Shadow banks and macroeconomic instability Roland Meeks*, Ben Nelson* and Pier Alessandri† *Bank of England and † Banca

d’Italia

U. Western Ontario London, June 27, 2012

The views expressed in this presentation are those of the authors alone, and not necessarily those of the Bank of England or the Bank of Italy.

Some terms used in this talk Pooling. The bundling together of many individual loans for the purpose of destroying private information possessed by informed parties (buyers or sellers). Tranching. The creation of prioritized claims to cash flows on a pool of assets. Securitization. The transformation of an illiquid loan into a tradeable security through the issuance of prioritized claims backed by loan pools. Claims labeled Asset-Backed Securities, ABS. Shadow banking. The activity of securitizing loans originated by commercial banks.

Background & Motivation Figure: Credit cycles in traditional and shadow banking 10% 8% 6% 4% 2% 0% ‐2% ‐4% ‐6% ‐8% ‐10% 1984Q1

1987Q1

1990Q1

NBER recession

1993Q1

1996Q1

1999Q1

Shadow Bank credit

2002Q1

2005Q1

2008Q1

2011Q1

Traditional Bank credit (ex MBS)

Note: HP filtered data from the United States flow of funds.

Background & Motivation Contribution of the paper... or, what we do, and why we do it. 1

Develop a quantitative general equilibrium model featuring securitization and shadow banking GE essential to capture feedback effects within financial system, and between financial system and the macroeconomy.

2

Demonstrate conditions under which securitization does and does not lead to macroeconomic instability Degree of risk sharing between issuers and investors in ABS is crucial; and the source of the shock matters.

3

Show that shocks to the collateral value of ABS affect bank and shadow bank balance sheets, and the macroeconomy Don’t have to rely on ‘big aggregate shocks’ to generate big recessions.

4

Show that government policies aimed at supporting the securitization market by purchasing ABS do not stabilize overall credit Such policies may appear to ‘work’ (by bringing down spreads) but ignore spillovers that squeeze total credit.

Background & Motivation Contribution of the paper... or, what we do, and why we do it. 1

Develop a quantitative general equilibrium model featuring securitization and shadow banking GE essential to capture feedback effects within financial system, and between financial system and the macroeconomy.

2

Demonstrate conditions under which securitization does and does not lead to macroeconomic instability Degree of risk sharing between issuers and investors in ABS is crucial; and the source of the shock matters.

3

Show that shocks to the collateral value of ABS affect bank and shadow bank balance sheets, and the macroeconomy Don’t have to rely on ‘big aggregate shocks’ to generate big recessions.

4

Show that government policies aimed at supporting the securitization market by purchasing ABS do not stabilize overall credit Such policies may appear to ‘work’ (by bringing down spreads) but ignore spillovers that squeeze total credit.

Background & Motivation Contribution of the paper... or, what we do, and why we do it. 1

Develop a quantitative general equilibrium model featuring securitization and shadow banking GE essential to capture feedback effects within financial system, and between financial system and the macroeconomy.

2

Demonstrate conditions under which securitization does and does not lead to macroeconomic instability Degree of risk sharing between issuers and investors in ABS is crucial; and the source of the shock matters.

3

Show that shocks to the collateral value of ABS affect bank and shadow bank balance sheets, and the macroeconomy Don’t have to rely on ‘big aggregate shocks’ to generate big recessions.

4

Show that government policies aimed at supporting the securitization market by purchasing ABS do not stabilize overall credit Such policies may appear to ‘work’ (by bringing down spreads) but ignore spillovers that squeeze total credit.

Background & Motivation Contribution of the paper... or, what we do, and why we do it. 1

Develop a quantitative general equilibrium model featuring securitization and shadow banking GE essential to capture feedback effects within financial system, and between financial system and the macroeconomy.

2

Demonstrate conditions under which securitization does and does not lead to macroeconomic instability Degree of risk sharing between issuers and investors in ABS is crucial; and the source of the shock matters.

3

Show that shocks to the collateral value of ABS affect bank and shadow bank balance sheets, and the macroeconomy Don’t have to rely on ‘big aggregate shocks’ to generate big recessions.

4

Show that government policies aimed at supporting the securitization market by purchasing ABS do not stabilize overall credit Such policies may appear to ‘work’ (by bringing down spreads) but ignore spillovers that squeeze total credit.

