FinObservatory

Housing cycles

Real house prices around a banking crisis

A five-years-before, five-years-after event study on the BIS real residential property price index, with the Laeven-Valencia systemic banking crises as the events. Run again on 150 years of Jorda-Schularick-Taylor data and split at 1945, the same test gives a median five-year real run-up of +5.5% and a median trough of -5.3% for the 29 measurable episodes dated 1882 to 1939, against +29.1% and -21.3% for the 25 dated 1974 to 2008.

29
Usable crisis episodes
26 countries, 1977 to 2011
+19.7%
Median 5-year real run-up
into the crisis year; 8 of 29 were negative
-20.6%
Median trough after
lowest of t to t+5, against t; worst -49.1%
19/29
Episodes dated 2007 to 2011
the rest: 10 episodes
Read this first. The BIS index is rebased to 2010 = 100 in every country separately, so a level comparison across countries is meaningless. Only growth and drawdown are compared here. 19 of the 29 usable episodes began between 2007 and 2011, so the cross-country median is not 29 independent draws. Every limitation is listed at the bottom of the page.

The same event study before and after 1945

Jorda-Schularick-Taylor carries annual house prices and a banking-crisis flag for 18 advanced economies, 1870 to 2020. The identical event study on it yields 54 episodes with a complete t-5 to t+5 window: 29 dated 1882 to 1939, and 25 dated 1974 to 2008. No usable episode falls between 1940 and 1973. The completeness rule bites hardest early: 29 of the 63 banking crises JST dates through 1945 have house prices five years either side of them, against 25 of 25 after 1945. The earlier median therefore describes 29 of the 63 crises JST dates through 1945, not all of them.

708090100crisis year-5-4-3-2-10+1+2+3+4+5years from the first year of the banking crisisThrough 1945, 1882-1939 (n=29)After 1945, 1974-2008 (n=25)

Source: Jorda-Schularick-Taylor Macrohistory Database Median across episodes of real house prices (hpnom / cpi), re-normalised to 100 in the crisis year (crisisJST = 1). Episodes require a complete t-5 to t+5 window. Methodology

EraCrises datedUsable episodesCountriesMedian run-up, t-5 to tMedian trough, t to t+5
Through 1945 (usable: 1882 to 1939)632912+5.5%-5.3%
After 1945 (usable: 1974 to 2008)252517+29.1%-21.3%

Source: Jorda-Schularick-Taylor Macrohistory Database Medians are taken over episodes, not over countries. Crises dated counts every crisisJST = 1 year in the era, usable or not. Methodology

The modern event study, 29 episodes, 26 countries

BIS real index, Laeven-Valencia banking crises, 1977 to 2011. The median path is highest at t-1 (102.3, against 100 at t) and stands at 80.6 five years after the crisis year. At t+5 the 25th to 75th percentile across episodes runs 73.3 to 93.3, and 5 of the 29 episodes are above their own crisis-year level.

708090100crisis year-5-4-3-2-10+1+2+3+4+5years from the first year of the banking crisisMedian (n=29)Mean

Source: BIS residential property prices (selected series) | Laeven and Valencia, Systemic Banking Crises Database (2025 update) Annual mean of the four quarterly real index values, re-normalised to 100 in the first crisis year. Complete t-5 to t+5 windows only. Methodology

Year from crisisEpisodesMedianMeanp25p75Above t2007-2011 median (n=19)All other (n=10)
t-52983.687.674.8100.9882.293.5
t-42985.491.980.0103.8985.399.9
t-32994.997.388.5103.41094.2104.4
t-229101.0101.494.6106.51599.4103.9
t-129102.3103.0100.1106.822102.3101.9
t (crisis)29100.0100.0100.0100.05100.0100.0
t+12993.393.388.599.1593.889.7
t+22990.389.781.798.5590.482.2
t+32984.187.379.396.4584.184.0
t+42981.384.774.793.4479.482.6
t+52980.684.373.393.3578.583.0

Source: BIS residential property prices (selected series) | Laeven and Valencia, Systemic Banking Crises Database (2025 update) Crisis year = 100 by construction. Above t counts the episodes strictly above their own crisis-year level at that horizon. The last two columns split the same 29 episodes: the 19 dated 2007-2011 stand at 82.2 at t-5, the 10 others at 93.5. Methodology

Every episode, one row each

Sorted by the five-year real run-up into the crisis. 8 of the 29 episodes had real house prices falling over the five years before the crisis year, and 3 never closed a year below the crisis-year level in the five years after it.

