Fiscal Crises
  • 1 0000000404811396https://isni.org/isni/0000000404811396International Monetary Fund

Contributor Notes

A key objective of fiscal policy is to maintain the sustainability of public finances and avoid crises. Remarkably, there is very limited analysis on fiscal crises. This paper presents a new database of fiscal crises covering different country groups, including low-income developing countries (LIDCs) that have been mostly ignored in the past. Countries faced on average two crises since 1970, with the highest frequency in LIDCs and lowest in advanced economies. The data sheds some light on policies and economic dynamics around crises. LIDCs, which are usually seen as more vulnerable to shocks, appear to suffer the least in crisis periods. Surprisingly, advanced economies face greater turbulence (growth declines sharply in the first two years of the crisis), with half of them experiencing economic contractions. Fiscal policy is usually procyclical as countries curtail expenditure growth when economic activity weakens. We also find that the decline in economic growth is magnified if accompanied by a financial crisis.

Abstract

A key objective of fiscal policy is to maintain the sustainability of public finances and avoid crises. Remarkably, there is very limited analysis on fiscal crises. This paper presents a new database of fiscal crises covering different country groups, including low-income developing countries (LIDCs) that have been mostly ignored in the past. Countries faced on average two crises since 1970, with the highest frequency in LIDCs and lowest in advanced economies. The data sheds some light on policies and economic dynamics around crises. LIDCs, which are usually seen as more vulnerable to shocks, appear to suffer the least in crisis periods. Surprisingly, advanced economies face greater turbulence (growth declines sharply in the first two years of the crisis), with half of them experiencing economic contractions. Fiscal policy is usually procyclical as countries curtail expenditure growth when economic activity weakens. We also find that the decline in economic growth is magnified if accompanied by a financial crisis.

I. Introduction

We fear fiscal crises, but do not know much about them. The term, coined by James O’Connor (1973), came to prominence in the wake of the oil crisis and the breakdown of the Bretton Woods system to denote a “structural gap” between public revenues and expenditures when growth plummeted, and unemployment and inflation surged. Yet, fiscal crises may be triggered by other imbalances in the economy or exogenous shocks.2 In addition, in the aftermath of the global economic and financial crisis of the late 2000s, there is greater interest in how to avoid fiscal crises, including via well-designed fiscal adjustments (Mauro, 2011). Regardless, there is little empirical research about fiscal crises. What is the frequency and duration of crisis episodes? What happens around them? How disruptive are they? Do they have a persistent impact on economic growth? These are just some of the questions that need answers to better understand and prevent these periods of heightened fiscal distress, which can be accompanied by large and abrupt declines in growth (Figure 1.1).

Figure 1.1.
Figure 1.1.

Ten of the Worst Crises Since 2005

(decline in the real GDP p.c. growth rate, percentage points)

Citation: IMF Working Papers 2017, 086; 10.5089/9781475592153.001.A001

Source: Authors’ own calculations, based on IMF WEO data and the authors’ newly created fiscal crises database.Note: Reported is the difference in simple averages of real GDP p.c. growth rate of the three years preceding a crisis and the three years after the onset of the crisis, including the onset year.

The literature, with rare exceptions, focuses on sovereign debt defaults. The research has been directed predominantly to identifying public debt crises triggered by external default episodes (e.g., Detragiache and Spilimbergo, 2001; Chakrabarti and Zeaiter, 2014). This has been later complemented by large-scale official financing (e.g., Manasse, Roubini, and Schimmelpfennig, 2003) and, to a very limited degree given data constraints, evidence on domestic public default (e.g., Reinhart and Rogoff, 2009 and 2011; Gourinchas and Obstfeld, 2012). While defaults are undoubtedly an important part of the story, they do not capture all periods of fiscal crisis.

In this paper, we take a broader approach to identify and build a new database of fiscal crises. We look at periods of extreme fiscal distress, when governments have not been able to contain large fiscal imbalances leading to the adoption of extreme measures (e.g., debt default and monetization of the deficit). We develop a new database building on the work by Baldacci et al. (2011) on identifying fiscal crises.3 However, our approach relies on enhanced criteria to better identify fiscal crises (e.g., implicit defaults or loss of market access).

Our database includes developing countries, which remain underrepresented in the literature and where fiscal crises have unique characteristics. We expand the country coverage to 188 countries, over 1970-2015, more than double the size of the sample relative to many other studies. We identify 436 fiscal crisis episodes with countries facing on average two crises between 1970 and 2015. While we would expect low-income developing countries (LIDCs) to suffer the most from a crisis—as they tend to have less capacity to manage shocks—the data does not confirm this.4 LIDCs do have the highest frequency of crises, while advanced markets (AMs) have the lowest. Surprisingly, however, the decline in growth around crisis periods is smaller in LIDCs and larger in AMs. Future research will need to identify the reasons, but this could be related to the fact that LIDCs receive international support in times of crisis.5

We use the database to better understand the patterns of fiscal crises and their economic impact. While not necessarily implying causality, our findings shed some light on the crisis periods. In terms of policies, we find that crises are accompanied by disruptive fiscal adjustments—countries have to curtail spending growth at a time when economic conditions are worsening. Importantly, taking exceptional measures (e.g., debt default, monetization of the deficit) does not appear to eliminate the need for a fiscal adjustment. This is consistent with another finding: public debt tends to rise in the first three years of the crisis and only gradually decline afterwards. However, when the crisis is associated with credit events that only involve official creditors, debt falls quickly. The data also suggests economic growth usually declines sharply at the onset of crises. AMs and emerging markets (EMs) face the deepest contraction in the growth rate (between 3 and 6 percentage points in the first two years of the crisis) and about half of them go through recessions. Lastly, we find that twin (fiscal and external) deficits usually deteriorate in the pre-crisis years.

We also study whether twin (fiscal-financial) crises are different. The data shows that twin crises experience a deeper decline in growth than stand-alone fiscal crises. Finally, we assess whether fiscal crises have a persistent effect on output and public debt using impulse response functions (IRFs). The evidence suggests a negative permanent impact on real GDP across all countries, while the impact on public debt varies.

The remainder of the paper is organized as follows. The next section presents the new database and criteria used to identify fiscal crises, as well as key patterns of the crisis episodes. Section III takes advantage of the database to study the macro-fiscal dynamics around fiscal crises. Section IV concludes.

II. Database of Fiscal Crisis Episodes

A. Conceptual Framework

We use the term fiscal crisis to describe a period of heightened budgetary distress, resulting in the sovereign taking exceptional measures. In normal times, a government collects tax and non-tax revenues and borrows to fund its expenditures. A country may experience fiscal distress, when large imbalances emerge between inflows (revenues and financing) and outflows (primary expenditures and debt service). These imbalances may lead to a fiscal crisis if the country does not (is not able to) sufficiently adjust its fiscal position—i.e., it may face an acute funding distress and need for exceptional and disruptive actions (e.g., default on its debt or print money to finance the deficit). More concretely, this can be thought of as a disruption in the normal debt dynamics:

dt=dt1*(rtg)/(1+gt)pt+SFt

where the debt-to-GDP ratio dt depends on the initial stock of debt in t-1, effective interest rate rt (derived from the interest expense in t divided by the debt stock in t-1), nominal GDP growth rate gt, the primary balance-to-GDP ratio pt, and a residual reflecting stock-flow adjustments SFt. (capturing e.g., exchange rate movements or materialization of contingent liabilities).

A fiscal crisis can happen for several reasons. As the debt equation shows, there may be several factors driving a country to unsustainable fiscal positions, including policies or economic shocks. First, the buildup of large budgetary imbalances (due to rising expenditures or cuts in revenues) may make the debt path unsustainable and lead to a loss of market access. Second, changes in key macro-financial variables (such as the cost of borrowing, exchange rate, and economic growth) trigger a fiscal crisis. Lastly, especially in already vulnerable countries, large fiscal imbalances can arise due to shocks, including the materialization of contingent liabilities (e.g., arising from a banking crisis or a natural disaster), or large drops in commodity prices (see IMF, 2015a).6

Our focus is on extreme and disruptive episodes of fiscal policy. We are not including cases where countries managed to address imbalances via large fiscal adjustments, while avoiding a fiscal crisis. While we recognize some of these periods may involve some distress, it would be very difficult to objectively separate “normal” fiscal adjustments from the more painful ones. In addition, even if a threshold was set, it would also be difficult to appropriately measure fiscal policy adjustments across countries and time, as variations in fiscal balances could reflect many factors that are not under policy control.7 As such, our empirical strategy focuses on identifying the extreme cases (culminating in a crisis), when countries adopt exceptional measures.

