This Time Is Different: Eight Centuries of Financial Folly Read online

  TABLE 10.7

  The effect of a capital flow bonanza on the probability of a banking crisis in a sixty-six country sample, 1960–2007

  Indicator

  Percentage of countries

  Probability of a banking crisis

  Conditional on a capital flow bonanza (three-year window)

  18.4

  Unconditional

  13.2

  Difference

  5.2*

  Share of countries for which the conditional probability is greater than the unconditional probability

  60.9

  Source: Reinhart and Reinhart (2009, tables 2 and 4), and authors’ calculations.

  Notes: The window encompasses three years before the bonanza (see Reinhart and Reinhart 2009, table 2), the year (or years if these are consecutive) of the bonanza, and the three years following the episode. The asterisk (*) denotes significance at the 1 percent confidence level.

  These findings on capital flow bonanzas are also consistent with other identified empirical regularities surrounding credit cycles. Mendoza and Terrones, who examine credit cycles in both advanced and emerging market economies using a very different approach from that just discussed, find that credit booms in emerging market economies are often preceded by surges in capital inflows. They also conclude that, although not all credit booms end in financial crisis, most emerging market crises were preceded by credit booms. They link credit booms to rising asset prices, an issue we turn to next.21

  Equity and Housing Price Cycles and Banking Crises

  In this section we summarize the literature on asset price bubbles and banking crises, extending it to incorporate new data on housing prices in emerging markets, as well as data on the crises that are currently unfolding in advanced economies.

  The now-infamous real estate bubble in the United States that began to deflate at the end of 2005 occupies center stage as a culprit in the recent global financial crisis. But the Second Great Contraction is far from unique in that regard. In an earlier work, we documented the trajectory in real housing prices around all the post–World War II banking crises in advanced economies, with particular emphasis on the “Big Five” crises (Spain, 1977; Norway, 1987; Finland and Sweden, 1991; and Japan, 1992).22 The pattern that emerges is clear: a boom in real housing prices in the run-up to a crisis is followed by a marked decline in the year of the crisis and subsequent years. Bordo and Jeanne, also studying the advanced economies during 1970–2001, found that banking crises tend to occur either at the peak of a boom in real housing prices or right after the bust.23 Gerdrup presented a compelling narrative of the links between Norway’s three banking crises from the 1890s through 1993 and the booms and busts in housing prices.24

  Table 10.8 illustrates the magnitudes and durations of the downturns in housing prices that have historically accompanied major banking crises in both advanced and emerging economies. Although the links between developed-country banking crises and the housing price cycle have been examined both in our earlier work and in numerous other papers (most frequently case studies), this is the first time systematic evidence has been provided on the behavior of housing prices in emerging market economies around some of their major banking crises. The crisis episodes include the “Big Six” Asian crises of 1997–1998 (Indonesia, Korea, Malaysia, the Philippines, Thailand, and the much-buffeted Hong Kong).

  Other episodes in emerging markets have included Argentina’s megacrisis in 2001–2002 and Colombia’s 1998 crisis, which produced the worst recession since the national income accounts began to be tabulated in the early 1920s. In the conjuncture of recent crises we include Hungary in addition to the advanced economies that have recently had housing market bubbles (Iceland, Ireland, Spain, the United Kingdom, and the United States).25

  TABLE 10.8

  Cycles of real housing prices and banking crises

  Two features stand out from the summary statistics presented in table 10.8. First is the persistence of the cycle in real housing prices in both advanced economies and emerging markets, typically for four to six years.26 The second feature that stands out from table 10.8 is that the magnitudes of the declines in real housing prices around banking crises from peak to trough are not appreciably different in emerging and advanced economies. This comparability is quite surprising given that most macroeconomic time series exhibit drastically greater volatility in emerging markets; therefore, it merits further attention.27 Certainly the first results presented here from comparisons of housing price booms and busts around the dates of banking crises appear to provide strong support for the contention that banking crises are an equal-opportunity menace.

  The prolonged housing price downturns following financial crises are in stark contrast to the behavior of real equity prices, as illustrated in figure 10.2, in which the pattern of decline and recovery is more V-shaped. (The figure shows only emerging markets, but, as we shall detail later in part V, equity prices exhibit a similar V-shaped recovery in advanced countries.)

  The figure shows the evolution of real equity prices from four years prior to a crisis to three years afterward. As the figure makes plain, equity prices typically peak before the year of a banking crisis and decline for two to three years as the crisis approaches and, in the case of emerging markets, in the year following the crisis. The recovery is complete in the sense that three years after the crisis, real equity prices are on average higher than at the precrisis peak. However, postcrisis Japan offers a sobering counterexample to this pattern, because in that country equity prices only marginally recovered to a much lower peak than the precrisis level and have subsequently continued to drift lower.

  Figure 10.2. Real equity prices and banking crises: Forty episodes in emerging markets, 1920–2007.

