Banking crises interventions, 1257-2019 - Andrew Metrick Paul Schmelzing April 2021 IMHOS seminar - Centre for ...
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Banking crises interventions, 1257-2019
Andrew Metrick
Paul Schmelzing
April 2021
IMHOS seminar
1GENERAL
• Extensive literature on banking crises chronologies
– Reinhart & Rogoff (2009).
– Schularick & Taylor (2012).
– Laeven & Valencia (2013, 2020).
– Baron, Verner, and Xiong (2021).
• General focus: systemic crises-year classifications from 1870.
• Broader intervention focus overwhelmingly on post-1980s, systemic resolution of banking crises: e.g. Laeven & Valencia, Igan et al. (2019, for
post-2007).
• What we ARE trying to do
– Specify exact crises responses to “canonical” (existing literature) and “candidate” crises (confirmed interventions outside existing literature).
– Detailing for the first time a large sample of global “would-be systemic crises” – partly averted because of early intervention measures.
– Specify (secular) probabilities of “hands-off ” responses, or private burden sharing in direct intervention costs.
– Using the latest long-run DM GDP datasets to specify direct intervention costs (fiscal, liquidity, guarantee) for hundreds of cases.
• What we are NOT trying to do
– “Reclassifying” existing chronologies.
– Judging intervention effectiveness over time.
– Calculating all secondary crises costs: wider output, bank-systemic, financial market costs.
– Documenting interventions outside of banking/financial sector (i.e. GM 2008).
2OUR APPROACH
• Classifying banking crises interventions across 5 major intervention groups, consisting of 21 intervention types.
• Including both private and public sector interventions, across “canonical” and new “candidate” crises.
• Incorporating latest long-run (current) GDP series (e.g. Malanima 2011; Ridolfi 2020) to provide direct intervention costs, wherever
available. Currently: 473 intervention cost datapoints (of which: 386 outside L/V’s 1980-2017). Prior to 1850, current GDP focus on
“core DM” group.
• Types
– Lending: ad hoc liquidity assistance (AHLA), broad-based liquidity assistance (BBLA), market liquidity assistance (MLA).
– Guarantees: deposit guarantees (AG), asset guarantee (ASG), blanket guarantees (BG), other liability guarantees (OLG –
existing/forward/combined).
– Capital injections: ad hoc capital injection (AHCI), broad-based capital injections (BBCI).
– Rules: debt and payment moratoria (DPM), stock market closures and bank holidays (SBH), credit rules (CRL), other rules
(ORL).
– Other: resolution or restructuring (RES), major communication (MC), stress test (ST), stakeholder bail-in (SPI), ad hoc asset
management (AHAM), broad-based asset management (BBAM), no intervention (NO/I), other.
3CRISES AND INTERVENTION UNIVERSE
C=No. of crisis country years (start year); I=No. of interventions identified for respective crisis
4
country start years. Canonical crises only record “unique” crises for non-Reinhart/Rogoff.IDENTIFICATION PROCESS
1. Investigation of all “canonical” crises in literature.
• Primary, secondary sources detailing public/private responses.
2. Wide investigation of financial, financial history primary and secondary literature, across space, time, and languages to identify “candidate”
crises.
• Micro-histories (e.g. Roover’s “Medici Bank”, Buist’s “Hope & Co.”), Institutional histories (e.g. Gilbart’s History of the Bank of Ireland,
Montaud’s “Banca de Emision en Cuba”).
• Sector histories, general (financial) history – including EM and DM (e.g. Bisschop’s, Rise of the London Money Market, 1640-1826 (1910),
Davidson’s Geschichte von Florenz (1896)).
• Pre-1945 literature, including Italian, Spanish, German sources (e.g. Ferrara’s “Documenti per servire alla Storia de’ Banchi Veneziani”
(1871)).
• Primary sources (e.g. Calendar of State Papers, Fugger archives Dillenburg, newspapers), here primarily DM.
• Minimum criteria:
• Intervention or balance sheet size of affected single institution during crisis event > 5m USD/GBP (from 1850), > 250,000
USD/GDP (from 1700), > 10,000 USD/GBP/RFL (from 1257).
OR
• Intervention affecting more than one institution at once.
• Where neither information is available – judgment call (few cases overall).
• In general, we estimate >90% of “core DM” printed material is covered since 1650s.