B. Identification Methodology and Data Issues

In order to empirically identify fiscal crises, our focus is on periods of extreme funding difficulties. Our identification strategy, partly building on Baldacci et al. (2011), employs a combination of four distinctive criteria (see Table 2.1): (1) credit events associated with sovereign debt (including outright defaults and restructuring); (2) recourse to large-scale IMF financial support; (3) implicit domestic public default (e.g., via high inflation rates); and (4) loss of market confidence in the sovereign. These criteria are complementary as countries may avoid outright default by, among other things, restructuring their debt, resorting to the IMF, or deflating their debt through high inflation. These criteria also supplement market indicators, which alone would not capture all fiscal crises or the right timing of the crisis.8

Table 2.1.

Identification Methodology and Data Sources

article image

This paper advances the identification methodology, and data quality, from the literature in a number of ways. We increase the country coverage to include 188 IMF member countries9 across all levels of development between 1970 and 2015. We expand the set of criteria and use new data sources.10 In particular, we add two new sub-criteria: accumulation of domestic arrears and, whilst the previous literature focuses solely on yields, a measure of loss of market access. The credit event criterion is not only more comprehensive, but also takes advantage of quantitative estimates of sovereign defaults.

Identification Criteria

(1) Credit events

This criterion captures any year in which the actions of the sovereign reduce the present value of its debt owed to official or other creditors. Due to data limitations, the focus is de facto mainly on external debt. This criterion relates the most to the literature on sovereign defaults (e.g., Detragiache and Spilimbergo, 2001; or Chakrabarti and Zeaiter, 2014). However, apart from publicly disclosed outright defaults, it can be difficult to determine if a sovereign debt operation (such as a debt equity swap, buybacks, or re-denomination of the currency) reduced the present value of debt and made creditors incur a loss. In addition, previous attempts at identifying crises also face the problem of dealing with technical defaults and their continued reporting.11

Unlike previous approaches, we take into consideration the defaulted amounts to better identify credit events. For this purpose, we use the Bank of Canada’s (BoC, 2016) annual database on the aggregated nominal stock of sovereign debt obligations in default, with the latter defined as any debt operation that inflicts an economic loss on creditors (e.g., outright default, restructuring, or rescheduling).12 The database includes sovereign defaults to both private and official creditors (e.g., Paris Club or international financial institutions, IFIs).13 With local currency defaults only reported sporadically, the database mainly contains external defaults on sovereign debt denominated in foreign currency.

Our approach refines the credit event criterion in two ways. First, we exclude small-scale technical defaults (default amounts below 0.2 percent of GDP).14 Moreover, we exclude cases of continued reporting of previously defaulted amounts (i.e., defaulted nominal amounts that grow by less than 10 percent per year). This helps avoid signaling a perpetuation of a crisis only because of factors such as lengthy regularization procedures and the accumulation of late interest and penalty fees on the previously defaulted amount outstanding.

(2) Exceptionally large official financing

An alternative to outright default or other exceptional measures for countries is recourse to large official financing, captured by financial support from the IMF.15 This support is usually for countries that are unable to pay their international bill and have associated balance of payments problems. In many occasions, fiscal distress is behind a country’s inability to keep its financial obligations. Importantly, IMF program data shows that all high-access financial arrangements had fiscal adjustment as an overarching program objective.

This criterion captures any year under an IMF financial arrangement with access above 100 percent of quota and fiscal adjustment as a program objective.16 The threshold is consistent with established IMF access rules and was also used by Baldacci et al. (2011). We exclude precautionary agreements, but include those that became active with access above our threshold (e.g., Macedonia, FYR in 2013 or Kosovo in 2012).

(3) Implicit domestic public debt default

Countries may also opt to default implicitly on domestic debt or their payment obligations. Data on these events are very scarce and the limited literature on this topic focuses almost exclusively on implicit defaults via inflation.17 We adopt a similar strategy, but complement it with data on domestic arrears (for a subset of our sample countries). Specifically, this criterion intends to capture periods where governments have difficulty meeting their obligations and resort either to running domestic payment arrears or printing money to finance the budget (leading to high inflation). We identify these episodes by looking at periods of (a) very high inflation (when the sovereign resorts to seigniorage to finance the fiscal deficit) and/or (b) accumulation of domestic arrears (for which the data is much more limited).

(3a) Very high inflation. Following Baldacci et al. (2011), we employ an inflation rate threshold of 35 percent per year for AMs, based on the average haircut on their public debt (Sturzenegger and Zettelmeyer, 2006).18 We apply the same criterion to small developing states (SDSs) as their inflation patterns are very similar to AMs. In contrast, we use a threshold of 100 percent yearly for EMs and LIDCs following Fischer, Sahay, and Vegh (2002). Those authors show that the relationship between the fiscal deficit and seigniorage is strong only in the high-inflation countries (inflation above 100 percent).19

(3b) Domestic arrears accumulation. In the absence of consistent and readily available data, we use a steep increase of “other account payables” (OAP) as a proxy. We require the OAP-to-GDP ratio to grow more than 1 percentage point per year. The threshold is in line with evidence from Checherita-Westphal et al. (2015), who show that increased delays in public payments affect private sector liquidity and profits, and ultimately economic growth. The OAP data is available for most OECD countries at least from the early 1990s onwards.

(4) Loss of market confidence

This criterion captures any year with extreme market pressures. Our two sub-criteria catch both periods of low/no volume and sovereign yield spikes.

(4a) Loss of market access. IMF (2015b) defines market access as “the ability to tap international capital markets on a sustained basis through the contracting of loans and/or issuance of securities across a range of maturities, regardless of the currency denomination of the instruments, and at reasonable interest rates.” Guscina, Sheheryar and Papaioannou (2016) compile an indicator of loss of market access (LMA): when sovereigns default or stop issuing bonds, controlling for financing needs and previous patterns of issuance. We complement the data with additional information from Gelos, Sahay, and Sandleris (2004) and rating agency reports. However, the loss of market access criterion is only applied to a country that regularly accesses international markets—that is, it has to have enjoyed two consecutive years of market access and maintained it for at least one fourth of the time (for the available sample period).20

(4b) Price of market access. We set an absolute threshold at 1,000 basis points (bps) for the spreads, which is widely seen in practice as market participants’ psychological barrier (Sy, 2004; Baldacci et al., 2011). It roughly corresponds to the 95th percentile of the spreads distribution. Any other abnormally high spreads for a country, given its history, are captured by the loss of market access criterion. Where available, we consider the JPMorgan Emerging Market Bond Index (EMBI). For a smaller number of cases, we fill the data gaps with spreads estimated as the 10-year local-currency bond yield spread to the 10-year US treasury adjusted for inflation. In addition, we look at 5- and 10-year credit default swaps (CDS) spreads.

Identifying a Crisis

We consider a year to be a fiscal crisis year when at least one of the four criteria (Table 2.1) is met.21 To separate crisis year episodes into distinct crisis events, we require at least two years of no fiscal distress between the distinct crisis events. If only one year of no fiscal distress lies between crisis year episodes, we lump them together in one event. This approach helps identify the start and end years of fiscal crisis episodes, which we later use to draw inferences on duration.22 Non-crisis years are all other years for which we have data for at least one of the criteria.

C. Key Characteristics of Fiscal Crises

We use our database to derive some new evidence on the key characteristics of fiscal crises across country groups, such as on frequency, triggers, duration, and overlap with financial crises.

