Ring-fencing, Banking Reform, and Taxpayer Exposure Oz Shy b Hanken
Rune Stenbackab School of Economics
Seminar Presentation Bank of Canada Ottawa, November 14, 2016 The views expressed in this presentation are those of the presenter and do not necessarily represent the views of the affiliated institutions.
O. Shy and R. Stenbacka
Ring-fencing and Banking Reform
Bank of Canada
1 / 24
Introduction
Bank failures and financial crises
Bank failures and financial crises Fractional-reserve banking system collapses and bailed out by taxpayers every 10 to 30 years (for over 700 years). Banks create money (Alchemy). Banks lend out 90% of deposits made into transaction accounts, and, often, 99% of deposits made into saving accounts. Depositors do not have control over the risk taken with their money. ⇒ Banks “bundle” account services with risk! Thereby creating a standard “moral hazard” conflict, associated with “managing other people’s money,” Assets and significant amount of leverage = Equity (see balance sheet):
O. Shy and R. Stenbacka
Ring-fencing and Banking Reform
Bank of Canada
2 / 24
Introduction
History of bank failures
Early history of large bank failures and taxpayer bailout In 1345, world’s biggest banks failed, led by the Bardi & Peruzzi companies of Florence (preceded by 30 years of fictitious financial practices) ⇒ total financial disintegration in Europe. Trade volume declined. In 1361, Venice’s Senate prohibited lending out depositors’ money (repealed later on). In 1584, Venice’s largest bank (House of Pisano & Tiepolo) closed because of inability to refund depositors. State gov’t turned it into a state bank that also failed 1619. In 1790 The Bank of Amsterdam was taken over by the City after depositors discovered that it broke its promise to maintain reserves. 1813 Napoleon takes possession of the Bank of Hamburg and finds 7.5m silver Marks in excess over deposits. Later on, French gov’t returns securities ⇒ bank fails. O. Shy and R. Stenbacka
Ring-fencing and Banking Reform
Bank of Canada
3 / 24
Introduction
History of bank failures
Modern history of large bank failures and taxpayer bailout London bank failures and bailouts: 1825, 1836, 1847, 1866, and 1914. 1986–1995: Collapse of 1,043 savings and loan (S&L) associations.
In 1992 the Swedish gov’t guaranteed all bank deposits of the nation’s 114 banks and assumed some of bad debts. 2008: Total collapse of financial systems in the U.S. and Europe
2012–2013 Cyprus’ banks collapse (some depositors eventually get 80 cents on e1). Bitcoin price rises sharply.
O. Shy and R. Stenbacka
Ring-fencing and Banking Reform
Bank of Canada
4 / 24
Introduction
History of bank failures
Major failure waves of U.S. banks
In 1792, the first bank bailout by the U.S. government (Alexander Hamilton, Treasury Secretary). O. Shy and R. Stenbacka
Ring-fencing and Banking Reform
Bank of Canada
5 / 24
Introduction
Taxpayer cost
A sample of taxpayer cost of bailing out banks By 1995, the S&L crisis had cost taxpayers $124 billion, plus additional $29 billion on the thrift industry. The most recent 2008 crisis cost the taxpayer $750b (TARP). The cumulative bailout commitment [asset purchases plus lending by Federal Reserve during 2007–2009 was $7.77 trillion to 407 banks (Bloomberg) and over $29 trillion (Felkerson)]. In March 2013, a e10 billion international bailout of Cyprus’ banks. U.K.’s 2009 gov’t support for the top 5 banks exceeded £100 billion: Therefore, lending (credit) to banks is equivalent to lending to governments (that will bail banks out).
O. Shy and R. Stenbacka
Ring-fencing and Banking Reform
Bank of Canada
6 / 24
Introduction
Deposit insurance
Deposit insurance during financial crises The FDIC’s deposit insurance fund has $72.6 billion (2015/12), which translates to 1.11% reserve ratio (of total U.S. deposits). JPMorgan Chase total deposit is $1.33 trillion (=$1,330b). Bank of America $1.24 trillion. Wells Fargo $1.22 trillion. Citibank $0.94 trillion. Each of the 24 largest U.S. banks holds deposits more than the entire FDIC fund.
Conclusion: Deposit insurance is not a feasible solution for large bank failures! O. Shy and R. Stenbacka
Ring-fencing and Banking Reform
Bank of Canada
7 / 24
Introduction
Lobbying
Lobbying expenditure by U.S. banks
which, in 2015, was over $1.2 million per week.
