Financial Crises
Chapter

Chapter 8. From Recession to Recovery: How Soon and How Strong?

Author(s):
Stijn Claessens, Ayhan Kose, Luc Laeven, and Fabian Valencia
Published Date:
February 2014
Share
  • ShareShare
Show Summary Details
Author(s)
Prakash Kannan, Alasdair Scott and Marco E. Terrones The authors are grateful to Olivier Blanchard, Charles Collyns, Jorg Decressin, Francis Diebold, Don Harding, and David Romer for their comments and suggestions. Gavin Asdorian and Emory Oakes provided outstanding research assistance.

This chapter was initially published in the April 2009 IMF World Economic Outlook when the world economy was experiencing one of the most severe recessions in modern times. The main text of the chapter is reproduced virtually unaltered as a test of its originality and the durability of its main findings.

At the time it was written, many felt that the recovery from the recession associated with the 2007–09 global financial crisis would be quick and robust. To test this belief, previous experiences of recessions and recoveries in advanced economies were examined. Crucially, the chapter made two distinctions. The first was to divide recessions into those following financial crises from those arising from other shocks (e.g., energy prices or monetary policy). The second was to look at globally synchronized recessions. The findings were stark: First, recessions associated with financial crises are typically severe and protracted and their recoveries slow and weak. Second, globally synchronized recessions are often long and deep, and recoveries from these recessions are generally weak and protracted. The implications of these findings for the global economy were that the ongoing recession was likely to be unusually long and severe and that the recovery would be weak and sluggish.

For policies, the message was that early intervention could make a material difference to the duration and severity of the recession and to the strength of the recovery. But that came with caveats: the bang per buck of monetary policy was found to be lower in the aftermath of a financial crisis, and fiscal policy was likely to be less effective when public debt was high.

In retrospect, the chapter was, if anything, not pessimistic enough. Even on the basis of the evidence presented, it would have seemed incredible that a crisis that began in 2007 would linger for at least five years. There are developments that at the time this chapter was written could not have been anticipated. The zero nominal bound on interest rates would prove more quickly binding than anticipated, and the deterioration of public finances in several advanced economies as a result of the downturn was substantial. The financial crisis and ensuing recession exposed fundamental weaknesses in the design of the euro area, which experienced its own distinct crisis.

Nonetheless, the chapter’s main findings, made early in 2009, hold up well at the end of 2012.

Introduction

The global economy experienced its deepest downturn in the post-World War II period as the 2007–09 financial crisis rapidly spread around the world. A large number of advanced economies fell into recession, and economies in the rest of the world slowed abruptly. Global trade and financial flows shrank, while output and employment losses mounted. Credit markets remained frozen as borrowers engaged in a drawn-out deleveraging process and banks struggled to improve their financial health.

Many aspects of the current crisis were new and unanticipated.1 Uniquely, the disruption combined a financial crisis at the heart of the world’s largest economy with a global downturn. But financial crises—episodes during which there is widespread disruption to financial institutions and the functioning of financial markets—are not new.2 Nor are globally synchronized downturns. Therefore, history can be a useful guide to understanding the present.

To put the recession caused by the 2007–09 global crisis in historical perspective, some broad questions about the nature of recessions and recoveries and the role of countercyclical policies are addressed in this chapter. In particular,

  • Are recessions and recoveries associated with financial crises different from other types of recessions and recoveries?

  • Are globally synchronized recessions different?

  • What role do policies play in determining the shape of recessions and recoveries?

To shed light on these questions, this chapter examines the dynamics of business cycles that occurred during the past half century. It complements existing literature on the business cycle along several dimensions.3 These dimensions include a comprehensive study of recessions and recoveries in 21 advanced economies,4 a classification of recessions based on their underlying sources, and an assessment of the impact of fiscal and monetary policies in recessions and recoveries. Similar to most other studies in this area, this chapter makes extensive use of event analysis and statistical associations.

The main findings related to common elements across business cycles are as follows:

  • Recessions in the advanced economies since 1990 have become less frequent and milder, whereas expansions have become longer, reflecting in part the “Great Moderation” of advanced economies’ business cycles.

  • Recessions associated with financial crises have been more severe and longer lasting than recessions associated with other shocks. Recoveries from such recessions have typically been slower, associated with weak domestic demand and tight credit conditions.

  • Recessions that are highly synchronized across countries have been longer and deeper than those confined to one region. Recoveries from these recessions have typically been weak, with exports playing a much more limited role than in less synchronized recessions.

The implications of these findings for the global financial crisis situation were sobering at the time. The downturn was highly synchronized and was associated with a deep financial crisis, a rare combination in the postwar period. Accordingly, the downturn was likely to be unusually severe, and the recovery was expected to be sluggish. It is not surprising, therefore, that many commentators looking for historical parallels for the 2007–09 crisis episode focused on the Great Depression of the 1930s, by far the deepest and longest recession in the history of most advanced economies.

Regarding policies, these are the main findings:

  • Monetary policy seems to have played an important role in ending recessions and strengthening recoveries. Its effectiveness, however, is weakened in the aftermath of a financial crisis.

  • Fiscal stimulus appears to be particularly helpful during recessions associated with financial crises. Stimulus is also associated with stronger recoveries; however, the impact of fiscal policy on the strength of the recovery is found to be smaller for economies that have higher levels of public debt.

These results suggested that to mitigate the severity of the recession following the global crisis and to strengthen the recovery, aggressive monetary and particularly fiscal measures were needed to support aggregate demand in the short term, but care needed to be taken to preserve public debt sustainability in the medium term. Even with such measures, a return to steady economic growth would depend on restoring the health of the financial sector. One of the most important lessons from the Great Depression, and from more recent episodes of financial crisis, is that restoring confidence in the financial sector is crucial for recovery to take hold.

The rest of the chapter is structured as follows. The next section presents key stylized facts on recessions and recoveries for the advanced economies during the past 50 years. The second section reviews the major differences across recessions and recoveries resulting from different types of shocks and different degrees of synchronization. Particular attention is paid to the influence of financial crises. The third section analyzes the effects of discretionary monetary and fiscal policies on the severity of recessions and on the strength of recoveries. It also examines how the level of public debt conditions the effectiveness of fiscal policy. The last section places the downturn associated with the 2007–09 crisis in historical perspective and discusses its policy implications.

Business Cycles in The Advanced Economies

To put the recession caused by the 2007–09 crisis in historical perspective, the features of previous cycles are first identified. Each cycle is divided into two main phases: a recession phase, characterized by a decline in economic activity, and an expansion phase. Following the long-standing tradition of Burns and Mitchell (1946), this chapter employs a “classical” approach to dating turning points in a large sample of advanced economies from 1960 to 2007. It focuses on quarterly changes in real GDP to determine cyclical peaks and troughs (Figure 8.1).5

Figure 8.1Business Cycle Peaks and Troughs

Source: Authors’ calculations.

The two main properties of the cycle are considered:

  • Duration: the number of quarters from peak to trough in a recession, or from trough to the next peak in an expansion.

  • Amplitude: the percentage change in real GDP, from peak to trough in a recession, or from trough to the next peak in an expansion.

The analysis also examines the slope of a recession (or expansion), that is, the ratio of amplitude to duration, which indicates the steepness of each cyclical phase.

Recessions and Expansions: Basic Facts

On average, advanced economies have experienced six complete cycles of recession and expansion since 1960.6 The number of recessions, however, varies significantly across countries, with some (Canada, Ireland, Japan, Norway, and Sweden) experiencing only three recessions and others (Italy, New Zealand, and Switzerland) experiencing nine or more.

Recessions are distinctly shallower, briefer, and less frequent than expansions. In a typical recession, GDP falls by about 2¾ percent (Table 8.1).7 In contrast, during an expansion, GDP tends to rise by almost 20 percent. This illustrates mainly the importance of trend growth; the higher the long-term growth rate of an economy, the shallower the recession and the greater the amplitude of expansions. Some recessions, however, are severe, with peak-to-trough declines in output exceeding 10 percent. These episodes are often called depressions (IMF, 2002). Since 1960, there have been six depression episodes in the advanced economies; the latest was observed in Finland in the early 1990s. In contrast, some expansions witness trough-to-peak output increases larger than 50 percent—the “Irish Miracle” during the first decade of the 2000s being a recent example.