Stylized Facts of Fiscal Crises

The database identifies 436 fiscal crisis episodes, implying that countries faced on average two crises since 1970 (Table 2.2). They occurred most often in LIDCs (with an average of 3.4 crises per country) and least often in AMs (with an average of less than 1 per country). In contrast, 37 countries experienced no fiscal crisis at all, of which the majority were AMs (20). For some countries, the absence of a crisis may reflect lack of sufficient data to make an assessment (e.g., Namibia, Palau, or Libya). In some cases, countries undertook large fiscal adjustments to prevent a fiscal crisis based on our criteria (e.g., oil exporters like Saudi Arabia and the U.A.E.). At both ends of the sample period, 32 crises were already ongoing at the beginning and 36 were still ongoing at the end. Crisis times are relatively frequent, with crisis years representing almost one third of all observations (Appendix Table 2).

Table 2.2.

Number of Identified Fiscal Crisis Episodes (1970-2015)

article image
Source: Authors’ calculations.

Compared to past studies, our sample is more than double the number of countries and crisis episodes (Table 2.3). To a large extent, this reflects the inclusion of LIDCs and SDSs, which have been excluded from previous analyses. In addition, past studies mainly focused on sovereign debt defaults, while our focus is on fiscal crises. Even when compared with Bruns and Poghosyan (2016), which are closer to our definition of crisis, we have a significantly larger number of events (436 versus 201) thanks to the larger sample of countries, new data, and a longer sample period.

Table 2.3.

Comparison of Identified Fiscal Crisis Episodes (1970-2015)

article image
Note: We consider events to overlap when the start date is within one year of our start date, or the event falls within our crisis period.

LIDCs and EMs have the longest fiscal crises. A fiscal crisis lasts on average almost 6 years, albeit with sizeable differences depending on countries’ development stage.23 LIDCs and EMs endure the longest crises, while SDSs face the shortest crises. Thirty of the countries in our sample have been in a crisis for more than half of the sample period. The majority of these cases are LIDCs (including e.g., Sierra Leone, Burkina Faso, or Burundi), but there are also some EMs (including e.g., Brazil, Peru, and Egypt). The common thread among many of these cases is a history of heightened political instability and weak institutions (e.g., fragile states).

The most turbulent decade was the 1990s (Figure 2.1). The 1980s also saw the largest rise in the number of crises, especially among EMs, possibly reflecting the large fall in commodity prices (with many EMs being commodity exporters), as well as the rise in global interest rates in the early part of the decade. The 1980s and 1990s were also periods of high inflation, especially among EMs and LIDCs, for which the average hovered around 100 percent before dropping to single digits during the first decade of the present century. The pattern of crises across decades is similar across the four criteria (Figure 2.2). Credit events are the most frequent in all decades, but less so in the latest years. Exceptional high-access financing from the IMF was the second most important criterion to identify crises. There was a spike in exceptional access to Fund programs between 2006 and 2010 owing to the global economic and financial crisis.

Figure 2.1.
Figure 2.1.

Number of Crisis Episodes per

Decade (1970-2015)

Citation: IMF Working Papers 2017, 086; 10.5089/9781475592153.001.A001

Figure 2.2.
Figure 2.2.

Number of Crisis Criteria Triggered per

Decade (1970-2015)

Citation: IMF Working Papers 2017, 086; 10.5089/9781475592153.001.A001

The majority of crises in non-AMs are associated with credit events (Table 2.4), which involve in most cases both official and private creditors. For these countries, credit events are the first criterion met almost three quarters of the time. Interestingly, both official and private creditors are affected in the majority of credit events—about 90 percent of the times involved official creditors, while about two-thirds of credit events included private creditors.24 Strikingly, credit events never signal the start of a crisis for AMs. This could be explained by the “the original sin” (Eichengreen and Hausmann, 1999), as most developing economies have difficulty borrowing in their own currency, especially for long maturities.25 In contrast, AMs do not have a dominant crisis identification trigger—it is broadly equally divided among the other three criteria (see also Appendix Table 2 on crisis years).

Table 2.4.

Triggering Criteria per Country Groups

article image
Source: Authors’ calculations.

Several of the identification criteria tend to overlap during the crisis period. When looking at crisis years, at least two criteria overlap more than one quarter of the time (Appendix Table 4).26 In contrast, only about 9 percent of fiscal crisis episodes start with more than one fulfilled criterion (Appendix Table 3). This highlights the relevance of using the different criteria in parallel. The most frequent combination is credit events and high-access IMF programs, which happens for more than a fifth of the crisis years that saw a credit event. Amongst non-AMs, EMs show the highest rates of overlap of different criteria in the sample. In 17 crises, their start years are identified by two or more criteria (Appendix Table 3). More importantly, crises in AMs and EMs show two or more active criteria around one third of the time, compared to a quarter of the time in LIDCs. This is relevant as the data suggests that economic growth was lower in crisis periods when more than one criterion was met (Figure 2.3).

Figure 2.3.
Figure 2.3.

Average Real GDP p.c. Growth

Citation: IMF Working Papers 2017, 086; 10.5089/9781475592153.001.A001

Fiscal and Financial Crises

At times, fiscal crises overlap with financial crises. The database contains a number of twin and triple crises (Figure 2.4).27 Close to a fifth of fiscal crises happen at the same time as either a banking or currency crisis, three percent of them even coincide with both.

Figure 2.4.
Figure 2.4.

Overlap between Fiscal, Banking, and Currency Crises

Citation: IMF Working Papers 2017, 086; 10.5089/9781475592153.001.A001

The vast majority of fiscal-banking crises occurred in AMs and EMs. This can indicate two effects. On one hand, in countries with large financial sectors, a fiscal crisis may be triggered by banking sector problems. For example, Laeven and Valencia (2012) find that the fiscal cost of banking crises net of recoveries averages 13⅓ percent of GDP. Gross fiscal outlays related to the restructuring of the financial sector can even be larger than 30 percent of crisis-year GDP (e.g., Indonesia 1997, Argentina 1980, Iceland 2008, and Jamaica 1996). In addition, IMF (2014) argues that banking crises have larger fiscal costs in countries with deeper and more leveraged banking sectors that rely more on external funding. On the other hand, a fiscal crisis could spill over to the banking system. For instance, Alter and Beyer (2014) find heightened risks of spillover from sovereign distress to the banking system for some countries in Europe. More broadly, it is possible both effects are present (Acharya et al. 2014) when government bailouts to the financial sector increase fiscal stress, in turn fueling sovereign credit risk and spreads.

More than half of the fiscal-currency crises occurred in EMs. Their public debt generally has a relatively high share of FX debt, exposing them to the risk of sudden stops. For example, in the Dominican Republic, that share amounted to 80 percent prior to the start of a fiscal crisis in 2002. A currency crisis followed in 2003, nearly doubling the public debt ratio. Another example is Malaysia (1998), which underwent an abrupt depreciation of roughly 30 percent and a surge in spreads despite major fiscal tightening (of almost 13 percentage points of GDP) between 1997 and 1999.

The triple crises are typically associated with especially turbulent times (Figure 2.5). AMs experienced three of those triple crises, while EMs experienced the other 11. These periods tend to be accompanied by substantial declines in growth. We noticed that in 6 of the cases, the fiscal crisis was signaled by more than one criterion. In 5 cases, they turn out to be some of the worst crises in terms of economic growth once the crisis started (Figure 2.5). The exception was Uruguay (2002), where the crisis was preceded by a severe recession. In several other cases, the economy contracted at a fast pace in the first years of the fiscal crisis: in Korea (1997), growth fell to -5.4 percent in 1998, and in Iceland, it fell from 9½ percent in 2007 to 1½ in 2008 and -4⅔ percent in 2009. This suggests the overlap of crises may be particularly damaging to the economy.28

Figure 2.5.
Figure 2.5.

Real GDP p.c. Growth—Before and During the Crisis

(in percentage points)

Citation: IMF Working Papers 2017, 086; 10.5089/9781475592153.001.A001

Note: Simple average of real GDP p.c. growth rates with t as the start year of the crisis.Source: Authors’ calculations.