CountryCrisisRun-up t-5 to tTrough vs tYears to troughPeak to trough
RUS Russia2008-2009+146.6%-34.1%3-34.1%
GBR United Kingdom2007-2011+47.5%-22.5%5-22.5%
DNK Denmark2008-2009+45.8%-20.6%4-27.2%
FRA France2008-2009+43.2%-6.2%1-7.9%
ISL Iceland2008-2012+41.2%-25.9%2-31.0%
SWE Sweden2008-2009+40.9%0.0%no fall-2.2%
ESP Spain2008-2012+35.6%-39.7%5-42.9%
SWE Sweden1991-1995+33.6%-24.9%2-26.6%
BEL Belgium2008-2012+33.1%-0.7%4-0.9%
MYS Malaysia1997-1999+30.7%-18.3%2-18.8%
ESP Spain1977-1981+27.5%-25.1%5-25.1%
USA United States2007-2011+25.6%-31.7%4-37.4%
CHE Switzerland2008-2009+21.7%0.0%no fall0.0%
USA United States1988+20.2%-8.0%5-8.0%
IRL Ireland2008-2012+19.7%-49.1%4-54.5%
GRC Greece2008-2012+17.2%-37.9%5-39.4%
ITA Italy2008-2009+16.3%-16.9%5-17.5%
CYP Cyprus2011-2015+16.1%-21.6%4-32.7%
FIN Finland1991-1995+16.0%-28.1%2-47.1%
NLD Netherlands2008-2009+11.6%-26.7%5-26.9%
AUT Austria2008-2012+1.5%0.0%no fall-2.1%
COL Colombia1998-2000-0.9%-29.0%5-38.9%
THA Thailand1997-2000-1.2%-18.7%4-24.4%
HUN Hungary2008-2012-3.0%-31.4%5-34.0%
DEU Germany2008-2009-8.4%-1.5%2-9.7%
PRT Portugal2008-2012-12.8%-19.4%5-29.7%
JPN Japan1997-2001-14.1%-15.4%5-27.3%
KOR Korea1997-1998-22.9%-17.8%4-36.6%
NOR Norway1991-1993-26.9%-8.0%2-40.6%

Source: BIS residential property prices (selected series) | Laeven and Valencia, Systemic Banking Crises Database (2025 update) Trough vs t = the lowest of the annual real index values in t to t+5, against the crisis year. Peak to trough measures from the highest annual value in t-5 to t. Years to trough is 0 when the crisis year is itself the low. Methodology

The long US view: 1890 to 2022

In real terms the index went from 100.0 in 1890 to 218.3 in 2022: 0.59% a year over 132 years. The same series in nominal terms multiplied by 83.9x (3.41% a year). Every figure here is real: Shiller’s real index is the nominal index divided by the CPI column of the same table and set to 100.0 at the first observation.

501001502001900192019401960198020002020Real US home price index, 1890 = 100.0, annual means

Source: Robert Shiller, Online Data (Irrational Exuberance, home price index) Complete years only, as annual means. The table is mixed-frequency: 63 annual rows (1890-1952) and 842 monthly rows (1953-2023), and the final partial year is dropped rather than plotted as a year. Methodology

Real index, monthly eraMonthLevel
Peak2005-12195.8
Trough of the deepest real drawdown2012-02125.9
All-time real high2022-05224.4
Last observation2023-01209.3

From the 2005-12 peak, real US house prices fell 35.7% to 2012-02. On complete-year annual means, the deepest fall in the whole record runs from 1894 (124.0) to 1921 (65.6), -47.1%, and the index did not close a year above the 1894 level again until 1987, 93 years later. The two are measured at different frequencies: the first monthly, the second on annual means.

The all-time real high is 2022-05, +14.6% above the 2005 peak. The last observation (2023-01, where the series ends) is -6.7% off that high.