Table 8.1Business Cycles in Advanced Economies: Summary Statistics
Duration (quarters)Amplitude (percent change in real GDP)
RecessionRecovery1ExpansionRecessionRecovery2Expansion
All
1. Mean3.643.2221.75−2.714.0519.56
2. Standard deviation2.072.7217.892.933.1217.50
3. Coefficient of variation (line 2 ÷ line 1)0.570.840.821.080.770.89
4. Number of events122109122122112122
By driver of recession
Financial crises
5. Mean5.67**5.64**26.40**−3.392.21***19.47
6. Standard deviation3.153.3224.743.251.1820.46
7. Coefficient of variation (line 6 ÷ line 5)0.560.590.940.960.531.05
8. Number of events151115151315
Other3
9. Mean3.36**2.95**21.09**−2.614.29***19.58
10. Standard deviation 1.71 2.5216.772.893.2217.15
11. Coefficient of variation (line 10 ÷ line 9)0.510.850.791.110.750.88
12. Number of events1079810710799107
By extent of synchronization
Highly synchronized
13. Mean4.54***4.19*19.97***−3.45*3.66**16.24*
14. Standard deviation2.503.5915.322.961.7211.85
15. Coefficient of variation (line 14 ÷ line 13)0.550.860.770.860.470.73
16. Number of events373237373437
Other4
17. Mean3.25***2.82*22.52***−2.39*4.21**21.01*
18. Standard deviation1.732.1618.942.883.5619.33
19. Coefficient of variation (line 18 ÷ line 17)0.530.770.841.210.850.92
20. Number of events857785857885
Memo item:
Recessions associated with financial crises that are highly synchronized
Mean7.336.7524.33−4.822.8218.83
Source: Authors’ calculations.Note: The symbols *, **, and *** indicate statistical significance at the 10, 5, and 1 percent levels, respectively. Statistical significance for recessions associated with financial crises (highly synchronized recessions) calculated versus other recessions.

Number of quarters after trough and before real GDP recovers its level achieved in the previous peak.

Percent increase in real GDP one year after the trough.

Recessions not associated with a financial crisis.

Recessions that are not highly synchronized.

Source: Authors’ calculations.Note: The symbols *, **, and *** indicate statistical significance at the 10, 5, and 1 percent levels, respectively. Statistical significance for recessions associated with financial crises (highly synchronized recessions) calculated versus other recessions.

Number of quarters after trough and before real GDP recovers its level achieved in the previous peak.

Percent increase in real GDP one year after the trough.

Recessions not associated with a financial crisis.

Recessions that are not highly synchronized.

A typical recession persists for about a year, whereas an expansion often lasts more than five years. As a result, advanced economies are in a recession phase of the cycle only 10 percent of the time. The longest episodes of recessions and expansions in these countries lasted more than 3 years and 15 years, respectively. Finland and Sweden experienced two of the longest recessions and Ireland and Sweden experienced two of the longest expansions.

Since the mid-1980s, recessions in advanced economies have become less frequent and milder, and expansions have become longer lasting, a development associated with the Great Moderation (Figure 8.2).8 A host of factors may explain this, including global integration, improvements in financial markets, changes in the composition of aggregate output toward the service sector and away from manufacturing, and better macroeconomic policies (Blanchard and Simon, 2001; Romer, 1999). Another possibility is that the Great Moderation was the result of good luck, primarily reflecting the absence of large shocks to the world economy.

Figure 8.2Business Cycles Have Moderated over Time

Source: Authors’ calculations.

The recovery phase of the cycle has been an object of constant interest in policy circles.9 An economy typically recovers to its previous peak output in less than a year (see Table 8.1). Perhaps more important, recoveries are typically steeper than recessions—the average growth per quarter during a recovery exceeds the rate of contraction during a recession by more than 25 percent. In fact, there is evidence of a bounce-back effect: output growth during the first year of recovery is significantly and positively related to the severity of the preceding recession. A number of factors can drive an economy to bounce back, including fiscal and monetary policies (this possibility is explored later in the chapter), technological progress, and population growth.10

Does The Cause of A Downturn Affect The Shape of The Cycle?

This section associates recessions and their recoveries with different types of shocks: financial, external, fiscal policy, monetary policy, and oil price shocks.11 The objective of this exercise is to determine whether there have been important differences between the recessions associated with financial crises and those associated with other shocks. In addition, this section examines whether there is a difference between highly synchronized and nonsynchronized recessions.

Different shocks are found to be associated with different patterns of macroeconomic and financial variables during recessions and recoveries. In particular, recessions associated with financial crises have typically been severe and protracted, whereas recoveries from recessions associated with financial crises have typically been slower, held back by weak private demand and credit. In addition, highly synchronized recession episodes are longer and deeper than other recessions, and recoveries from these recessions are typically weak. Moreover, developments in the United States play a pivotal role both in the severity and duration of these highly synchronized recessions.

Categorizing Recessions and Recoveries

The analysis begins categorizing recessions and recoveries by first defining financial crises as episodes during which there is widespread disruption to financial institutions and the functioning of financial markets. Financial crises are identified using the narrative analysis of Reinhart and Rogoff (2008a, 2008b, 2009),12 which in turn draws on the work of Kaminsky and Reinhart (1999).13 Next, a recession is said to be associated with a financial crisis if the recession episode starts at the same time as or after the beginning of the financial crisis.14 Of the 122 recessions in the sample, 15 are associated with financial crises (Table 8.2).15 The other disturbances are identified using simple statistical rules of thumb (see Appendix 8A).16 More than half of the 122 recessions in the sample are associated with one or more of these shocks.17 Oil shocks are the most widespread type, affecting 17 economies in the sample. Monetary and fiscal policy shocks are less common, and external demand shocks are the least common of all, affecting only a handful of the smaller and more open economies (see Table 8A.2 in the appendix to this chapter). Although recessions have become less common overall during the Great Moderation, those associated with financial crises have become more common (Figure 8.3).

Table 8.2Financial Crises and Associated Recessions
Australia1990:Q2–1991:Q2
Denmark1987:Q1–1988:Q2
Finland1990:Q2–1993:Q2*
France1992:Q2–1993:Q3
Germany1980:Q2–1980:Q4
Greece1992:Q2–1993:Q1
Italy1992:Q2–1993:Q3
Japan1993:Q2–1993:Q4*
Japan1997:Q2–1999:Q1
New Zealand1986:Q4–1987:Q4
Norway1988:Q2–1988:Q4*
Spain1978:Q3–1979:Q1*
Sweden1990:Q2–1993:Q1*
United Kingdom1973:Q3–1974:Q1
United Kingdom1990:Q3–1991:Q3
Source: Authors’ calculations.* The Big Five financial crises (Reinhart and Rogoff, 2008a).
Source: Authors’ calculations.* The Big Five financial crises (Reinhart and Rogoff, 2008a).

Figure 8.3Temporal Evolution of Recessions by Type of Shock

Source: Authors’ calculations.

Summaries of the stylized facts of these different categories of recessions and recoveries are presented in Table 8.1 and Figure 8.4. With the notable exception of oil shocks, the amplitude of a recession is closely related to its duration.18 Recessions associated with financial crises are longer and generally more costly than others; those associated with the Big Five financial crises identified by Reinhart and Rogoff (2008a) were particularly costly (Figure 8.4, panel a).19 Financial crises are also followed by weak recoveries: the time taken to recover to the level of activity reached in the previous peak is as long as the recession itself, whereas cumulative GDP growth in the four quarters after the trough is typically lower than following other types of recessions (Figure 8.4, panel b).20 Note that the cumulative growth one year after the trough for a financial crisis is up to 2½ percentage points lower than in other cases, after controlling for the severity and duration of the previous recession.

Figure 8.4Average Statistics for Recessions and Recoveries

Source: Authors’ calculations.

Note: The Big Five financial crises are Finland (1990–93), Japan (1993), Norway (1988), Spain (1978–79), and Sweden (1990–93—Reinhart and Rogoff, 2008a).

Why Are Financial Crises Different?

What are the mechanisms that differentiate recessions and recoveries associated with financial crises? An answer to this question needs to take into account the nature of the expansions that preceded these recessions. Narrative evidence indicates that financial crisis episodes have often been associated with credit booms involving overheated goods and labor markets, house price booms, and frequently, a loss of external competitiveness.21 This can be seen in Figure 8.5, which shows median values of macroeconomic variables during the eight quarters before the peak in GDP. Credit growth during the expansions preceding financial crises is higher than during other expansions, and this is associated with higher-than-usual consumption as a share of GDP leading up to the peak. Relative to other expansions, labor market participation is high, nominal wage growth is high, and unemployment is low. Price increases—for example, the GDP deflator, house prices, and equity prices—are all noticeably higher than usual. Credit booms have frequently followed financial deregulation.22 There is some evidence of asset price bubbles: in the period leading up to financial crisis episodes, the ratio of house prices to housing rental rates rises above that during other recession episodes, starting from levels well below (Figure 8.6).

Figure 8.5Expansions in the Run-Up to Recessions Associated with Financial Crises and Other Shocks

Source: Authors’ calculations.

1 Data in real terms.

Figure 8.6House Price-to-Rental Ratios for Recessions Associated with Financial Crises and Other Shocks

Source: Authors’ calculations.