III. Economic Outcomes and Policies Around Fiscal Crises

In order to better understand the nature of fiscal crises, it is worth looking at the behavior of some economic variables around fiscal crises.29 At first glance, not surprisingly, they deteriorate with the onset of a crisis (Table 3.1). The deterioration is especially large for economic growth and the fiscal deficit, while the current account deficit remains large but stable. The changes around fiscal crises vary considerably across country groups. For example, AMs suffer the largest growth decline. Their pre-crisis years show large external imbalances, which improve afterwards.30 In contrast, EMs experience the largest increases in the fiscal deficits around the onset of the crisis, and their current account balance deteriorates the most. LIDCs have the smallest decline in economic growth, albeit starting from the lowest pre-crisis level.

Table 3.1.

Key Macro Variables Before and During Crises

(average, in percent of GDP, except for real GDP p.c. growth which is in percentage points)

article image
Note: *t-3 to t-1. **t to t+2.Source: Authors’ calculations.

We now use our new database to scrutinize economic developments around fiscal crises in a more formal way. First, we study the behavior of macro-fiscal outcomes around fiscal crises using an event study approach. Although not necessarily implying causality, our results help to understand the context in which fiscal crises occur, particularly when differentiating along country groups and crisis identification triggers. We also study what happens to growth and debt around twin crises using a panel fixed-effect difference-in-difference (DID) model with interaction terms. Finally, we assess the long-term impact of fiscal crises on economic growth and public debt using panel impulse response functions (IRF).

A. The Dynamics Around Fiscal Crises

We start by analyzing the behavior of key fiscal (e.g., primary balance and expenditures) and macro (e.g., growth, inflation) variables that—as discussed in Section II.A—can undermine fiscal and debt sustainability, in turn triggering a crisis. Our aim is to quantitatively measure the different dynamics between the crisis period and the out-of-window normal (non-crisis) period.

Model Specification

Following the literature, we apply an event study to analyze the behavior of key variables during an 11-year window around the start of the crisis by comparing the dynamics of variables within the (pre- and post-crisis) time window with that of an out-of-window normal period.31 Following most closely Gourinchas and Obstfeld (2012), we specify fixed-effects panel regressions with a discrete-choice time window around the crisis start year as specified in:32

yi,t=αi+Σj=55βjDt+j+ɛi,t

where y is a list of variables, αi the crisis fixed-effect, Dt+j the 11 dummy variables taking the value of 1 in period t+j (where t is the crisis start year), and βj the conditional effects of a crisis over the event window relative to tranquil times. Below we discuss the main findings, but the appendix shows detailed results for the sample as a whole (Appendix Figure 1), across levels of development (Appendix Figure 23), and across identification triggers (Appendix Figure 4).

Results

A fiscal crisis tends to be preceded by loose fiscal policy

In the run-up to the crisis, there is robust (above normal times) real expenditure growth (Appendix Figure 1). Once the crisis begins, governments make an effort to contain expenditure growth aggressively, indicating fiscal policy is procyclical as economic conditions are weaker during this period. The adverse environment likely explains why budget revenues grow at a weaker pace and the fiscal balance continues to worsen. However, this varies considerably across country groups. Expenditure policy tends to be more procyclical in AMs and EMs and less so in LIDCs in the period around the crisis. The primary balance deterioration in crisis years is strongest among EMs (Appendix Figure 3.a) despite a significant reduction in expenditure growth (Appendix Figure 3.c). AMs and SDSs face a more temporary deterioration in the primary balance around the time of the crisis, which they are able to undo at a faster pace thanks to a large upfront deceleration in expenditure growth. Crises signaled by credit events and implicit defaults are the ones with the largest increases in the post-crisis budget deficit (Appendix Figure 4.c).

Economic growth falls sharply at the onset of the crisis, with AMs suffering the worst contraction

In the crisis run-up, economic growth is generally higher than in normal times (Appendix Figure 2.d). As the crisis starts, it declines sharply. In particular, economic growth tends to slow down in the year preceding the crisis and fall significantly with the onset of the crisis. Fiscal crises triggered by implicit default come with a more severe economic deterioration than all others (Appendix Figure 4.b). This may be because high inflation, and a rise in domestic arrears, can be particularly damaging for domestic activity. We also find that the growth pattern for credit events is mainly driven by the cases when private creditors are involved (whether official creditors also participate or not; Appendix Figure 5.b). When only public creditors are involved, the decline in growth from pre-crisis years is much smoother and growth converges to normal times (Appendix Figure 5.d).

Surprisingly, AMs experience the largest fall in real growth in the first two years of the crisis. EMs have the second largest fall. The fall in growth for these two groups varies between 3 and 6 percentage points in the first two years of the crisis. Remarkably, almost half of AMs and EMs experience negative growth in the first and second year of the crisis (Appendix Table 5). While LIDCs have the highest frequency of crisis and would be expected to be more vulnerable (together with SDSs), the observed adverse effects on the real economy are milder, although a third of the countries still face negative growth in the first year of the crisis. A possible explanation may be that LIDCs receive more international support to weather the crisis,33 but more research will be needed to explore the reasons.

Public debt rises and remains elevated during the first years of the crisis

Somewhat unexpectedly, public debt levels remain higher than in normal times throughout the crisis period (Appendix Figure 2.a). At the crisis onset, public debt ratios rise substantially, especially in AMs and EMs. These countries experience a large increase in their debt burden, which only falls very gradually several years after. In contrast, in LIDCs, debt falls with the start of the crisis (from above-average levels), in many cases reflecting debt defaults.

Crises associated with credit events where official creditors dominate show the most pronounced fall in public debt. While public debt tends to remain elevated (above normal years) during the first years of the crisis (Appendix Figure 4.a), this appears to be mainly driven by events involving private creditors (Appendix Figure 5.a). Most credit events involve both private and public creditors, but in the cases where only public creditors are affected (Appendix Figure 5.c), the debt levels usually fall quickly—possibly showing that debtor countries manage to obtain better conditions in such cases. On the other hand, debt levels remain significantly higher when crises are signaled by a loss of market confidence and only start declining a few years later.

Countries seek IMF support to help manage the crisis when facing twin (fiscal and external) deficits

In cases where crises are identified by access to IMF programs or a loss of market confidence, they are usually accompanied by larger external imbalances. IMF programs involve a substantial upfront effort to reduce public expenditure growth (Appendix Figure 4.d) and there is a large improvement in the external current account (Appendix Figure 4.f). After rising at the crisis onset, post-crisis public debt ratios stabilize around the average for normal periods (Appendix Figure 4.a). Possibly reflecting the larger imbalances (twin deficits), the drop in economic growth is larger than for credit events, but less pronounced relative to other triggers (Appendix Figure 4.b). Crises associated with a loss of market confidence tend to be similar. However, when market confidence falters, economic growth declines more sharply (Appendix Figure 4.b).

Twin deficits also worsen in the crisis run-up (Appendix Figure 1). This pattern dominates in AMs and EMs, with the post-crisis adjustment in the current account being largest in AMs (Appendix Figure 2.f). Among EMs, many are resource-rich countries that suffer from dramatic losses in resource exports, making their current account adjustment more difficult. In addition, except for SDSs, countries tend to experience a sharp exchange rate depreciation peaking in the year after the start of the crisis (Appendix Figure 2.c).

Was the Recent Wave of Fiscal Crises Different?

Worldwide, 80 fiscal crises occurred after the onset of the global financial crisis (GFC) in 2008. All country groups were affected to a high degree (except SDSs).1/

Fiscal crises among AMs were preceded by larger twin deficits and followed by much larger increases in public debt and lower growth than in other crises periods. While all AMs faced serious challenges posed by the GFC, a fifth of them even went into fiscal crisis—all of which occurred between 2008 and 2012. On average, those crises seem to have been somewhat shorter in duration, but stronger in impact than their predecessors. AMs saw their public debt levels rise, peaking three years after the onset of their fiscal crisis at more than 50 percentage points of GDP higher than previous fiscal crises. The crisis was triggered at a time when expenditure growth was stronger than past experience. Another particularity, the recent crises were preceded by a much larger deterioration in the external current account. The performance of economic growth, relative to past crises, deteriorated significantly only after a few years—by more than 4 percentage points less at its lowest point. In contrast, inflation dropped well below previous crises’ levels at the crisis onset. Exchange rates experienced a short-lived sizable stronger depreciation, partly fueled by quantitative easing (QE) by central banks, shortly after the onset of the fiscal crisis.

