Systemic Banking Crises Revisited
Author: Mr. Luc Laeven and Mr. Fabian Valencia

Contributor Notes

Author’s E-Mail Address: Luc.Laeven@ecb.europa.eu and Fvalencia@imf.org.
Publication Date:
14 Sep 2018
eISBN:
9781484376379
Language:
English
Keywords:
WP; banking crisis; crisis; country; financial crisis; dating database; crisis sequencing pattern; banking policy intervention; crisis dating database; systemic banking crisis duration
Download PDF (3.2 MB)

This paper updates the database on systemic banking crises presented in Laeven and Valencia (2008, 2013). Drawing on 151 systemic banking crises episodes around the globe during 1970-2017, the database includes information on crisis dates, policy responses to resolve banking crises, and the fiscal and output costs of crises. We provide new evidence that crises in high-income countries tend to last longer and be associated with higher output losses, lower fiscal costs, and more extensive use of bank guarantees and expansionary macro policies than crises in low- and middle-income countries. We complement the banking crises dates with sovereign debt and currency crises dates to find that sovereign debt and currency crises tend to coincide or follow banking crises.

Abstract

This paper updates the database on systemic banking crises presented in Laeven and Valencia (2008, 2013). Drawing on 151 systemic banking crises episodes around the globe during 1970-2017, the database includes information on crisis dates, policy responses to resolve banking crises, and the fiscal and output costs of crises. We provide new evidence that crises in high-income countries tend to last longer and be associated with higher output losses, lower fiscal costs, and more extensive use of bank guarantees and expansionary macro policies than crises in low- and middle-income countries. We complement the banking crises dates with sovereign debt and currency crises dates to find that sovereign debt and currency crises tend to coincide or follow banking crises.

I. Introduction

Systemic banking crises are highly disruptive events which lead to sustained declines in economic activity, financial intermediation, and ultimately in welfare. It is then no surprise that academics and policymakers devote significant efforts to develop models to attempt to predict crises and to design policies to resolve them and mitigate their economic impacts. But much of these efforts crucially depend on the proper identification of the dates when these crises occur. The use of an inappropriate crisis dating measure may obscure a true relationship between a crisis event and other economic variables or create the appearance of a causal link when there is none.

To facilitate this endeavor, this paper updates the comprehensive global database on systemic banking crises in Laeven and Valencia (2008, 2013), which has become the standard reference for information on banking crises worldwide, to cover all episodes during the period 1970–2017. As in our previous versions of the database, we date systemic banking crises based on the intensity of the policy response to reduce the use of subjective criteria to identify crisis episodes. As in Laeven and Valencia (2013), the database on banking crises episodes during the period 1970–2017 is further complemented with dates of sovereign debt and currency crises during the same period. In total, we identify 151 banking crises, 236 currency crises, and 74 sovereign crises.

The database also includes information about policy responses, fiscal costs, output losses, and other stylized facts about banking crises. When comparing banking crises episodes across countries of different income levels, we find significant differences. In terms of policy responses, we find that the use of financial intervention policies in high-income countries tends to be similar to that in low and middle-income economies, except for guarantees on bank liabilities. The use of the latter has been relatively more common in high-income countries, arguably due to a higher quality of institutions and/or larger fiscal space which rendered the guarantees relatively more credible. Moreover, we document a more extensive use of expansionary monetary and fiscal policies in banking crises episodes in high-income economies than in low- and middle-income ones. Availability of fiscal and monetary space and/or ability to finance larger deficits allowed high-income countries to act countercyclically to mitigate the impact of the crisis on the real economy. In contrast, low-and middle-income countries may have faced binding borrowing constraints that forced them to act procyclically during crisis episodes.

We also find that direct fiscal costs of banking crises—defined as fiscal outlays directly related to government intervention measures in the financial sector—tend to be larger in low-and middle-income countries than in high-income countries. However, using a broader definition of fiscal costs that includes fiscal outlays not directly targeting the financial sector—measured as the increase in public debt-to-GDP ratios around banking crises—we find the exact opposite: increase in public indebtedness tend to be more pronounced for high-income countries. This result follows from a combination of a greater ability of high-income countries to use fiscal stimulus during banking crises, which increases public debt, and larger output losses in high-income countries in the aftermath of banking crises.2

The literature on banking crisis dating has attracted increased attention since the global financial crisis with notable contributions including Reinhart and Rogoff (2009), Schularick and Taylor (2012), Romer and Romer (2017), and Baron and others (2018). Several of these studies document similarities and differences in outcomes with our earlier versions of the database (Laeven and Valencia, 2008, 2013). Relative to these other papers, the main advantage of our database is the dating of banking crises for a comprehensive sample of countries and the documentation of policy responses during such crises.3 This distinction is important, particularly for drawing implications of banking crises beyond advanced economies and large emerging markets.

The remainder of the paper is organized as follows. Section II presents our definition of banking crises. Section III shows the resulting list of crises during the period 1970–2017. Section IV complements our banking crises dates with those for currency and sovereign debt crises. Section V presents the policy responses and Section VI presents the crisis outcomes, including fiscal costs and output losses. Section VII concludes.

II. Definition of a Banking Crisis

We follow in this paper the same definition adopted in Laeven and Valencia (2013), reproduced below for convenience, where we define a banking crisis as an event that meets two conditions:

  • 1) Significant signs of financial distress in the banking system (as indicated by significant bank runs, losses in the banking system, and/or bank liquidations).

  • 2) Significant banking policy intervention measures in response to significant losses in the banking system.

We consider the first year that both criteria are met to be the year when the crisis became systemic. This is to ensure that we date the crisis at the first signs of major problems in the banking system.