  Sources: Global Financial Data (n.d.) and the authors’ calculations.

  Notes: Four of the forty episodes were from before World War II (1921–1929). The year of the crisis is indicated by t; t − 4 = 100.

  One can conjecture that one reason major banking crises are such protracted affairs is that these episodes involve the real estate market’s persistent cycle in a way that “pure stock market crashes”—for instance, Black Monday in October 1987 or the bursting of the information technology (IT) bubble in 2001—do not.28

  Overcapacity Bubbles in the Financial Industry?

  Philippon has analyzed the expansion of the financial services sector (including insurance) in the United States, which averaged 4.9 percent of GDP during 1976–1985 and rose to 7.5 percent during 1996–2005.29 In his paper he argues that this gain was not sustainable and that a decline of at least 1 percent of GDP was probable. In the wake of the subprime crisis, the shrinkage of the financial sector during 2008 and 2009 is proving to be significantly larger. The precrisis explosion and postcollapse implosion of the financial sector surrounding a banking crisis are also not new or unique to the United States.

  Figure 10.3 plots the number of banks in the United States in the run-up to and the aftermath of the Great Depression. Perhaps the bubble in equity and real estate prices also extended to the number of financial institutions. This expansion in the number of financial institutions in the run-up to a crisis and contraction in its aftermath have been evident during other banking crises—especially in those cases in which financial liberalization preceded the crisis.

  The Fiscal Legacy of Financial Crises Revisited

  Looking at the fiscal and growth consequences of banking crises, we again find some surprising parallels between developed countries and emerging markets. Our analysis of the fiscal consequences, in particular, is a sharp departure from the previous literature, which has focused almost entirely on imputed “bailout costs” to the government, which, as we will argue, are extremely difficult to measure. Instead, we focus on the fiscal costs to the central government, particularly the huge buildup in debt that follows banking crises. We are able to do so by tapping the extensive new cross-country data set on annual domestic debt that underlie
s the research for this book, data we have already exploited in earlier chapters. These data allow us to show the remarkable surge in debt that occurs in the wake of a crisis.

  Figure 10.3. The number of banks in the United States, 1900–1945.

  Source: Carter et al. (2006).

  The Elusive Concept of Bailout Costs

  As we have noted, much of the literature on episodes of banking crisis focuses on estimating the ultimate fiscal costs of the bailouts (see, for example, an excellent discussion by Frydl and various papers published by Norges Bank).30 However, estimates of bailout costs vary markedly across studies, depending on the methodology, and vary even more across time, depending on the length of the horizon used to calculate the fiscal impact of the crisis, a point stressed by Frydl.31

  Table 10.9 presents the upper and lower bounds of estimates of the bailout costs for some of the better-known banking crises in both advanced and emerging economies in nearly all regions. The discrepancies across estimates are large and, in some cases, staggering. Among the “Big Five” crises in advanced economies since World War II, the differences in estimated bailout costs for Japan and Spain, for instance, are 16 and 11 percent of GDP, respectively. Furthermore, as noted by Vale, if the costs are calculated over a longer time horizon after the crisis, the picture that emerges is even more at odds with the higher-end estimates; it shows that the Norwegian government actually made a small profit on the banking resolution due to the later sale of shares in the nationalized banks.32

  TABLE 10.9

  Creative accounting? Bailout costs of banking crises

  In what follows, we argue that this nearly universal focus on opaque calculations of bailout costs is both misguided and incomplete. It is misguided because there are no widely agreed-upon guidelines for calculating these estimates. It is incomplete because the fiscal consequences of banking crises reach far beyond the more immediate bailout costs. These consequences mainly result from the significant adverse impact that the crisis has on government revenues (in nearly all cases) and the fact that in some episodes the fiscal policy reaction to the crisis has also involved substantial fiscal stimulus packages.

  Growth in the Aftermath of Crises

  The fact that most banking crises, especially systemic ones, are associated with economic downturns is well established in the empirical literature, although the effects on some key variables, such as housing and government debt and fiscal finances, more broadly, are much less studied.33 Figure 10.4 shows output for the advanced economies as a group, as well as those that have experienced the “Big Five” crises (Japan, the Nordic countries, and Spain), while figure 10.5 augments this analysis with a comparable summary of the postwar banking crises in emerging markets. As before, t denotes the year of the crisis. Interestingly, the figures show a steeper decline but a somewhat faster comeback in growth for emerging markets than for the advanced economies. It is beyond the scope of this book to ascertain the longer-run growth consequences of banking crises (it is too difficult to delineate the end of banking crises, and growth is simply too complex a subject to mix in here). Nevertheless, this postcrisis pattern is noteworthy because growth (important in its own right) has nontrivial implications for fiscal balances, government debt, and the broader cost and consequences of any financial crisis.

  Figure 10.4. Real GDP growth per capita (PPP basis) and banking crises: Advanced economies.