Note: Peak in output at t = 0.

Rapid credit growth has typically been associated with shifts in household saving rates and a deterioration of the quality of balance sheets.23 Panel a of Figure 8.7 shows that household saving rates out of disposable income are noticeably lower in expansions before financial crises. However, after a financial crisis strikes, saving rates increase substantially, especially during recessions. In the Big Five episodes, the turnaround in household saving rates was larger still. Data for net lending paint a complementary picture (Figure 8.7, panel b). Although these data cover only a few of the financial crisis episodes under consideration here, patterns from some of the most relevant episodes—Denmark (1985–89), Finland (1988–92), Norway (1986-90), and the United Kingdom (1988–92)—show that households’ net lending balances increased substantially during recessions.

Figure 8.7Household Saving Rate and Net Lending before and after Business Cycle Peaks

Source: Authors’ calculations.

Note: Peak in output at t = 0. The Big Five financial crises are Finland (1990–93), Japan (1993), Norway (1988), Spain (1978–79), and Sweden (1990–93—Reinhart and Rogoff, 2008a).

Taken together, the behavior of these variables suggests that expansions associated with financial crises may be driven by overly optimistic expectations for growth in income and wealth.24 The result is overvalued goods, services, and in particular, asset prices. For a period, this overheating appears to confirm the optimistic expectations, but when expectations are eventually disappointed, restoring household balance sheets and adjusting prices downward toward something approaching fair value require sharp adjustments in private behavior. Not surprisingly, a key reason recessions associated with financial crises are so much worse is the decline in private consumption.

Turning to the recovery phase, the weakness in private demand tends to persist in upswings that follow recessions associated with financial crises (Figure 8.8). Private consumption typically grows more slowly than during other recoveries. Private investment continues to decline after the recession trough; in particular, residential investment typically takes two years merely to stop declining. Thus, output growth is sluggish, and the unemployment rate continues to rise by more than usual. Credit growth falters, whereas in other recoveries it is steady and strong. Asset prices are generally weaker; house prices in particular follow a prolonged decline. However, although the recovery of domestic private demand from financial crises is weaker than usual, economies hit by financial crises have typically benefited from relatively strong demand in the rest of the world, which has helped them export their way out of recession.

Figure 8.8Recessions and Recoveries Associated with Financial Crises and Other Shocks

Source: Authors’ calculations.

Note: Data in real terms unless noted otherwise. Peak in output at t = 0.

1 Difference for level at t = 0, in percentage points.

What do these observations reveal about the dynamics of recovery after a financial crisis? First, households and firms either perceive a stronger need to restore their balance sheets after a period of overleveraging or are constrained to do so by sharp reductions in credit supply. Private consumption growth is likely to be weak until households are comfortable that they are more financially secure. It would be a mistake to think of recovery from such episodes as a process in which an economy simply reverts to its previous state.

Second, expenditures with long planning horizons—notably real estate and capital investment—suffer particularly from the aftereffects of financial crises. This appears to be strongly associated with weak credit growth. The nature of these financial crises and the lack of credit growth during recovery indicate that this is a supply issue. Furthermore, industries that conventionally rely heavily on external credit recover much more slowly after these recessions.

Third, given the below-average trajectory of private demand, an important issue is the extent to which public and external demand can contribute to growth. In many of the recoveries following financial crises examined in this section, an important condition was robust world growth. This raises the question of what happens when world growth is weak or nonexistent.

Are Highly Synchronized Recessions and Their Recoveries Different?

The 2007–09 downturn was global, implying that the recovery could not in the aggregate be driven by a turnaround in net exports (although this could be true for individual economies). Therefore, an examination of the features of synchronized recessions may help in gauging the evolution of the recession and prospective recovery.

To address this issue, highly synchronized recessions are defined as those during which 10 or more of the 21 advanced economies in the sample were in recession at the same time.25 In addition to the cycle kicked off by the 2007–09 crisis, there were three other episodes of highly synchronized recessions: 1975, 1980, and 1992 (Figure 8.9).26 As seen in Table 8.1, highly synchronized recessions are longer and deeper than others: the average duration of a synchronous recession is 40 percent greater than that of other recessions, and the amplitude is 45 percent greater.

Figure 8.9Highly Synchronized Recessions

Source: Authors’ calculations.

Note: Shaded areas denote U.S. recession.

What are the distinctive features of highly synchronized recessions? The most obvious is that they are severe, as seen in Figure 8.10. Moreover, recoveries from synchronous recessions are, on average, very slow, with output taking 50 percent longer on average to recover its previous peak than after other recessions. Credit growth is also weak, in contrast to recoveries from nonsynchronous recessions, during which credit and investment recover rapidly. As with financial crises, investment and asset prices continue to decline after the trough in GDP. However, a key difference from the recoveries following localized financial crises is that net trade is much weaker. When compared with nonsynchronous recessions, exports are typically more sluggish in synchronous recessions.

Figure 8.10Are Highly Synchronized Recessions Different?

Source: Authors’ calculations.

Note: Data in real terms unless noted otherwise. Peak in output at t = 0.

1 Difference from level at t = 0, percentage points.

The United States has typically been at the center of synchronous recessions. Three of the four synchronous recessions (including the one associated with the 2007–09 crisis) were preceded by, or coincided with, a recession in the United States. During both the 1975 and 1980 recessions, sharp drops in U.S. imports caused significant contractions in world trade.27 In addition to strong trade linkages, downward movements in U.S. credit and equity prices are likely to be transmitted to other economies.

Does Bad Plus Bad Equal Worse?

Recessions that are associated with both financial crises and global downturns have been unusually severe and long lasting. Since 1960, there have been only 6 recessions out of the 122 in the sample that fit this description: Finland (1990), France (1992), Germany (1980), Greece (1992), Italy (1992), and Sweden (1990). On average, these recessions lasted almost two years (Table 8.1, final row). Moreover, during these recessions GDP fell by more than 4¾ percent. Reflecting in part the severity of these recessions, recoveries from synchronized recessions are weak.

Can Policies Play A Useful Countercyclical Role?

Up to this point, this chapter has examined the dynamics of recessions and recoveries without accounting for economic policy responses. Policymakers, however, generally try to reduce fluctuations in output. Narrative studies of the policy decision-making process, such as Romer and Romer (1989, 2007), show that concerns about the state of the economy are a key input to the formulation of policy.

This section examines how monetary and fiscal policies have been used as countercyclical tools during business cycle downturns. The effectiveness of policy interventions in smoothing the business cycle is a topic of long debate in the academic literature. Much of the debate centers on the impact of active, or discretionary, policies rather than on the component of policies that automatically responds to the business cycle. The debate about the role of fiscal policy has been particularly intense, and estimates of how output responds to discretionary changes in policy vary dramatically depending on the methodology used, the sample of countries, and the period examined. Indeed, there is evidence that the multipliers can at times be negative. The consensus, however, is that discretionary fiscal policy does have a positive impact on growth, though the magnitude is fairly small.28

The effectiveness of policy interventions in smoothing the business cycle is a topic of long debate in the academic literature. Much of the debate centers on the impact of active, or discretionary, policies rather than the component of policies that automatically responds to the business cycle. The debate over the role of fiscal policy has been particularly intense, and estimates of how output responds to discretionary changes in policy vary dramatically depending on the methodology employed, the sample of countries, and the time period examined. Indeed, there is evidence that the multipliers can at times be negative. The consensus, however, is that discretionary fiscal policy does have a positive impact on growth, though the magnitude is fairly small.

A common challenge faced in empirical research on macroeconomic policies is the appropriate measurement of discretionary policy. In general, any measure of macroeconomic policy is interrelated with output, making causal inference difficult. To address this problem, this section distinguishes the automatic response of policy (which depends on economic activity) from the discretionary one by using a simple regression framework. The discretionary component of fiscal policy is proxied by the cyclically adjusted primary fiscal balance as well as by cyclically adjusted real government consumption.29 Similarly, the discretionary component of monetary policy is proxied by the nominal interest rate and real interest rate deviations from a Taylor rule, which attempts to capture how the central bank responds to fluctuations in the output gap and to deviations from an explicit, or implicit, inflation target. For each recession phase, the baseline measure of policy response is the peak-to-trough change, a cumulative measure of the degree of loosening or tightening of policy over the whole recession.30

Discretionary fiscal and monetary policies have typically been expansionary during recessions (Figure 8.11).31 The mean increase in the discretionary component of government consumption during a recession is about 1.1 percent a quarter, whereas the average decline in real interest rates, beyond that implied by a Taylor rule, is about 0.2 percentage point a quarter. 32 The advanced economies have historically responded more aggressively using monetary policy than other countries.33 However, some European economies are unable to lower interest rates independently during recessions because of their commitment to the European Exchange Rate Mechanism and membership in the euro area.