O. Shy and R. Stenbacka
Ring-fencing and Banking Reform
Bank of Canada
8 / 24
Introduction
Purpose and motivation
Purpose and motivation We analyze structural reforms that aim to separate traditional banking activities and investment banking. Often referred to as ring-fencing. US: “Volcker” Rule (in the framework of the 2010 Dodd-Frank Act). UK: The 2011-2 “Vickers” report (also under consideration by the Bank of England). EU: The 2012 “Liikanen” Report. Ring-fencing separates deposit and loan activities from proprietary trading, investment banking activities, and other wholesale activities. Ring-fencing may prevent the transmission of disastrous contagion within banking organizations. O. Shy and R. Stenbacka
Ring-fencing and Banking Reform
Bank of Canada
9 / 24
Introduction
Our analysis
We analyze 3 levels of ring-fencing: No ring-fencing (NR): Banks are not required to separate retail activities (deposits and loans) from risky investment activities. Deposit insurance and bailout programs cover all realized bank losses (loan defaults and investment losses).
Weak ring-fencing (WR): Banks are not required to separate retail activities from risky investment activities (same as NR). However, deposit insurance and bailout programs do not cover losses incurred by trading and investment activities. Problem: Time inconsistency (we don’t analyze).
Strong ring-fencing (SR): Full separation of retail banking and investment banking. Investment banking is ruled out by the terms specified in a license for commercial banking. Resembles the 1933 Glass-Steagall Act (repealed 1999). O. Shy and R. Stenbacka
Ring-fencing and Banking Reform
Bank of Canada
10 / 24
Introduction
Main results
Main results Starting from a regime with no ring-fencing (NR): (1) Weak ring-fencing (WR) does not affect the equilibrium lending rate. (2) Equilibrium lending rate is lower under strong ring-fencing (SR), hence, (3) banks earn lower profits under strong ring-fencing (SR), hence (4) Strong ring-fencing (SR) generates the highest borrowers’ and consumer surplus = = depositor surplus + borrower surplus − government bailout tax. (5) SR intensifies competition in the loans’ market relative to NR and WR. (6) Weak ring-fencing (WR) generates the lowest expected bailout cost (banks’ investment risks are transferred to depositors). (7) No ring-fencing (SR) generates the highest expected bailout cost.
O. Shy and R. Stenbacka
Ring-fencing and Banking Reform
Bank of Canada
11 / 24
The model
The economy
A model of bank competition with financial crises Two competing banks, k = 1, 2. Each bank maintains D dollars (exogeneous) worth of deposits (2D total value of deposits in this economy). Pay a fixed interest rate on deposits R (exogeneous). Banks compete for borrowers by setting lending rates i1 and i2 . Banks allocate depositors’ money that is not loaned out (2D − B) to risky investments, with rate of return ρI (if no crisis). φ (0 < φ < 1) is probability of a financial crisis, where under (NR) and (WR): (i) Borrowers default on loans (B dollars loaned out are lost), (ii) Bank investments fail (2D − B) dollars investments are lost).
However, under (SR), with probability φ, only borrowers default on loans (B dollars loaned out are lost) as there are no investments. O. Shy and R. Stenbacka
Ring-fencing and Banking Reform
Bank of Canada
12 / 24
The model
Lending market competition
Lending market competition Bank 1
Borrow from bank 1
0
Borrow from bank 2
Bank 2 -
1
xb
x
Borrower x’s expected benefit (return on a loan, or benefit) , x ∈ [0, 1], ( (1 − φ)B(ρL − i1 ) − τ x if borrows from bank 1 Ux = L (1 − φ)B(ρ − i2 ) − τ (1 − x) if borrows from bank 2. We say that lending market competition is intensified or enhanced if τ decreases, whereas competition is reduced or softened if τ increases. Let `1 and `2 be the dollar aggregate value of loans made by each bank: `1 = B xb = B
(1 − φ)B(i2 − i1 ) + τ 2τ `2 = B (1 − xb) = B
O. Shy and R. Stenbacka
Ring-fencing and Banking Reform
(1 − φ)B(i1 − i2 ) + τ . 2τ Bank of Canada
13 / 24
The model
No ring-fencing (NR)
No ring-fencing (NR): Lending market competition Each bank k = 1, 2 chooses its lending rate ik to solve: max πk = `k (1 − φ)(ik − R) + (D − `k )(1 − φ)(ρI − R) {z } | {z } | ik expected profit on loans
expected investment profits
yielding the equilibrium lending rate and expected bank profits ikNR = ρI +
τ B(1 − φ)
and πkNR = D(1 − φ)(ρI − R) | {z } return less deposit rate
τ + | {z 2}
.