5GENERAL FINDINGS, I
• Bank interventions have a long pre-Bagehot, pre-central bank history, often featuring substantial outlays across categories, e.g.:
– 1595 Bolognese Monte fund (“early SPV”) to support banking sector: 8.9% of GDP
– 1739 Venetian recapitalization of Bancogiro: 26.9% of GDP
– 1815 Prussian guarantee for Koeniglich Preussische Bank: 2.6% of GDP
– 1875 Brazilian emergency loans to Banco de Brazil, Banco Rural e Hypothecario et al.: 3.1% of GDP
• Among the most popular pre-Bagehot tools: debt moratoria/mandated bank suspensions (DPM, SBH), liquidity (AHLA).
• Over time, liquidity assistance interventions have been the most frequently-used tool to respond to crises, though receding over 2000s.
• Clear differences between DM and EM crises response: DMs with meaningfully higher shares of liquidity, guarantee use (c.f. slide 12).
• As countries “graduate” towards higher per capita GDP levels, “NO/I” and “rules” use appears to decrease (c.f. slide 12).
• There are clear historical trends in tools used:
– Capital injections represent a comparatively new, 20th century tool – much more scarcely used prior to 1945.
– “Guarantees” have seen a notable revival in their frequency, and are the 2nd most prominent tool as of the 2000s.
– “Rules” represent a key traditional intervention tool until the 1980s. But since then, their importance has sharply diminished.
6GENERAL FINDINGS, II
• Further:
– Frequencies of “hands-off ” (“NO/I”), or relying on private sector response (“PRI-PRI”, “PRI-PRI (partly)”) are becoming
scarcer. Private burden-sharing, in fact, appears to have reached the lowest levels ever in the period since 1980.
– While overall crises frequency increases gradually over the (very) long-term (for “core DM” by a factor of ca. 3.5x over 1700-
2019), crises frequency for key historical periods does not necessarily imply more aggressive responses (intervention costs).
– The pre-classical gold standard period’s (pre-1880) fiscal response was on average more aggressive than the classical gold standard
period, with comparable aggressiveness on liquidity measures.
– While the Bretton Woods Period (1945-71) ranks low on crises frequency (c.f. Bordo et al. 2001), it already shows substantial rise
in crises costs. The interwar period shows high crises frequency, but surprisingly low fiscal and liquidity crises costs.
Intervention costs to NGDP Pre-1880 1880-1945 1914-1945 1945-1971 1945-1980 Post-1980
Fiscal 2.2% 1.5% 1.7% 7.1% 15.8% 11.5%
Liquidity 1.6% 1.8% 1.8% 3.6% 8.2% 14.2%
Guarantees 0.9% 7.9% 11.2% - 0.2% 15.2%
Crisis frequency, % of “core
DM” GDP affected
5.4% 14.4% 20.4% 5.1% 8.6% 21.2%
Share of NO/I (all countries) 5.0% 16.3% 12.6% 14.7% 12.4% 6.6%
Share of “PRI-PRI”, or “PRI-
PRI (partly)” (all countries)
14.2% 16.0% 13.7% 9.4% 12.1% 4.9%
7INCREASING SENSITIVITY OF PUBLIC SECTOR DURING CRISIS EVENTS
– Conditional on the existence of a crisis event, the combined probabilities of either a “hands-off ” approach, or a substantial
burden-sharing of the private sector in the direct intervention costs appear to have decreased substantially post-1980s – while it
was unusually high during the classical Gold Standard Era (1880-1914).
Frequencies of “non-public” response categories, by historical regime (all countries)
35.0%
30.0%
25.0%
20.0%
15.0%
10.0%
5.0%
0.0%
Pre-1880 1880-1945 1914-1945 1945-1971 1945-1980 Post-1980
NO/I PRI-PRI PRI-PRI (partly) 8DATABASE STRUCTURE Monthly dates wherever available Private sector involvement? State-owned
banks?
Total no. of countries: 133
Crisis start year,
associated with
one or more
interventions
9EXAMPLE: LIQUIDITY COST CALCULATION – SWEDEN 1907 CRISIS.
• “The Riksbank provided this foreign exchange to Swedish banks through the mechanism of rediscounting bills of exchange,
presented by the banks in the autumn of 1907 (Söderlund, 1964, pp. 303–304; Fregert, 2018, p. 117). Between August and
December 1907, the value of domestic bills rediscounted by the Riksbank increased by 53% and returned to August 1907 levels
as late as May 1909”.