When the losses in the banking sector and/or liquidations are severe, we treat the first criterion as a sufficient condition to date a systemic banking crisis. We operationalize this definition by considering that losses are severe when either (i) a country’s banking system exhibits significant losses resulting in a share of nonperforming loans above 20 percent of total loans or bank closures of at least 20 percent of banking system assets or (ii) fiscal restructuring costs of the banking sector are sufficiently high, exceeding 5 percent of GDP.4 However, relying exclusively on the first criterion is problematic because it is not always straightforward to quantify the degree of financial distress in a banking system, particularly in low- and middle-income countries, and also because losses can be mitigated by policy responses. To address this problem, we also rely on the second criterion, if policy intervention meets the requirement of being significant. We consider policy interventions in the banking sector to be significant if at least three out of the following six measures have been used:5

  • 1) deposit freezes and/or bank holidays;

  • 2) significant bank nationalizations;

  • 3) bank restructuring fiscal costs (at least 3 percent of GDP);

  • 4) extensive liquidity support (at least 5 percent of deposits and liabilities to nonresidents);

  • 5) significant guarantees put in place; and

  • 6) significant asset purchases (at least 5 percent of GDP);

The above categories cover all policy interventions that have been employed to resolve a banking crisis (see Honohan and Laeven, 2005, and Laeven and Valencia, 2008). Since not all policies are used in all crises, we require that at least three measures have been put in place. It is worth noting that setting thresholds sufficiently high helps us avoid labeling a non-systemic event or the preemptive use of some of these policies as a systemic banking crisis.6 For interventions that can be quantified more easily, such as liquidity support, asset purchases, and financial restructuring costs, we also adopt quantitative thresholds to define what significant intervention means.

The policy variables we used in our crisis definition are more specifically defined as follows:

  • Deposit freeze and bank holidays: indicates whether the government introduced restrictions on deposit withdrawals or a bank holiday. If implemented, we also collect information on the duration of the deposit freeze and bank holiday, and the affected instruments.

  • Significant nationalizations: takeovers by the government of systemically important financial institutions, including cases where the government takes a majority stake in the capital of such financial institutions.

  • Significant bank guarantees: a significant government guarantee on bank liabilities, indicating that either a full protection of liabilities has been issued by the government or that government guarantees have been extended to non-deposit liabilities of banks.7 Actions that only raise the level of deposit insurance coverage are not included.8

  • Liquidity support: It is measured as central bank claims on other depository institutions (from IFS) and liquidity support directly provided by the Treasury. We normalize this variable by the total deposits and bank liabilities to non-residents. We consider liquidity support to be extensive when this ratio exceeds 5 percent and more than doubles relative to its pre-crisis level.9

  • Bank restructuring costs: defined as gross fiscal outlays directed to the restructuring of the financial sector, with the most important component being recapitalization costs. We consider restructuring costs to be significant if they exceed 3 percent of GDP, excluding liquidity assistance provided directly from the treasury. We focus on gross fiscal costs instead of net because it takes time to record recoveries. However, wherever data on recoveries were available we report also net fiscal costs.

  • Asset purchases: This variable refers to purchases of assets from financial institutions implemented by the central bank, the treasury, or a government entity (such as an asset management company). We define significant asset purchases as those exceeding 5 percent of GDP.

The logic for choosing this approach to date banking crises is to reduce the use of subjective criteria in identifying these events, which gives our database a clear advantage over existing databases such as Caprio and Klingebiel (1996) and Reinhart and Rogoff (2009). Moreover, the chosen thresholds for policy intervention help us focus only on systemic events, where subjectivity in the identification of crises is further reduced. And finally, it is a relatively simple definition that allows a consistent implementation across time periods and countries of different income levels. In Laeven and Valencia (2013) we showed that many episodes in our dataset can be replicated by a simple alternative definition based on credit and real GDP growth, particularly in high-income countries.

More recent studies have explored alternative crisis dating strategies, such as Romer and Romer (2017), who rely on a narrative approach to identify episodes of financial distress in 24 OECD countries; Baron and others (2018), who identify crises in 46 countries by looking at large declines in banks’ stock prices; and Chaudron and de Haan (2014), who study four crises for which the timing strongly differs across databases. Chaudron and de Haan (2014) conclude that using information on the number and size of bank failures allows determining the timing of banking crises more precisely. Their dating for these four episodes corresponds closely with ours. More generally, all these studies note important similarities with our crisis dating to the extent that the samples overlap. However, our approach allows a more comprehensive coverage of countries.

III. Banking Crises Episodes During 1970–2017

Our definition identifies 151 banking crises since 1970, of which 4 episodes started since 2011: Cyprus (2011), Guinea Bissau (2014), Moldova (2014), and Ukraine (2014). The complete dataset is included in the accompanying data file with the main variables reported in the appendix. The banking crises dates—years for all cases, and year and month whenever feasible—include borderline systemic crises, defined as cases where our definition is close to being met. Most countries have experienced at least one systemic banking crisis during 1970–2017, with many going through multiple episodes (Figure 1). However, only three countries experienced more than two systemic banking crises during the past 48 years: Argentina (4), the Democratic Republic of Congo (3), and Ukraine (3).

Figure 1.
Figure 1.

Frequency of Systemic Banking Crises Around the World, 1970–2017

Citation: IMF Working Papers 2018, 206; 10.5089/9781484376379.001.A001

Source: Authors’ calculations.

Following the World Bank’s historical income classification, we group episodes according to the income level of the affected country at the start of the crisis (Figure 2). Figure 2 shows that systemic banking crises are rarely single-country events, with waves of crises clearly visible in the figure, starting with the episodes in Latin America in the early 1980s, the crises in the aftermath of the breakup of the Soviet Union, the Tequila Crisis, the Asian crisis, and more recently the global financial crisis. The period around the mid-2000s was unusual in terms of the low incidence of crises, which was disrupted by the global financial crisis. Since then, some episodes have taken place in low- and middle-income countries, but in general we are facing again a period of relative calm in what pertains to systemic banking crises. The figure also shows that the late eighties and nineties included some episodes in high-income countries, reflecting the savings and loans crisis in the United States, the crises in the Nordic countries in the early 1990s, and the one in Japan in the late 1990s. However, prior to the 2008 global financial crises, banking crises had predominantly been a low and middle-income country phenomenon, at least since 1970. As noted by Reinhart and Rogoff (2009), the global financial crisis made it clear that “financial crises are an equal opportunity menace” for high-and low and middle-income countries.

Figure 2.
Figure 2.

Systemic Banking Crises Episodes by Income Level 1970–2017

Citation: IMF Working Papers 2018, 206; 10.5089/9781484376379.001.A001

Source: Authors’ calculations.