Outline

1

Model overview

2

Equilibrium in the ABS market

3

Model properties

4

Securitization crises

5

Summary & Discussion

Model set-up and assumptions

High-level assumptions I Two constraints mark a departure from the complete markets, real business cycle baseline: 1

Participation. Financial institutions assumed to have vital intermediation role. Households cannot loan funds directly to goods producers (e.g. inability to enforce contracts; inability to verify cash flows; etc.).

2

Pledgeability. Financial institutions are unable to completely pledge income generated by their assets; creditors limit extent of funding.

Model set-up and assumptions Institutional framework Assumption 1 retreats from an ‘institution free’ macroeconomics. Further allow that commercial banks and shadow banks are institutionally distinct, matching their distinct economic functions: Commercial banks originate loans (purchase equity stakes or ‘primary securities’) from goods-producing firms. Shadow banks (‘brokers’ for short) use raw loans to ‘manufacture collateral’ acceptable to commercial bank creditors through securitization. Financial frictions Assumption 2 leads financial intermediaries to earn rents (incentive payments), driving a wedge between returns earned by savers (households) and costs incurred by borrowers (firms).

Model set-up and assumptions

High-level assumptions II Two additional assumptions are specific to our model of securitization: 3

Valuable securitization. Securitization is assumed to augment net aggregate liquidity. Put another way, by transforming illiquid loans into tradeable assets, securitization allows collateral to be used more efficiently. Shadow banks have a comparative advantage in holding loan pools.

4

Risk transfer. The ‘hot potato’ of risk assumed to remain within the financial system, rather than being transferred to unlevered ‘real money’ investors.

Model set-up and assumptions Valuable securitization Existence of shadow banking does not depend on avoidance of regulation. Assumption 3 means that securitization allows the financial system to exploit ‘gains from trade’. ? Suggestive evidence for collateral/liquidity value of securitized assets from changes to bankruptcy code (Perotti, 2010). Risk transfer When commercial banks sell assets, they transfer risk to the shadow banking system (transfers can be complete or partial). Assumption 4 means that aggregate risk concentrates on the balance sheets of levered financial intermediaries. ? More true for some ‘shadow’ banking entities than others (e.g. ABCP conduits; Acharya, Schnabl & Suarez, forthcoming).

Model set-up and assumptions Figure: Real and financial claims in the shadow banking model Commercial banks

Firms

Sb

Nc Mc

K D Sc

Sc

Brokers Nb Sb

Mb

K – physical capital S – primary securities (loans) N – net worth M – asset-backed securities D – deposits

Note: Balance sheet of the three financially constrained sectors. Height of LH column represents assets. Height of RH column represents liabilities. Balance sheets are always valued at market prices.

Model set-up and assumptions Securitization market architecture The distribution of aggregate risk amongst financial intermediaries matters. We consider two alternative assumptions: ‘Risk-sharing’ shadow banking or ‘pass through’ securitization investors get pass through exposure to a broad collateral pool. ABS returns depend on the performance of underlying assets, and so risks are shared between investors and ABS issuers. ‘Risk-taking’ shadow banking, or ‘bank-like’ shadow banks - risk intolerant investors demand insured deposit alternatives (institutional cash pools, Pozsar, 2011; GSG investors, Bernanke, 2011). ABS are fixed (state non-contingent) claims and shadow banks retain risk from assets securitized.

Outline

1

Model overview

2

Equilibrium in the ABS market

3

Model properties

4

Securitization crises

5

Summary & Discussion

Commercial bank portfolio choice Commercial bank’s objective function: c Vtc = Et Λt,t+1 [(1 − σ)nct+1 + σVt+1 ]

Commercial bank’s constraints: Qt sct + qt mct = nct + dt Vtc ≥ θc (Qt sct + [1 − ωc ]qt mct ) := Gct plus the law of motion nct = (Rst − Rmt )Qt−1 sct−1 + (Rmt − Rt ) dt−1 + Rmt nct−1 Linearity ensures the value function is also linear in coefficients vc :
Vtc = µcst Qt sct + vcmt /qt − vct dt + vcmt nct where µcst := vsst /Qt − vcmt /qt .

Commercial bank portfolio choice Problem: max V c (sc , d, nc ) s.t. V c (sc , d, nc ) ≥ Gc (sc , d, nc ). The first order necessary conditions for optimal {sct , dt , λct } are: µcst

λct ≤ θc ωc 1 + λct

λct vcmt − vct ≤ θc (1 − ωc ) qt 1 + λct (µct − θc ωc )Qt sct + (vcmt /qt − vct − θc [1 − ωc ])dt + (vcmt /qt − θc [1 − ωc ])nct ≥ 0 with complementary slackness. The coefficients turn out to depend on expected rates of return: µcs is a function of Rs − Rm ; vcm is a function of Rm ; and so on.

