only generates λe[θ S t ], λ < 1

Model Setup: Credit Markets Risk-neutral, competitive lenders All promised interest rates are endogenous depend on aggregate maturity structure Debt contracts specifies maturity and face value: can match project maturity: D 0,T or shorter maturity D 0,t, then rollover D t,t+τ etc. lenders make uncoordinated rollover decisions All debt has equal priority in default: proportional to face value

Model Setup: Credit Markets (2) Main Friction: borrower cannot commit to maturity structure particularly relevant for financial institutions opaqueness of balance sheet and maturity structure desire for financial flexibility Borrower: simultaneously offers debt contracts to creditors each bilateral contract does not condition on other creditors contracts An equilibrium maturity structure must satisfy two conditions: 1. Break even: all creditors must break even 2. No deviation: no incentive to change one creditor s maturity

Analysis with One Rollover Date For now: focus on only one possible rollover date, t < T Outline of thought experiment: Conjecture an equilibrium in which all debt has maturity T Calculate break-even face values At break-even interest rate, is there an incentive do deviate? Denote fraction of short-term debt by α

A Simple Example: News about Default Probability θ only takes two values: θ H with probability p ( no default) θ L with probability 1 p ( default) p random, revealed at date t If all financing has maturity T : (1 p 0 ) θ L + p 0 D 0,T = 1, D 0,T = 1 (1 p 0) θ L p 0 Break-even condition for first t-rollover creditor: (1 p t ) D t,t D 0,T θ L + p t D t,t = 1, D t,t = 1 (1 p 0) θ L θ L p 0 + (1 θ L ) p t

Illustration: News about Default Probability Deviation payoff: Π α α=0 = E [p t D 0,T ] E [p t D t,t ] > 0? Product of two quantities matters: Promised face value under ST and LT debt (left) Probability that face value is repaid (right) Face Value 3.0 2.5 D t,t S t Repayment Probability 1.0 0.8 2.0 1.5 1.0 D 0,T 0.6 0.4 p t 0.5 0.2 0.0 0.2 0.4 0.6 0.8 1.0 p 0.2 0.4 0.6 0.8 1.0 p

Illustration: News about Default Probability Multiplying promised face value and repayment probability: Marginal Cost 1.5 Long term financing 1.0 A Rollover financing 0.5 B 0.0 0.2 0.4 0.6 0.8 1.0 p Note: A > B implies rolling over cheaper in expectation

A Simple Example: News about Recovery Value θ only takes two values: θ H with probability p = 1/2 θ L with probability 1 p Low cash flow θ L random, revealed at date t If all financing has maturity T : 1 2 D 0,T + 1 2 E [ θ L] = 1, D 0,T = 2 E [ θ L] Break-even condition for first t-rollover creditor: 1 2 D t,t + 1 D t,t θ L ( = 1, D t,t θ L ) 2 E [ θ L] = 2 2 D 0,T 2 E [θ L ] + θ L

Illustration: News about Recovery Value Deviation payoff: Π α α=0 = 1 2 D 0,T 1 2 E[D t,t (θ L )] > 0? Product of two quantities matters: Promised face value under ST and LT debt (left) Probability that face value is repaid (right) Face Value Repayment Probability 2.0 1.0 1.8 0.8 1.6 D t,t S t D 0,T 0.6 1.4 0.4 1.2 0.2 0.0 0.2 0.4 0.6 0.8 1.0 ΘL 0.2 0.4 0.6 0.8 1.0 ΘL

Illustration: News about Recovery Value Multiplying promised face value and repayment probability: Marginal Cost 1.1 1.0 0.9 0.8 B' Rollover financing Long term financing 0.7 A' 0.6 0.0 0.2 0.4 0.6 0.8 1.0 ΘL Note: A < B implies rolling over more expensive in expectation

What is going on? Interim Information Matters! Rollover face value D t,t (promised interest rate) is endogenous adjusts to interim information Effect on LT creditors: Interim Signal D t,t default no default Negative high LT creditors lose no effect Positive low LT creditors gain no effect LT creditors lose on average if default sufficiently more likely after negative signals

General Proposition Extend to: general payoff distribution start from any conjectured equilibrium that involves some amount of LT debt Condition 1: D t,t (S t ) D T (S t) df (θ S t ) is weakly increasing in S t }{{} repayment probability Guarantees signal has sufficient effect on default probability Proposition: Under Condition 1, the unique equilibrium is all short-term financing (α = 1).

