Discussion Liquidity requirements, liquidity choice and financial stability by Doug Diamond Guillaume Plantin Sciences Po Plantin Liquidity requirements 1 / 23
The Diamond-Dybvig model Summary of the paper Comments Plantin Liquidity requirements 2 / 23
The Diamond-Dybvig model Plantin Liquidity requirements 3 / 23
Transformation A bank is a firm whose current operations consist in granting loans and receiving deposits. Transformation: refinancing illiquid claims with highly liquid ones More than 50% of transformation in the U.S. is not performed by licensed banks This is an important source of financial fragility So, what is desirable about transformation? Plantin Liquidity requirements 4 / 23
Benefits from extreme maturity transformation by banks 1 None? 2 Disciplining fragility? 3 Diamond-Dybvig Plantin Liquidity requirements 5 / 23
Benefits from extreme maturity transformation by banks None? Narrow banking view (Friedman, Cochrane) Transformation is a historical accident Private benefits, social costs Should get rid of it Bad reasons for transformation: maximizing the value of explicit and implicit taxpayer guarantees Plantin Liquidity requirements 6 / 23
Benefits from extreme maturity transformation by banks Disciplining device General idea that short-term debt is a governance tool that helps firms outsiders keep insiders on a short leash mitigate a free cash flow problem Because informational asymmetries are particularly strong in banking this device needs to be particularly strong The credible threat of a run provides incentives for banks to screen and monitor borrowers The threat is credible because it is difficult to renegotiate with many panicked depositors Calomiris and Kahn, The Role of Demandable Debt in Structuring Optimal Banking Arrangements, The American Economic Review, 1991 - Diamond and Rajan, A Theory of Bank Capital, Journal of Finance, 2000 Plantin Liquidity requirements 7 / 23
Benefits from extreme maturity transformation by banks The most influential/widely accepted view: Diamond-Dybvig model Banks are insurance companies that insure idiosyncratic shocks to agents time preferences Two differences with other insurance branches: Policyholders pre-pay the maximum potential claim and get a refund if they do not file Claims are not verifiable This comes at the benefits that the bank can invest large amounts in non-marketable projects This creates fragility: filing a claim is optimal under the expectation that other policyholders do so What type of liquidity regulations best address this fragility? Plantin Liquidity requirements 8 / 23
Summary of the paper Plantin Liquidity requirements 9 / 23
Optimal asset-liability management of a bank Suppose the net wealth of a bank ν(w, λ) depends on w the fraction of anticipated early withdrawals λ the fraction of the bank s assets that are liquid (=available to meet withdrawals at no liquidation costs) Optimal asset-liability structure: for every w there exists λ(w) that maximizes net wealth: λ(w) increases in w ν(λ(w), w) decreases in w λ(w) = max λ {ν(λ, w)} Plantin Liquidity requirements 10 / 23
Optimal asset-liability management of a bank ν(λ,w) λ(w) λ Inefficient liquida9on of long-term assets Unnecessary hoarding of low-return liquid assets Plantin Liquidity requirements 11 / 23
Exogenous versus endogenous withdrawals Suppose there are two types of early withdrawals: w F are exogenous early withdrawals that do not depend on the bank s balance sheet Endogenous withdrawals. An additional fraction of the depositors withdraw early if they believe that the bank will be insolvent if it has to meet a mass w F + of early withdrawals A bank is stable if its liquid holdings λ deter such endogenous withdrawals: ν(λ, w F + ) 0 Denote λ S (w F ) the smallest liquid holdings λ such that a bank with exogenous withdrawals w F is stable Depositors are very risk-averse: they are willing to make deposits only if they believe that the bank is stable Plantin Liquidity requirements 12 / 23
Symmetric information Suppose depositors (or a regulator acting in their interest) observe λ and w F Then the bank optimally chooses λ = Max{λ(w F ), λ S (w F )} Since an unstable bank cannot collect deposits, the bank fully internalizes the need to be stable Plantin Liquidity requirements 13 / 23
ν(λ,w) ν(λ,w+δ) λ Automa)cally stable Not automa)cally stable λ(w) λ S (w) Plantin Liquidity requirements 14 / 23
