Capital Flow Bonanzas: An Encompassing View of the Past and Present
2009; University of Chicago Press; Volume: 5; Issue: 1 Linguagem: Inglês
10.1086/595995
ISSN2150-8372
AutoresCarmen Reinhart, Vincent Reinhart,
Tópico(s)Monetary Policy and Economic Impact
ResumoPrevious articleNext article FreeCapital Flow Bonanzas: An Encompassing View of the Past and PresentCarmen M. Reinhart and Vincent R. ReinhartCarmen M. ReinhartUniversity of Maryland and NBER Search for more articles by this author and Vincent R. ReinhartAmerican Enterprise Institute Search for more articles by this author PDFPDF PLUSFull Text Add to favoritesDownload CitationTrack CitationsPermissionsReprints Share onFacebookTwitterLinked InRedditEmailQR Code SectionsMoreI. IntroductionA pattern has often been repeated in the modern era of global finance. Foreign investors turn with interest toward some developing country. Capital flows in volume into small and shallow local financial markets. The exchange rate tends to appreciate, asset prices to rally, and local commodity prices to boom. These favorable asset price movements improve national fiscal indicators and encourage domestic credit expansion. These, in turn, exacerbate structural weaknesses in the domestic banking sector even as those local institutions are courted by global financial institutions seeking entry into a hot market. At the same time, local authorities resort to large‐scale foreign exchange sales of the local currency to cushion the effects on the exchange rate of the capital inflow bonanza.Other policy interventions, such as increases in reserve requirements and transactions taxes, usually follow to insulate the domestic economy from the accumulation of reserves. An inherent tension emerges: Local authorities take such changes as a global vote of approval that might encourage them to delay the difficult task of structural adjustment.This pattern is etched sharply in the experience of the exchange rate mechanism, the exchange rate mechanism squared, and the Latin convergence associated with the North American Free Trade Agreement and its regional successors. In the run‐up to a more perfect union, potential entrants are increasingly looked on favorably by global investors. Those investors appreciate that close integration with a strong anchor country or group of countries will ultimately discipline policy makers in the periphery, which will narrow exchange rate fluctuations and country risk spreads and buoy local equity prices. But these same dynamics also play out in commodity‐exporting emerging market economies when the prices of their output surge on world markets or when very low interest rates and sluggish growth in the advanced countries turn the attention of investors there outward. Across countries, a different rising tide raises all boats.But tides also go out when the fancy of global investors shifts and the "new paradigm" looks shopworn. Flows reverse and asset prices give back their gains, often forcing a painful adjustment on the economy.This experience has provided a fertile testing ground for international economists. A varied flora has blossomed that will be reviewed in Section II. Given double‐entry bookkeeping and the zero‐sum nature of global trade, these issues of global adjustment have been described in terms of the current or financial accounts and as deterioration in some countries or improvement in others. Moreover, the time windows have been chosen to isolate the buildup or rundown of any of these measures.This paper attempts to be encompassing in its examination of these issues by simplifying the definition of the key event that is studied and by widening the time window around that event. We investigate what happens before, during, and after a capital flow bonanza. That is, we ask, how do economies perform in and around periods when capital inflows are relatively large (or, equivalently, when their financial account surpluses are relatively large)? Owing to data limitations discussed below, we make this operational by examining episodes of large current account deficits.We study 181 countries from 1980 to 2007, a subset of 66 countries from 1960 to 2007 for which more detailed information on economic variables is available, and a smaller group of 18 countries for which house price data are available; the samples include all regions and income groups. Our primary aim is to quantitatively define and date capital inflow bonanza episodes so as to study their various aspects. In Section III, we document several features of these bonanza periods, including their incidence and duration. In Section IV, we examine the evidence on potential links between capital flow bonanzas and debt, currency, inflation, and banking crises. In Section V, we systematically illustrate the behavior of a variety of macroeconomic, financial, and policy indicators on the eve and aftermath of these episodes.Our main findings can be summarized as follows: With nearly 50 years of data, it is evident that bonanzas have become more frequent as restrictions on international capital flows have been relaxed worldwide. Although the approaches differ, this finding is in line with the evidence presented in Eichengreen and Adalet (2005).The heavy inflow episode can persist, often lulling policy makers and investors into treating the bonanza as a permanent phenomenon rather than a temporary shock. Episodes end, more often than not, with an abrupt reversal or "sudden stop" à la Calvo (as in Calvo [1998]). The current account path around bonanzas is distinctly V-shaped, irrespective of whether the broader, but more recent, sample or the less inclusive, but longer, sample is the benchmark.Capital inflow bonanza periods are associated with a higher incidence of banking, currency, and inflation crises in all but the high‐income countries (using some of the crises indicators developed in Kaminsky and