competition effect
Interpretation: bank k sets ikNR so that: (1 − φ)(ikNR − R) = (1 − φ)(ρI − R) + τ /B ⇒ Under perfect competition (ruled out), ikNR = ρI . Otherwise, if ikNR < ρI , banks would completely abandon making loans and become investment banks. O. Shy and R. Stenbacka
Ring-fencing and Banking Reform
Bank of Canada
14 / 24
The model
No ring-fencing (NR)
No ring-fencing (NR): Analysis of lending rates and profits
ikNR = ρI +
τ B(1 − φ)
and πkNR = D(1 − φ)(ρI − R) | {z } return less deposit rate
τ + | {z 2}
.
competition effect
(a) Key result: ρI ↑ ⇒ ikNR ↑ ⇒ borrowers become worse off! (higher return on banks’ investment reduces lending competition!) (b) Other effects on lending rates: (τ ↑ ⇒ ikNR ↑), (φ ↑ ⇒ ikNR ↑), (B ↑ ⇒ ikNR ↓). (c) Effects on bank profit: ( ρI ↑ ⇒ πkNR ↑), ( D ↑ ⇒ πkNR ↑), ( τ ↑ ⇒ πkNR ↑), ( R ↑ ⇒ πkNR ↓), O. Shy and R. Stenbacka
Ring-fencing and Banking Reform
Bank of Canada
15 / 24
The model
No ring-fencing (NR)
No ring-fencing (NR): Bailout cost and welfare Expected government bailout cost g NR =
φB(1 + R) | {z }
expected loan losses
+ φ(2D − B)(1 + R) = {z } | expected investment losses
φ2D(1 + R) {z } |
.
loss of deposits plus interest
Expected borrower welfare and depositor welfare: Z xˆ Z Aggreate borrowers’ “travel” cost t = τ x dx + 0
bw NR = (1−φ)B(ρL −ikNR )−t = (1−φ)B(ρL −ρI )−
1
τ (1 − x) dx =
xˆ
5τ 4
τ . 4
and dw NR = 2DR.
Expeced aggregate consumer welfare: cw NR = bw NR +dw NR −g NR = (1−φ)B(ρL −ρI )+2D [(1 − φ)R − φ]−
5τ . 4
Key result: ρI ↑ ⇒ bw NR ↓ ⇒ cw NR ↓ ⇒ consumers become worse off! O. Shy and R. Stenbacka
Ring-fencing and Banking Reform
Bank of Canada
16 / 24
The model
Weak ring-fencing (WR)
Weak ring-fencing (WR) Bank investments are excluded from DI and gov’t bailouts, (time-inconsistency issues may arise). Bailouts apply only to loan default crises, hence, no change in bank behavior: ikNR = ikWR , πkNR = πkWR , bwkNR = bwkWR . Risk of bank investment failures is shifted from gov’t to depositors: g WR = φB(1 + R). dw WR =
BR |{z}
insured interest
+ (1 − φ)(2D − B)R − | {z } uninsured interest
φ(2D − B) | {z }
.
bank investment loss
No change in aggregate consumer surplus: cw WR = bw WR + dw WR − g WR = cw NR . | {z } risk reallocation
O. Shy and R. Stenbacka
Ring-fencing and Banking Reform
Bank of Canada
17 / 24
The model
Strong ring-fencing (SR)
Strong ring-fencing (SR) Banks are prohibited from using depositors’ money to fund their risky investments. Depositors’ money can be used to fund bank loans. Resembles the 1933 Glass-Steagall Act (repealed in 1999). The law was not always fully enforced since the 1960s when National City Bank (now, Citibank) introdueced “negotiable” CDs. Banks solve: max πk = `k (1 − φ)ik − | {z } ik profit on loans
(1 − φ)DR | {z }
.
interest paid on deposits
Note: If there is no crisis, banks are committed to paying interest R on the entire deposits D they hold (including on D − `k that they hold as reserves). O. Shy and R. Stenbacka
Ring-fencing and Banking Reform
Bank of Canada
18 / 24
The model
Strong ring-fencing (SR)
Strong ring-fencing (SR): Equilibrium ikSR =
τ B(1 − φ)
τ 2 |{z}
and πkSR =
−
competition effect
g SR =
φB(1 + R) | {z }
+
expected loan losses
depositors’ interest
φ(2D − B)R {z } |
.
interest on reserved deposits
5τ bw SR = (1 − φ)B ρL − ikSR − t = (1 − φ)BρL − 4 dw NR = dw SR = 2DR
(1 − φ)DR , | {z }
(lower ikSR < ikNR ).
