Discussion of Allen, Carletti, Goldstein & Leonello

„Government Guarantees and Financial Stability“

Gerhard IllingLMU Munich University/CESifo

Norges Bank Workshop on Understanding Macroprudential Regulation

29 November, 2012

Central issues

• How to cope with Moral Hazard effects of public interventions (deposit guarantee schemes)?

• Optimal design of Financial Safety Nets?

• Challenge: Distinguish between fundamental and panic driven runs (runs due to coordination failure)

• Insolvency vs. illiquidity• Panic driven runs: Multiple equilibria ~

how to handle indeterminacy?

• Elegant model. Tractable Structure But only first step – some key issues not yet solved

Summary – Model setup

• Modeling Strategy: Analyze Public Guarantuee Schemes in Goldstein /Pauzner version of Diamond/Dybvig model

• Model allows for both fundamental and panic driven bank runs

• Model determines strategies of depositors and banks endogenously

• Indeterminacy of multiple equilibria solved by Global Game approach (Goldstein /Pauzner)

• Depositors receive noisy signals about fundamentals• Inefficiency if runs are panic driven;

Public support improves outcome, but may increase region with fundamental runs beyond “efficient” level

Summary – Model setup• Diamond Dybvig type Deposit contract

High return R>1 with p(θ) at date 2 θ: state of the economyDepositors get noisy signal: xi= θ+εi

• θ high: Good fundamentals - no run (upper dominance); θ≤θ low: bad fundamentals - always run (lower dominance)intermediate range: multiple equilibria; panic runs

• Goldstein/Pauzner Global games solution:• Critical θ*: no run above some threshold θ*!

Both θ and θ* are increasing in c1In the range θ≤θ≤θ* panic driven runs Interventions can prevent panic runsencourage insurance (higher c1) Moral Hazard: Support may induce „excessive risk“ - shifting θ(c1) upward beyond some optimal level.

Comments• Laissez Faire solution:

Banks determine θ*(c1) such that • Marginal gain from better risk sharing

(higher c1 for early consumers) equals Marginal loss from increased probability of runs (higher θ*(c1) ) c1

D

• (Constrained) efficient solution: prevent panic runs only fundamental runs; threshold θ(c1) c1

SP>c1D

• Problem: How to avoid panic runs? Costless insurance against panic runs?

Implementation mechanism left unclear in the paper: Insure depositors only for θ<θ(c1). Resources needed?

• Announcement to repay depositors only if θ <θ(c1) won’t help if private agents cannot observe θ

• General Critique: Clear-cut regions of fundamental and panic runs implausible ~~ Too simplified view: In reality, signals provide noisy information about true state of the world alpha error vs. beta error

Comments• Social planner allows transfer of resources from some public good

• Idea: Real deposit insurance in period 1: Guarantee c1

SPI>1 in the case of fundamental runs (θ<θ(c1))• Paid out from funds g available for public goods

• Ad hoc modeling strategy

Since risk averse agents prefer some insurance,why not insure depositors with c1

SPI>1 in all states θ?• Why not also insure against bad realization in period 2?

• Crucial issue: Resources g modeled as exogenously given; corner solutions g not properly modeled (deus ex machina): Partial equilibrium! Determine investment in g endogenously ex ante (distortionary taxes)Strong incentives to provide insurance pool against systemic risks Why no private insurance (investment in safe assets; equity funds)?

CommentsInefficiencies from public guarantee schemes• Guarantuees induce moral hazard (excessive risk taking):

c1GG >c1

SP θ(c1GG)>θ(c1

SP). • Externality: • Government provides insurance funds without adequate „pricing,“

taking private deposit contracts c1GG as given;

overinsurance

• In line with intuition, but not worked out properly: Characterise efficient pricing strategy as benchmark case~ not done convincingly in the paper (only a first step)

Key argument: Cannot prevent banks to offer contracts c1GG >c1

SPI

• Simple mechanism: Provide deposit insurance only for banks offering contracts with payout c1 ≤ c1

SPI

• Other available options : capital adequacy; liquidity requirementsNo role in your set-up ~ strong limitation

Comments

• Comparison of different public deposit insurance schemesAll transfer resources from some given public good g to depositors1) Pay out c1

D to depositors only at t=1• 2) Pay out c1

D to depositors both at t=1 and t=2• 3) Insure all deposit claims fully at t=1 and t=2 • Key insight: Optimal scheme depends on size of g

If g is large, full insurance more efficient than moderate intervention

• With tight budget (small g), limited intervention allowing panic runs is preferred

• Limited insight - Puzzle: How to determine optimal size g? • Very preliminary work

Suggestions• Key problem:

Dynamic inconsistency of conditional guarantee schemes:Incentives to renege on commitment not to intervene

• Cao/Illing (2011), JICB Endogenous exposure to systemic risk Banks have incentives to invest excessively in activities prone to systemic risk

• Allows to model different regulatory designs Liquidity (and capital adequacy) requirements can address these incentivesDiamond/Dybvig framework less suitable – Sequential Service constraint: Optimality of deposit contracts?

Minor comments: Analysis incomplete: Compare c1

SPI relative to c1SP ?

Upper dominance region: Same return R at date 1 and 2 ~ contradicts initial claims