Commercial bank portfolio choice Figure: Determining the asset portfolio of the commercial bank system

Note: On-balance-sheet holdings of loans satisfying bank ICC indicated by hatched area.

Commercial bank portfolio choice Figure: Shadow value of internal funds

Note: Figure shows effect of relaxing the incentive constraint by one unit. µ is the spread Rs − Rm .

Shadow bank portfolio choice Broker’s objective function: b Vtb = Et Λt,t+1 [(1 − σ)nbt+1 + σVt+1 ]

Broker’s constraints: Qt sbt = nbt + qt mbt Vtb ≥ θb (qt mbt + nbt ) := Gbt where θb < θc (Assumption 3), and the law of motion nbt = (Rst − Rmt )Qt−1 sbt−1 + Rmt nbt−1 Linearity ensures the value function is also linear in coefficients vb : Vtb = µbst Qt sbt + (vbmt /qt )nbt where µbst := vbst /Qt − vbmt /qt .

Shadow bank portfolio choice Problem: max V b (sb , nc ) s.t. V b (sb , nc ) ≥ Gb (sb , nb ). The first order necessary conditions for {sbt , λbt } are: µbst ≤ θb

λbt

1 + λbt !   vbmt b b b µst − θb Qt st + n ≥0 qt t with complementary slackness. The coefficients again turn out to depend on expected rates of return: µbs is a function of Rs − Rm ; vbm is a function of Rm .

ABS market equilibrium

Solving the commercial bank program tells us that ABS demand is a function of deposits (+) and net worth (–) ( c ) vst /Qt − θc c 1 c dt − nt qt mt = ωc θc ωc − µcst Solving the shadow bank program tells us ABS supply is a function of broker net worth (+): vb /Qt − θb b qt mbt = st nt θb − µbt

ABS market equilibrium Figure: The ABS market clears when ABS demand and supply are equated

Note: The loan to ABS spread Rs − Rm adjusts to clear the market.

Outline

1

Model overview

2

Equilibrium in the ABS market

3

Model properties

4

Securitization crises

5

Summary & Discussion

Parameter values used in simulations Target Data Expected financier survival time Gross financial wedge Rs − R ABS spread to safe rates Rm − R Share of assets securitized Commercial bank loan to equity ratio Shadow bank loan to equity ratio

Parameter σ θc ωc θb ξc ξb

Value 0.90 0.2216 0.50 0.1224 0.0134 0.0083

Value 10 quarters 100 bps 50 bps 0.3 4.5 10

Description Survival probability for financiers Divertibility of bank loans Relative divertibility of ABS Divertibility of broker loans Fraction of assets transferred to new banks Fraction of assets transferred to new brokers

Aggregate shocks Figure: A decline in total factor productivity I

-1

-2

-1.5 0

-4 0

20 QS

c

-0.2

-0.8 0

20 QS

b

-5

-10

-10

-10 0

-20 0

-20 0

c

N

% 2 1 0 20

-1 0

N

0

0

Risk sharing Risk taking

%

20

-20 20

3

b

10 %

20

20 E[Rm - R]

c

%

0

%

0

%

0

-4 0

20 qM

10

-20 0

-2

-0.6

10

-10

0

-0.4

5

20

Q 2

%

0

C

%

-0.5 %

2

ppt

Y 0

-40 0

20

20

Aggregate shocks Table: Correlation between credit and output in the model and the data

Output Investment Total credit Traditional Bank credit Shadow bank credit Bank leverage Shadow bank leverage

Risk Sharing 1 0.90 0.75 0.71 -0.46 -0.19 -0.38

Risk Taking 1 0.95 0.82 -0.80 0.83 -0.81 0.82

Data 1 0.80 0.12 0.51 -0.35 -0.34 -0.23

Note: Theoretical correlations are conditional on productivity shocks only. Data correlations are on HP filtered series taken from the United States Flow of Funds, 1984:1–2007:2. Shadow bank leverage is for broker-dealer data only.