Many Rollover Dates: The Maturity Rat Race Up to now: focus on one potential rollover date Assumed everyone has maturity of length T Showed that there is a deviation to shorten maturity to t This extends to multiple rollover dates Assume all creditors roll over for the first time at some time τ < T By same argument as before, there is an incentive to deviate In proof: For τ < T replace final payoff by continuation value Successive unraveling of maturity structure

The Maturity Rat Race: Successive Unraveling t 0 1 2 T-2 T-1 T

The Maturity Rat Race: Successive Unraveling Condition 2: D t 1,t (S t 1 ) dg (S t S t 1 ) is increasing in S t 1 t S } t {{} prob of rollover at t Guarantees signal has sufficient effect on rollover probability at next rollover date Proposition: Sequential Unraveling. Under Condition 2, successive application of the one-step deviation principle results in unraveling of the maturity structure to the minimum rollover interval.

Rat Race Causes Inefficiencies Excessive Rollover Risk Project could be financed without any rollover risk Rat race leads to positive rollover risk in equilibrium Underinvestment Creditors rationally anticipate rat race NPV of project must outweigh eqm liquidation costs some positive NPV projects don t get financed

Relation to Banking Literature Banking literature highlights positive role of short-term debt liquidity services (Diamond and Dybvig, 1983) discipline (Calomiris and Kahn, 1991; Diamond and Rajan, 2001) Point of this paper... is not to argue that ST debt hat no benefits but that despite benefits, too much ST debt may be used Example: if debt is a disciplining device, trade off benefits against rollover costs our model suggests financial institution may go beyond optimal ST debt amount

Rat Race Strongest During Crises Rat race stronger when more information about default probability is released at interim dates ability to adjust financing terms becomes more valuable Volatile environments, such as crises, facilitate rat race Explains drastic shortening of unsecured credit markets in crisis e.g. commercial paper during fall of 2008

Can This be Solved via Covenants? Extension of model allows for commitment via covenants Covenants are costly: direct costs (e.g., monitoring costs of covenants) loss in financial flexibility Predictions: firms with low covenant costs (corporates) eliminate rat race firms with high covenant costs (financials) do not eliminate rate race sharpens cross-sectional predictions of model Inefficiency likely remains: social and private incentives to write covenants may differ law may allow firms to bind themselves more efficiently than through covenants

Seniority Seniority for LT debt can reduce externality of ST debt on LT debt if default occurs at T, LT creditors are senior and immune to higher face values of ST creditors However: ST creditors can still withdraw their funding early (i.e., at t) hence, ST creditors may still have de facto seniority Seniority unlikely to eliminate externality completely

Summary Under non-commitment, equilibrium maturity structure may be efficiently short-term: Contractual externality between ST and LT creditors Particularly relevant for financial institutions Maturity Rat Race successively unravels long-term financing This leads to too much maturity mismatch excessive rollover risk underinvestment Not easily fixed through covenants or seniority for LT debt

Background: Super-Seniority of Derivatives Derivatives benefit from privileges in bankruptcy: not subject to the automatic stay netting, collateral, and closeout rights can keep eve-of-bankruptcy payments To the extent that their net exposure is collateralized, derivative counterparties get paid before anyone else... This is an example of a race to priority via the bankruptcy code industry groups have lobbied for years for the special treatment of derivatives But why should derivatives be (effectively) senior?

Why We Should be Interested Role of derivatives in the demise of Lehman Brothers This caused a massive destruction of value. Harvey Miller (2009) Treatment of Qualified Financial Contracts (QFCs) which include derivatives under Dodd-Frank Ex-ante distortions of seniority for derivatives It s plausible to wonder whether Bear s financing counterparties would have so heavily supported Bear s short-term repo financings were they unable to enjoy the Code s advantages. Mark Roe (2010)

The Model (Simplified Setup) Firm raises F using debt contract with (endogenous) face value R Absent derivative: default in low cash-flow state C L 1 (with prob. 1 θ) no default in C H 1 state When hedging with derivative: Derivative pays out notional X with probability γ in low cash-flow state, against endogenous fair premium of x Derivative is costly: Counterparty has to post collateral ζ(x ) at opportunity cost of foregone investment Γ 1 Required collateral increasing in promised payment X : ζ (X ) > 0 This is derived endogenously in paper (à la Holmström and Tirole)

Senior Derivatives To eliminate default, with probability (1 θ)γ, need to set: X S = R J C L 1 R J is determined by creditor breakeven condition: [θ + (1 θ) γ] R J + (1 θ) (1 γ) ( C L 1 x S) = F x S determined by derivative counterparty breakeven condition: θx S = (1 θ) [X S + (Γ 1) ζ(x S )]