(depositors are fully insured).
h i 5τ cw SR = bw SR + dw SR − g SR = B (1 − φ)ρL − φ + (1 − φ)2DR − . 4 O. Shy and R. Stenbacka
Ring-fencing and Banking Reform
Bank of Canada
19 / 24
The model
Comparison of 3 ring-fence regimes
Comparing 3 ring-fence regimes: Lending rates and profits (SR) generates the most intense bank competition in the lending market: ikNR = ikWR > ikSR
and ikNR − ikSR = ikWR − ikSR = ρI .
Hence, allowing banks to invest (NR) reduces lending market competition, (ρI = difference in lending rates).
(SR) generates lower expected profits for banks: πkNR = πkWR > πkSR .
Bank profit: πkNR , πkWR , πkSR
← lower (investment risk) higher →
2
πkNR = πkWR (τ = 0.1) πkNR = πkWR (τ = 0.4) πkSR (τ = 0.1) πkSR (τ = 0.4)
1
0
φ 0
O. Shy and R. Stenbacka
0.1
0.2
Ring-fencing and Banking Reform
0.3
0.4
0.5
Bank of Canada
20 / 24
The model
Comparison of 3 ring-fence regimes
Comparing ring-fencing: Depositor and borrower welfare (WR) generates the lowest expected depositor welfare: dw NR = dw SR > dw WR . (SR) generates the highest expected borrower welfare: bw SR > bw NR = bw WR . ← lower (investment risk) higher →
1 0 dw NR = dw SR dw WR
−1 −2 −3 −4
φ 0
0.1
0.2
O. Shy and R. Stenbacka
0.3
0.4
0.5
Borrower welfare: bw NR , bw WR , bw SR
Depositor welfare: dw NR , dw WR , dw SR
← lower (investment risk) higher →
bw NR = bw WR (τ = 0.4) bw NR = bw WR (τ = 0.6) bw SR (τ = 0.4) bw SR (τ = 0.6)
5 4 3 2 0 0.1
Ring-fencing and Banking Reform
φ 0.2
0.3
0.4
Bank of Canada
0.5
21 / 24
The model
Comparison of 3 ring-fence regimes
Comparing ring-fencimg: Bailout cost, consumer welfare (WR) generates the lowest expected bailout coste: g NR > g SR > g WR . (SR) generates the highest expected aggregate consumer welfare: cw SR > cw NR = cw WR . Consumer welfare: cw NR , cw WR , cw SR
← lower (investment risk) higher → cw NR = cw WR (τ = 0.1) cw NR = cw WR (τ = 0.4) cw SR (τ = 0.1) cw SR (τ = 0.4)
4 2 0
−2
O. Shy and R. Stenbacka
φ 0
0.1
Ring-fencing and Banking Reform
0.2
0.3
0.4
Bank of Canada
0.5 22 / 24
Model extension
Bank with unequal size
Robustness: Unequal deposit dollar amounts Bank 1 maintains D1 = σ2D deposits (0.5 < σ < 1). Bank 2 maintains D2 = (1 − σ)2D dollar deposits. Assumption: Each bank has sufficient deposits to cover the equilibrium demand for loans: D1 > D2 = (1 − σ)2D > B/2. Results: The larger bank realizes a larger drop in profit compared with the smaller bank if strong ring-fencing is enforced: π1NR − π1SR > π2NR − π2SR > 0
and π1WR − π1SR > π2WR − π2SR > 0,
Expected bailout cost of the larger bank decreases by a greater amount than that of the smaller bank when shifting to strong ring-fencing: g1NR − g1SR > g2NR − g2SR > 0
and g1SR − g1WR > g2SR − g2WR > 0.
Note: The effect of SR of aggregate consumer welfare would depend on the tax strucutre and distribution of deposit amounts. O. Shy and R. Stenbacka
Ring-fencing and Banking Reform
Bank of Canada
23 / 24
Conclusion and extensions
Concluding remarks and extensions Weak ring-fencing minimizes taxpayer exposure [only if goverments can overcome pressures to bail out banks (inconsistency problem)]. Strong ring-fencing (the 1933 GS Act) generates higher consumer welfare and borrower welfare than no and weak ring-fencing. Final remark: The paper underestimates the distortion generated by allowing banks to channel deposits toward trading and investments. Why? Because we assumed that bank investments bear a real expected return (1 − θ)ρI . However, banks also engage in bets such as credit-default swaps [e.g. bets on exchange-rate, interest rate and housing price movements] where a gain to one bank is followed by the loss to another bank that bets in the opposite direction. O. Shy and R. Stenbacka
Ring-fencing and Banking Reform
Bank of Canada
24 / 24