INTRODUCTION

Are U.S. Banks Safer?Reading the Fed’s New Dashboard

Andrew G. Atkeson Adrien d’Avernas

UCLA Economics, FRB MPLS Stockholm School of Economics

INTRODUCTION

The Fed Board’s New Dashboard

Accounting Signals

Book value of equity to assets

Bank profitability, loan quality, and asset growth

Market Signals

Market value of equity to assets

CDS and bond spreads and equity volatility—distance to default

How to interpret these signals for:

regulators, holders of bank equity, bank bonds, and U.S. Taxpayers

regulators, holders of bank equity, bank bonds, and U.S. Taxpayers

INTRODUCTION

Conflicting Signals

Accounting Signals . better

Book value of equity to assets

Bank profitability, loan quality, and asset growth

Market Signals . worse

Market value of equity to assets

CDS and bond spreads and equity volatility

How to interpret these signals for:

regulators and U.S. Taxpayers

holders of bank equity and bank bonds

INTRODUCTION

Today’s Talk

Unified model of banks’ accounting and market signals

Sarin and Summers (2016)

Atkeson, d’Avernas, Eisfeldt, and Weill (2019)

Market signals of size of crisis shock to bank assets

Begenau, Bigio, Majerovitz, and Vieyra (2019)

Begenau, Piazzesi, Schneider (2015)

Interpret changes in average signals pre- to post-crisis

Interpret accounting and market signals in the cross-section

Meiselman, Nagel, and Purnanandam (2018)

Bailout expectations in bond spreads

Berndt, Duffie, Zhu (2019)

INTRODUCTION

Outline

1. Fed Dashboard and Stress Test Framework

2. Market Signals of Size of Crisis Shock

3. Unified Model of Banks’ Accounting and Market Signals

4. Interpret Accounting and Market Signals

5. Conclusion

INTRODUCTION

Fed Dashboardand Stress Test Framework

INTRODUCTION

Accounting Data: Book Equity Up, Small Drop in 2007-2008

32  

5. Capital Adequacy and Asset Growth Note: CET1, tier 1 and total capital is reported instead of the components of tier 1 common equity and tier 1 and total risk-based capital by advanced approaches firms starting in 2014:Q1, and by all other firms starting in 2015:Q1, causing series breaks in some capital ratios in those quarters. Changes in the measurement of RWA starting in 2013:Q1 and 2015:Q1 also affect measurement of risk-weighted capital ratios and the ratio of RWA to total assets starting in those quarters. See ''Caveats and Limitations'' for details. See data notes for definition of tier 1 common equity.

7

8

9

10

11

12

13

14

15

1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013 2015 2017 2019

All Institutions BHCs >$500bn

BHCs $50bn-500bn Banks and BHCs <$50bn

Tier 1 capital as % of risk-weighted assets

Tier 1 Capital Ratio

INTRODUCTION

Accounting Data: Profitability Improving

12  

2. Earnings and Pre-Provision Net Revenue

-2

-1.5

-1

-.5

0

.5

1

1.5

1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013 2015 2017 2019

All Institutions BHCs >$500bn

BHCs $50bn-500bn Banks and BHCs <$50bn

Annualized net income as % of total assets

Return on Assets

-20

-10

0

10

20

1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013 2015 2017 2019

All Institutions BHCs >$500bn

BHCs $50bn-500bn Banks and BHCs <$50bn

Annualized net income as % of equity

Return on Equity

INTRODUCTION

Accounting Data: Loan Quality Improving

17  

3. Loan Performance Note: Non-performing loans include loans that are (1) 90 days or more past due and still accruing or (2) non-accrual.

0

1

2

3

4

5

6

7

8

1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013 2015 2017 2019

All Institutions BHCs >$500bn

BHCs $50bn-500bn Banks and BHCs <$50bn

Total non-performing loans as % of total loans

Non-performing Loans

0

2

4

6

8

10

12

14

1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013 2015 2017 2019

All Institutions BHCs >$500bn

BHCs $50bn-500bn Banks and BHCs <$50bn

Non-performing residential real estate loans as % of residential real estate loans

Non-performing Residential Real Estate Loans

INTRODUCTION

Market Data: Market Equity Down, Big Drop in 2007-2008

2000 2002 2004 2006 2008 2010 2012 2014 2016 20180.06

0.08

0.1

0.12

0.14

0.16

0.18

0.2

0.22

0.24

0.26Market Equity to Risk-Weigthed Assets

INTRODUCTION

Market Data: CDS Spreads Up

0

50

100

150

200

250

300Ja

n-06

Oct-0

6Ju

l-07

Apr-0

8Ja

n-09

Oct-0

9Ju

l-10

Apr-1

1Ja

n-12

Oct-1

2Ju

l-13

Apr-1

4Ja

n-15

Oct-1

5Ju

l-16

Apr-1

7Ja

n-18

Oct-1

8

Basis

Poin

tsAverage CDS Spread for Large Banks

INTRODUCTION

Stress Test Framework and Results

Stress test framework can be summarized as:

CET1t+1

RWAt+1=

1

(1 + gAt)1/4CET1t+1

RWAt

CET1t+1

RWAt=CET1tRWAt

+PPNRt

RWAt− Provisionst

RWAt− TaxestRWAt

− DividendstRWAt

. model responses of income statement items to severe macro shock

. cumulative impact on accounting capital over a fixed horizon

Results:

aggregate drop in CET1/RWA of about 6 percentage points

much larger (smaller) than the loss of book (market) capital

INTRODUCTION

Market Signals of Size of Crisis Shock

. observed drop in market value of equity

. Merton’s distance-to-default

. crisis losses on corporate bond portfolios

INTRODUCTION

Observed losses on market value of bank equity

Banks lost about $1 Trillion in market cap 2006-2009

Drop in ME/RWA of 20 percentage points

Compared to drop in BE/RWA of less than 2 percentage points

Compared to stress test drop in BE/RWA of 6 percentage points

. Lower bound on the drop in bank value

INTRODUCTION

Distance-to-Default: Equity Volatility and Market Leverage

Merton’s distance-to-default

Moody’s expected default frequency

Berndt, Douglas, Duffie, and Ferguson (2018)

Atkeson, Eisfeldt, and Weill (2017)

Problem: volatility jumps in a crisis

d’Avernas (2018)

Nagel and Purnanandam (forthcoming)

. Have to model the probability of a jump in volatility

INTRODUCTION

Bond returns as a measure of crisis losses

Begenau, Piazzesi, Schneider (2015);Atkeson, d’Avernas, Eisfeldt, and Weill (2019)

corporate bonds include MBS

look at total returns on diversified bond portfolios

Have to match bank loan portfolios on maturity and credit risk

maturity risk

BBB credit risk

High Yield credit risk

losses of 15-20 percent of portfolio value in 2 years or less frequentcan be as high as 40 percent

INTRODUCTION

Unified Model of Banks’ Accountingand Market Signals

. bank franchise value and market value of government guarantees

. feed in crisis losses of 16% of bank value

. can match many accounting and market signals pre- and post-crisis

INTRODUCTION

Gordon Growth Model for Accounting and Valuation

Time period: time to reset book equity or close bank

States s ∈ S are i.i.d. under risk-neutral probability q(s)

For calibration two states: s ∈ {n, c} normal and crisis.

. q(c) risk-neutral crisis probability

Constant risk-free rate i

Assets: loans L

gov’t guaranteed deposits D, sub. debt B, and book equity BE

Assets and liabilities grow at the same rate g(s)

INTRODUCTION

Franchise Value of Equity (FE)

Fair value of a one-dollar loan

PV of: interest - servicing costs + principal payments - default

vL > book value = 1

Fair value of a one-dollar deposit

PV of: interest + servicing costs + principal payments

vD < book value = 1

Franchise value of equity per dollar of loans

FE = (vL − 1)× L− (vD − 1)× D

lend at high rates, borrow at low rate

INTRODUCTION

Market Value of Equity (ME)

Market value of equity with default decision

ME =1

1 + i

∑s

q(s) max {0, divE(s) + (1 + g(s))ME}

What happens upon default?

gov’t seizes the bank and injects cash to assist sale

full recovery on deposits, loss `(s) on subordinated debt

size of gov’t bailout depends on bank losses and leverage

INTRODUCTION

Gov’t Guarantees and the Market Value of Equity

Define the market value of gov’t guarantees

MG = PV of all future cash injections

valuation multiple times expected bailout in a crisis

MG =q(c)

i− gT (c)

Modigliani Miller with gov’t as negative stakeholder

ME = BE + FE + MGGOV

BE

INTRODUCTION

Equity Valuation and Bond Spreads

Market value of equity is multiple times dividend in normal time

ME =q(n)

(1 + i)− q(n)(1 + g(n))divE(n)

Yield spread is risk neutral crisis probability times loss given crisis:

yB − i = q(c)`(c)

vB

INTRODUCTION

Accounting Profitability, Franchise Value, and Risk Taking

Benchmark for accounting profitability

q(n)ROE(n) + q(c)ROE(c) = i+ (i− g)

(FE

BE

)Profitability ROE(n) high either due to

high franchise value FE

risk-taking, that is low ROE(c)

INTRODUCTION

Interpret Accounting and MarketSignals

INTRODUCTION

Identification: what drives the change in signals?