Junior Derivatives To eliminate default, with probability (1 θ)γ, need to set: X J = R S C L 1 R S determined by creditor breakeven condition: [θ + (1 θ) γ] R S + (1 θ) (1 γ) C L 1 = F x J determined by derivative counterparty breakeven condition: [θ (1 θ) (1 γ)]x J = (1 θ) [ X J + (Γ 1) ζ(x J ) ]

Key Observation: Senior Derivatives Raise Cost of Debt Face value of debt is lower when debt is senior: R S R J Therefore: Required derivative position is smaller when debt senior: R S C L 1 }{{} X J R J C1 L }{{} X S Senior debt is efficient because it reduces the deadweight cost of hedging: ζ(x J ) ζ(x S ) Key Takeaway: At the level of the individual firm, senior debt and junior derivatives is efficient.

When Can Seniority for Derivatives be Efficient? Can the previous result be overturned when counterparty enters derivatives with many firms? Yes, but this requires quite specific conditions Risks must be such that, even though individual derivative positions are larger, the counterparty s aggregate collateral requirement is lower when derivatives are senior The key requirement: basis risk of the derivative must be idiosyncratic Seniority for derivatives allows counterparty to use premia received by some (defaulted) firms to offset obligations to other firms, lowering required collateral. How likely is this in practice? One might argue that firms use similar derivatives (e.g., ABX during housing boom/bust), so basis risk correlated...

Hedging or Speculation? Up to now we assumed that firm will take derivative position that is required for hedging, X = R C L 1. Suppose that it is privately optimal for the firm to hedge default risk and that derivative is senior to debt. Marginal payoff to increasing derivative beyond X S = R J C L 1 : 1 θ }{{} marginal derivative payoff [ 1 1 θ ] [ (1 γ) (1 θ)[1 + ζ (X S )(Γ 1)] ] 0 θ }{{}}{{} marginal cost of derivative 1 The firm s privately optimal derivative position ex post coincides with the optimal derivative position only if γ γ. When γ < γ, the firm enters a derivative position that is inefficiently large.

Hedging or Speculation? Therefore, under non-commitment, seniority for derivatives can induce risk shifting. Leads to deadweight collateral costs. These incentives are not there when derivatives are junior: Marginal payoff to increasing derivative beyond X J = R S C L 1 : 1 θ }{{} (1 θ)[1 + ζ (X S )(Γ 1)] < 0 }{{} marginal derivative payoff marginal cost of derivative However, junior derivatives make it more likely that firm does not hedge at all (see paper). classic tension between debt overhang and risk shifting

Discussion: Financial Firms Automatic stay exemption for derivatives may have particular bite for financial firms Exemption from automatic stay particularly hard to undo: costly to assign cash as collateral to all creditors/depositors ex-ante but then hard to shield cash from derivative counterparties initial margins margin calls once drained of cash, financial firm ceases to operate See, e.g., Duffie (2010): Failure mechanics of dealer banks

Where To From Here? We have discussed in detail two instances of the race for priority: the maturity rat race the privileged treatment of derivatives in bankruptcy In the remainder of the lecture, I will briefly sketch a number of other applications and related research. If you are interested, the relevant papers are listed in the syllabus

Collateral Note the role of collateral in what we have discussed: Derivatives are effectively senior to debt mainly because they are collateralized The least dilutable claim in the maturity rat race is short-term and collateralized (repo) So what about secured debt more generally? Donaldson, Gromb, and Piacentino (2017) develop a model in which ex-post collateralization can hamper financing ex ante. So should firms prevent ex-post collateralization via negative pledge clauses? See Ayotte and Bolton (2011).

Dynamic Models There are a number of papers that explicitly investigate the dynamic aspects of financing under non-commitment: DeMarzo and He (2017) investigate leverage choices in a dynamic model without commitment to future debt issues: Leverage follows and endogenous mean-reverting process Firms never repurchase debt He and Milbradt (2016) study the dynamics of maturity structure under non-commitment to future maturity choices: Debt maturity shortens when fundamentals deteriorate Self-fulfilling shortening spirals can arise

Sovereign Debt There are many other applications of financing choices in the absence of commitment. To give one interesting example outside of banking/corporate finance, consider sovereign debt: Bolton and Jeanne (2007, 2009) show that, in equilibrium, sovereign debt may be inefficiently hard to restructure. Intuition: creditors of each issue want there issue to be harder to restructure than other issues by the same sovereign. In equilibrium, this becomes self-defeating.

THANK YOU! If you have questions, please feel free to contact me on: m.oehmke@lse.ac.uk