Figure 8.11Average Policy Response during a Recession

Source: Authors’ calculations.

Note: G7 economies include Canada, France, Germany, Italy, Japan, the United Kingdom, and the United States.

Do Policies Help Mitigate the Duration of Recessions?

The impact of discretionary monetary and fiscal policies on the duration of recessions is examined by looking at the cross-country experience across various recession episodes using duration analysis. Duration analysis models the probability that an event, such as the end of a recession, will occur. Previous studies have used these models to address the question of whether recessions are more likely or less likely to end as they continue.34 This chapter adds to this analysis by looking at the impact of policies on the likelihood that an economy exits a recession.

Across all types of recessions, there is evidence that expansionary monetary policy is typically associated with shorter recessions, whereas expansionary fiscal policy is not. A 1 percent reduction in the real interest rate beyond that implied by the Taylor rule increases the probability of exiting a recession in a given quarter by about 6 percent. However, fiscal policy, measured either by changes in the primary balance or in government consumption, is not found to have a significant impact on the duration of recessions when examined across all recessions.

However, both expansionary fiscal and monetary policies tend to shorten the duration of recessions associated with financial crises, although the effect of monetary policy is not statistically significant (Table 8.3). During these episodes, a 1 percent increase in government consumption is associated with an increase in the probability of exiting a recession of about 16 percent. The stronger impact of fiscal policy in these events is consistent with evidence that fiscal policy is more effective when economic agents face tighter liquidity constraints (Tagkalakis, 2008).35 The lack of a statistically significant effect from monetary policy could be a result of the stress experienced by the financial sector during financial crises, which hampers the effectiveness of the interest-rate and bank-lending channels of the monetary policy transmission mechanism.36

Table 8.3Impact of Policies on the Probability of Exiting a Recession
(1)(2)(3)(4)
Recession associated with financial crisis1−1.275***−2.238***−0.454−1.391**
(0.381)(0.602)(0.612)(0.763)
Government consumption2−0.110***−0.131***
(0.027)(0.029)
Government consumption × financial crisis0.278**0.284**
(0.143)(0.139)
Real rate3−0.024***−0.033***
(0.008)(0.009)
Real rate × financial crisis−0.028−0.024
(0.031)(0.031)
Constant−3.224***−3.269***−3.571***−3.742***
(0.449)(0.459)(0.499)(0.514)
Ln p40.900***0.983***0.960***1.070***
(0.069)(0.069)(0.072)(0.072)
Fixed effectsYesYesYesYes
Number of observations121120117117
Source: Authors’ calculations.Note: The baseline hazard function is assumed to follow a Weibull distribution. Coefficient values of the individual covariates in the hazard function are reported. Standard errors are reported in parentheses. The symbols ***, **, and * indicate significance at the 1, 5, and 10 percent levels, respectively.

Recession associated with financial crisis is an indicator variable that takes on a value of 1 when the recession is identified as being related to a financial crisis as described in the text.

Government consumption refers to the change in discretionary government consumption during a recession.

Real rate refers to the cumulative deviations of real interest rates from a Taylor rule during a recession.

Ln p reports the value of the (logged) Weibull parameter that governs the shape of the hazard function.

Source: Authors’ calculations.Note: The baseline hazard function is assumed to follow a Weibull distribution. Coefficient values of the individual covariates in the hazard function are reported. Standard errors are reported in parentheses. The symbols ***, **, and * indicate significance at the 1, 5, and 10 percent levels, respectively.

Recession associated with financial crisis is an indicator variable that takes on a value of 1 when the recession is identified as being related to a financial crisis as described in the text.

Government consumption refers to the change in discretionary government consumption during a recession.

Real rate refers to the cumulative deviations of real interest rates from a Taylor rule during a recession.

Ln p reports the value of the (logged) Weibull parameter that governs the shape of the hazard function.

A useful way of visualizing the impact of monetary and fiscal policies on the duration of recessions is to look at estimates of the probability that an economy will stay in a recession beyond a certain number of quarters (Figure 8.12, panel a). The estimated probabilities are significantly higher for recessions associated with financial crises relative to the average recession, indicating that the former type lasts longer than the latter. The implementation of expansionary policies clearly helps reduce the median duration of the recession (Figure 8.12, panel b). For instance, a one-standard-deviation increase in government consumption reduces the median duration of a recession associated with financial crisis from 5.1 quarters to 4.1 quarters. In contrast, the effect of monetary policy, although still helping to reduce the duration of a recession associated with financial crisis, is insignificant.

Figure 8.12Impact of Policies during Financial Crisis Episodes

Source: Authors’ calculations.

Note: Recessions associated with financial crises, as described in the text.

1 Survivor functions show the probability of remaining in a recession beyond a certain number of quarters.

2 Refers to a one-standard-deviation increase in government consumption or decrease in real interest rates, respectively.

Do Policies Help Boost Recoveries?

As noted previously, recessions are typically followed by swift recoveries. Although, as discussed earlier, factors such as technological progress and population growth help the economy eventually recover, this section investigates whether fiscal and monetary policies undertaken during the recession also contribute to the strength of the economic recovery, using an event study to exploit the cross-country variation in the data. The variable of interest in this case is the cumulative output growth one year after the cyclical trough, which is used as a proxy for the strength of the recovery. An economy emerging from recession has typically surpassed its previous peak output by this time. The measures of policy used are the same as in the duration analysis, which were measured as cumulative changes during the recession phase. In addition to the policy variables, both the duration and amplitude of the preceding recession are included as controls.

The results suggest that both fiscal and monetary expansions undertaken during a recession are associated with stronger recoveries (Table 8.4). In particular, increases in government consumption and reductions in both nominal and real interest rates, beyond that implied by the Taylor rule, have a positive effect on the strength of economic recovery (Figure 8.13).37Table 8.4 shows the quantitative impact of each policy measure separately and in combination. The coefficient on the government consumption variable, which is about 0.2, implies that a one-standard-deviation increase in government consumption during a recession is associated with an increase in the cumulative growth rate during the recovery phase of about 0.7 percent. The response to a one-standard-deviation reduction in real interest rates, beyond that implied by the Taylor rule, is about 0.4 percent. However, changes in the cyclically adjusted primary balance during a recession are not significantly associated with output growth during recovery.38

Table 8.4Impact of Policies on the Strength of Recoveries
(1)(2)(3)(4)(5)(6)(7)(8)
Recession duration−0.0440.111−0.248−0.208−0.201*−0.056−0.406−0.342
(0.121)(0.126)(0.156)(0.211)(0.110)(0.144)(0.251)(0.286)
Recession0.1550.0920.446***0.426***0.415***0.353***0.358***0.323**
amplitude(0.116)(0.102)(0.082)(0.103)(0.069)(0.082)(0.117)(0.137)
Government0.201**0.173**0.252**0.236*
consumption1(0.080)(0.082)(0.119)(0.131)
Government−0.437**−0.415*
consumption × debt(0.186)(0.209)
Primary balance2−0.040−0.041−0.567**−0.575**
(0.070)(0.071)(0.247)(0.236)
Primary balance × debt1.029***1.056***
(0.354)(0.340)
Real rate3−0.035***−0.010−0.028*−0.015
(0.011)(0.025)(0.016)(0.025)
Public debt4−1.505**−1.468**−3.890***−3.755***
(0.647)(0.670)(0.797)(0.885)
Fixed effectsYesYesYesYesYesYesYesYes
Number of observations112109757596937272
R20.100.130.340.340.120.160.460.46
Source: Authors’ calculations.Notes: Dependent variable is the cumulative growth one year into the recovery phase. Robust standard errors clustered by country are reported in parentheses. The symbols ***, **, and * indicate significance at the 1, 5, and 10 percent levels, respectively.

Government consumption refers to the change in discretionary government consumption during the preceding recession.

Primary balance refers to the change in the cyclically adjusted primary balance during the preceding recession.

Real rate refers to the cumulative deviations of real interest rates from a Taylor rule during a recession.

Public debt refers to the ratio of public debt to GDP at the start of the recession.

Source: Authors’ calculations.Notes: Dependent variable is the cumulative growth one year into the recovery phase. Robust standard errors clustered by country are reported in parentheses. The symbols ***, **, and * indicate significance at the 1, 5, and 10 percent levels, respectively.

Government consumption refers to the change in discretionary government consumption during the preceding recession.

Primary balance refers to the change in the cyclically adjusted primary balance during the preceding recession.

Real rate refers to the cumulative deviations of real interest rates from a Taylor rule during a recession.

Public debt refers to the ratio of public debt to GDP at the start of the recession.

Figure 8.13Effect of Policy Variables on the Strength of Recovery

Source: Authors’ calculations.