Financial shocks Figure: Commercial bank funding shock Y

I

0.5

-0.5

0

0.2

-2

-6 0

c

%

0.4 0 20 QS

-0.2 0

b

-4 -6 0

20 qM

6

-2

6

6

4

-4

4

4

ppt

8 %

8

2

-6

2

2

0

-8 0

0 0

0 0

-2 0

20 Nc

20

20

Nb

0

10

-20

-10

%

0

With Securitization Without Securitization

-20 -40 0

20

20 E[Rm - R]

c

0 %

%

0 -2 -4 20 QS

%

Q 2

%

%

%

0

-1 0

C

2

-30 0

20

Note: One-off transfer of 10% of steady state bank capital to households.

20

Importance of heterogeneity? Figure: Redistribution of net worth from brokers to commercial banks I

-1 -1.5 0

20

QS -10

-10

-30 0

-30 0

%

0.2

20

-10 %

0

4 2 0 0

-20 20

-30 0

0 -0.2

Nb

6

20 E[Rm - R]

c

-20 20

-0.2 0

20 qM

0

Nc

%

0

0

-20 20

0

b

%

%

5

0.2

-0.2 0

20

Q

0.2

%

QS

c

10

0 0

%

%

%

-0.5

-0.2 -0.4 0

C

0

ppt

Y 0

20

Note: One-off transfer of 25% of steady state broker capital to banks.

-0.4 0

20

Outline

1

Model overview

2

Equilibrium in the ABS market

3

Model properties

4

Securitization crises

5

Summary & Discussion

Securitization crises Figure: Spreads on Auto Securitizations 1200

1000

800

600

400

200

0 04/01/1994 01/10/1996 01/07/1999 02/04/2002 04/01/2005 01/10/2007 01/07/2010

Note: Scale is in basis points (hundredths of a percentage point).

Securitization crises Figure: Credit cycles in traditional and shadow banking (repeat) 10% 8% 6% 4% 2% 0% ‐2% ‐4% ‐6% ‐8% ‐10% 1984Q1

1987Q1

1990Q1

NBER recession

1993Q1

1996Q1

1999Q1

Shadow Bank credit

2002Q1

2005Q1

2008Q1

2011Q1

Traditional Bank credit (ex MBS)

Note: HP filtered data from the United States flow of funds.

Securitization crises Challenge: how to generate very large increases in ABS spreads without assuming a ‘really big’ aggregate shock (which we didn’t see). Ingredients of a crisis: capital and collateral Aim to capture in a reduced form way the main elements of the securitization crisis of 2007-2009. 1

Shadow bank assets become less effective for raising funding. Proxies for investor doubts about the quality of loan pools.

2

Shadow bank liabilities become less valuable collateral for leveraged investors (here, commercial banks). Proxies for illiquidity of ABS.

3

Commercial bank net worth is impaired. A one-off shock to proxy for the effect of defaults on loans. This has the effects already discussed.

These are fundamentally financial shocks, as they affect the terms under which funding can be obtained.

Government backstops

Government asset purchases Consolidated government sector includes fiscal authorities, central bank, GSE’s. Asset purchases involve an ‘efficiency cost’ of τ per unit Lump sum taxes Tt are available g

Asset purchases are funded by issuing risk-free debt Dt

Government debt partly substitutes for deposits in households’ asset portfolio Consider direct funding for firms (loan purchases) and shadow banks (ABS purchases)

Government backstops With active asset purchases the government’s budget constraint becomes: g

g

g

g

g

g

Gt + Qt St + qt Mt + Rt Dt−1 = Tt + Dt + Rst Qt−1 St−1 + Rmt qt−1 Mt−1 The presence of a real resource cost associated with asset purchases gives rise to non-zero public expenditure, parameterized by τ = 0.002 (2/10 of a cent on the dollar). g g Gt = τ(St + Mt ) The share of assets purchased (as a proportion of steady state credit) follow simple rules: ϕst = γ0s + γ1s {Et (Rs,t+1 − Rt+1 ) − (Rs − R)} ϕm t = γ0m + γ1m {Et (Rm,t+1 − Rt+1 ) − (Rm − R)} g

where e.g. ϕst = St /K. Set γ0i = 0.025; implies government holds about 10% of the steady state stock of ABS.

Securitization crisis Figure: Direct loan purchases Y

I

0.5

-5

c

-10

qM

-20 0

c

N

20

20

20

S shr

0

400

No intervention Loan purchases

%

600

-40 0

0 0

g

20

-20

-40

-30 0

20

%

-20

5

-20

N

0

10

-10

b

20

20 E[Rm - R]

c

0

-10 20

20

-10 0

10 %

0

b

%

0 %

20 QS

10

-0.5 0

ppt

-8 0

20 QS

%

0

0

-6

10

-60 0

5 %

-4

%

%

% -0.5

-20 0

Q

0.5

-2

0

-1 0

C

0

200 20

0 0

20

Note: Share of government loan holdings goes to 14% of ss stock (from 2.5%).