EMs experienced softer fiscal crises. Two-fifths of EMs went into a fiscal crisis between 2008 and 2014, mostly financially interconnected countries in Emerging Europe and resource-rich countries in Latin America and the Middle East. Before the fiscal crises hit, on average, those EMs had enjoyed a period of strong growth, resulting in both lower public debt levels and smaller fiscal deficits than previous episodes. However, at the onset of the fiscal crisis, debt levels had already converged to the average levels in other crisis periods and outpaced them a few years after the onset of the fiscal crisis (albeit to a much lesser degree than AMs). In EMs, partly reflecting the AMs’ QE, there was a short-lived but large dip in both effective interest rates and exchange rate depreciation in the first year after the onset of the fiscal crisis. Economic growth tended to remain somewhat above levels of previous crises.

LIDCs also fared better than during previous crisis periods. Two-thirds of LIDCs went into fiscal crisis between 2008 and 2014. They generally started from a stronger position—especially much lower debt levels (reflecting in part previous debt relief) and higher growth. Despite some acceleration in expenditure growth when the last crisis hit, LIDCs were able to contain fiscal deterioration, partly reflecting stronger real GDP growth relative to other crises. Consequently, debt levels remained well below those of previous crisis episodes.

1/ The findings are based on event studies using a fixed-effect difference-in-difference regression, in which time dummies are interacted with a “this time is different” dummy. The latter takes the value of 1 if a fiscal crisis happens in 2008 or later. The resulting coefficients give the additional effect of those recent crises over previous fiscal crises.

B. Twin Crises

Building on the evidence in Section II.B, we investigate the existence of an amplification effect of twin fiscal-financial crises on macro-fiscal variables.

Model Specification

We apply a fixed-effect model à la Gourinchas and Obstfeld (2012) and Reinhart and Rogoff (2011) with two separate crisis dummies to identify the interactions. Specifically, we follow the fixed-effect difference-in-difference (DID) analysis with the banking crisis dummy Bi and the currency crisis dummy Ci:

yi,t=αi+Σj=55βjDt+j+Σj=55γjDt+jBi+Σj=55δjDt+jCi+ɛi,t

where, during an 11-year time window around a fiscal crisis, γi indicates an additional effect as a result of twin fiscal-banking crises, and δi a similar effect caused by twin fiscal-currency crises (as captured in footnote 27 and defined in Annex 1).

Results

The evidence suggests that fiscal crises that overlap with other crises are accompanied by a more pronounced decline in economic growth and leave a country more indebted than fiscal crises alone.

Sovereigns’ indebtedness levels rose more steeply during twin fiscal-financial crises than during stand-alone fiscal crises

In fiscal-banking crises, debt ratios increase by an additional 10 percentage points of GDP, with the impact felt over time (Appendix Figure 6.a). The main drivers of higher debt in fiscal-banking crises are AMs and EMs, mainly reflecting the associated fiscal costs (e.g., bank asset purchase and bank capital injections).

In fiscal-currency crises, the additional increase in debt is more immediate, likely reflecting a sizeable exchange rate depreciation. Debt rises by around an additional 10 percent of GDP relative to a fiscal crisis (Appendix Figure 7.a), but the effect tends to dissipate after some years. While the rise in public debt is common to all country groups, the pattern and size vary significantly (Appendix Figure 7.c). LIDCs usually see a large buildup of public debt in the pre-crisis years, reflecting to some extent periods of high political instability and economic turbulence.34

Also the decline in economic growth was magnified in twin fiscal-financial crises

Growth is lower by an additional 2½ percentage points on average for fiscal-banking crises in the year after the onset of the crisis (Appendix Figure 6.b). This amplification effect is especially strong in AMs, possibly reflecting a greater dependence on the banking sector. The deterioration in economic growth around the crisis period is somewhat stronger for fiscal-currency crises. They present an additional growth contraction of about 2 to 3.5 percentage points in the first two years of the crisis (Appendix Figure 7.b). This is driven by the experience of EMs and some AMs (Appendix Figure 7.d).35 In LIDCs, the economic behavior is quite different, as twin crises tend to be preceded by economic turbulence as discussed above.

C. The Long-Term Growth and Debt Impact

Fiscal crises are associated with lower economic growth and public debt tends to be above normal periods, but are these trends persistent or are they transitory? Some countries indeed do not go back to their pre-crisis trend (Figure 3.1). Some have not only faced a deep recession, but have also remained unable to recover to pre-crisis growth rates in the aftermath of fiscal crises (e.g., Indonesia), while others have recovered partially from the crisis and returned to similar growth rates over time (e.g., Turkey). In addition, countries appear to struggle to contain (and reverse) the rise in public debt. We now examine if fiscal crises have long-term effects.

Figure 3.1.
Figure 3.1.

Actual Deviation of Real GDP p.c. from Trend

(normalized to 100 in the crisis start year)

Citation: IMF Working Papers 2017, 086; 10.5089/9781475592153.001.A001

Note: The Figure plots actual deviations of real GDP p.c. from the long-term trend over the 7-year time window after the onset of 2 fiscal crises episode (i.e. Indonesia 1997 and Turkey 1998). The projected long-term trend is based on the pre-crisis growth rate that is the average growth during [t-5, t-1] where t is the start year of a fiscal crisis. The value of real GDP p.c. is normalized to 100 in year t.

Model Specification

We specify an impulse response function (IRF) analysis à la Cerra and Saxena (2008)36 to estimate a fixed-effect AR(p) model that accounts for both lagged dependent variables yi,t−j and lagged exogenous shocks Di,t−s:

yi,t=αi+Σj=1pβjyi,tj+Σs=0qδsDi,ts+ɛi,t

Note that yi,t denotes the real GDP p.c. growth (or, in an alternative specification, the public debt-to-GDP ratio), so that the cumulative deviations in the growth rate from the pre-crisis path is the permanent loss in real GDP p.c. We derive the IRFs with a one-standard-error band drawn from a thousand Monte Carlo simulations. For our exercise, we set p = 1 and q = 4.37

Results

The impact on public debt ratios varies across country groups, but on average, public debt returns to similar levels as the pre-crisis trend (Appendix Figure 8.a). This average masks differences across country groups (Appendix Figure 8.c). In AMs, which have fewer defaults, their public debt ratio rises by around 10 percentage points of GDP at the onset of a fiscal crisis. In the long-term, the ratio remains about 4 percentage points of GDP above the pre-crisis level. In contrast, EMs experience a rise in public debt, but eventually converge to the pre-crisis level. Finally, LIDCs and SDSs’ public debt first falls but eventually returns to levels similar to pre-crisis periods.

The results suggest fiscal crises are associated with a permanent loss of real GDP of around 2 percent (Appendix Figure 8.b). The impact varies across country groups (Appendix Figure 8.d). While AMs and EMs experience greater output losses during a fiscal crisis, they tend to recover half of the initial output losses. LIDCs and SDSs, on the other hand, face a more gradual fall in GDP, but no recovery, converging to pre-crisis growth rates over time (Appendix Figure 8.d).38

IV. Conclusion

This paper provides a new comprehensive database of fiscal crises and analyzes economic developments around them. This is an area that has not been researched systematically in the past. We advance the research in two key areas. With few exceptions, the literature has focused on sovereign debt defaults. While this is an important feature of fiscal crisis, it is not the only one. We advance the identification methodology by compiling a more comprehensive set of criteria. Second, we extend significantly the coverage to all IMF member countries, which allows us to identify more crisis episodes than other studies.

