American Economic Review 101 (August 2011): 1676–1706
http://www.aeaweb.org/articles.php?doi
=
10.1257/aer.101.5.1676
1676
The economics profession has an unfortunate tendency to view recent experience
in the narrow window provided by standard datasets.
1
It is particularly distressing
that so many cross-country analyses of nancial crises rely on debt and default data
going back only to 1980, when the underlying cycle can be a half century or more
long, not just 30 years.
2
This paper attempts to address this deciency by employing a comprehensive new
long-term historical database for studying debt and banking crises, ination, and
currency crashes.
3
To construct our dataset, we build on the work of many scholars
as well as a considerable amount of new material from diverse primary and second-
ary sources. The data covers 70 countries in Africa, Asia, Europe, Latin America,
North America, and Oceania.
4
The range of variables encompasses external and
domestic debt, trade, GNP, ination, exchange rates, interest rates, and commodity
prices.
5
Our analysis spans over two centuries, going back to the date of indepen-
dence or well into the colonial period for some countries.
1
That is why an exception such as Milton Friedman and Anna Jacobson Schwartz’s (1963) monumental mon-
etary history of the United States still resonates almost one-half century after publication.
2
For a longer perspective on crises, see the work of Michael Bordo, Barry Eichengreen, Peter H. Lindert and
Peter J. Morton, and Moritz Schularick and Alan M. Taylor (2009).
3
This is the rst formal application of the core dataset described in Reinhart and Rogoff (2009a), and the scope
of the dataset has been expanded signicantly as well.
4
See Appendix Table A1 for the full list of countries.
5
External debt refers to debt that is legally governed by foreign law, in contrast to debt governed by the law of
the issuing country. This is not the only way to parse the data, but it is a useful one empirically.
From Financial Crash to Debt Crisis
†
By C M. R K S. R*
Newly developed historical time series on public debt, along with
data on external debts, allow a deeper analysis of the debt cycles
underlying serial debt and banking crises. We test three related
hypotheses atboth“world” aggregate levels and on an individual
country basis. First, external debt surges are an antecedent to bank-
ing crises. Second, banking crises (domestic and those in nancial
centers) often precede or accompany sovereign debt crises; we nd
they help predict them. Third, public borrowing surges ahead of
externalsovereign default, as governments have “hidden domestic
debts” that exceed the better documented levels of external debt.
(JEL E44, F34, F44, G01, H63, N20)
* Reinhart: Peterson Institute for International Economics, 1750 Massachusetts Avenue NW, Washington,
DC 20036, NBER, and CEPR; Rogoff: Department of Economics, Harvard University, 1875 Cambridge Street,
Cambridge, MA 02138 (e-mail: krogoff@harvard.edu) and NBER. The authors are grateful to Vincent Reinhart,
Jane Trahan, and Wei Xiong, seminar participants at Columbia, Princeton, and Johns Hopkins universities, and two
anonymous referees for useful suggestions, and the National Science Foundation Grant No. 0849224 for nancial
support.
†
To view additional materials, visit the article page at http://www.aeaweb.org/articles.php?doi=10.1257/
aer.101.5.1676.
1677
REINHART AND ROGOFF: FROM FINANCIAL CRASH TO DEBT CRISIS
VOL. 101 NO. 5
Exploiting the multicentury span of the data, we study the role of repeated
extended debt cycles in explaining the observed patterns of serial default and bank-
ing crises that characterize the economic history of so many countries—advanced
and emerging alike. We test three related hypotheses atboth“world” aggregate lev-
els and on an individual country basis. First, external debt surges are a recurring
antecedent to banking crises. Second, banking crises (both domestic and those ema-
nating from international nancial centers) often precede or accompany sovereign
debt crises. Indeed, we nd they help predict them. Third, public borrowing surges
ahead of an externalsovereign debt crisis, as governments often have “hidden debts”
that far exceed the better documented levels of external debt. These hidden debts
include domestic public debt (which was largely undocumented prior to our data)
and private debt that becomes public (and publicly known) as the crisis unfolds.
6
A
fourth related hypothesis (which we document but do not test) is that during the nal
stages of the private and public borrowing frenzy on the eve of banking and debt
crises and (most notoriously) bursts of hyperination, the composition of debt shifts
distinctly toward short-term maturities.