IV. Crises Sequencing

To assess the sequencing of crisis, we complement the database with currency and sovereign crises dates. We follow the same definitions employed in Laeven and Valencia (2008, 2013), which in turn build on Frankel and Rose’s (1996) approach. We define a currency crisis as a “sharp” nominal depreciation of the currency vis-a-vis the U.S. dollar. We consider two thresholds for a depreciation to meet this definition: i) a year-on-year depreciation of at least 30 percent; and ii) of at least 10 percentage points higher than the rate of depreciation observed in the year before.10 Under this definition, there were 236 currency crises during the period 1970–2017.11 We choose bilateral exchange rates because we are interested in the loss of value relative to a reserve currency. Admittedly, the identified episodes can vary with the thresholds, as noted in Laeven and Valencia (2013). However, it is a simple definition that can be implemented easily across countries.

We also date episodes of sovereign debt default and restructuring by relying on information from Beim and Calomiris (2001), World Bank (2002), Sturzenegger and Zettelmeyer (2006), IMF Staff reports, and reports from rating agencies and the media. The compiled data on sovereign debt crises reported in our database include the year of sovereign default to private creditors and/or restructuring. If public debt was restructured without a suspension of payments, the sovereign crisis year is recorded as the year of the restructuring. Using this approach, we identify 75 episodes of sovereign debt crises during 1970–2017, 11 of which took place since 2007. Figure 3 shows the frequency of currency and sovereign debt crises episodes by year and income level.

Figure 3.
Figure 3.

Currency and Sovereign Debt Crises Episodes by Income level

Citation: IMF Working Papers 2018, 206; 10.5089/9781484376379.001.A001

Source: Authors’ calculations.Note: We distinguish high income countries from low and middle-income countries, based on the level of GNI per capita in US$ in the year in which the banking crisis episode started. The classification is assigned by comparing such level of GNI per capita to the income thresholds defined by the World Bank for that same year.

Currency crises are a rare phenomenon among high-income countries, including during the global financial crisis, in part due to the reserve currency status of some of these economies. The global financial crisis brought about sovereign debt crises in high-income countries: Greece with its 2012 restructuring and the 2015 default to the IMF, and Cyprus with the 2013 debt exchange.

Banking and sovereign debt crises can coincide, either because the entire economy is hit by a large shock, or because there are sizeable spillovers from the public to the banking sector (i.e., through banks’ sovereign exposures) or from the banking to the public sector (i.e. through sovereign bailouts of banks) (IMF, 2015; Dell’Ariccia and others, 2018). And analogous connections can be drawn between banking and currency crises: for instance, when a sharp depreciation of the currency wipes out banks’ capital due to large open foreign exchange positions of their own or their borrowers or when significant bank failures lead depositors to seek shelter in foreign assets, simultaneously provoking a run on the currency.

Figure 4 shows the incidence of banking, currency, and sovereign debt crises over the sample period covered in our database. We find that all three types of crises, not just banking crises, come in waves. The number of sovereign debt crises peaked in the mid-1980s, driven predominantly by Latin America, with recent episodes including both high and low and middle-income economies. The frequency of currency crises peaked in the mid-1990s and saw surges around the global financial crisis. Their incidence increased in 2015 due to the large currency depreciations in many commodity-exporter countries triggered by a decline in commodity prices (Kohlscheen and others, 2017). The figure also reports the number of standalone crises as well as those that coincided with other types of crises.12 In total we document 11 triple crises (i.e., simultaneous banking, currency, and sovereign debt crises in a given country) over the period 1970–2017. Among twin crises, the currency/banking and currency/debt crisis pairs tend to be more common than the banking/debt crisis pair.

Figure 4.
Figure 4.

Financial Crises by Type

Citation: IMF Working Papers 2018, 206; 10.5089/9781484376379.001.A001

Source: Authors’ calculations.

To better identify a crisis sequencing pattern, we show in Figure 5 the incidence of currency and sovereign debt crises along a time scale (in years) in countries that experienced a banking crisis in year T. A clearer pattern now emerges. Currency and sovereign debt crises, on average, tend to coincide or follow banking crises, with currency crises peaking at one year after the beginning of the banking crisis. This pattern is in line with findings in earlier studies that have examined the causes as well as the sequencing of crises (e.g., Kaminsky and Reinhart, 1999; Fratzscher and others, 2011; Reinhart and Rogoff, 2011; Gourinchas and Obstfeld, 2012). Although they covered different sample periods and relied on different definitions of crises, the similarity in the conclusion is quite clear: it is common for banking crises to happen at the same time or precede currency and sovereign debt crises. This provides a clear rationale for our emphasis on banking crises.

Figure 5.
Figure 5.

Sequencing of Crises

Citation: IMF Working Papers 2018, 206; 10.5089/9781484376379.001.A001

Source: Authors’ calculations.Note: The figure is constructed by selecting banking crises episodes and plotting the percentage of them that were followed, coincided, or were preceded by a sovereign or currency crisis, with T denoting the start of the banking crisis.

V. Policy Response to Banking Crises

To complement our crisis dating database, we collect information on policy responses deployed during these episodes and directed towards containing and/or resolving a banking crisis. While our focus is predominantly on financial sector policy interventions, we also look at crude measures of monetary and fiscal policy to offer a broad perspective on policy responses during banking crises.

A. Financial Sector Interventions during Crises

Initially, a country’s policy response to bank distress typically includes the deployment of liquidity support to the banking sector, particularly in response to bank runs. The provision of extensive liquidity support during systemic banking crises is pervasive in our sample. We measure liquidity support as the ratio of central bank claims on the financial sector to deposits and foreign liabilities.13 We report two measures of liquidity support: the peak of this ratio, labeled as peak liquidity support in Table 2 in the appendix, and the change between the peak and the average of the ratio during the year before the start of the crisis, labeled as liquidity support. The median peak liquidity support ratio reaches 20.2 percent, with 100 out of 151 episodes recording a positive value of up to 28 percent. However, the median peak liquidity at 12 percent for high-income countries is less than half of the 23 percent recorded for low-and middle-income countries. The second measure, liquidity support, shows up with a median of 10.8 percent, with 95 out of 151 episodes exhibiting a positive value of up to 16 percent. Again, the median value for high-income countries, at 6.4 percent, is less than half of the 14.8 percent corresponding to low and middle-income countries.