Our database allows analyzing countries across all levels of development, including developing countries, which remained mostly ignored in the literature. This is especially relevant as LIDCs are the most prone to crisis. While countries on average have had two crises since 1970, LIDCs suffered at least three. Surprisingly, the decline in growth around crisis periods is lower in LIDCs and larger in advanced economies (AMs). This is something that will need to be further studied in the future. We also find that crises are predominantly linked to credit events, with the majority involving both private and official creditors.

We also used the data to highlight some other patterns associated with a fiscal crisis. While the findings do not necessarily imply causality, they reveal how policies and key economic variables evolve around these exceptional periods. First, we find that fiscal policy is procyclical, especially in AMs and EMs. Crises are preceded by loose fiscal policy, as expenditures grow above average. Once the crisis starts, countries tighten expenditure growth as economic conditions deteriorate. Second, economic growth tends to sharply decline at the onset of the crisis and, at least partially, there seems to be a permanent decline in GDP. AMs and EMs face the deepest contraction in growth and about half of them experience recessions. The decline in economic growth is particularly large when crises are triggered by high inflation. Third, public debt tends to rise and remain elevated during the first years of the crisis. An exception is when the crisis is identified by credit events that only involve official creditors. Fourth, fiscal crises are usually associated with both fiscal and external imbalances. We find that twin deficits usually deteriorate in the pre-crisis years. Finally, the data also shows that fiscal-financial twin crises experience a deeper decline in growth than stand-alone fiscal crises.

Overall, the evidence supports the adoption of prudent macro-fiscal policies to avoid the need for a disruptive fiscal adjustment as the crisis starts. Namely, defaulting on debt or financing the deficit by printing money should not be seen as a costless alternative to good policies. While a crisis may be either because of complacency or difficulty in implementing a necessary fiscal adjustment, the evidence suggests countries will be forced to tighten fiscal policy once the crisis arrives in a difficult economic environment.

Finally, we hope that the new database will contribute to future research and better understanding of fiscal crises, including to improve policy advice on how to prevent and manage them. Our paper raised several issues that will require further analysis, but future research could also focus on better identifying the economic and policy factors that lead to a crisis, including developing early warning indicators, and how to avoid them.

References

  • Acharya, V., I. Drechsler, and P. Schnabl, 2014, “A Pyrrhic Victory? Bank Bailouts and Sovereign Credit Risk,” The Journal of Finance, 69, pp. 26892739.

    • Search Google Scholar
    • Export Citation
  • Alter, A. and A. Beyer, 2014, “The dynamics of spillover effects during the European sovereign debt turmoil,” Journal of Banking and Finance, 42, pp. 134153.

    • Search Google Scholar
    • Export Citation
  • Baldacci, E., J. McHugh, and I. Petrova, 2011, “Measuring Fiscal Vulnerability and Fiscal Stress: A Proposed Set of Indicators,” IMF Working Paper No. 11/94 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Baldacci, E., I. Petrova, N. Belhocine, G. Dobrescu, and S. Mazraani, 2011, “Assessing Fiscal Stress,” IMF Working Paper No. 11/100 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Bank of Canada (BoC), 2016, “Database of Sovereign Defaults, 2016,” available via the internet: http://www.bankofcanada.ca/2014/02/technical-report-101/.

    • Search Google Scholar
    • Export Citation
  • Beers, D. and J. Mavalwalla, 2016, “Database of Sovereign Defaults,” Bank of Canada Technical Report No. 101.

  • Bruno, M. and W. Easterly, 1995, “Inflation Crises and Long-Run Growth,” NBER Working Paper No. 5209, Cambridge.

  • Bruns, M. and T. Poghosyan, 2016, “Leading Indicators of Fiscal Distress: Evidence from the Extreme Bound Analysis,” IMF Working Paper No. 16/28 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Catão, L. and G. M. Milesi-Ferretti, 2014, “External Liabilities and Crises,” Journal of International Economics, 94, pp. 1932.

  • Cavallo, M., 2005, “Understanding the Twin Deficits: New Approaches, New Results,” FRBSF Economic Letter, No. 2005-16.

  • Cerra, V. and S. Saxena, 2008, “Growth Dynamics: The Myth of Economic Recovery,” American Economic Review, 98(1), pp. 439457.

  • Chakrabarti, A., and H. Zeaiter, 2014, “The Determinants of Sovereign Default: A Sensitivity Analysis,” International Review of Economics and Finance, 33, pp. 300318.

    • Search Google Scholar
    • Export Citation
  • Checherita-Westphal, C., A. Klemm, and P. Viefers, “Governments’ Payment Discipline: The Macroeconomic Impact of Public Payment Delays and Arrears,” IMF Working Paper No. 15/13 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Cottarelli, C., 2011, “The Risk Octagon: A Comprehensive Framework for Assessing Sovereign Risks” presented in University of Rome “La Sapienza,” January. Available via the internet: www.imf.org/external/np/fad/news/2011/docs/Cottarelli1.pdf.

    • Search Google Scholar
    • Export Citation
  • Cruces J.J. and C. Trebesch, 2013, “Sovereign Defaults: The Price of Haircuts,” American Economic Journal: Macroeconomics, 5(3), pp. 85117.

    • Search Google Scholar
    • Export Citation
  • Detragiache, E. and A. Spilimbergo, 2001, “Crises and Liquidity: Evidence and Interpretation,” IMF Working Paper No. 01/02 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Eichengreen, B., A. Rose, and C. Wyplosz, 1995, “Exchange Market Mayhem: The Antecedents and Aftermath of Speculative Attacks,” Economic Policy: A European Forum, 21, pp. 249296.

    • Search Google Scholar
    • Export Citation
  • Eichengreen, B. and R. Hausmann, 1999, „Exchange Rates and Financial Fragility,” NBER Working Paper No. 7418.

  • Fischer, S., R. Sahay, and C. Vegh, 2002, “Modern Hyper- and High Inflations,” Journal of Economic Literature, 40, pp. 837880.

  • Gelos, R., R. Sahay, and G. Sandleris, 2004, “Sovereign Borrowing by Developing Countries: What Determines Market Access?,” IMF Working Paper No. 04/211 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Ghosh, A. and U. Ramakrishnan, 2006, “Do Current Account Deficits Matter?”, Finance and Development, 43(4).

  • Gourinchas, P.-O. and M. Obstfeld, 2012, “Stories of the Twentieth Century for the Twenty-First,” American Economic Journal: Macroeconomics, 4(1), pp. 226265.

    • Search Google Scholar
    • Export Citation
  • Guscina, A., M. Sheheryar, and M. Papaioannou, 2016, “Assessing Loss of Market Access: Conceptual and Operational Issues,” IMF Working Paper (forthcoming) (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • IMF, 2015a, “The Commodities Roller Coaster—A Fiscal Framework for Uncertain Times,” Fiscal Monitor, October 2015 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • IMF, 2015b, “The Fund’s Lending Framework and Sovereign Debt—Further Considerations,” MCM Board Paper (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • IMF, 2014, “From Banking to Sovereign Stress: Implications for Public Debt”, MCM Board Paper (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • IMF, 2013, “Macroeconomic Issues in Small States and Implications for Fund Engagement,” Board Paper (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • IMF, 2011, “Shifting Gears: Tackling Challenges of the Road to Fiscal Adjustment,” Fiscal Monitor, April 2011 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Kaminsky, G. and C. Reinhart, 1999, “The Twin Crises: The Causes of Banking and Balance-of-Payments Problems,” American Economic Review, 89(3), pp. 473500.

    • Search Google Scholar
    • Export Citation
  • Khan, M. and A. Senhadji, 2001, “Threshold Effects in the Relationship Between Inflation and Growth,” IMF Working Paper No. 00/110 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Laeven, L. and F. Valencia, 2012, “Systemic Banking Crises Database; An Update,” IMF Working Paper No. 12/163 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Laeven, L. and F. Valencia, 2008, “Systemic Banking Crises: A New Database,” IMF Working Paper No. 08/224 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Manasse, P., N. Roubini, and A. Schimmelpfennig, 2003, “Predicting Sovereign Debt Crises,” IMF Working Paper No. 03/221 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Mauro, P., 2011, Chipping away at Public Debt (Hoboken: Wiley & Sons, Inc.).