Reinhart [1999] and codified in Reinhart and Rogoff [2008b]). This result is not the artifact of a few extreme cases; in more than 60% of the countries, the probability of a crisis around the dates of a capital flow bonanza is higher than for the entire sample. Capital flow bonanzas systematically precede sovereign default episodes.In developing countries (those designated by the World Bank as middle and low income), the stance of fiscal policy, as measured by the growth in real government spending, is notoriously procyclical during capital inflow bonanzas. This is consistent with the earlier observation that temporary "good times" are often treated as permanent. In effect, our preliminary results also suggest that fiscal policy plays a destabilizing role around capital flow bonanzas—and possibly more generally.For the advanced economies, the results are not as stark, since there is no systematic cross‐country evidence over 1960–2007 that the probability of a financial crisis increases during bouts of heavy capital inflows. The crisis‐prone Nordic countries in the early 1990s and the Icelandic, U.K., and U.S. crises at present would appear to be important departures from this general result (as in Reinhart and Rogoff [2008a]). Nonetheless, capital flow bonanzas are associated with more volatile macroeconomic outcomes for real GDP growth, inflation, and the external accounts.Real GDP growth tends to be higher in the run‐up to a bonanza and then systematically lower. The imprint of bonanzas is evident in asset markets. Equity prices rise when capital flows in and retreat when capital flows out. A similar pattern is evident in house prices for our small sample. A bonanza is not to be confused with a blessing.The last section (Sec. VI) turns to some of the policy implications of our analysis and discusses possible future research in this area.II. Concepts and Data IssuesA. Reviewing the Existing LiteratureThe existing literature has studied multiple manifestations of international adjustment in the balance‐of‐payments data. Double‐entry bookkeeping and the global summing to zero of trade flows produce the four alternative frames of reference laid out in Table 1. The main issues of adjustment can be described in terms of either an improvement or a deterioration (along the rows of the matrix) in either the current or capital accounts (along the columns).Table 1. Frames of Reference in the LiteratureChangeBalance‐of‐Payments AccountCurrent AccountCapital AccountImprovementCurrent account reversalCapital inflow problemDeteriorationTwin deficitsSudden stop View Table Image There is a rich empirical literature on current account reversals, the upper‐left cell mostly documenting the macroeconomic consequences of a marked improvement in a sample of many countries. Many features of these studies follow the path laid out in the pioneering paper by Milesi‐Ferretti and Razin (2000). As summarized in Table 2, they established three criteria to identify a current account reversal that are now the norm: The change in the balance must be large relative to nominal GDP, must be large absolutely in dollar terms, and must not be the product of a spike in a single year. Focusing on low‐ and middle‐income countries, they find that the adjustment experience is heterogeneous and depends importantly on whether the currency crashes on the foreign exchange market.Table 2. Current and Capital Account Reversals: Some DefinitionsStudyCriteria Used to Select the Episodes of InterestCurrent account reversals: Milesi‐Ferretti and Razin (2000); also Edwards (2005), Eichengreen and Adalet (2005), and Freund and Warnock (2005)Their underlying idea is that "large" events provide more information on determinants of reductions in current account deficits than short‐run fluctuations. These events have to satisfy three requirements:1. There must be an average reduction in the current account deficit of at least 3 (or 5) percentage points of GDP over a period of 3 years with respect to the 3 years before the event.2. The maximum deficit after the reversal must be no larger than the minimum deficit in the 3 years preceding the reversal.3. The average current account deficit must be reduced by at least one‐third. The first and second requirements should ensure that we capture only reductions of sustained current account deficits rather than sharp but temporary reversals. The third requirement is necessary so as to avoid counting as a reversal a reduction in the current account deficit from, say, 15% to 12%. Events are based on 3‐year averages.Capital account—sudden stops: Calvo, Izquierdo, and Mejía (2004)A sudden stop is defined as a phase that meets the following conditions:1. It contains at least one observation in which the year‐on‐year fall in capital flows lies at least two standard deviations below its sample mean (this addresses the "unexpected" requirement of a sudden stop).2. The sudden stop phase ends once the annual change in capital flows exceeds one standard deviation below its sample mean. This will generally introduce persistence, a common fact of sudden stops.3. Moreover, for the sake of symmetry, the start of a sudden stop phase is determined by the first time the annual change in capital flows falls one standard deviation below the mean.Capital account—sudden stops: Calvo, Izquierdo, and Loo‐Kung (2006)1. In addition to the criterion of large capital flow reversals exceeding two standard deviations from the mean (for their capital flow proxy), Calvo et al. have the following requirement:2. These reversals must be accompanied by a spike in some external aggregate measure of the cost of funds in order to capture systemic effects. More specifically, Calvo et al. use the (log of the) J. P. Morgan Emerging Market Bond Index spread over U.S. Treasury bonds for emerging markets, the Merrill Lynch euro area Government Index spreads for euro area countries (as well as Nordic countries such as Denmark, Norway, and Sweden), and G7 Government Index spreads for all remaining developed countries. Calvo et al. construct aggregate high‐spread episodes in a fashion analogous to the Calvo, Izquierdo, and Mejía (2004) measure of large capital flow reversals (i.e., Calvo et al. consider spikes in spreads exceeding two standard deviations from the mean) and determine that a sudden stop occurs when the measure of the fall‐in‐capital‐flows phase overlaps (on a yearly basis) with the aggregate high‐spread phase. Episodes that lie within a 6‐month interval are considered part of the same sudden stop phase. View Table Image Eichengreen and Adalet (2005) extended the sample to include the pre‐1970 experience, thereby providing historical context. In particular, large current account reversals appear to be the product of open trade in goods, services, and assets. Reversals have been frequent only in the two heydays of global capital markets—the recent period and the 1920s and 1930s. Large adjustments were much rarer under the pre–World War I gold standard and during the Bretton Woods years.An important fuel to the study of current account reversals has been the U.S. experience of sustained large deficits. The intent is to find rules of thumb that will be informative about the U.S. experience when the presumed "day of reckoning" comes and the unsustainable is no longer sustained. The search for such lessons appears in important papers by Edwards (2005, 2007) and Freund and Warnock (2005). They find an important role for the textbook forces thought to rein in a current account imbalance—a slowing in income growth and a real depreciation of the currency.Similar interest in the U.S. experience produced work in the 1980s on why the current account deteriorated, which is the subject of the lower‐left cell of Table 1. The main culprit at that time was identified to be the large budget deficit, which through national income accounting was mirrored in its twin, the current account. Contemporaneous discussions of this can be found in Federal Reserve Bank of Kansas City (1985), and a later review has been provided by Bosworth (1993). This line of argument petered out in the late 1990s when the U.S. federal budget went into surplus but the current account remained deeply in red.Those researchers focusing on the right cells of the contingency table typically take the perspective of emerging market economies. In particular, they view the portfolio investment decisions of investors at the center of the global financial system as somewhat fickle. Assets in some emerging markets may be in fashion for a time. Those inflows tend to appreciate the exchange rate, lead to reserve accumulation as authorities attempt to offset that force, and push up prices in asset markets. Altogether, this presents a "capital inflow problem" as described by Calvo, Leiderman, and Reinhart (1993), an issue also studied by Fernandez‐Arias and Montiel (1996).When capital no longer flows into an emerging market, the nation can no longer support an excess of spending over income. The result, in the phrase of Calvo and his coauthors, is a "sudden stop," forcing current account adjustment. The empirical application of this insight can be found in Calvo, Izquierdo, and Mejía (2004) and Calvo, Izquierdo, and Loo‐Kung (2006), both of which are described in more detail in Table 2.B. Defining a Capital Flow BonanzaThe decision to adopt a particular algorithm to date and catalog capital inflow bonanzas naturally involves trade‐offs. An advantage of casting our net wide to all large capital inflow episodes is that it does not predispose us to episodes that inevitably ended in a marked reversal. In this sense, there is a lower predisposition to tilt the analysis toward economic crises. An inflow bonanza can end with a bang or with a whimper. In this sense, our approach parallels the analysis of Goldfajn and Valdes (1999), who, rather than starting their analysis with currency crises dates, began by documenting episodes of cumulative real exchange rate appreciations of varying degrees and then sorted out which episodes unwound through an abrupt nominal exchange rate crash and which did so through reductions in inflation versus their trading partners.1We began with the presumption that the best indicator of capital flows would be reserve accumulation less the current account balance, since it measures the resources acquired (or dispersed) through issuance (or retirement) of home country liabilities. This indirect measurement of the change in liabilities seemed more likely to be available for a longer time span and for more countries than direct information from financial accounts. In the event, data on reserves tend to be published only on a delayed basis in many countries. To keep our efforts topical, the current account balance as a percentage of GDP is our benchmark indicator. It is measured more consistently across time and international boundaries than its capital account and financial account counterpart.2 For the more recent period, the same filter rules are applied to the other measures as a robustness check as is reported in the appendix.We began by applying the three‐step approach proposed by Milesi‐Ferretti and Razin to our data set with a suitable revision that does not enforce a current account reversal. This approach, however, raised some issues about dating the bonanzas of many well‐known episodes. In some countries in which the deterioration in the current account (and hence the rise in capital inflows) was a relatively smooth process over several years, this algorithm did not flag these episodes as bonanzas even though the current account deficits were large by historical standards. Heavy inflow cases, such as the United States since 2004 and Australia in several cycles since 1960, were missed altogether. In other