yield spreads imply still have risk of default

risk free rate i plus dashboard data

and price-dividend or asset growth in normal times pin down

risk to equity q(c) and sub. debt `(c)

does not help with measuring risk to taxpayers

Two strategies to identify risk to taxpayers MG

measure franchise value directly

calibrate risk in bank assets to infer franchise value

INTRODUCTION

Risk-Taking, ROE, and ME in the cross-section

Meiselman, Nagel, and Purnanandam (2018)

high ROE pre-crisis predicts high systematic tail risk in the crisis

Model implications: higher exposure to crisis shock goes with

higher ROE in normal times

higher market to book with gov’t guarantees

no impact on market to book without guarantees

In the data

pre-crisis high ROE predicts high market to book ratio

high market-to-book ratio of equity predicts high systematic tail risk

INTRODUCTION

rL(c) 0.9 0.88 0.86 0.84 0.82

ME/BE 1.63 1.63 1.83 2.02 2.19ROE(n) 0.14 0.15 0.16 0.17 0.19

Pre-Crisis Cross-Section BE = 9%

rL(c) 0.9 0.88 0.86 0.84 0.82

ME/BE 1.06 1.06 1.09 1.19 1.32ROE(n) 0.07 0.07 0.08 0.09 0.10

Post-Crisis Cross-Section BE = 13%

INTRODUCTION

rL(c) 0.9 0.88 0.86 0.84 0.82

ME/BE 1.63 1.63 1.83 2.02 2.19ROE(n) 0.14 0.15 0.16 0.17 0.19

Pre-Crisis Cross-Section BE = 9%

rL(c) 0.9 0.88 0.86 0.84 0.82

ME/BE 1.04 1.04 1.04 1.04 1.04ROE(n) 0.05 0.06 0.06 0.07 0.07

Post-Crisis Cross-Section BE = 20%

INTRODUCTION

(1) (2) (3) (4)log(ME/BE) tail risk measure tail risk measure tail risk measure

ROE(n) 0.534∗∗∗ -0.413∗∗∗ -0.171∗∗∗

(14.25) (-10.22) (-3.94)

log(ME/BE) -0.544∗∗∗ -0.452∗∗∗

(-14.59) (-10.40)Observations 510 510 510 510R2 0.286 0.171 0.295 0.316

Standardized beta coefficients; t statistics in parentheses∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001

INTRODUCTION

Bailout Expectations in Bond Spreads

In our model, bond spreads are given by:

yB − i = q(c)`(c)

vB

Berndt, Duffie, Zhu 2018

proxies for risk-neutral probability of default q(c)

infer changing expected loss on bonds `(c)

model calibration of q(c) and `(c)

interpret BDZ estimates in common units

yB − i = q(c)× (1− π)× (c)

INTRODUCTION

Small Change in Expected Default Losses

Pre-Crisis

π cds `(c) q(c)

0.51 75 0.29 2.55%0.46 75 0.32 2.31%0.43 75 0.34 2.19%0.39 75 0.37 2.05%

- 75 0.15 5.00%

Post-Crisis

π cds `(c) q(c)

0.30 158 0.42 3.76%0.20 158 0.48 3.29%0.10 158 0.54 2.93%0.00 158 0.60 2.63%

- 158 0.32 5.00%

INTRODUCTION

Conclusion

INTRODUCTION

Main takeaways

Model to interpret accounting and market signals

feasible to develop a quantitatively plausible model

useful for framing debates over the meaning of signals

Faith in CCAR Stress Tests likely misplaced

book capital is not the problem in a crisis

should use market measures of crisis shocks to banks

and try to measure bank franchise value

What’s missing from our model?

Duffie 2019: securities broker dealers were the problem in 2008

Will derivatives be the next crisis?

INTRODUCTION

Securities Broker Dealers Total Assets Have Shrunk

0

10

20

30

40

50

6019

52Q

119

55Q

119

58Q

119

61Q

119

64Q

119

67Q

119

70Q

119

73Q

119

76Q

119

79Q

119

82Q

119

85Q

119

88Q

119

91Q

119

94Q

119

97Q

120

00Q

120

03Q

120

06Q

120

09Q

120

12Q

120

15Q

120

18Q

1

Perc

enta

ge P

oint

sSecurity Broker Dealer Total Assets (Balance Sheet) over GDP

INTRODUCTION

Bank Interest Rate Derivatives Have Shrunk

0

5

10

15

20

25

30

35

4019

95:Q

119

96:Q

219

97:Q

319

98:Q

420

00:Q

120

01:Q

220

02:Q

320

03:Q

420

05:Q

120

06:Q

220

07:Q

320

08:Q

420

10:Q

120

11:Q

220

12:Q

320

13:Q

420

15:Q

120

16:Q

220

17:Q

320

18:Q

4

Perc

enta

ge Po

ints

Fair Value of Interest Rate Derivatives over GDP

Positve Negative

INTRODUCTION

Bank Credit Derivatives Have Shrunk

0

1

2

3

4

5

6

7

820

02:Q

120

02:Q

420

03:Q

320

04:Q

220

05:Q

120

05:Q

420

06:Q

320

07:Q

220

08:Q

120

08:Q

420

09:Q

320

10:Q

220

11:Q

120

11:Q

420

12:Q

320

13:Q

220

14:Q

120

14:Q

420

15:Q

320

16:Q

220

17:Q

120

17:Q

420

18:Q

3

Perc

enta

ge Po

ints

Fair Values of Credit Derivatives Positive and Negative Protection Bought and Sold