  • O’Connor, J., 1973, The Fiscal Crisis of the State (New York: St. Martin’s Press).

  • Reinhart, C. and K. Rogoff, 2011a, “From Financial Crash to Debt Crisis,” American Economic Review, 101(5), pp. 1676706.

  • Reinhart, C. and K. Rogoff, 2011b. “The Forgotten History of Domestic Debt.” Economic Journal 121 (552), pp. 319350.

  • Reinhart, C. and K. Rogoff, 2009, This time is different: Eight centuries of financial folly (Princeton, NJ: Princeton University Press).

    • Search Google Scholar
    • Export Citation
  • Sturzenegger, F. and J. Zettelmeyer, 2006, Debt Defaults and Lessons from a Decade of Crises, Table 1, Chapter 1 (Cambridge: MIT Press).

    • Search Google Scholar
    • Export Citation

Annex

Annex 1. Data Sources

The database covers 188 countries (all IMF members as of end-2015) and 46 years (1970-2015). We group countries in advanced economies (AMs, 35), non-small emerging economies (EMs, 70), non-small low-income developing economies (LIDCs, 50), and small developing states (SDSs, 33) for analytical purposes. There are, however, large differences in coverage among variables. Most variables come from the WEO and its previous vintages. We also use FAD’s Historical Public Debt Database to expand the gross debt time series retrieved from WEO into the past. See Appendix Tables 1 and 5.

The rest of the variables are from:

Bank of Canada’s CRAG sovereign defaults database.

  • As discussed in Beers and Chambers (2006), BoC (2016) considers that a sovereign default occurs “when debt service is not paid on the due date (or within a specified grace period), payments are not made within the time frame specified under a guarantee, or, absent an outright payment default, in any of the following circumstances where creditors incur material economic losses on the sovereign debt they hold: (i) agreements between governments and creditors that reduce interest rates and/or extend maturities on outstanding debt; (ii) government exchange offers to creditors where existing debt is swapped for new debt on less-economic terms; (iii) government purchases of debt at substantial discounts to par; (iv) government redenomination of foreign currency debt into new local currency obligations on less-economic terms; (v) swaps of sovereign debt for equity (usually relating to privatization programs) on less-economic terms; (vi) conversion of central bank notes into new currency of less-than-equivalent face value.” This is also in line with much of the literature on sovereign defaults (e.g., Cruces and Trebesch, 2011), and the practice of credit-rating agencies.

  • The database covers 136 countries from 1970-2015. We close gaps first by identifying countries that never defaulted using rating agency reports (especially Moody’s and S&P), and second by deriving lower bound estimates of sovereign debt in default from separately available components of BoC’s (2016) data. Those include, for instance, restructured amounts from Cruces and Trebesch (2013), rescheduled and relieved amounts from WB IDS and WDI.

IMF financial program data are from an IMF database that covers all members from 1952 to 2015.

The other account payables (OAP) data (our proxy for arrears) are from Eurostat and the OECD’s data portal; we prioritize the use of Eurostat data over the OECD’s. These data are available for most OECD countries starting in the early 1990s.

Yields data are from IFS. EMBI are from Reuters Datastream. CDS Spreads are from Bloomberg. For criterion 4 we prioritize the use of EMBI spreads when available.

The loss of market access dummies are from Guscina, Sheheryar, and Papaioannou (2016), for 57 countries from 1990 onwards. For years prior to 1990 and missing observations, we complement with Gelos, Sahay, and Sandleris (2004), who provide a database of market access for 140 developing countries from 1980 through 2000.

The indicators for currency and banking crises are taken from Laeven and Valencia (2012). They signal a banking crisis if at least 3 of the following 6 criteria are met: (i) extensive liquidity support; (ii) bank restructuring gross costs; (iii) significant bank nationalization; (iv) significant guarantees; (v) significant asset purchases; and (vi) deposit freezes. They signal a currency crisis if the nominal rate of depreciation is at least 30 percent and it is also 10 percentage points larger than the previous year’s rate of depreciation.

Appendix

Appendix Table 1.

Sample Countries

article image
Source: AMs as defined by the IMF WEO, SDSs as defined by IMF (2013), LIDCs as defined by the group of currently PRGT-eligible IMF members adding Zimbabwe but excluding SDSs.
Appendix Table 2.

Number of Fiscal Crisis Years

article image
Note: The sum of (sub-)criteria counts may not equal the criteria count, because several (sub-)criteria may be active at the same time.
Appendix Table 3.

Onset of Fiscal Crises—Confluence of Criteria

article image
Appendix Table 4.

Fiscal Crisis Years—Overlap between Fulfilled Criteria

article image
Appendix Figure 1.
Appendix Figure 1.

Event Study—Key Macro-Fiscal Variables

Citation: IMF Working Papers 2017, 086; 10.5089/9781475592153.001.A001

Note: The Figure plots the estimates of βj for each variable during the 11-year time window (solid line), together with the 95 percent confidence interval (dotted lines). Following the event study approach in Gourinchas and Obstfeld (2012), we measure the difference between values during the 11-year time window and “normal” period average. The x-axis is the time distance to the start of fiscal crises. We drop one outlier case in the sample for our event study (i.e. Zimbabwe 2000), because the primary balance ratio was reduced by more than 1600 percentage points 4 years after the onset of the fiscal crisis.
Appendix Figure 2.
Appendix Figure 2.
Appendix Figure 2.

Event Study along Country Groups—Public Debt, Growth, and Prices

Citation: IMF Working Papers 2017, 086; 10.5089/9781475592153.001.A001

Note: The Figure plots the estimates of βj for 6 variables (i.e. public debt ratio, effective interest rate, FX depreciation rate, real GDP p.c. growth rate, inflation rate, and current account ratio) along four country groups (i.e. AMs, EMs, LIDCs, and SDSs) during the 11-year time window (solid line), together with the 95 percent confidence interval (dotted lines). Note that, following the event study approach in Gourinchas and Obstfeld (2012), we measure the difference between values during the 11-year time window and “normal” period average. The x-axis is the time distance to the start of fiscal crises.
Appendix Figure 3.
Appendix Figure 3.

Event Study along Country Groups—Key Fiscal Indicators

Citation: IMF Working Papers 2017, 086; 10.5089/9781475592153.001.A001

Note: The Figure plots the estimates of βj for 3 variables (i.e. primary balance ratio, real public revenue growth rate, and real public expenditure growth rate) along four country groups (i.e. AMs, EMs, LIDCs, and SDSs) during the 11-year time window (solid line), together with the 95 percent confidence interval (dotted lines). Note that, following the event study approach in Gourinchas and Obstfeld (2012), we measure the difference between values during the 11-year time window and “normal” period average. The x-axis is the time distance to the start of fiscal crises.
Appendix Table 5.

Percent of Episodes with Negative Real GDP p.c. Growth

article image
Source: Authors’ calculations.
Appendix Figure 4.
Appendix Figure 4.
Appendix Figure 4.

Event Study along Crisis Criteria—Key Macro-Fiscal Variables

Citation: IMF Working Papers 2017, 086; 10.5089/9781475592153.001.A001

Appendix Figure 5.
Appendix Figure 5.

Disaggregation of Credit Events Criterion—Public Debt and Growth

Citation: IMF Working Papers 2017, 086; 10.5089/9781475592153.001.A001

Appendix Figure 6.
Appendix Figure 6.

Twin Fiscal-Banking Crises—Public Debt and Growth

Citation: IMF Working Papers 2017, 086; 10.5089/9781475592153.001.A001

Note: The Figure plots the estimates of γi, the additional effect on public debt ratio and real GDP p.c. growth rate in twin fiscal-banking crises, during an 11-year time window (solid line), together with the 95 percent confidence interval (dotted lines). Specifically, we extend the event study approach in Gourinchas and Obstfeld (2012) with difference-in-difference estimations. A twin fiscal-banking crisis occurs if a banking crisis happens during [t-1, t+1] where T is the start year of a fiscal crisis. The x-axis is the time distance to the start of the fiscal crisis.
Appendix Figure 7.
Appendix Figure 7.