  Sources: Maddison (2004); International Monetary Fund (various years), World Economic Outlook; Total Economy Database (2008), and the authors’ calculations.

  Notes: Episodes of banking crisis are listed in appendix A.3. The year of the crisis is indicated by t.

  Figure 10.5. Real GDP growth per capita (PPP basis) and banking crises: Emerging market economies (112 episodes).

  Sources: Maddison (2004); International Monetary Fund (various years), World Economic Outlook; Total Economy Database; and the authors’ calculations.

  Notes: Episodes of banking crisis are listed in appendixes A.3 and A.4. The year of the crisis is indicated by t.

  Beyond Bailout Costs: The Impact of a

  Crisis on Revenues and Debt

  Since World War II the most common policy response to a systemic banking crisis (in both emerging and advanced economies) has been to engineer (with varying degrees of success) a bailout of the banking sector, whether through purchases of bad assets, directed mergers of bad banks with relatively sound institutions, direct government takeovers, or some combination of these. In many cases such actions have had major fiscal consequences, particularly in the early phases of the crisis. However, as we have emphasized repeatedly, banking crises are protracted affairs with lingering consequences in asset markets—notably real estate prices and the real economy. It is no surprise, then, that government revenues are adversely and significantly impacted by crises.

  As noted, several studies have traced the adverse impacts of banking crises on economic activity; what these studies have left unexplored are the direct consequences of the recession on government finances—specifically, tax revenues. Figure 10.6 plots the average pattern in annual real revenue growth three years before, during, and three years after a crisis for a total of eighty-six banking crises during 1800–1944 for which we have complete revenue data.34

  Figure 10.6. Real central government revenue growth and banking crises: All countries, 1800–1944.

  Sources: Revenues are from Mitchell (2003a, 2003b). For the numerous country-specific sources of prices, see Reinhart and Rogoff (2008a).

  Notes: The figure shows that the toll on revenue from crises is not new. Central government revenues are deflated by consumer prices. There were a total of eighty-six episodes of banking crisis during 1800–1940 for which we have revenue data. The year of the crisis is indicated by t.

  A comparable exercise is shown in figure 10.7 for all 138 banking crises since World War II. The patterns of the pre- and postwar samples have not been identical but have been strikingly similar. Annual revenue growth was robust in the years leading up to the banking crisis, weakened significantly in the year of the crisis, and subsequently posted declines in the years immediately following the onset of the crisis. For the prewar episodes, revenues declined on average for two years, while for the postwar crises the revenue slump has extended to the third year.

  Parallels in Revenue Losses between

  Emerging Markets and Developed Economies

  Again, the parallels in revenue losses between developed countries and emerging markets have been striking. Figure 10.8 shows the revenue declines surrounding banking crises for the advanced countries across the entire sample, with the “Big Five” postwar crises listed separately. Generally revenue growth resumes (from a lower base) starting in the third year after a crisis. Advanced economies exhibit a stronger inclination to resort to stimulus measures to cushion economic activity, as seen most spectacularly in the aggressive use of infrastructure spending in Japan during the 1990s. Emerging markets, more debt intolerant and more dependent on the vagaries of international capital markets for financing, are far less well poised to engage in countercyclical fiscal policy. Nevertheless, the effect of a crisis on the trajectory of taxes is broadly similar between the types of countries. Figure 10.9 shows revenue declines around banking crises for emerging markets for the entire sample. The average revenue drop is actually quite similar to that of the “Big Five” crises, although the recovery is faster—in line with a swifter recovery in growth, as discussed in the preceding section.

  Figure 10.7. Real central government revenue growth and banking crises: All countries, 1945–2007.

  Sources: Revenue information is taken from Mitchell (2003a, 2003b). For the numerous country-specific sources of prices, see Reinhart and Rogoff (2008a).

  Notes: The figure shows that bailout costs are only part of the story of why public debt surges after a crisis. Central government revenues are deflated by consumer prices. There were a total of 138 banking crises during 1945–2008 for which we
have revenue data. The year of the crisis is indicated by t.

  Figure 10.8. Real central government revenue growth and banking crises: Advanced economies, 1815–2007.

  Sources: Revenue information is taken from Mitchell (2003a, 2003b). For the numerous country-specific sources of prices, see Reinhart and Rogoff (2008a).

  Notes: Central government revenues are deflated by consumer prices. The year of the crisis is indicated by t.

  Government Debt Buildup in the Aftermath of Banking Crises

  To obtain a rough approximation of the impact of a crisis on government finances, we use the historical data on central government debt compiled in appendix A.2, as discussed earlier. It is important to note that these data provide only a partial picture, because the entire country, including states and municipalities (not just the central government), is affected by the crisis. Also, typically during these episodes, government-guaranteed debt expands markedly, but this tendency does not show up in the figures for central governments.