Source: Grodecka-Messi, Kenny, and Oegren (2021). 10GENERAL BREAKDOWN – 21 TYPES
Intervention types, 1257-2019 (all countries)
250
Total: 1,624 crisis interventions By major intervention group:
221
Lending: 433
200 196
Guarantees: 268
Rules: 227
169
# of occurences
159
150
115
106 105 104
98
100
60 58
50 44 43
25 22 19 16 11 6 3 3
0
AHLA
RES
SBH
DPM
ASG
AHA
SPI
BG
ST
AHCI
AG
OLG
BBCI
ORL
other
CRL
MLA
BBLA
BBAM
MC
NO/I
MINTERVENTION DISTRIBUTION, ALL COUNTRIES, 1750-2019
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
1778
1750
1764
1792
1806
1820
1834
1848
1862
1876
1890
1904
1918
1932
1946
1960
1974
1988
2002
2016
Guarantees Lending - BBLA Lending - AHLA Rules
Capital injection - BBCI Capital injection - AHCI other NO/I
12
15-year centred average.GENERAL BREAKDOWN – INTERVENTION GROUPS BY PERIOD
100%
12.1% 12.9% 13.4% 15.6%
90% 20.9% 20.3% 17.5%
24.4%
5.2%
3.4% 9.1% 8.2%
80% 5.8%
2.7% 4.7% 18.8%
70% 8.1% 5.2% 14.7% 17.1%
4.7%
36.2% 28.0% 31.1%
60%
24.4% 12.1%
22.1% 28.1% 21.8%
50%
2.8%
40%
6.4%
17.4% 35.8% 9.4% 18.5%
30% 38.2% 33.7%
37.9% 20.8%
20%
21.9%
23.3% 24.3%
10% 13.5%
11.0% 15.9%
5.2% 9.1% 6.3%
0%
1257-1699 (total 1700-1799 (total 1800-1879 (total 1880-1914 (total 1915-1945 (total 1946-1971 (total 1972-1998 (total 1999-2019 (total
n=86) n=65) n=203) n=176) n=238) n=32) n=467) n=325)
Guarantees Lending Rules Capital injections RES All other
13GENERAL BREAKDOWN – INTERVENTION GROUPS BY COUNTRY TYPE
100%
90% 17.9% 18.1%
21.6%
80% 5.6% 6.3%
70% 6.9%
14.6% 16.7%
60%
17.2%
12.4%
10.5%
50%
40% 11.1%
30.7% 30.1%
30%
21.6%
20%
10% 18.8% 18.4%
10.0%
0%
DM total (n=844) DM since 1800 (n=702) EM total (n=745)
Guarantees Lending Rules Capital injections NO/I All other 14“CANONICAL” VERSUS “CANDIDATE” CRISES RESPONSE
100%
90% 19.1%
33.4%
80%
16.4%
70%
60% 15.8%
15.9%
50%
11.0%
40%
31.3%
30% 23.9%
20%
10% 17.3% 16.0%
0%
Candidate crises (n=424) Canonical crises (n=571)
Guarantees Lending Rules Capital injections other
• greater propensity to use “lending” and “rules” in candidate (“would-be systemic”) crises –
versus non-standard tools in systemic crises.
15CRISES COST OVERVIEW: FISCAL, LIQUIDITY, GUARANTEES.
70.0%
Example fiscal+liquidity: at present, 407 cases 1257-2019, mean cost:
7.9% of current GDP.
60.0%
Range: 0.02% (“FRA-1810”) – 56.8% (“ID-1997”).
50.0%
Cost to current GDP (capped axis)
40.0%
30.0%
20.0%
10.0%
0.0%
1280
1579
1878
1257
1303
1326
1349
1372
1395
1418
1441
1464
1487
1510
1533
1556
1602
1625
1648
1671
1694
1717
1740
1763
1786
1809
1832
1855
1901
1924
1947
1970
1993
2016
ALG ANG ARG ARM AT AUS AZ BA BEL BEN BOL BOS BRL BUG BUK BUR CAD CAF CB
CDI CH CM CN CL COB COGR COS CR CY CZK DJI DK DOR DRC EC EG ELS EQG
ER ES EST ETH EU FIN FRA G GER GG GH GI GIB GRE GUA GY HK HO HT
HUG ICE ID IN IR IRL IS IT JAC JOR JP KOR KW KY KYR KZ LAT LB LIB
LIT LUX MAC MAG MAL MAU MD MN MO MOG MOZ MRT MT MX MY NE NG NIC NL
NOR NP NZL PA PG PHP PK PR POL PT QT ROM RUS SAF SEN SG SIR SLO SLV 16
SRL SWA SWE SZ TG TH TN TK TW TZ UA UG UK UKR US VN VTCRISES FREQUENCY, “CORE DM”, 1665-2019.