cases, the inflow bonanza persisted after the peak current account deficit had been reached. For instance, the Thai and Malaysian current account deficits peaked in the early 1990s; however, while the deficits remained large by historical standards well into 1996, these years are not classified as bonanzas by this algorithm. Many of the important (but less persistent) surges in capital inflows of the late 1970s and early 1980s also go undetected.We ultimately settled on an alternative algorithm that provided uniform treatment across countries but was flexible enough to allow for significant cross‐country variation in the current account. As in Kaminsky and Reinhart (1999), we select a threshold to define bonanzas that is common across countries (in this case the 20th percentile).3 This threshold included most of the better‐known episodes in the literature but was not so inclusive as to label a bonanza more "routine" deteriorations in the current account. Because the underlying frequency distributions vary widely across countries, the common threshold produces quite disperse country‐specific cutoffs. For instance, in the case of relatively closed India, the cutoff to define a bonanza is a current account deficit/GDP ratio in excess of 1.8%, whereas for trade‐oriented Malaysia the comparable cutoff is a deficit/GDP ratio of 6.6%.4Figure 1, which plots the frequency distribution for 181 countries, highlights these differences across both countries and major income groups. As the figure makes clear, the range of experience is wide, but large deficits appear more frequently in lower‐income countries.Fig. 1. Distribution of current account cutoffs, used in defining bonanzas: 181 countries, 1980–2008. Sources: International Monetary Fund, World Economic Outlook, and authors' calculations.View Large ImageDownload PowerPointC. Sample Coverage and DataWe employ three samples to analyze the capital bonanza phenomenon. The broadest sample includes the 181 countries covered in the International Monetary Fund's World Economic Outlook for 1980–2007. Information is available on the current account, real GDP, inflation, and the real exchange rate. This allows us to examine the recent country experiences in a truly global setting.We will refer to the second data set as the "core" sample, which spans 1960–2007 and covers 64 countries across all regions. This sample is dominated by high‐ and middle‐income countries, where data availability poses less of a constraint. It is for this sample that we are able to examine in greater depth the macroeconomic features of the bonanzas. Also, for the core countries, we have a sufficiently complete dating of economic crises (debt, banking, etc.) that allows us to assess whether a capital inflow bonanza predisposes countries to financial crises.The third set is a small sample of 18 industrial countries for which we have data on house prices from the Bank for International Settlements. Otherwise, the data coverage for this group is the same as for the core group. Appendix tables A1 and A2 list the countries (and the income group they belong to) that make up the three samples.5All data are annual. In addition to including time series on the current account, capital and financial accounts, and nominal GDP (all in U.S. dollars), we employ a variety of macroeconomic time series. These include country‐specific variables: international reserves, nominal and real exchange rates, real GDP, consumer prices, export, imports, government expenditure, revenue, deficits, equity, and (in the case of some advanced economies) real estate prices. In addition, we have dichotomous variables that date external debt crises, currency crashes, and inflation and banking crises. Global variables, such as commodity prices, international interest rates, growth in the world's largest economies, measures of macroeconomic volatility, and the global incidence of capital flow bonanzas and various "types" of economic crises, round out the analysis. Appendix table A3 provides a full list of the variables as well as their respective sources.The availability of long time series on various aspects of macroeconomic performance was important in deciding on the design principle of our key indicator—a capital flow bonanza. Because we had gathered a sufficiently rich data set, we could be somewhat general in defining events, because we will be able to characterize behavior in a wide window around those events. That is, we can see the run‐up and the wind‐down in a manner that encompasses the definitions of earlier work.III. Capital Flow Bonanzas: Global Cycles and Country EpisodesIn what follows, we provide a sketch of country‐specific and global capital flow cycles, including incidence, by region and income group; duration; and links to global indicators.A. The Big PictureIt is relatively well known that international capital flows have an important cyclical component.6 The fact that capital (contrary to the predictions of the neoclassical growth paradigm) does not flow from rich to poor countries has also received considerable attention.7 Both of these stylized facts are illustrated in the two panels of figure 2, which plot the incidence (i.e., the percentage of countries) of capital inflow bonanzas for the broad sample consisting of 181 countries. The specific dates of the bonanza episodes on a country‐by‐country basis are listed in the four‐part appendix table A4, for high, middle‐high, middle‐low, and low‐income groupings. Column 2 of this table also provides the dates of sovereign external debt crises (defaults or restructuring).8 For our core sample of 66 countries, which account for about 90% of world GDP, the bonanza dates for 1960–2007 are listed in Table 3.Fig. 2. Incidence of bonanzas by region and income group: 181 countries, 1980–2007. Sources: International Monetary Fund, World Economic Outlook (April 2008), and a
Referência(s)