While both high and low and middle-income countries have relied extensively on liquidity support when hit by a crisis, Laeven and Valencia (2010), Claessens and others (2011), and Stone and others (2011), have noted the wider array of instruments used by high-income countries when experiencing a crisis, including through the coordinated deployment of central bank swap facilities during the global financial crisis. These studies have also pointed out that low and middle-income countries tend to rely on liquidity provision as a containment tool for much longer than high-income countries, on average, before introducing bank recapitalization and restructuring measures. Weaker institutions, including non-independent central banks and regulators, in some low and middle-income countries particularly in the 1980s and 1990s, may have led to the prolonged reliance on liquidity support and a delayed recognition of bank solvency problems. Bank recapitalization measures, such as the Trouble Asset Relief Program (TARP) in the United States, were deployed much quicker during the global financial crises, compared to episodes prior to the global financial crisis (predominantly in low and middle-income countries). The more prolonged reliance on liquidity support in low and middle-income economies may explain why it tended to be higher in these episodes than in high-income countries.

During the early stages of banking crises, and often in combination with liquidity support, governments have also resorted to limited or full guarantees on some or most bank liabilities, to help stem bank runs and alleviate liquidity pressures on these entities. They typically buy policymakers time to develop more comprehensive resolution and restructuring plans. Laeven and Valencia (2012) examine the experience of 42 crisis episodes, of which 14 made use of explicit guarantees on bank liabilities and find that these guarantees do help to reduce liquidity pressures on banks. Altogether, we report in our database 34 crisis episodes where blanket guarantees were announced, of which 19 cases corresponded to high-income countries, mostly during the 2008 global financial crisis. Guarantees are often left in place for many years and are only gradually removed. The blanket guarantees announced in Mexico in 1993 and in Malaysia in 1998 were fully removed only in 2003 and 2005, respectively. At end-2016, European Union governments collectively still had 120 billion euros in outstanding guarantees issued in support of the financial system, according to the European Commission’s 2017 State Aid Scoreboard. While this amount represents a sharp decline from its peak of 835 billion euros in 2009, it remains non-trivial.

In cases where liquidity pressures have been significant, countries have in some cases resorted to administrative measures, suspending the convertibility of deposits into cash and restricting foreign payments. These “deposit freezes” have often been preceded by bank holidays—the temporary closure of banks—often by design as banks need some time to adapt their IT systems and procedures to the new regime. However, bank holidays and deposit freezes have been rarely used. We report in our database only 8 episodes were deposit freezes were imposed. The most recent cases include Cyprus in 2013, Ukraine in 2014, and Greece in 2015. In Cyprus, restrictions to domestic payments were removed in May 2014, while those on external payments remained in place until April 2015. For Ukraine, cash withdrawals from domestic currency bank accounts were lifted in September 2016 and those from FX accounts in August 2017, although some restrictions on FX transactions remained in place as of early 2018. Similarly, in Greece, the restrictions on deposits have been gradually relaxed since their introduction in July 2015, but there were restrictions still in place as of early 2018, including a monthly limit on cash withdrawals and limits on cross-border bank transfers.

We report 6 bank holidays, with Cyprus and Greece being the only recent cases. In 5 of the 6 cases, the bank holiday was in place for a length between 4 and 8 days. The exception is Greece where the bank holiday was in place for 21 days. In all the 6 reported instances, the bank holiday was followed by a deposit freeze.

The above policies are intended to contain liquidity pressures. However, banks experiencing significant drains in liquidity often see a deterioration in their capital position as they are forced into asset disposals at fire sale prices to meet liquidity needs. Compounded by a deterioration in asset quality as financially weakened borrowers fall delinquent on their loans, additional measures are often needed to restore solvency of affected banks. These may include private or public recapitalization of viable institutions, resolution of insolvent ones, and even outright nationalization. The appropriateness and effectiveness of these tools in situations of severe financial distress have been widely studied in the literature. There is theoretical research showing that in those circumstances recapitalizing banks with public money can increase welfare (e.g., Philippon and Schnabl, 2013; and Sandri and Valencia, 2013) and there is empirical evidence suggesting that recapitalizing banks with public money can alleviate the real effects of banking crises (e.g., Homar and others, 2017; Giannetti and Simonov, 2013; and Laeven and Valencia, 2013). Implementation, however, may take many forms (Laeven and Valencia, 2008; Claessens and others, 2014).

Bank recapitalization is a tool that has been used in most crises we report in our database, and it is also the most important component of direct fiscal costs from government intervention in the financial sector. Government capital injections, encompassing often a combination of preferred and common equity, have also been accompanied by conditions or restrictions, for instance requiring board seats for government representatives, and limiting or prohibiting dividend payments (Laeven and Valencia, 2008). These recapitalizations can often lead governments to own a majority share of a bank’s capital, in which cases we classify the intervention as a nationalization, together with outright nationalization cases. Finally, we also report if the treasury or the central bank engaged in asset purchases to support the banking system and whether an asset management company was established to administer or resolve these assets.

The differences in financial policy mix to resolve banking crises between high income and low-and middle-income economies is shown in Figure 6. The figure makes it clear that countries of both income groups resort broadly to the same types of policies to resolve systemic banking crises, except for guarantees. Significant guarantees on bank liabilities are more common among high-income countries, arguably because of generally better institutions or fiscal space that make the guarantees more credible. However, as noted in Claessens and others (2011), guarantees during the global financial crises were on average less comprehensive (i.e., more targeted) than in countries of lower income levels. In those countries, governments tended to announce blanket guarantees of banks’ liabilities. In many cases, limited protection of deposits was introduced after a banking crisis (Laeven and Valencia, 2013). The absence of these schemes in many episodes in low and middle-income countries may have prompted policymakers to announce comprehensive guarantees of bank liabilities.

Figure 6.
Figure 6.

Containment and Resolution Policies

Citation: IMF Working Papers 2018, 206; 10.5089/9781484376379.001.A001

Source: Authors’ calculations.

B. Macro Policies

In addition to using financial sector intervention measures to resolve banking crises, policymakers often use monetary and fiscal policy to mitigate their economic consequences. But as we report in this section, there is a difference between the use of these tools among high-income and low- and middle-income countries.