120 GFC, 35.0%
European debt crisis
Share of advanced economy GDP (11-year centered avg)
30.0%
100
Great Depression
Asian/EM crisis 25.0%
80 Napoleonic Wars
# of events
European Revolutions
20.0%
60 Long
War of the Depression
Spanish Succession
15.0%
40
10.0%
20
5.0%
0 0.0%
1793
1905
2017
1665
1681
1697
1713
1729
1745
1761
1777
1809
1825
1841
1857
1873
1889
1921
1937
1953
1969
1985
2001
Guarantees Lending Rules Capital injection other Share of DM GDP showing bank stress (RHS)*
*bank stress frequency: combining Bordo et al. (2001), Reinhart/Rogoff (2009), Schularick and Taylor (2012), Laeven and Valencia (2020), Baron/Verner/Xiong (2021), Metrick and Schmelzing (2021) banking crises or bank
intervention chronologies, for eight country DM sample. Frenquency=(no. of country years with stress event in any database)/(total no. of country years). Includes systemic and non-systemic events. GDP weights based on
Schmelzing (2020).FREQUENCY-ADJUSTED COSTS: INTERVENTION COSTS AS A % OF GDP,
PER ANNUM.
Arithmetically-weighting country-level GDP outlays, frequency-adjusted
1.4%
1.2%
As % of current GDP, 15-year centered avg
1.0%
0.8%
0.6% Floating era: 0.55%
0.4%
0.2% All-time average year: 0.15%
0.0%
1887
1700
1711
1722
1733
1744
1755
1766
1777
1788
1799
1810
1821
1832
1843
1854
1865
1876
1898
1909
1920
1931
1942
1953
1964
1975
1986
1997
2008
2019
Fiscal authority outlays, p.a. Monetary authority outlays, p.a.
Total intervention cost outlays, all countries, p.a. Long-run average (all countries, fiscal+liquidity)
Floating regime avg total 18TAKEAWAYS
• New and nuanced panorama of banking sector intervention patterns over multiple centuries, drawing on wide source basis.
• Interventions per se existed long before Bagehot and modern central banks – across fiscal, liquidity, and guarantee types.
• Intervention bias shows dynamic picture, over space and time: increasing focus on guarantee, and capital injection measures.
• Increasing bias to opt for capital injections and guarantees over “rules” measures (esp. DPM, SBH) as countries graduate, financial systems
mature.
• “Crises” (canonical + candidate) appear to secularly increase in frequency.
• “Crises” (canonical + candidate) appear to secularly become more costly for authorities (as a share of GDP and in absolute terms), with a first
notable break around 1945 (especially pronounced for fiscal costs).
• Globally, fiscal authorities are “back in play”.
• Frequencies of “NO/I”, “PRI-PRI”, “PRI-PRI (partly)” have all secularly decreased: since 1980s, crises appear more likely than ever to
witness(exclusive or near-exclusive) public shouldering of (direct) intervention costs.
• Future uses/research
– Playbook: how to avoid graduation of “non-systemic” into “systemic” event. Use of early-stage liquidity interventions?
– Assess macroeconomic and sectoral dynamics, by intervention group.
– Assess links to capital mobility, and other systemic features and intervention bias (e.g. gold standard era vs. “deregulation era”)
– Assess public-private cost sharing dynamics.
– Assess NO/I dynamics: when and why are authorities becoming more “sensitive” to crises.
– Assess drivers for secular shifts in intervention bias: EM vs DM, pre-1980s vs. post-1980s.
19APPLICATIONS: SECULARLY RISING PRIVATE SECTOR RISK COULD BE
BEHIND BROADER TRENDS IN PRIVATE-PUBLIC RISK PREMIA.
30.0% 4.0%
3.0%
25.0%
2.0%
20.0%
1.0%
15.0% 0.0%
-1.0%
10.0%
-2.0%
5.0%
-3.0%
0.0% -4.0%
1713
1985
1665
1681
1697
1729
1745
1761
1777
1793
1809
1825
1841
1857
1873
1889
1905
1921
1937
1953
1969
2001
2017
Share of DM GDP showing bank stress* (LHS) Spread private - public long-term rates (RHS)
Private-public spread based on identical “Core DM” country sample, using long-term Continental European
secured mortgage rates for private, GDP-weighted sovereign long-term yields for public, described in more 20
detail in Schmelzing (2020).You can also read