Note: Scatter plots shown here are conditional plots that take into account the effect of several other controlling variables, as noted in Appendix 8A.

The aggressive use of discretionary fiscal policy raises concerns about the sustainability of public finances. For instance, Perotti (1999), using a sample of 19 OECD countries, finds that a fiscal stimulus reduces private consumption in periods during which the level of government debt is particularly high.39 Do concerns about fiscal sustainability detract from the effectiveness of fiscal stimulus during recoveries? To address this question, the levels of public debt relative to GDP that were prevalent at the beginning of the recession are introduced into the benchmark regression framework interacted with the proxy of fiscal policy. The results, shown in Table 8.4, suggest that the degree of public indebtedness reduces the effectiveness of fiscal policy. Thus, fiscal stimulus in economies that have low levels of public debt has a higher impact on the strength of the recovery relative to economies that have higher levels of public debt. For public debt levels in excess of 60 percent of GDP, the fiscal stimulus becomes ineffective and even detrimental; however, there is high uncertainty in the estimation of this threshold debt level.

These findings point to the need for a commitment to medium-term fiscal sustainability to accompany any short-term fiscal stimulus. Doubts about debt sustainability can slow the recovery process through lower consumer spending and higher long-term real interest rates. To ensure policy effectiveness, it is crucial that the implementation of temporary stimulus measures occur in a framework that guarantees fiscal sustainability.40

This section has focused on fiscal and monetary policy; however, previous experiences of recessions associated with financial crises strongly suggest that the effectiveness of monetary and fiscal policies is substantially reduced without the implementation of prompt and well-targeted financial policies. Many observers consider the policies undertaken by Sweden in the early 1990s to have been highly effective in restoring the health of the financial sector, paving the way for strong recovery (Jackson, 2008, and references therein). A key component of those measures was the establishment of independent asset management companies, which removed bad assets from the balance sheets of banks so that the banks could resume normal lending activities. In Japan, slow recognition of the extent of the bad-loan problem contributed to the slow recovery from the financial crises of the 1990s (see, for instance, Hoshi and Kashyap, 2008).

Financial sector support typically has fiscal costs. However, a substantial part of the up-front gross cost is usually recovered, through asset sales, in the medium term. For example, in the Scandinavian countries and in Japan, the gross cost of recapitalization averaged some 5 percent of GDP, whereas the average recovery rate in the first five years was about 30 percent.41 The speed of the economic recovery and associated improvement in financial conditions are important factors in determining the recovery rate. In Sweden, for example, more than 90 percent of the initial outlay was recovered within the first five years. The equivalent rate for the Japanese recession in the late 1990s, however, was just about 10 percent; it reached almost 90 percent by 2008.

Lessons for The Recession Following The 2007–09 Crisis and Prospects for Recovery

Data through the fourth quarter of 2008 indicated that 15 of the 21 advanced economies considered in this chapter were in recession. Based on output turning points, Ireland was in decline for seven quarters; Denmark for five; Finland, New Zealand, and Sweden for four; Austria, Germany, Italy, Japan, the Netherlands, and the United Kingdom for three; and Portugal, Spain, Switzerland, and the United States for two (although the U.S. recession was already four quarters old at the end of 2008 using NBER dating).42 This section looks at the prospects for recovery from these recessions in light of the findings of this chapter.

Many of the economies in recession after the global crisis saw expansions that closely resembled those preceding previous episodes of financial stress, as discussed in the chapter, exhibiting similarly overheated asset prices and rapid expansions in credit.43 At end-2008, there were clear signs that, consistent with previous experiences of financial stress (IMF, 2008b), these recessions were already more severe and longer than usual. Figure 8.14 plots median growth rates of key macroeconomic variables for all 122 previous recessions, along with upper and lower quartile bands. Overlaid on each are data for the latest U.S. recession and the median for all other concurrent recessions.44 GDP data indicate that these economies had been deteriorating at a relatively rapid pace. In particular, declines in goods, labor, and asset markets in the United States were steep. Three aspects of these developments are especially notable.

Figure 8.14Economic Indicators around Peaks of Current and Previous Recessions

Source: Authors’ calculations.

Note: Median log differences from one year earlier unless noted otherwise. Peak in output at t = 0. Data in real terms unless noted otherwise.

1 Median percentage point difference from one year earlier.

First, there was evidence of negative feedback between asset prices, credit, and investment, which, as seen in the previous sections, is common in severe recessions associated with financial crises. The evidence showed exceptional reductions in credit. The deterioration in financial wealth, as represented by equity prices, was sharp. The decline in U.S. house prices was as steep as those in the Big Five episodes discussed previously. Residential investment clearly showed exceptional declines compared with previous recessions.

Second, the evidence indicated that the sharp falls in household wealth seen in several economies and the need to rebuild household balance sheets would result in larger-than-usual declines in private consumption. The reduction in U.S. consumption in the most recent quarters was clearly atypical. Consumer confidence in all economies had been steadily weakening, suggesting that declines in private demand and confidence would make for a protracted recovery.

Finally, the recessions following the global crisis were also highly synchronized, further dampening prospects of a normal recovery. In particular, the rapid drop in consumption in the United States represented a large decline in external demand for many other economies.

Hence, it seemed unlikely that overleveraged economies would be able to bounce back quickly via strong growth in domestic private demand—fundamentally, a prolonged period of above-average saving would have been required. In many previous cases of banking system stress, net exports led the recovery, facilitated by robust demand from the United States and by exchange rate depreciations or devaluations. But that option would not be available in this episode, given the unique stress at the heart of the world’s largest economy.

Given the likely shortfalls in both domestic private demand and external demand, policy needed to be used to arrest the cycle of falling demand, asset prices, and credit. Monetary policy had been loosened quickly in most advanced economies, much more so than in previous recessions, and extraordinary measures had been taken to provide liquidity to markets. Further effective easing was possible, even as nominal interest rates approached zero. However, evidence from the analysis indicates that interest rate cuts would likely have less of an impact during a financial crisis. In view of the continued distress in the financial sector, authorities would have been wise not to rely solely on standard policy measures.

The evidence in this chapter shows that fiscal policy can make a significant contribution to reducing the duration of recessions associated with financial crises. In effect, governments can break the negative feedback between the real economy and financial conditions by acting as “spender of last resort.” But this presupposes that public stimulus can be delivered quickly. Moreover, as the chapter shows, the sustainability of the eventual debt burden constrains the scope of expansionary fiscal policy, and it will not be possible to support demand for an extended period in economies that have entered recession with weak fiscal balances and large levels of public debt. In the event of severe and prolonged recessions during which deflation is an important risk, fiscal and monetary policies should be tightly coordinated to contain downward demand pressures. Furthermore, given the globally synchronized nature of the recession following the 2007–09 crisis, fiscal stimulus should be provided by a broad range of countries with fiscal room to do so, so as to maximize the short-term impact on global economic activity.

Restoring the health of the financial sector is an essential component of any policy package.45 Experiences with previous financial crises—especially those involving deleveraging, such as in Japan in the 1990s—strongly signal that coherent and comprehensive action to restore financial institutions’ balance sheets, and to remove uncertainty about funding, is required before a recovery will be feasible. Even then, recovery is likely to be slow and relatively weak.

Appendix 8A. Data Sources and Methodologies

This appendix provides details on the data and briefly reviews the methodologies used to identify “large shocks” and discretionary fiscal and monetary policies. The appendix also reports robustness exercises on the measure of fiscal policy.

Data Sources

The main data source for this chapter is Claessens, Kose, and Terrones (2008), from here on denoted as CKT. Other sources are listed in Table 8A.1.

Table 8A.1Data Sources
VariableSource
OutputCKT, Haver Analytics
Real private consumptionCKT, Haver Analytics
Real government consumptionCKT, Haver Analytics
Real private capital investmentCKT
Real residential investmentCKT, Haver Analytics
Real exportsCKT
Real net exportsOrganization for Economic Cooperation and Development (OECD) Analytical Database
GDP deflatorOECD Analytical Database
Consumer price indexCKT, IMF International Financial Statistics (IFS) database
Oil pricesIMF Primary Commodity Prices database
Real house pricesCKT, Bank for International Settlements (BIS), OECD
Stock pricesCKT, IFS database
CreditCKT, IFS database
Nominal interest rateCKT, IFS database, Thomson Datastream
Unemployment rateCKT, Haver Analytics
Labor force participation rateOECD Analytical Database
Nominal wagesIFS database, OECD Analytical Database
House-price-to-rental ratioOECD
Household saving rateOECD Analytical Database
Household net lendingOECD Analytical Database
Public debtInternational Monetary Fund
Note: House prices; stock prices, credit, and interest rates are deflated using consumer price indices.
Note: House prices; stock prices, credit, and interest rates are deflated using consumer price indices.