We trace the median evolution of short-term interest rates around systemic banking crises to gauge whether countries tended to ease or tighten monetary policy. Figure 7 shows that in high-income countries, short-term interest rates declined to a median level very close to zero in the year after the start of the crisis, from a median of about 5 percent. In contrast, the median short-term interest rate increases in low and middle-income countries, reflecting the often-limited space to conduct countercyclical monetary policy at times of heightened financial distress in these countries.14 Concerns about sharp currency depreciations and the resulting impact on private balance sheets exposed to exchange rate risk often force these countries to raise interest rates, ultimately leading also to sharper deterioration in banks’ asset quality.

Figure 7.
Figure 7.

Short-term Interest Rates and Fiscal Balances around Banking Crises

Citation: IMF Working Papers 2018, 206; 10.5089/9781484376379.001.A001

Source: World Economic Outlook, IMF, IFS, and authors’ calculations.

A similar outcome emerges when comparing the evolution of primary fiscal balances. While the median primary balance tends to deteriorate sharply in high-income countries, it improves in low and middle-income countries. The latter group is forced to adopt a procyclical fiscal policy as these countries tend to face limited financing options in those circumstances.

VI. Crisis Outcomes

We collect and report data on the following outcomes for banking crises: i) the direct fiscal costs, measured as fiscal outlays linked to government intervention policies in the banking system; ii) a broader measure of fiscal costs, determined by the increase in public debt; iii) peak nonperforming loans (NPLs); iv) crisis duration, measured in number of years between the start and end of the crisis; and v) output losses.

A. Fiscal Costs of Banking Crises

We measure fiscal costs of banking crises as the sum of all fiscal outlays directly linked to government interventions to stabilize the banking system since the start of the crisis. These interventions include capital injections in financial institutions, operating costs of agencies or entities such as asset management companies, exercised public guarantees, and any other fiscal cost directly attributable to the rescue of financial institutions.

In reporting the fiscal costs of a banking crisis episode, we normalize the outlays by the nominal gross domestic product of the year in which they are incurred and sum them up. We also report these fiscal costs in percent of financial system assets, where the latter are measured as of the year before the start of the banking crisis. In reporting fiscal costs, we do not include government guarantees of bank liabilities or assets because they do not represent an outlay,15 although they are critical if one wanted to measure the total ex-ante risk taken by the public sector during the early stages of a banking crisis. Our ex-post analysis focuses on the actual fiscal costs of a banking crisis episode.16 Data on fiscal costs are collected from official country publications, supranational agencies, and IMF staff reports.17

We collect recoveries of government outlays for a subset of episodes using the same data sources from which we collect fiscal costs. Data on recoveries allow us to report the net fiscal cost (i.e., outlays minus recoveries) of a banking crisis episode. We define recoveries as proceeds from sales of financial assets—acquired to resolve a banking crisis—revenues from fees on guarantees, dividends, interest, and any other cash inflow for the government that can be directly attributable to unwinding financial sector intervention measures. Our definition of recoveries means that we exclude unrealized capital gains on assets that are still on the government balance sheet, which implies that over a longer horizon, recoveries can exceed what we report in our database.18

The histograms in Figure 8 show substantial variation in the fiscal costs of systemic banking crises episodes, both in high-income and low and middle-income economies. Still, the median cost for crises in high-income countries is 6.7 percent of GDP and 10 percent of GDP for low and middle-income countries. The difference in fiscal costs between the two groups of countries increases to slightly above 6 percentage points of GDP after subtracting recoveries: The median net fiscal cost reaches 3.3 percent of GDP for high-income countries and 9.6 percent of GDP for low- and middle-income countries.

Figure 8.
Figure 8.

Gross and Net Fiscal Costs of Banking Crises

Citation: IMF Working Papers 2018, 206; 10.5089/9781484376379.001.A001

Source: Authors’ calculations.Note: Fiscal costs corresponds to government outlays directly linked to policies to stabilize the financial system. Net fiscal costs refer to fiscal costs minus recoveries whenever there were available data on them. Samples differ as recoveries are collected for a subset of episodes.

The difference in fiscal costs between the two groups of countries becomes even more pronounced when fiscal costs are measured relative to the size of the financial system, as shown in Figure 9.19 Relative to the size of financial systems, banking crises appear to have been much costlier, in terms of direct fiscal costs, in low- to middle-income economies. But these differences may also be the outcome of the greater reliance on macroeconomic policy tools, as noted in the previous section, which reduces the burden on financial sector policies to resolve the crisis.20

Figure 9.
Figure 9.

Fiscal Costs in Percent of Financial System Assets

Citation: IMF Working Papers 2018, 206; 10.5089/9781484376379.001.A001

Source: Authors’ calculations.

The use of fiscal space leads also to larger increases in public debt—our broader measure of fiscal costs of banking crises—in high-income countries compared to low and middle-income countries. Discretionary fiscal policy and automatic stabilizers affect directly this broader measure of fiscal costs of crises. These factors play a much smaller role in driving up public debt after a banking crisis in low- and middle-income countries. The median increase in public debt, measured over T-1, T+3, where T is the starting year of the banking crisis, reaches 21.1 percent of GDP in high-income countries compared to 16.4 percent of GDP in low- and middle-income countries (Figure 10).21

Figure 10.
Figure 10.

Increases in Public Debt around Banking Crises

Citation: IMF Working Papers 2018, 206; 10.5089/9781484376379.001.A001

Source: Authors’ calculations.

B. Peak Nonperforming Loans

The sharper deterioration in asset quality of banks in low- and middle-income countries can be noted by looking at the peak nonperforming loans (NPLs) across crisis episodes. Figure 11 shows the distribution of peak NPLs in the two groups of countries. In both income groups there is quite a bit of dispersion in the distribution, although in about 70 percent of crises in high-income countries, NPLs never surpassed 20 percent of total loans. The median peak NPL among crises in countries within this income bracket slightly exceeds 11 percent. In contrast, the median peak NPL reaches 30 percent among crises episodes in low and middle-income economies. While cross-country differences in the definition of NPLs makes it difficult to directly compare levels of NPLs across countries, the systematic and sizable difference between the two groups is unlikely to be entirely driven by differences in definitions.

Figure 11.
Figure 11.

Peak NPLs in Banking Crises Episodes

Citation: IMF Working Papers 2018, 206; 10.5089/9781484376379.001.A001

Source: Authors’ calculations.