Private information If w F is private information to the bank then If, for all exogenous withdrawals w, the bank is stable for λ = λ(w): ν(λ(w), w + ) 0 which is true if long-term assets sufficiently profitable/liquid, then the bank can truthfully announce w F and choose liquidity λ(w F ) If ν(λ(w), w + ) < 0 for some w, then asking the bank to announce w F and then implement λ = Max{λ(w F ), λ S (w F )} is not incentive-compatible. Suppose otherwise. Depositors believe the bank is stable, so a bank with exogenous withdrawals w F will announce w < w F such that λ S (w ) = λ(w F ) < λ S (w F ), which contradicts depositors beliefs Plantin Liquidity requirements 15 / 23
Solution 1- Net Stable Funding Ratio (NSFR) The bank can choose an initial level of liquidity λ = Max{λ(w), λ S (w)}, where w is the maximum exogenous withdrawal Non contingent: Leads to large hoarding of illiquid assets even if w has an arbitrarily small probability of occurence Plantin Liquidity requirements 16 / 23
Solution 2- Liquidity Coverage Ratio (LCR) The bank commits to keep a sufficiently large fixed fraction ρ of its deposits backed by liquid assets after the early withdrawals have occured and the remaining deposits are (in equilibrium) 1 w F. This is ex-post useless but makes under-reporting w F costly and elicits truthful reports: If reports w < w F the gain in liquid assets would be λ(w F ) λ(w ) ρ(w w F ). The need to plan for a lot of future idle liquidity to back the incorrect claim of a small w offsets the gain from under-reporting Plantin Liquidity requirements 17 / 23
Comparing NSFR and LCR The NSFR does great if the beliefs about the exogenous withdrawal are (correct and) very accurate because there is no idle liquidity left in the bank after the withdrawal, which translates into more profitable assets ex ante. Otherwise it may lead to overly conservative liquidity levels because it is based on a worst-case scenario The LCR comes at the cost of leaving unused cash in the balance sheet. But it is contingent, and thus better suited to the case in which exogenous withdrawals are more uncertain Plantin Liquidity requirements 18 / 23
Optimal mechanism The optimal mechanism consists in eliciting reports of the bank s information, and punishing under-reporting through regulating dividends or compensations or acting as a LLR at punitive rates in case of larger than expected withdrawals Same idea as in LCR of making under-reporting costly, but the improvement is that this does not lead to excessive liquidity hoarding: more liquidity can be released ex-post Optimal mechanism design leads to an interesting insight: very much like designing the details of capital requirements while being silent on resolution is not satisfactory, a discussion of liquidity ratios without discussing LLR policies is incomplete Plantin Liquidity requirements 19 / 23
Comments Plantin Liquidity requirements 20 / 23
Comment 1 My gut feeling is that the NSFR approach is unlikely to be close to optimal in practice Priors on extreme liquidity crises must be pretty diffuse for all agents Especially worst-case scenarios see 2008 Plantin Liquidity requirements 21 / 23
Comment 2 Assuming fixed exogenous and endogenous withdrawals and risk-free illiquid assets are useful simplifications In more general models and in practice, the magnitude of withdrawals would likely depend on the riskiness of illiquid assets This implies that LCR should be risk-based. But it is complicated: What should matter is the riskiness of the senior tranche of the illiquid assets that backs deposits Perhaps a mix of risk-based capital requirements and non-risk based liquidity requirements is a good approximation Plantin Liquidity requirements 22 / 23
Comment 3 Liquidity requirements strike me as potentially much more pro-cyclical than capital requirements with a much faster feedback mechanism It might also be highly disruptive to the transmission of monetary policy Ex-ante efficient punishment might be difficult to commit to Macro-prudential approaches seem much more crucial for liquidity regulation than for solvency regulation because the time scale is very different very hard to switch expectations quickly in a panic. This is actually why there are no episodes of contagion in traditional insurance/reinsurance markets: long-term solvency issues without liquidity problems Plantin Liquidity requirements 23 / 23