Methodology Used to Categorize Recessions and Recoveries

The statistical rules for the nonfinancial shocks identify large changes in macroeconomic variables, as follows:

  • Oil shocks. An indicator of oil price movements records, at a given date and for each country, the maximum change in nominal local oil prices in the preceding 12 quarters.46 Oil shocks are defined as those in which the indicator is greater than the mean plus 1.75 standard deviations of this index.

  • External demand shocks. The indicator of external demand is constructed as percentage deviations from trend of the trade-weighted GDP for each economy.47 External demand shocks are defined as those in which the indicator is less than the mean minus 1.75 standard deviations of the indicator.

  • Fiscal policy shocks. For the indicator of discretionary fiscal policy, a measure of the cyclically adjusted primary balance is constructed.48 Fiscal contractions are those in which the year-over-year difference of the cyclically adjusted primary balance is greater than the mean plus 1.75 standard deviations of the cyclically adjusted primary balance.49

  • Monetary policy shocks. For the indicator of discretionary monetary policy, the residuals from estimated Taylor rules are employed. Monetary policy contractions are those in which the residual is greater than 1.75 standard deviations. The analysis also examines term spreads (the difference between yields on 3-month government bills and 10-year government bonds), recording as contractionary those instances in which the spread is greater than 1.75 standard deviations above trend.

  • The next step is to associate recessions with these shocks. A shock in the four quarters preceding a peak in GDP is attributed one point for correctly calling the downturn ahead. This leads to the results in Table 8A.2. Finally, Table 8A.3 provides some evidence on the association between financial crises and the deregulation of mortgage markets.

Table 8A.2Results from Categorizing Recessions
NumberPercent
Episodes with positive overall “pre-peak” scores (total of all indicators; at least one indicator is > 0 during pre-peak period)5646
Episodes with scores greater than zero (by indicator)
Oil2319
External demand65
Fiscal policy87
Monetary policy1512
Financial crisis1512
Number of recessions with positive “pre-peak” score by country and type of shock
Number of recessionsOilExternal demandFiscal policyMonetary policyFinancial crisis
Australia601011
Austria611010
Belgium710120
Canada310010
Denmark710111
Finland500201
France420101
Germany820021
Greece820211
Ireland300000
Italy910001
Japan300002
Netherlands521020
New Zealand1211011
Norway310011
Portugal411110
Spain410001
Sweden311001
Switzerland910000
United Kingdom520002
United States620010
Source: Authors’ calculations.
Source: Authors’ calculations.
Table 8A.3Financial Crises and Deregulation in the Mortgage Market
CountryYearMeasure
Australia1986Removal of ceiling on mortgage interest rates
Denmark1982Liberalization of mortgage contract terms; deregulation of interest rates
Finland1986-87Deregulation of interest rates; removal of guidelines on mortgage lending
France1987Elimination of credit controls
Germany1967Deregulation of interest rates
Italy1983-87Deregulation of interest rates; elimination of credit ceilings
Japan1993-94Reduction of bank specialization requirements; deregulation of interest rates
New Zealand1984Removal of credit allocation guidelines; deregulation of interest rates
Norway1984-85Abolition of lending controls; deregulation of interest rates
Sweden1985Abolition of lending controls for banks; deregulation of interest rates
United Kingdom1980-86Credit controls eliminated; banks allowed to compete with building societies for housing finance; building societies allowed to expand lending activities; guidelines on mortgage lending removed

Methodology Used to Identify Fiscal and Monetary Policies

Two measures of fiscal policy are used: cyclically adjusted government consumption and cyclically adjusted primary balances. In instances in which only one measure is discussed or presented, it is cyclically adjusted government consumption. In all cases, changes in policy are measured as changes in the respective variable from the peak of a particular cycle to the trough.

The cyclically adjusted primary balance is computed using OECD elasticities on the different tax and expenditure components. For government consumption, however, such elasticities are not readily available and thus have to be estimated. The elasticity of government consumption with respect to the business cycle is computed as follows:

in which gct is government consumption at time t, gapt is a measure of the output gap at time t (“potential output” is measured using the Hodrick-Prescott [H-P] filter) and trend is a time trend. In estimating the equation above, the lagged value of the output gap is used as an instrument. Cyclically adjusted government consumption (cagct) is then computed as

Two measures of monetary policy are used: nominal and real interest rates. Both of these variables are measured as deviations from a “policy rule.” When only one measure is used, it is the real rate. The policy response over the course of a recession is measured as the sum of the impulse relative to the policy rule for each quarter of the recession period. A policy rule of the following form is estimated:

in which it is the nominal interest rate, dummy_85 is a dummy for periods after 1985 (to allow for a shift in the equilibrium real rate), πt is the inflation rate, and gapt is a measure of the output gap (“potential GDP” is measured using the H-P filter). The measure of monetary policy that is used in the analysis is

in which î is the fitted value of the regression.

Real rates are simply measured as it – πt, and the steps taken to get the measure of monetary policy are the same as above.

Robustness Test Using Government Consumption as A Proxy for Fiscal Policy

Apart from the two measures of fiscal policy presented in the chapter, the same set of regressions were also run using changes in real government consumption during preceding recessions, without any cyclical adjustment. Table 8A.4 contains the results of regressions using the alternative measure of fiscal policy. Although most of the main results in the chapter are preserved, the interaction term with public debt is statistically significant only at the two- and three-quarter horizon during the recovery phase. The limitations of the data may be one possible cause.

Table 8A.4Impact of Policies on the Strength of Recoveries Using an Alternative Measure of Fiscal Policy
Dependent variable
Cumulative growth four quarters into recovery phaseCumulative growth three quarters into recovery phase
(1)(2)(3)(4)(5)(6)(7)(8)
Recession duration−0.027−0.209−0.1790.090−0.076−0.0400.0150.009
(0.110)(0.194)(0.217)(0.123)(0.092)(0.145)(0.174)(0.107)
Recession0.203**0.439***0.421***0.154*0.217*0.283***0.254**0.176**
amplitude(0.083)(0.080)(0.096)(0.086)(0.085)(0.093)(0.103)(0.077)
Government0.289***0.2030.1770.269**0.261***0.489***0.414***0.229***
consumption1(0.088)(0.157)(0.178)(0.098)(0.042)(0.129)(0.117)(0.050)
Public debt2−2.066**−2.047**−0.801−0.807
(0.829)(0.851)(0.672)(0.694)
Government−0.224−0.200−0.714***−0.638***
consumption × debt(0.285)(0.302)(0.180)(0.175)
Real rate3−0.009−0.026*−0.022−0.022*
(0.026)(0.013)(0.018)(0.012)
Fixed effectsYesYesYesYesYesYesYesYes
Number of observations11275751091178080114
R20.120.330.330.140.140.400.420.15
Source: Authors’ calculations.Note: Robust standard errors clustered by country are reported in parentheses. The symbols ***, **, and * indicate significance at the 1, 5, and 10 percent levels, respectively.

Government consumption refers to the change in government consumption during preceding recessions.

Public debt refers to the ratio of public debt to GDP at the start of the recession.

Real rate refers to the cumulative deviations of real interest rates from a Taylor rule during a recession.

Source: Authors’ calculations.Note: Robust standard errors clustered by country are reported in parentheses. The symbols ***, **, and * indicate significance at the 1, 5, and 10 percent levels, respectively.

Government consumption refers to the change in government consumption during preceding recessions.

Public debt refers to the ratio of public debt to GDP at the start of the recession.

Real rate refers to the cumulative deviations of real interest rates from a Taylor rule during a recession.

References

    BernankeBen S. and MarkGertler1989Agency Costs, Net Worth, and Business FluctuationsAmerican Economic Review Vol. 79 (March) pp. 1431.

    BernankeBen S. and MarkGertler1995Inside the Black Box: The Credit Channel of Monetary Policy TransmissionJournal of Economic Perspectives Vol. 9 (Autumn) pp. 2748.

    BlanchardOlivier and RobertoPerotti2002An Empirical Characterization of the Dynamic Effects of Changes in Government Spending and Taxes on OutputQuarterly Journal of Economics Vol. 107 (November) pp. 132968.

    BlanchardOlivier and JohnSimon2001The Long and Large Decline in U.S. Output VolatilityBrookings Papers on Economic Activity Vol. 32 No. 1 pp. 13574.

    BrunnermeierMarcus2009Deciphering the 2007–08 Liquidity and Credit CrunchJournal of Economic Perspectives Vol. 23 No. 1 pp. 77100.

    BurnsArthur F. and Wesley C. Mitchell1946Measuring Business Cycles (New York: National Bureau of Economic Research).