C. Systemic Banking Crisis Duration

Following the same definition as in Laeven and Valencia (2013), we also report end dates for each crisis episode, defined as the year before both real GDP growth and real credit growth are positive for at least two consecutive years.22 The rationale for identifying the end of a banking crisis through this approach hinges on the notion that a deterioration in bank solvency can disrupt the supply of credit (e.g. Bernanke and Gertler, 1987; Van Den Heuvel, 2006; Valencia, 2014; Abbasi and others, 2016) and these disruptions to the supply of credit can have real effects (e.g. Peek and Rosengren, 1997; Ashcraft, 2005; Kroszner and others, 2007; Dell’Ariccia and others, 2008; and Alfaro and others, 2017). Therefore, we look for evidence of a reversal in the negative effects of a banking crisis.

In all cases, we truncate the duration of a crisis at five years, starting from the first year of the crisis. The rationale for this truncation is twofold: first, our metric is based on credit stocks not flows (new lending), and stocks are affected by write-offs and restructurings. Therefore, a potential measurement error in the recovery of new lending could bias upwards the duration of the crisis episode. Second, as the length of time increases, our simple metric may start picking up the impact of other shocks. Therefore, whenever we report a crisis lasting five years, it should be read as five years or more. Figure 12 shows the distribution of the estimated duration of banking crisis episodes. The chart on the left shows that, according to our definition of end dates, about two-thirds of crises ended in less than five years. But these aggregate statistics mask some important differences among countries of different income levels. More than half of the episodes we record in high-income countries experienced crises that were quite persistent, lasting five years or more. In contrast, most crises in low and middle-income countries lasted four years or less.

Figure 12.
Figure 12.

Banking Crises Duration

Citation: IMF Working Papers 2018, 206; 10.5089/9781484376379.001.A001

Source: Authors’ calculations.

Crisis severity may be an important factor explaining these differences in duration as many crises in high-income countries corresponded to the global financial crisis. At the same time, larger financial systems and institutions in these countries adds a layer of complexity to the resolution of the crisis, which could help explain the longer duration of crises. Finally, the ability of high-income countries to rely also on monetary and fiscal policy to mitigate the real effects of banking crises may also discourage more active bank restructuring which could ultimately prolong the duration of a crisis (Claessens et al., 2011).

D. Output Losses

We report output losses associated with banking crises episodes, computed as deviations of actual GDP from its trend. 23 The output losses are reported in cumulative terms over [T, T+3], with T denoting the starting year of the crisis, and expressed in percent of one year’s trend GDP. It is important to note that these losses should not be interpreted as solely stemming from banking crises, as they may include the impact of other shocks happening around crises. They should instead be read as what happens to output in the aftermath of a banking crisis. While admittedly the level of output losses is sensitive to how the trend is calculated, Laeven and Valencia (2013) showed that the ranking of crises is robust to using alternative sample periods when computing the trend. Therefore, the metric is primarily adequate to capture the relative size and heterogeneity of output losses across crises.

Figure 13 shows that the output losses in high-income countries tend to be much larger than those in low and middle-income countries. As with the earlier result on crisis duration, the larger output losses in high-income countries could be explained by the presence of larger and deeper financial systems, whose disruption has stronger effects on the real economy.

Figure 13.
Figure 13.

Output Losses around Banking Crises

Citation: IMF Working Papers 2018, 206; 10.5089/9781484376379.001.A001

Source: Authors’ calculations.

The evolution of output in the aftermath of banking crises suggest that these episodes tend to be followed by a very persistent decline in the level of real output, as highlighted in Figure 14. This stylized fact is consistent with new and old empirical work assessing the real consequences of banking crises which has highlighted the persistent real effects of these episodes (see for instance Cerra and Saxena, 2008, 2017; Abiad and others, 2014; Jorda and others, 2015; and Romer and Romer, 2017, 2018).24 However, this persistence in the decline of output in the aftermath of banking crises appears to be much more pronounced, on average, in high-income countries than in low- and middle-income countries, as suggested by Figure 14.25 Aslam and others (forthcoming) look at the recovery in the aftermath of banking crises and find that output remains below trend for longer in advanced economies than in emerging economies, consistent with the simple stylized fact presented here.

Figure 14.
Figure 14.

Output and Export Volume Paths around Banking Crises

Citation: IMF Working Papers 2018, 206; 10.5089/9781484376379.001.A001

Source: Authors’ calculations.

In addition to differences in the size of financial systems between high- and low and middle-income countries, one additional element that could explain the difference in output paths in the aftermath of crises is the evolution of export volumes. Consistent with the slowdown in trade volumes documented in IMF (2016), Figure 14 shows a sluggish evolution in export volumes in the aftermath of banking crises in high-income countries, comprising mostly episodes during the global financial crisis. In contrast, the median path among crisis episodes in low-and middle-income countries does not show a slowdown. Countries in this income group, comprising episodes mostly prior to the global financial crisis, often benefited from a boost from external demand that resulted in a faster recovery in the aftermath of the banking crisis.

VII. Conclusions

A decade since the start of the global financial crisis has allowed sufficient time for some crisis episodes to end. However, many countries have been left with important legacy issues in terms of permanent output losses, elevated levels of public debt, policy support still to be fully unwound, and significant government ownership of financial assets. While these crisis episodes have enriched our experience, much remains to be learned regarding how to predict banking crises, how to prevent them, and how best to resolve them. To make progress in such an ambitious endeavor, a key prerequisite is the availability of high-quality data on banking crises. To help in this direction, this paper provides a comprehensive database on systemic banking crises during the period 1970–2017, reflecting updates to outcomes from banking crises reported in our earlier releases (Laeven and Valencia, 2008, 2010, and 2013) and new events that occurred since then.

It is our hope that these data will assist academics and policymakers in improving our understanding of the causes and consequences of banking crises, and how best to resolve them. While only a few countries have experienced a crisis in recent years, this period may just be the lull before the storm.

References

  • Abbassi, Puriya, Raj Iyer, and Francesc R. Tous, 2016, “Securities Trading by Banks and Credit Supply: Micro-Evidence,” Journal of Financial Economics, Vol. 121(3), 569– 594.