7
This paper is organized as follows. Section I describes our approach toward cata-
loging, dating, and connecting the various manifestations of economic crises. Here
we dene the key concepts in our analysis: serial default and the “this time is differ-
ent” syndrome. Section II presents the big picture on global cycles of debt, nancial
crises and sovereign debt crises. We use representative country histories to elaborate
on and complement some of the patterns seen in the global aggregates. The robust-
ness of the descriptive analysis is grounded in a related Chartbook
8
that spans more
than two centuries of data and documents the crisis experience and debt history
of each and every one of the 70 countries that make up our sample. We empha-
size describing the broad phases of the debt cycle, the sequencing of crises, and
some of their features—such as the duration and frequency of default spells. History
suggests that policymakers should not be overly cheered by the absence of major
external defaults from 2003 to 2009 after the wave of defaults in the preceding two
decades. Given that international waves of defaults are typically separated by many
years, if not decades, there is no reason to suppose that serial default is dead.
Section III discusses some alternative theoretical frameworks that might help
explain the observed patterns discussed in the preceding section with a special
emphasis on serial default and the this-time-is-different syndrome. Section IV com-
plements the descriptive big-picture analysis in Section II by exploiting the rich
panel dimension of our data to test for temporal causal patterns across crises and the
role of public and private debts in the run-up to sovereign debt and nancial crises.
I. Crisis Denitions and Other Concepts
We begin by developing working denitions of what constitutes a nancial cri-
sis, as well as the methods—quantitative where possible—to date the beginning
6
Quantifying public contingent liabilities is beyond the scope of this paper.
7
This is closely related to the themes in Dani Rodrik and Andres Velasco (2000).
8
See Reinhart (2010), “This Time Is Different Chartbook: Country Histories on Debt, Default, and Financial
Crises,” which henceforth will be referred to as the Chartbook.
1678
THE AMERICAN ECONOMIC REVIEW
AUGUST 2011
and end of a crisis. The boundaries drawn are generally consistent with the exist-
ing empirical economics literature, which by and large is segmented across the
various types of crises considered (e.g., sovereign debt, exchange rate, etc.). Two
approaches are used to identify crisis episodes. One, which can be applied to ina-
tion and exchange rates crises, is quantitative in nature, while the other, which
we apply to debt and banking crises, is based on a chronology of events. The
crisis markers discussed in this section refer to individual countries as opposed to
global events.
A. Ination, Hyperination, and Currency Crises
Expropriation takes various forms, beyond outright default, repudiation, or the
restructuring of domestic or external debts. Indirect routes to achieving the same
end—ination and currency debasement—can also erode the value of some types
of existing debts. Thus, we date both the beginning of an ination or currency crisis
episode and its duration. Many of the high-ination spells can be best described as
chronic, in that they last many years.
Reinhart and Rogoff (2004), which classied exchange rate arrangements for
the post–World War II period, used a 12-month ination threshold of 40 percent
or higher to dene a “freely falling” episode. Our current work spans a much
longer period, before the widespread creation of at currency. Median ination
rates before World War I were well below those of the more recent period: 0.5
percent for 1500–1799 and 0.7 percent for 1800–1913 versus about 5 percent for
1914–2009. Accordingly, we dene an ination crisis using a threshold of 20 per-
cent per annum. Hyperinations, which are dened as episodes where the annual
ination rate exceeds 500 percent, are of modern vintage.
9
Hungary 1946 holds
the sample’s record despite the recent challenge from Zimbabwe, which comes
in second.
10
To date currency crashes, we follow a variant of Jeffrey A. Frankel and Andrew
K. Rose (1996) and focus exclusively on exchange rate depreciation. This de-
nition is the most parsimonious, as it does not rely on other variables, such as
reserve losses (data that many central banks guard jealously) and interest rate
hikes.
11
Mirroring our treatment of ination episodes, an episode is counted for
the entire period in which annual depreciations exceed the threshold of 15 percent
per annum.
Hardly surprising, currency crashes and ination crises go hand in hand. Figure 1
plots the incidence of the two varieties of monetary, or at-money, crises—i.e.,
exchange rate and ination crises. The “honor” for the record annual currency
crash goes to Greece in 1944, also a year of hyperination (see Reinhart and
Rogoff 2009a).
9
Note that this denition of hyperination (unlike Philip Cagan’s (1956) classic denition of a monthly ination
rate of 50 percent or greater) does not require monthly readings of ination, which are scarce prior to the twentieth
century.
10
See Figure 70 (Zimbabwe) in the Chartbook for a comparison of hyperination episodes.
11
See Graciela L. Kaminsky and Carmen M. Reinhart (1999) for a more detailed discussion of indices that
measure exchange market turbulence.
1679
REINHART AND ROGOFF: FROM FINANCIAL CRASH TO DEBT CRISIS
VOL. 101 NO. 5
B. Debt Categories and Debt Crises
External debt crises involve outright default on payment of debt obligations incurred
under foreign legal jurisdiction, including nonpayment, repudiation, or the restructur-
ing of debt into terms less favorable to the lender than in the original contract.
12
These events have received considerable attention in the academic literature
from leading modern-day economic historians, such as Bordo, Eichengreen, Marc
Flandreau, Lindert and Morton, and Taylor.