    ChauvetMarcelle and James D. Hamilton2005Dating Business Cycle Turning PointsNBER Working Paper No. 11422 (Cambridge, Massachusetts: National Bureau of Economic Research).

    ClaessensStijnM. Ayhan Kose and Marco E. Terrones2008What Happens during Recessions, Crunches, and Busts?IMF Working Paper 08/274 (Washington: International Monetary Fund).

    DebelleGuy2004Macroeconomic Implications of Rising Household DebtBIS Working Paper No. 153 (Basel: Bank for International Settlements).

    DecressinJörg and DouglasLaxton2009Gauging Risks for DeflationIMF Staff Position Note 09/01 (Washington: International Monetary Fund).

    DieboldFrancis and GlennRudebusch1990A Nonparametric Investigation of Duration Dependence in the American Business CycleJournal of Political Economy Vol. 98 pp. 596616.

    DieboldFrancis and GlennRudebusch and DanielSichel1993Further Evidence on Business Cycle Duration Dependence” in Business Cycles Indicators and Forecasting ed. by James H. Stock and Mark W. Watson (Chicago: University of Chicago Press).

    EstrellaArturo and Frederic S. Mishkin1996The Yield Curve as a Predictor of U.S. RecessionsCurrent Issues in Economics and Finance Vol. 2 No. 7 pp. 16.

    GavinMichael and RobertoPerotti1997Fiscal Policy in Latin AmericaNBER Macroeconomics Annual Vol. 12 pp. 1172.

    GreenlawDavidJanHatziusAnilKashyap and HyunSong Shin2008Leveraged Losses: Lessons from the Mortgage Market MeltdownProceedings of the 2008 Monetary Policy ForumFeb. 29New York.

    HamiltonJames D.2003What Is an Oil Shock?Journal of Econometrics Vol. 113 (April) pp. 36398.

    HardingDon and AdrianPagan2002Dissecting the Cycle: A Methodological InvestigationJournal of Monetary Economics Vol. 49 No. 2 pp. 36581.

    HellerPeter S. Richard D. Haas and Ahsan S. Mansur1986A Review of the Fiscal Impulse MeasureIMF Occasional Paper No. 44 (Washington: International Monetary Fund).

    HoshiTakeo and Anil K. Kashyap2008Will the U.S. Bank Recapitalization Succeed? Lessons from JapanNBER Working Paper No. 14401 (Cambridge, Massachusetts: National Bureau of Economic Research).

    International Monetary Fund (IMF)2002World Economic OutlookApril (Washington).

    International Monetary Fund (IMF)2008aGlobal Financial Stability Report: Financial Stress and Deleveraging—Macro-Financial Implications and PolicyOctober (Washington).

    International Monetary Fund (IMF)2008bWorld Economic OutlookOctober (Washington).

    JacksonJames2008The U.S. Financial Crisis: Lessons from SwedenCRS Report for Congress (Washington: Congressional Research Service).

    KaminskyGraciela L. and Carmen M. Reinhart1999The Twin Crises: The Causes of Banking and Balance of Payments ProblemsAmerican Economic Review Vol. 89 No. 4 pp. 473500.

    KilianLutz2006Not All Oil Price Shocks Are Alike: Disentangling Demand and Supply Shocks in the Crude Oil MarketCEPR Discussion Paper No. 5994 (London: Centre for Economic Policy Research).

    KindlebergerCharles1978Manias Panics and Crashes: A History of Financial Crises (Hoboken, New Jersey: John Wiley & Sons).

    LaevenLuc and FabianValencia2008aSystemic Banking Crises: A New DatabaseIMF Working Paper 08/224 (Washington: International Monetary Fund).

    LaevenLuc and FabianValencia2008bThe Use of Blanket Guarantees in Banking CrisesIMF Working Paper 08/250 (Washington: International Monetary Fund).

    LanePhillip2003The Cyclical Behaviour of Fiscal Policy: Evidence from the OECDJournal of Public Economics Vol. 87 (December) pp. 266175.

    LeamerEdward2008What’s a Recession, Anyway?NBER Working Paper No. 14221 (Cambridge, Massachusetts: National Bureau of Economic Research).

    McConnellMargaret and GabrielPerez-Quiros2000Output Fluctuations in the United States: What Has Changed Since the Early 1980s?American Economic Review Vol. 90 No. 5 pp. 146476.

    MendozaEnrique and Marco E. Terrones2008An Anatomy of Credit Booms: Evidence from Macro Aggregates and Micro DataNBER Working Paper No. 14049 (Cambridge, Massachusetts: National Bureau of Economic Research).

    PeersmanGert and FrankSmets2001Are the Effects of Monetary Policy Greater in Recessions than in Booms?ECB Working Paper No. 52 (Frankfurt: European Central Bank).

    PerottiRoberto1999Fiscal Policy in Good Times and BadQuarterly Journal of Economics Vol. 114 (November) pp. 1399436.

    RameyValerie A.2009Identifying Government Spending Shocks: It’s All in the TimingNBER Working Paper No. 15464 (Cambridge, Massachusetts: National Bureau of Economic Research).

    ReinhartCarmen and KennethRogoff2008aBanking Crises: An Equal Opportunity MenaceNBER Working Paper No. 14587 (Cambridge, Massachusetts: National Bureau of Economic Research).

    ReinhartCarmen and KennethRogoff2008bIs the 2007 U.S. Sub-Prime Crisis So Different? An International Historical ComparisonNBER Working Paper No. 13761 (Cambridge, Massachusetts: National Bureau of Economic Research).

    ReinhartCarmen and KennethRogoff2009The Aftermath of Financial CrisesNBER Working Paper No. 14656 (Cambridge, Massachusetts: National Bureau of Economic Research).

    RomerChristina1999Changes in Business Cycles: Evidence and ExplanationsJournal of Economic Perspectives Vol. 13 No. 2 pp. 2344.

    RomerChristina and DavidRomer1989Does Monetary Policy Matter? A New Test in the Spirit of Friedman and Schwartz” in NBER Macroeconomics Annual Vol. 4 ed. by OlivierJean Blanchard and StanleyFischer (Cambridge, Massachusetts: National Bureau of Economic Research).

    RomerChristina and DavidRomer1994What Ends Recessions?” in NBER Macroeconomics Annual 1994 ed. by StanleyFischer and JulioRotemberg (Cambridge, Massachusetts: MIT Press).

    RomerChristina and DavidRomer2007The Macroeconomic Effects of Tax Changes: Estimates Based on a New Measure of Fiscal ShocksNBER Working Paper No. 13264 (Cambridge, Massachusetts: National Bureau of Economic Research).

    SichelDaniel1994Inventories and the Three Phases of the Business CycleJournal of Business and Economic Statistics Vol. 12 (July) pp. 26977.

    SpilimbergoAntonioSteveSymanskyOlivierBlanchard and CarloCottarelli2008Fiscal Policy for the CrisisIMF Staff Position Note 08/01 (Washington: International Monetary Fund).

    TagkalakisAthanasios2008The Effects of Fiscal Policy on Consumption in Recessions and ExpansionsJournal of Public Economics Vol. 92 (June) pp. 1486508.

    WynneMark A. and Nathan S. Balke1993Recessions and RecoveriesEconomic Review Federal Reserve Bank of Dallas (First Quarter).

For detailed accounts of the financial aspects of this crisis, see IMF (2008), Greenlaw and others (2008), and Brunnermeier (2009).

A classic analysis of financial crises is Kindleberger (1978). Reinhart and Rogoff (2008a) show that financial crises have occurred with “equal opportunity” in advanced and less advanced economies.

In particular, this work builds on IMF (2002, 2008b) and Claessens, Kose, and Terrones (2008).

The sample includes the following countries: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Japan, the Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, the United Kingdom, and the United States.

The procedure used to date business cycles in this chapter has been referred to as BBQ (Bry-Boschen procedure for quarterly data; see Harding and Pagan, 2002). It identifies local maximums and minimums of a given series, here the logarithm of real GDP, that meet the conditions for a minimum duration of a cycle and of each phase (in this chapter, these are set at five and two quarters, respectively). Alternative dating algorithms, such as those developed by Chauvet and Hamilton (2005) and Leamer (2008), are more difficult to implement for a large sample of countries. The National Bureau of Economic Research (NBER), which dates business cycles in the United States, uses several measures of economic activity to determine peaks and troughs. These measures include—in addition to real GDP—employment, real income, industrial production, and sales. NBER dating is, however, subjective and not replicable internationally.

In the sample period, there were 122 completed and 13 ongoing recessions at the time of the writing of this chapter in 2009.

Related findings are reported in the April 2002 World Economic Outlook.

This phenomenon has been documented in several papers, including McConnell and Perez-Quiros (2000) and Blanchard and Simon (2001). During this period the average slope of a recession—a proxy for how steep or abruptly output contracts—is about –0.6 percent, which is lower in absolute value than the average –1.0 percent for other recession periods.