    • Search Google Scholar
    • Export Citation

  • Abiad, Abdul, Ravi Balakrishnan, Petya Koeva Brooks, Daniel Leigh, and Irina Tytell, 2014, “What’s the Damage? Medium-Term Output Dynamics after Financial Crises, In: Claessens, Stijn, M. Ayhan Kose, Luc Laeven, and Fabian Valencia (eds.). Financial Crises: Causes, Consequences, and Policy Responses, International Monetary Fund, Washington D.C.

    • Search Google Scholar
    • Export Citation

  • Alfaro, Laura, Manuel Garcia-Santana, and Enrique Moral-Benito, 2017, “Credit Supply Shocks, Network Effects, and the Real Economy,” Harvard Business School manuscript.

    • Search Google Scholar
    • Export Citation

  • Aslam, Aqib, Patrick Blagrave, Eugenio Cerutti, Sung Eun Jung, and Carolina Osorio-Buitron, forthcoming, “Recessions and Recoveries: Are EMs different from AEs?” IMF Working Paper.

    • Search Google Scholar
    • Export Citation

  • Ashcraft, Adam, 2005, “Are Banks Really Special? New Evidence from the FDIC-Induced Failure of Healthy Banks,” American Economic Review, Vol. 95, pp. 1712–1730.

    • Search Google Scholar
    • Export Citation

  • Baron, Matthew, Emil Verner, and Wei Xiong, 2018, “Identifying Banking Crises,” Princeton University manuscript.

  • Beim, David and Charles Calomiris, 2001, “Emerging Financial Markets.” Appendix to Chapter 1. New York: McGraw-Hill/Irwin Publishers.

    • Search Google Scholar
    • Export Citation

  • Bernanke, Ben and Mark Gertler, 1987, “Banking and Macroeconomic Equilibrium,” In: New Approaches to Monetary Economics, ed. W. Barnett and K. Singleton, New York: Cambridge University Press.

    • Search Google Scholar
    • Export Citation

  • Caprio, Gerard, and Daniela Klingebiel, 1996, “Bank Insolvencies: Cross-Country Experience,” Policy Research Working Paper No. 1620 (Washington: World Bank).

    • Search Google Scholar
    • Export Citation

  • Cerra, Valerie and Sweta Saxena, 2008, “Growth Dynamics: The Myth of Economic Recovery,” American Economic Review, Vol. 98(1), pp. 439–57.

    • Search Google Scholar
    • Export Citation

  • Cerra, Valerie and Sweta Saxena, 2017, “Booms, Crises, and Recoveries: A New Paradigm of the Business Cycle and its Policy Implications,” IMF Working Paper 2017/250.

    • Search Google Scholar
    • Export Citation

  • Chaudron, Raymond and Jakob de Haan, 2014, “Dating banking crises using incidence and size of bank failures: Four crises reconsidered,” Journal of Financial Stability, Vol. 15, pp. 63–75.

    • Search Google Scholar
    • Export Citation

  • Claessens, Stijn, Ceyla Pazarbasioglu, Luc Laeven, Marc Dobler, Fabian Valencia, Oana Nedelescu, Katharine Seal, 2011, “Crisis Management and Resolution: Early Lessons from the Financial Crisis,” IMF Staff Discussion Note No. 11/05.

    • Search Google Scholar
    • Export Citation

  • Claessens, Stijn, M. Ayhan Kose, Luc Laeven, and Fabian Valencia (eds.), 2014, Financial Crises: Causes, Consequences, and Policy Responses, International Monetary Fund, Washington D.C.

    • Search Google Scholar
    • Export Citation

  • Dell’Ariccia, Giovanni, Enrica Detragiache, and Raghuram Rajan, 2008, “The Real Effects of Banking Crises,” Journal of Financial Intermediation, Vol. 17, pp. 89–112.

    • Search Google Scholar
    • Export Citation

  • Dell’Ariccia, Giovanni, Maria Soledad Martinez Peria, Deniz O Igan, Elsie Addo Awadzi, Marc Dobler, and Damiano Sandri, 2018, “Trade-offs in Bank Resolution, IMF Staff Discussion note 18/02.

    • Search Google Scholar
    • Export Citation

  • Demirgüç-Kunt, Asli and Enrica Detragiache, 1998, “The Determinants of Banking Crises: Evidence from Developed and Developing Countries”, IMF Staff Papers, Vol. 45 (1).

    • Search Google Scholar
    • Export Citation

  • Fratzscher, Marcel, Arnaud Mehl, and Isabel Vansteenkiste, 2011, “130 Years of Fiscal Vulnerabilities and Currency Crashes in Advanced Economies,” IMF Economic Review, Vol. 59, pp. 683–716.

    • Search Google Scholar
    • Export Citation

  • Frankel, Jeffrey and Andrew Rose, 1996, “Currency Crashes in Emerging Markets: An Empirical Treatment,” Journal of International Economics, Vol. 41, pp. 351–366.

    • Search Google Scholar
    • Export Citation

  • Giannetti, Mariassunta and Andrei Simonov, 2013, “On the Real Effects of Bank Bailouts: Micro Evidence from Japan,” American Economic Journal: Macroeconomics, Vol. 5, pp.135–167.

    • Search Google Scholar
    • Export Citation

  • Gourinchas, Pierre-Olivier and Maurice Obstfeld, 2012, “Stories of the Twentieth Century for the Twenty-First,” American Economic Journal: Macroeconomics, Vol. 4(1), pp 226–65.

    • Search Google Scholar
    • Export Citation

  • Homar, Timotej, and Sweder J.G. van Wijnbergen, 2017, “Bank Recapitalization and Economic Recovery after Financial Crises,” Journal of Financial Intermediation, Vol. 32, pp. 16–28.

    • Search Google Scholar
    • Export Citation

  • Honohan, Patrick, and Luc Laeven, eds., 2005, Systemic Financial Crises: Containment and Resolution (Cambridge, UK: Cambridge University Press.

    • Search Google Scholar
    • Export Citation

  • IMF, 2014, “From Banking to Sovereign Stress: Implications for Public Debt,” IMF Board paper.

  • IMF, 2016, “Global Trade: What’s behind the Slowdown?” IMF World Economic Outlook, Chapter 2, October.

  • IMF, 2017, “Spain—Publication of Financial Sector Assessment Program Documentation—Technical Note on Impaired Assets and Nonperforming Loans,” IMF country report No. 17/343.