Source: Robert Shiller, Online Data (Irrational Exuberance, home price index) The nominal index behind this is spliced from 4 sources (Grebler 1890-1933, Five-CityMedian 1934-1952, PHCPI 1953-1974, S&P/CoreLogic/Case-Shiller 1975-2023) and the CPI deflator from 2 (Warren&Pearson 1890-1912, BLS 1913-2023). Methodology

Where each of the 57 BIS countries stands, and when its data starts

The BIS index starts as early as 1947 and as late as 2010, a 63-year spread, and 2 of the 57 countries start in 2010. 26 of 57 report through 2026-Q1 and the rest stop earlier (31 at 2025-Q4), so the Latest column is not a common date. Off own real peak is measured against each country’s own highest observed real quarter, so it is conditional on when that country’s index begins.

CountryStartsQuartersLatestReal y/yNominal y/yOff own real peakEpisodes (usable)
ITA Italy19473162025-Q4+2.9%+4.1%-27.8%1 (1)
JPN Japan19552842025-Q4+2.3%+5.0%-35.9%1 (1)
NZL New Zealand19622552025-Q4-4.2%-1.3%-27.2%0 (0)
ZAF South Africa19662412026-Q1+2.7%+6.0%-26.8%0 (0)
GBR United Kingdom19682322026-Q1-2.2%+0.9%-11.0%1 (1)
AUS Australia19702252026-Q1+5.7%+10.0%at peak0 (0)
BEL Belgium19702242025-Q4+1.3%+3.5%-4.9%1 (1)
CAN Canada19702252026-Q1-6.8%-4.8%-29.3%0 (0)
CHE Switzerland19702252026-Q1+4.5%+4.7%at peak1 (1)
DEU Germany19702242025-Q4+0.9%+3.0%-19.5%1 (1)
DNK Denmark19702242025-Q4+5.5%+7.6%-3.1%1 (1)
FIN Finland19702242025-Q4-3.2%-3.2%-24.7%1 (1)
FRA France19702252026-Q1-0.8%+0.2%-12.2%1 (1)
IRL Ireland19702252026-Q1+3.6%+6.7%-5.2%1 (1)
NLD Netherlands19702242025-Q4+3.1%+6.1%at peak1 (1)
NOR Norway19702252026-Q1+0.5%+3.8%-6.8%1 (1)
SWE Sweden19702242025-Q4+0.7%+1.2%-20.1%2 (2)
USA United States19702252026-Q1-2.1%+0.6%-2.9%2 (2)
ESP Spain19712202025-Q4+9.6%+12.9%-13.9%2 (2)
KOR Korea19752052026-Q1-0.3%+1.8%-30.2%1 (1)
HKG Hong Kong SAR19791862026-Q1+6.2%+7.8%-31.0%0 (0)
ISL Iceland19811812026-Q1-2.0%+3.2%-2.3%1 (1)
AUT Austria19861582025-Q4-1.8%+2.1%-17.3%1 (1)
COL Colombia19881522025-Q4+7.2%+12.9%-0.3%2 (1)
MYS Malaysia19881532026-Q1+0.1%+1.7%-1.2%1 (1)
PRT Portugal19881522025-Q4+16.3%+18.9%at peak1 (1)
THA Thailand19911412026-Q1+1.9%+1.3%at peak2 (1)
HUN Hungary19921362025-Q4+17.0%+21.4%at peak2 (1)
ISR Israel19941282025-Q4-2.7%-0.3%-3.7%1 (0)
GRC Greece19971172026-Q1+2.6%+5.7%-18.9%1 (1)
HRV Croatia19971162025-Q4+12.0%+16.0%at peak1 (0)
BGR Bulgaria19981122025-Q4+7.0%+12.6%-2.7%1 (0)
LTU Lithuania19981092025-Q4+6.8%+10.8%at peak1 (0)
PER Peru19981122025-Q4+4.6%+6.1%-27.7%1 (0)
SGP Singapore19981122025-Q4+2.1%+3.3%-0.1%0 (0)
MKD North Macedonia20001052026-Q1+12.6%+16.7%at peak1 (0)
SRB Serbia20001052026-Q1+2.7%+5.3%-15.1%0 (0)
BRA Brazil20011012026-Q1+0.2%+4.3%-26.0%2 (0)
RUS Russia20011012026-Q1+7.4%+13.7%-27.4%2 (1)
CHL Chile2002962025-Q4+2.0%+5.5%at peak2 (0)
CYP Cyprus2002962025-Q4+7.6%+7.1%-20.6%1 (1)
IDN Indonesia2002972026-Q1-3.2%+0.6%-33.3%1 (0)
CHN China2005842026-Q1-7.1%-6.3%-24.7%1 (0)
EST Estonia2005852026-Q1+8.1%+11.9%at peak1 (0)
MEX Mexico2005852026-Q1+4.4%+8.7%at peak2 (0)
MLT Malta2005842025-Q4+3.4%+6.1%at peak0 (0)
LVA Latvia2006812026-Q1+7.8%+10.9%-21.7%2 (0)
SVK Slovakia2006812026-Q1+9.3%+13.4%at peak1 (0)
LUX Luxembourg2007762025-Q4-2.8%+0.1%-21.1%1 (0)
MAR Morocco2007762025-Q4+0.4%+0.3%-25.2%1 (0)
SVN Slovenia2007762025-Q4+3.0%+5.8%at peak2 (0)
CZE Czechia2008722025-Q4+8.0%+10.4%-3.2%1 (0)
PHL Philippines2008722025-Q4-0.1%+1.6%-6.4%2 (0)
IND India2009682025-Q4+3.0%+3.6%-8.0%1 (0)
ROU Romania2009682025-Q4-2.5%+7.0%-40.0%1 (0)
POL Poland2010642025-Q4+1.8%+4.3%at peak1 (0)
TUR Türkiye2010652026-Q1-3.3%+26.7%-17.9%2 (0)