There is no common definition of recovery. Some define it as the time it takes for the economy to return to the peak level before the recession; others measure it by the cumulative growth achieved after a certain period, say a year, following the trough. In this chapter, both definitions are used. These two definitions are complementary and display a sort of duality—the first one determines the time it takes to achieve a given amplitude, and the second one determines the amplitude observed after a given time.

Wynne and Balke (1993) and Sichel (1994) provide evidence of a bounce-back effect in U.S. business cycles. Romer and Romer (1994) report that monetary policy has been instrumental in ending U.S. recessions and helping recoveries during the postwar period.

Term spreads, which have often been used as an indicator of monetary policy stance and as a predictor of short-term output growth—see, for example, Estrella and Mishkin (1996)—were also analyzed and found to give results very similar to those for monetary policy shocks.

An alternative method of defining financial crises is to use a time series or some combination of series as an indicator, based on some threshold (the method used for the other shocks). An advantage of using a narrative-based method is that it avoids having to define episodes according to characteristics in the very factors one is interested in—for example, a financial crisis could be defined as an episode in which there is a large reduction in credit, but that would preclude assessing the behavior of credit during and following financial crises.

Banking crises, which are defined by Kaminsky and Reinhart (1999, p. 476) as episodes leading to bank runs or large-scale government assistance to financial institutions, are of particular interest.

On these grounds, Reinhart-Rogoff episodes not immediately associated with recessions—for example, the savings and loan crisis of the early 1980s in the United States—are omitted.

In principle, there is a potential endogeneity problem here, because the financial crisis could lead to a recession and vice versa. To address this issue, the dating of crises and cyclical turning points is done using two different methods, as explained in the chapter.

These rules have the advantage of being transparent and of being easily and consistently applied to the GDP series for the 21 countries in the sample. There will always be cases that are not well identified by simple rules. However, a more thorough analysis of the nonfinancial shocks for each country is outside the scope of this chapter.

The scores often coincide, with 105 scores for the 65 recessions that are associated with these shocks, which indicates how misleading it can be to talk about a recession as a result of a single “cause.”

Overall, oil shocks typically lead to recessions that are very costly but relatively short lived. This is particularly true of the 1973–74 oil shocks, after which GDP growth bounced back relatively quickly.

The Big Five financial crises episodes include Finland (1990–93), Japan (1993), Norway (1988), Spain (1978–79), and Sweden (1990–93).

Recessions and recoveries are clearly different in their severity depending on the type of shock they are associated with. But, for the same shock, they are also roughly symmetric—the slope of the recession phase is closely matched by the slope of the recovery phase.

For a comprehensive analysis of credit booms in the advanced and emerging market economies, see, for instance, Mendoza and Terrones (2008).

For example, Table 8A.3 in the appendix shows that almost all of the 15 financial crises considered here followed deregulation in the mortgage market.

Unfortunately, comprehensive balance sheet data are not available for most of the financial crisis episodes. But, as an example, analysis of data for the United Kingdom shows a pronounced deterioration in the ratio of total household liabilities to liquid assets in the years before the recession of 1990–91, with a gradual recovery in the quality of household balance sheets during and after the recession.

In fact, real GDP growth rates before recessions associated with financial crises have not been exceptionally high compared with those before other recessions. Similarly, the relationship between the average level of the output gap in the four quarters before the peak and the output loss in the ensuing recession is positive, but financial crises do not stand out.

Alternatively, synchronized recessions could be defined as recession events whose peaks coincide within a given time window, say a year. The results reported in the text are robust to this definition.

Note that the global crisis recessions were excluded from this analysis. Almost one-third of all recessions were highly synchronized.

In these two recessions, U.S. imports fell by 11 percent and 14 percent, respectively. In the other five U.S. recessions, imports contracted by 3 percent, on average. These cases are picked up as recessions associated with external demand shocks for some countries, but not all, owing to the threshold that the identification imposes (see Appendix 8A).

See Chapter 5 of IMF (2008) for a summary. See also Blanchard and Perotti (2002), Romer and Romer (2007), and Ramey (2009) for recent attempts at identifying the impact of discretionary fiscal policy.

To check for the robustness of these results, an alternative measure of fiscal policy is also used. This measure—the percentage change in non–cyclically adjusted real government consumption—is based on the premise that changes in real government expenditures are largely independent of the cyclical fluctuations in output. As discussed in Appendix 8A, most of the results are preserved. Public investment spending would have been another option. However, its size is much smaller than that of government consumption, and its association with economic recovery is often limited, owing to significant implementation lags (Spilimbergo and others, 2008).

Details are presented in Appendix 8A. For the measures of monetary policy, the policy stimulus is computed as the sum of the deviations in each quarter that the economy is in recession. Most empirical studies, including those cited previously, do not discriminate among the various phases of the business cycle. Exceptions include Peersman and Smets (2001) and Tagkalakis (2008), who show, respectively, that monetary policy and fiscal policy tend to have larger effects during recessions than during expansions

Lane (2003) finds that current government spending, excluding interest payments, is countercyclical for a sample of Organization for Economic Cooperation and Development (OECD) countries, though he claims that automatic stabilizers are the main driving force behind the countercyclicality.

Note that these figures show the measures of the discretionary component of policy. Direct measures of policy, such as changes in interest rates or the primary balance, show more marked reductions during recessions.

The advanced economies comprise Canada, France, Germany, Italy, Japan, the United Kingdom, and the United States.

Previous studies find that postwar recessions in the United States are more likely to end the longer they progress (Diebold and Rudebusch, 1990; and Diebold, Rudebusch, and Sichel, 1993).

Bernanke and Gertler (1989) suggest that liquidity constraints are more prevalent in recessions than expansions.

See Bernanke and Gertler (1995) for a detailed discussion of the credit channel of the monetary policy transmission mechanism.

This positive impact of policy continues to remain statistically significant even after policies that were undertaken in the early stages of recovery are included.

It should be noted that no systematic relationship was found linking monetary or fiscal policy with the strength of recoveries associated with financial crises.

The procyclicality of fiscal policy in emerging markets is also largely attributable to the fact that constraints on the financing of government debt are usually tighter during recessions (see Gavin and Perotti, 1997, for a discussion on Latin America).

See Spilimbergo and others (2008) for further details on the design of appropriate policies that address sustainability concerns. Reinhart and Rogoff (2008a) find that financial crisis episodes are often associated with sharp increases in the level of public debt, potentially raising concerns about medium-term debt sustainability. However, they do not examine the behavior of long-term interest rates following such crises.

This rate is relatively low compared with the 55 percent recovery rate that advanced economies typically experience from the sale of assets acquired through interventions. Detailed data on financial policy responses for several of the financial crisis episodes studied in this chapter are available in Laeven and Valencia (2008a, 2008b).

The NBER has declared that the peak in U.S. output before the 2007–09 crisis was in December 2007.

Notable exceptions include Germany and Japan, although their economies were also experiencing financial stress.

The calculation of the median is limited to at least four observations, which is why the series for recent recessions did not extend to six quarters at the time the analysis was undertaken.

See, for instance, Decressin and Laxton (2009) for a discussion of unconventional monetary policy options, fiscal policy, synergies with financial sector policy, and lessons from the experience of Japan.

This is a version of Hamilton’s (2003) proposed filter for identifying oil shocks in the United States. The local price is defined as the world average spot price in U.S. dollars times the nominal exchange rate for the country in question. In addition, results using year-over-year changes in real and nominal local-currency oil prices and vector-autoregression-based identifications of oil supply shocks were also examined (Kilian, 2006).

The trend is implemented using the Hodrick-Prescott (H-P) filter with λ set to 1,600. Two key assumptions are, first, that domestic absorption is well approximated by GDP, and second, that the trade weights are of the other advanced economies alone. Some economies have significant trade relationships with nonadvanced economies that have suffered sharp declines in demand (e.g., New Zealand exports to east Asia during 1997–98). Robustness to using terms of trade and world GDP has been explored.

This follows standard IMF methodology (Heller, Haas, and Mansur, 1986). The H-P (1,600) filter is used to estimate potential. OECD estimates of income elasticities for revenues and expenditures are used to construct measures of discretionary changes in the fiscal stance and to filter out passive changes from preset targets and automatic stabilizers. There are a number of important assumptions, notably that the H-P filter estimates potential output well; that the income elasticities of expenditures and revenues are constant; that revenue shares (used to construct aggregate income elasticity of revenues) are constant; and that the GDP deflator (used to deflate nominal government expenditures) is a good proxy for the true government expenditures deflator.

A positive value corresponds to fiscal tightening because the primary balance is defined as tax revenues minus expenditures.

    Other Resources Citing This Publication