Twin Fiscal-Currency Crises—Public Debt and Growth

Citation: IMF Working Papers 2017, 086; 10.5089/9781475592153.001.A001

Note: The Figure plots the estimates of δj, the additional effect on public debt ratio and real GDP p.c. growth rate by twin fiscal-currency crises, during the 11-year time window (solid line), together with the 95 percent confidence interval (dotted lines). Specifically, we extend the event study approach in Gourinchas and Obstfeld (2012) with difference-in-difference estimations. A twin fiscal-currency crisis occurs if a currency crisis happens during [t-1, t+1] where T is the start year of a fiscal crisis. The x-axis is the time distance to the start of fiscal crises.
Appendix Figure 8.
Appendix Figure 8.

Fiscal Crisis Response—Changes in Public Debt and Output

Citation: IMF Working Papers 2017, 086; 10.5089/9781475592153.001.A001

Note: The Figure plots the response of public debt ratio and real GDP p.c. to a one period exogenous fiscal crisis shock after Monte Carlo simulations. In particular, we follow Cerra and Saxena (2008) in estimating a fixed-effect AR(p) model in order to simulate the impulse response functions. The figure presents the mean of simulations, together with a one standard deviation corridor.
1

Yizhi Xu was a summer intern at the IMF when this paper was being prepared; he is a PhD candidate at the University of California, Los Angeles (UCLA), United States. We are grateful to Tamon Asonuma, Jamshid Mavalwalla, and IMF seminar participants for useful comments and suggestions on earlier versions of the paper.

2

For example, a financial crisis may put the budget under pressure either directly due to a need to bail out banks (e.g., Ireland in 2010 or Brazil in 1998) or via a sharp economic deterioration and subsequent fall in tax revenues (e.g., Paraguay in 2002). It may also involve a loss of market confidence (e.g., Turkey in 1998 or Zambia in 1999), and extreme measures to address the imbalances, including defaulting on debts (e.g., Greece in 2010 or Ecuador in 1982) or printing money to finance the deficit (e.g., Zimbabwe in 2001).

3

Baldacci et al. (2011) compile a set of criteria to identify fiscal crises over 1970-2010. Bruns and Poghosyan (2016) extend the dataset through 2015. These papers focus on assessing the likelihood of entering a crisis, and do not study the crisis period itself. Also, their sample is limited to 81 advanced and emerging market economies.

4

LIDCs are usually seen as having less policy space to weather shocks especially as they have limited access to international capital markets and their domestic financial markets are underdeveloped. Many of these countries also have less diversified economies (depend on a few commodities), making them more vulnerable to shocks.

5

There are other possible explanations. For example, it may be that fiscal crises affect the informal economy less or that AMs accumulate larger imbalances preceding the crisis.

6

For a discussion of factors that may help predict a period of fiscal distress see also Cottarelli (2011), IMF (2011), or Baldacci, McHugh and Petrova (2011).

7

For example, the ex-post adjustment in the fiscal balance could reflect large variations in revenue due to economic growth or commodity prices (for resource-rich countries). Trying to correct for these would imply having to make an assessment of potential GDP and identify policy measures (including temporary versus permanent), among others, which would be very difficult given data limitations. Variations in exchange rate or interest rate could also have an impact.

8

Various factors can make market indicators overshoot, fail to react, or react with a lag. For instance, they incorporate the probability of non-market financing (i.e., IMF and other official support); involve biased thresholds (e.g., debt tolerance which has generally been lower in non-AMs); and are susceptible to contagion effects.

9

The database only excludes one (new) member of the IMF, Nauru, given lack of data.

10

Data availability varies significantly over time (e.g., limited coverage in the 1970s) and across countries, especially for many developing countries. See Annex 1 for details on the data, coverage, and sources.

11

Technical defaults include those related to debt payment delays due to administrative procedures and debt management capacity issues. Continued reporting of default captures those related to delayed regularizations due, for instance, to legal and negotiation reasons. This is indeed a common phenomenon among many countries (see also Reinhart and Rogoff, 2011a).

12

We complement the database with other sources (rating agencies) when needed. See Annex 1 for more details.

13

The dataset also includes information on relief operations, but they are not used to identify a fiscal crisis. This is because these operations have provided relief on debt arrears. Nevertheless, the debt relief would help reduce the debt stock and manage the fiscal crisis.

14

The threshold is relatively robust. Lowering it to 0.1 percent of GDP increases the number of fiscal crisis episodes by 11. In contrast, increasing it to 0.25 percent of GDP lowers the number by 6.

15

IMF programs can have a catalytic effect, i.e. other governments and international agencies will join efforts to provide additional official financing. IMF loans typically involve lower borrowing costs than market-based loans.

16

Available financing—i.e. total access and phasing—under an IMF program depends on the size and nature of a country’s financing need over the course of the program period, the strength of the reform program, and the access limits under the requested IMF program type. With the latter being restricted both per program request and on a cumulative basis, access also becomes a strategic, intertemporal choice.

17

A rare exception is Reinhart and Rogoff (2011b), who list 42 cases of explicit domestic default. Those include not only credit events (e.g., Russia’s local currency debt default in 1989-99, or Ghana’s default on central bank notes in 1979 and 1982), but also other forms of default (e.g., Mexico’s forcible conversion of USD deposits to pesos in 1982 or Panama’s arrears to domestic suppliers in 1988-89). Our database covers all of their 42 cases for our sample, using our criteria.

18

This threshold is roughly confirmed by the 95th percentile of the inflation distribution and lies between the 20 and 40 percent thresholds otherwise used in the literature (see e.g., Reinhart and Rogoff, 2011; Khan and Senhadji, 2001; or Bruno and Easterly, 1995).

19

The high threshold also reduces the likelihood of picking up high inflation triggered by supply side shocks.

20

Under the precondition of maintaining market access for one fourth of the time covered by the sample for sub-criterion (4a), criterion 4 is triggered 82 times. Changing to one half of the time reduces the number to 69. Eliminating the precondition gives 84 cases.

21

The database (available at http://www.imf.org/~/media/Files/Publications/WP/2017/datasets/wp1786.ashx) gives a dummy for the fiscal crisis years for the entire set of criteria, but also shows which individual criteria are met in any given year.

22

We do not assign definite start and end dates to episodes ongoing at both ends of the sample period. The exception is if (i) previous indicators identified start dates in 1970-71 (Baldacci et al., 2011; Cruces and Trebesch, 2013; or sovereign defaults from Laeven and Valencia, 2012) or (ii) recent information confirms end dates in 2014.

23

LIDCs (6.6 years), EMs (6 years), AMs (3.8 years), and SDS (averaging 3.2 years). When assessing duration, we do not consider the crisis periods that are ongoing at the start (or end of the sample period), because of a failure to determine the exact start (or end) date if it precedes (or outlasts) our sample period.

24

For example, excluding defaults involving official creditors would reduce the number of crises from 436 to 337. The fall in the number of crises is mitigated because other identification triggers would still help detect some of the crises. The official creditor-led events affect mainly SDSs and to a lesser degree LIDCs and EMs.

25

Also, developing economies often have weaker capacity (including in the areas of public financial and debt management), which could contribute to fiscal distress.

26

For example, out of the 10 crisis episodes in AMs that start with high-access IMF programs, 3 led to credit events within the same crisis period (Greece, Ireland, and Portugal).

27

We use Laeven and Valencia’s (2012) dummies for banking and currency crises (see more detail in Annex 1). We only consider fiscal crises up to 2012 to match their sample. We define a twin fiscal-banking (or fiscal-currency) crisis in year t as a fiscal crisis in year t combined with a banking (currency) crisis during the period [t -1, t +1]. We define a triple fiscal-banking-currency crisis in year t as a fiscal crisis in year t, combined with a currency crisis and a banking crisis during the period [t -1, t +1]. Our results are robust across different definitions of currency crises, such as Reinhart and Rogoff’s (2011) lower depreciation threshold.