Source: BIS residential property prices (selected series) | Laeven and Valencia, Systemic Banking Crises Database (2025 update) Real and nominal y/y are the BIS year-on-year series read directly, not differenced from the index. Methodology

What this data cannot tell you

Levels are not comparable. The BIS index is 2010 = 100 in every country independently. It is an index of each country’s own price against its own base year, not a price, so it cannot say whether housing is dearer in one country than another. Growth and drawdown are comparable; levels are not, and no level comparison across countries appears in this module.

19 of the 29 episodes began in the same five years. They began between 2007 and 2011. The median path is therefore not 29 independent observations. The event-study table carries the 19 and the 10 others in separate columns, so the split is visible at every horizon.

Coverage kills 34 of the 63 crises. Laeven-Valencia dates 63 banking crises across 49 of the 57 BIS countries. 29 of those 63 have a real index reaching five years before AND five years after, in 26 countries; the other 34 cannot be measured here. 30 of the 34 are in the 23 countries that have a crisis on record but no usable episode (BGR, BRA, CHL, CHN, CZE, EST, HRV, IDN, IND, ISR, LTU, LUX, LVA, MAR, MEX, MKD, PER, PHL, POL, ROU, SVK, SVN, TUR), and every one of those 30 is missing the year five years before the crisis. The remaining 4 sit in countries that do contribute at least one other episode. Every one of them is listed on its country page and marked not measurable.

8 countries have no crisis to study at all. Australia (AUS), Canada (CAN), Hong Kong SAR (HKG), Malta (MLT), New Zealand (NZL), Singapore (SGP), Serbia (SRB), South Africa (ZAF) carry no systemic banking crisis in Laeven-Valencia. They still get a page for their price history, and they enter no median here. Absence from the chronology is the chronology’s judgement, not proof that no crisis happened.

The aggregates are excluded. The BIS file also carries 4 aggregate series (Advanced economies, Emerging market economies (aggregate), Euro area, World) with no ISO3 code. They are not countries, they get no page, and they enter no median.

Shiller is the US, and it is spliced. One country. Its 1890-1952 segment is annual rather than monthly, and its nominal input over those years comes from Grebler and Five-CityMedian. The long US comparisons on this page (1894 to 1921, 1890 to 2022) are only as good as that splice and the CPI splice under it.

Timing, not causation. The crisis dates come from Laeven-Valencia, not from this price data, and nothing here separates cause from effect. This module reports what the two series do around each other in time.

Source: BIS residential property prices (selected series) | Laeven and Valencia, Systemic Banking Crises Database (2025 update) | Jorda-Schularick-Taylor Macrohistory Database | Robert Shiller, Online Data (Irrational Exuberance, home price index) Methodology