13
Relative to early banking crises, much
is known about the causes and consequences of these rather dramatic episodes. For
post-1824, the dates come from several Standard and Poor’s studies. However, these
are incomplete, missing numerous postwar restructurings and early defaults. This
source has been supplemented with additional information from Lindert and Morton
(1989), Christian Suter (1992) and Michael Tomz (2007). Of course, required read-
ing in this eld includes Max Winkler (1933) and William H. Wynne (1951).
While the time of default is accurately classied as a crisis year, in a large num-
ber of cases the nal resolution with the creditors (if it ever does take place) seems
interminable. Russia’s default following the revolution holds the record, lasting 69
years. Greece’s default in 1826 shut it out from international capital markets for
12
The Appendix provides a brief glossary of the major categories of debt studied in this paper.
13
This is not meant to be an exhaustive list of the scholars that have worked on historical sovereign default.
Closely related contributions include Bordo et al (2001), Eichengreen (1992), Eichengreen and Lindert (1989),
Flandreau and Frederic Zumer (2004), and Maurice Obstfeld and Taylor (2003).
F 1. T T C C C I C:
E M, 1865–2009
Notes: An ination crisis is dened as a year when ination exceeds 20 percent, while a currency crash is an annual
depreciation (devaluation) greater than or equal to 15 percent per annum. The correlations of ination and exchange
rate crises are contemporaneous.
Sources: Reinhart and Rogoff (2008, 2009a), sources cited therein, and authors’ calculations.
1800–2009 0.607
1800–1940 0.240
1950–2009 0.754
0
10
20
30
40
50
60
1866 1876 1886 1896 1906 1916 1926 1936 1946 1956 1966 1976 1986 1996 2006
0
10
20
30
40
50
60
Share of countries with an annual inflation rate above 20% (shaded bars)
Share of countries with
a currency crash
(a 15% or greater annual
depreciation/devaluation,
solid line)
Emerging markets:
Correlations of the share
of countries with inflation
and currency crises
1680
THE AMERICAN ECONOMIC REVIEW
AUGUST 2011
53 consecutive years, while Honduras’s 1873 default had a comparable duration.
Looking at the full default episode is, of course, useful for characterizing the bor-
rowing/default cycles, calculating hazard rates, etc. But it is hardly credible that a
spell of 53 years could be considered a crisis. Thus, in addition to constructing the
country-specic dummy variables to cover the entire episode, we also employ one
where only the rst year of default enters as a crisis.
Information on domestic debt crises is scarce, but it is not because these crises do
not take place.
14
Indeed, as Reinhart and Rogoff (2009a) show, domestic debt crises
typically occur against much worse economic conditions than the average external
default. Domestic debt crises do not usually involve external creditors, which may
help explain why so many episodes go unnoticed. Another feature that characterizes
domestic defaults is that references to arrears or suspension of payments on sover-
eign domestic debt are often relegated to the footnotes of data tables. Lastly, some
of the domestic defaults that involved the forcible conversion of foreign currency
deposits into local currency have occurred during banking crises, hyperinations, or
a combination of the two; deposit freezes are also numerous. Our approach toward
constructing categorical variables follows that previously described for external
debt default. Like banking crises and unlike external debt defaults, the endpoint of
domestic default is not always known.
C. Banking Crises
Due to the paucity of quantitative information, our analysis stresses events when
dating banking crises. For example, the relative price of bank stocks (or nancial
institutions relative to the market) would be a logical indicator to examine, but such
time series are not readily available, particularly for the earlier part of our sample as
well as for developing countries (where many banks are not publicly traded).
If the beginning of a banking crisis is marked by bank runs and withdrawals, then
changes in bank deposits could be used to date the crisis. This indicator would cer-
tainly have done well in dating the numerous banking panics of the 1800s. Often,
however, the banking problems do not arise from the liability side, but from a pro-
tracted deterioration in asset quality, be it from a collapse in real estate prices or
increased bankruptcies in the nonnancial sector. In such cases, a large increase
in bankruptcies or nonperforming loans would better mark the onset of the crisis.
Unfortunately, indicators of business failures and nonperforming loans are also usu-
ally available only sporadically; the latter are also made less informative by banks’
desire to hide their problems for as long as possible.
Given these data limitations, we mark a banking crisis by two types of events:
(i) bank runs that lead to the closure, merging, or takeover by the public sector of
one or more nancial institutions; or (ii) if there are no runs, the closure, merging,
takeover, or large-scale government assistance of an important nancial institution
(or group of institutions) that marks the start of a string of similar outcomes for
other nancial institutions.
14
Domestic debt refers to public debts issued under domestic law. Most often, such debts have been denomi-
nated in the domestic currency and largely held by residents.