    • Search Google Scholar
    • Export Citation

  • Jorda, Oscar, Moritz Schularick, and Alan M. Taylor, 2015, “Leveraged Bubbles,” Journal of Monetary Economics, Vol. 76, pp. S1–S20.

  • Kaminsky, Graciela and Carmen Reinhart, 1999, “The Twin Crises: The Causes of Banking and Balance-of-Payments Problems,” American Economic Review, Vol. 89, pp. 473–500.

    • Search Google Scholar
    • Export Citation

  • Kohlscheen, Emanuel, Fernando Avalos, and Andreas Schrimpf, 2017, “When the Walk Is Not Random: Commodity Prices and Exchange Rates,” International Journal of Central Banking, Vol. 13(2), pp. 121–158.

    • Search Google Scholar
    • Export Citation

  • Kroszner, Randall, Luc Laeven, and Daniela Klingebiel, 2007, “Banking Crises, Financial Dependence, and Growth,” Journal of Financial Economics, Vol. 84, pp. 187–228.

    • Search Google Scholar
    • Export Citation

  • Laeven, Luc and Fabian Valencia, 2008, “Systemic Banking Crises: A New Database,” IMF Working Paper No. 08/224, (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

  • Laeven, Luc and Fabian Valencia, 2010, “Resolution of Banking Crises: The Good, the Bad, and the Ugly,” IMF Working Paper No. 10/44, (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

  • Laeven, Luc and Fabian Valencia, 2012, “The Use of Blanket Guarantees in Banking Crises”, Journal of International Money and Finance, Vol. 1(5), pp. 1220–248.

    • Search Google Scholar
    • Export Citation

  • Laeven, Luc, and Fabian Valencia, 2013a, “Systemic Banking Crises Database,” IMF Economic Review, Vol. 61 (2), pp. 225–270.

  • Laeven, Luc and Fabian Valencia, 2013b, “The Real Effects of Financial Sector Interventions during Crises,” Journal of Money, Credit, and Banking, Vol. 45(1), pp. 147–177.

    • Search Google Scholar
    • Export Citation

  • Mishkin, Frederic, 1996, “Understanding Financial Crises: A Developing Country Perspective”, NBER Working Paper No. 5600.

  • Peek, Joe, and Erik Rosengren, 1997, “The International Transmission of Financial Shocks: The Case of Japan,” American Economic Review, Vol. 87(4), pp. 495–505.

    • Search Google Scholar
    • Export Citation

  • Philippon, Thomas and Philipp Schnabl, 2013, “Efficient Recapitalization,” Journal of Finance.

  • Reinhart, Carmen and Kenneth Rogoff, 2009, This Time is Different: Eight Centuries of Financial Folly, Princeton University Press.

  • Reinhart, Carmen and Kenneth Rogoff, 2011, “From Financial Crash to Debt Crisis,” American Economic Review, Vol. 101, pp. 1676–1706.

    • Search Google Scholar
    • Export Citation

  • Romer, David and Christina Romer, 2017, “New Evidence on the Aftermath of Financial Crises in Advanced Countries,” American Economic Review, Vol. 107(10), pp. 3072– 3118.

    • Search Google Scholar
    • Export Citation

  • Romer, David and Christina Romer, 2018, “Phillips Lecture – Why Some Times are Different: Macroeconomic Policy and the Aftermath of Financial Crises,” Economica, Vol. 85, pp. 1–40.

    • Search Google Scholar
    • Export Citation

  • Sandri, Damiano and Fabian Valencia, 2013, “Financial Crises and Recapitalizations,” Journal of Money, Credit, and Banking, Vol 45(S2), pp. 59–86.

    • Search Google Scholar
    • Export Citation

  • Schularick, Moritz, and Alan M. Taylor, 2012, “Credit Booms Gone Bust: Monetary Policy, Leverage Cycles, and Financial Crises, 1870–2008,” American Economic Review, Vol. 102(2), pp. 1029–1061.

    • Search Google Scholar
    • Export Citation

  • Stone, M., K. Fujita, and K. Ishi, 2011, “Should Unconventional Balance Sheet Policies Be Added to the Central Bank Toolkit?” IMF Working Paper 2011/145.

    • Search Google Scholar
    • Export Citation

  • Sturzenegger, Federico and Jeromin Zettelmeyer, 2006, Debt Defaults and Lessons from a Decade of Crises. Cambridge: MIT Press.

  • Valencia, Fabian, 2014, “Banks’ Precautionary Capital and Credit Crunches,” Macroeconomic Dynamics, Vol. 18(8), pp. 1726–1750.

  • Van Den Heuvel, Skander, 2006, “The Bank Capital Channel of Monetary Policy,” University of Pennsylvania unpublished manuscript.

  • World Bank, 2002, Global Development Finance. Appendix on Commercial Debt Restructuring. Washington, D.C.: World Bank.

Appendix

Table 1.

Crisis Dates

article image
article image
article image
article image
Table 2.

Banking Crises Resolution and Outcomes

article image
article image
article image
article image
article image

In percent of GDP. Output losses are computed as the cumulative sum of the differences between actual and trend real GDP over the period [T, T+3], expressed in percent of trend real GDP, with T denoting the starting year of the crisis. The trend is computed by applying an HP filter (λ=100) to the GDP series over [T-20, T-1]. No output losses are reported for crises in transition economies that took place during the period of transition to market economies.

Fiscal costs refer to outlays directly related to the restructuring of the financial sector.

Liquidity is measured as the ratio of central bank claims on deposit money banks (line 12 in IFS) and liquidity support from the Treasury to total deposits and liabilities to non-residents. Total deposits are computed as the sum of demand deposits (line 24), other deposits (line 25), and liabilities to non-residents (line 26).

In percent of total loans.

In percent of GDP. For episodes starting in 2007 and later, the increase in public debt is measured as the change in debt projections, over [T-1, T+3], relative to the pre-crisis debt projections, where T is the starting year of the crisis.

Credit data missing. For these countries, end dates are based on GDP growth only.

We truncate the duration of crises at 5 years, starting with the first crisis year.

Borderline cases.

Source: WEO, IFS, IMF Staff reports, IMF Financial Soundness Indicators, Laeven and Valencia (2013), and authors’ calculation.