Securitization crisis Figure: ABS purchases

-1 -1.5 0

-5

0

-10 0

20 Q Sc

-15 0

0

-10 0

20 c

N

20

20

400

0 0

20

No intervention ABS purchases

%

0

-40 0

5

M shr 600

%

%

-60 0

20 E[Rm - R]

g

20

-20

-40

-100 0

N

0 -20

20

-10 0 10

-50

b

20

20

%

50

0 -5

-10

0 -5

q Mc

5 %

%

0

Q 5

-1 0

20 Q Sb

5 -5

%

%

%

-0.5

C 1 %

I 0

ppt

Y 0

200 20

0 0

20

Note: Share of government ABS holdings goes to 50% of ss stock (from 10%).

Outline

1

Model overview

2

Equilibrium in the ABS market

3

Model properties

4

Securitization crises

5

Summary & Discussion

Summary & Relation to Literature Discussion of main insights We develop a model where banks and shadow banks are interdependent, and interact with the macroeconomy in general equilibrium Macro shocks drive credit activity in and out of the regulatory perimeter (the ‘disintermediation’ phenomenon has cyclical as well as structural characteristics) Asset price spillovers (‘pecuniary externalities’) create an additional channel between regulated and unregulated sectors When shadow banking is ‘bank like’, the result is macroeconomic instability (in line with FSB thinking) Capture effect of securitization crisis through exogenous changes in liquidity ABS purchases following a financial shock does not stabilize overall credit, even if policy succeeds in bringing down spreads and in stabilizing the traditional banking system

Summary & Relation to Literature How does our work relate to the wider literature on financial frictions in macroeconomics, and other work on shadow banking? Closely related to work of Gertler/Karadi/Kiyotaki (2011/12) – focus on supply side of credit market, rather than the demand side emphasis in financial accelerator literature (Bernanke, Gertler & Gilchrist, 1999; etc.) Shadow banking model of Gennaioli, Shleifer & Vishny (2011) – similar motivation for securitization, but qualitative rather than quantitative results; lots of special assumptions. Goodhart, Kashyap, Tsomocos & Vardoulakis (2012). Similar structure of financial sector balance sheets, but completely different motivation for existence of shadow banks. Focus on regulation (which is absent from our model).

Summary & Relation to Literature

How does our work relate to the wider literature on financial frictions in macroeconomics, and other work on shadow banking? (ctd...) The DSGE literature: Verona, Martins & Drummond (2011) – ad hoc approach, in which banks and shadow banks fail to interact Hobijn & Ravenna (2010) – adverse selection model with endogenous sorting of loans into securitization pools, but no role for intermediary capital Faia (2010) – two-sided moral hazard with unobserved liquidity shocks. Ability to securitize doesn’t depend on health of shadow banks.

End of presentation.

Additional Material Figure: Intermediation cycles

Total FinancialAssets Shares 100%

BD GSE Pool ABS FC CU SI CB

90% 80% 70%

60% 50%

40% 30% 20% 10% 1960Q1 1962Q1 1964Q1 1966Q1 1968Q1 1970Q1 1972Q1 1974Q1 1976Q1 1978Q1 1980Q1 1982Q1 1984Q1 1986Q1 1988Q1 1990Q1 1992Q1 1994Q1 1996Q1 1998Q1 2000Q1 2002Q1 2004Q1 2006Q1 2008Q1 2010Q1

0%

Note: HP filtered data from the United States flow of funds.

Regulation of shadow banking system

The Financial Stability Board issued a report Shadow banking: Strengthening oversight and regulation in October 2011. The FSB identifies four factors that lead shadow banks to pose systemic risks: 1

Maturity transformation

2

Liquidity transformation

3

Credit risk transfer

4

High leverage

Shadow banks in the complete risk transfer version of our model tick all these boxes.

Regulation of shadow banking system

The Financial Stability Board issued a report Shadow banking: Strengthening oversight and regulation in October 2011. We do not capture some key theory behind the FSB’s list: Runs, rollover risk and freezes. Volatility and risk pricing. Risk shifting and correlated strategies. All would make good topics for future work.