Flexing Your Muscles: Abandoning a Fixed Exchange Rate for Greater Flexibility

Previous articleNext article FreeFlexing Your Muscles: Abandoning a Fixed Exchange Rate for Greater FlexibilityBarry Eichengreen and Andrew K. RoseBarry EichengreenUniversity of California, Berkeley, and NBER Search for more articles by this author and Andrew K. RoseUniversity of California, Berkeley, and NBER Search for more articles by this author PDFPDF PLUSFull Text Add to favoritesDownload CitationTrack CitationsPermissionsReprints Share onFacebookTwitterLinked InRedditEmailQR Code SectionsMoreI. IntroductionThe maintenance of currency pegs by fast-growing emerging markets has become a flash point in discussions of economic policy. Not a few observers concerned for the stability of the international economic and financial system argue that international balance could be better maintained and global financial stability enhanced if emerging markets like China abandoned their pegs in favor of regimes of greater flexibility. In addition, more flexible exchange rates would give emerging markets greater ability to tailor policy to domestic conditions. Where growth is strong and inflation is a problem, currency appreciation would help damp down inflationary pressures and avoid asset bubbles and overheating. It would facilitate efforts in these countries to rebalance away from exports in favor of domestic spending. It would give them an additional instrument—a more flexible exchange rate—with which to cope with volatile capital flows as the capital account of the balance of payments becomes more open.1 Greater flexibility on the part of countries like China, implying less foreign exchange market intervention, would also slow the accumulation of reserves in the form of US treasury and other advanced-country securities. It would help the United States grow its exports. Insofar as exchange rates fixed at inappropriate levels contributed to global imbalances and thereby helped to plant the seeds for the global financial crisis, this is an issue of not just national but international significance. Thus, not only the IMF but high officials in both the United States and Europe regularly make the case for greater exchange-rate flexibility to emerging markets like China.Spokesmen for emerging markets counter that abandoning pegs for greater flexibility would damage their growth prospects. Currency stability against the dollar, they argue, has been integral to their successful economic development. Greater variability would make doing international business more difficult, given that banks and enterprises lack experience in dealing with currency fluctuations and the relevant hedging markets are often missing. Rapid currency appreciation might cause export growth to slow. This could precipitate financial problems for firms heavily invested in the production of tradables and, in turn, for their banks. Cut loose from its anchor, the exchange rate could grow dangerously unstable. Asset prices would react badly, compounding these other economic and financial problems. In a period when a majority of the growth of global demand emanates from emerging markets, problems for this set of countries are the last thing the world needs.2 Economic theory and logic have been deployed on both sides of this debate. Authors like Chinn (2007) have invoked the Mundell-Fleming model to show how the shift toward a more flexible exchange rate regime (assuming that this is accompanied by currency appreciation) would help countries in China's position to restore internal and external balance. On the other hand, authors like McKinnon and Schnabl (2007) have invoked elasticity pessimism and the specter of deflationary slumps to argue that currency appreciation would do little to correct global imbalances and could lead to falling wages and prices and, in the worst case, a Japanese-style deflationary syndrome. This leads them to conclude that if China adopted a more flexible exchange rate regime it would be undesirable.This arena, populated by competing models and conflicting priors, has not exactly been informed by copious empirical analysis. Evidence on the effects of abandoning currency pegs for greater flexibility has largely been limited to estimates of import and export demand elasticities, which are used to project prospective changes in the net exports of China and other countries on the assumption that the shift in regime will occasion appreciation. (For surveys see Marquez and Schindler 2007 and Garcia-Herrero and Kiovu 2010.) McKinnon and Schnabl, to inform their analysis of China, lean on the behavior of macroeconomic variables in the aftermath of Japan's exit from its dollar peg in 1971–1973. Using mainly graphical methods, Eichengreen et al. (1998) study a handful of emerging markets that had voluntarily abandoned currency pegs in favor of regimes of greater flexibility. More recently, Kappler et al. (2011) have studied 25 substantial nominal and real appreciations, but a number of their observations are step revaluations rather than appreciations that occurred in conjunction with a change in the exchange rate regime.Our approach is different. We examine a comprehensive data set covering over 200 countries and territories, both developing and advanced, since 1957.3 We focus on 51 instances where countries abandoned currency pegs for regimes of greater flexibility with a reasonable expectation that their currencies would appreciate.4 Thus, we consider changes in exchange rate regime in the direction of greater flexibility but rule out cases where the change was followed by sharp currency depreciation. In other words, we are not interested in "crisis" and "speculative attack" cases where a currency peg collapses under pressure, resulting in devaluation or sharp depreciation (having considered these cases elsewhere; see Eichengreen, Rose, and Wyplosz 1995).5 Rather, we focus on instances where a country shifts from a fixed to flexible exchange rate regime and either sees its exchange rate remain relatively unchanged or experiences an appreciation.6 In spirit, our paper is closest to the Eichengreen et al. (1998) study of exits from pegged exchange rates to greater flexibility. But our sample is larger and, we would argue, our methods are more systematic.We examine the impact of these events, which we call "flexes," on a range of macroeconomic and financial variables, including GDP growth, export growth, consumption, investment, and inflation. We compare the behavior of these variables in the "flexers" and a control group of countries maintaining a pegged exchange rate throughout. We look for and, where necessary, correct for selectivity bias by searching for differences in country circumstances in the period before the flex occurred.7 Although our set of "treatment cases" is not large, the fact that it is larger and more comprehensive than in earlier studies undertaking similar analyses allows us to utilize more systematic empirical methods.There is, of course, no single definition of what constitutes a "flex." In our benchmark results we consider cases where there was a change in the de facto exchange rate regime in the direction of greater flexibility and the exchange rate either appreciated or remained broadly unchanged (neither appreciated nor depreciated by more than 5%). Fortunately, similar results in fact obtain under a number of alternative definitions of what constitutes "flexing." We then analyze the behavior of a range of macroeconomic and financial variables over the periods three years before and after the change in exchange rate regime. We find similar results when we consider shorter periods. There is less evidence of a significant impact on the variables of interest when we compare longer periods, but this is plausible, since one would not expect a change in a nominal variable or in the regime governing its behavior to have implications for real variables over long horizons.8 Some of the cases we examine are likely to be dismissed as special. For example, a disproportionate number are clustered around the time of the collapse of the Bretton Woods System; they are therefore not truly independent flexes insofar as they can also be understood as reflecting the US decision to float the dollar and allow it to depreciate. Other cases occur in developing countries whose historical experience has not attracted much attention. But experience following the collapse of the Bretton Woods System is directly relevant to the future prospects of countries like China, insofar as their decisions to flex could result in the end of the regime that has been widely referred to as Bretton Woods II (Dooley, Folkerts-Landau, and Garber 2003).9 And while the experience of those few low-to-middle-income countries that have appreciated out of fixed exchange rate regimes with limited capital mobility in the past may have not received much prior attention, it is directly relevant to the prospects for middle-income countries contemplating moving to a more flexible exchange rate today, like China. In any case, the advantage of constructing as large a sample as possible is that the supposed special nature of a subset of those cases need not dominate the results.10The results reveal a wide range of responses of macroeconomic and financial variables. Another way of putting this is that the very wide dispersion of results makes it hard to identify significant differences in the behavior of the variables of interest before and after flexing, or between the flexers and other countries, in the wake of the event. This suggests that our 51 cases are heterogeneous. In a subset of cases, however, the decision to flex is followed by a discernible slowdown in the rate of economic growth. Slowdowns are most likely, we show, when the investment ratio is high, consumption, investment, exports, and imports are growing rapidly, and money growth is fast. Since we only have 51 observations, confidence intervals tend to be wide, but the finding that slowdowns following flexes tend to occur in high-investment and rapid-export-growth economies is robust across samples and specifications. The implication is that China, where both conditions prevail, may have some basis for worrying about the growth effects of appreciating out of its fixed exchange rate regime. In a subset of cases, in addition, the decision to flex was followed by a significant decline in the rate of inflation. Slower inflation is most likely, we show, in countries that are relatively open to trade (where the reduction in the rate of import price inflation presumably has the greatest impact) and in countries with high foreign reserves (which had presumably been sterilizing capital inflows with less than complete success prior to the change in exchange rate regime). These results also have obvious implications for China, which is currently characterized by inflation and the other macroeconomic characteristics in question.Section II describes our data, definitions, and 51 cases. Section III discusses the determinants of flexes and the possibility of selection bias. The main results are in Section IV, which presents event studies and some simple regression analysis. In concluding, Section V draws out the implications for China and the global-rebalancing debate. An appendix presents some case studies of flexing that were and were not accompanied by significant growth slowdowns. II. Data and DefinitionsExchange rate regimes come in many flavors. Many countries, while not attempting to maintain a peg, manage their exchange rates heavily (they may declare a commitment to flexibility but in practice prevent the currency from moving). Others may refrain from intervening but still see their currency display broad stability. Some countries have multiple exchange rates; they may regulate one according to official policy but have a different exchange rate (often on the black market) that moves differently and is used for unofficial transactions.The first and, in some sense, most important step in our analysis involves identifying shifts from regimes of pegged exchange rates to regimes of greater flexibility. For this purpose we use the Reinhart-Rogoff (2004, hereafter "RR") de facto classification of exchange rate regimes as extended by Reinhart, Rogoff, and Ilzetzki. This taxonomy distinguishes 15 exchange rate regimes by degree of flexibility for 218 "countries" (some of which are, in practice, territories) and is available monthly from 1946m1 through 2007m9. In constructing their index, RR utilize information on both official and black market exchange rates. We treat the first four of their categories—no separate legal tender, preannounced peg or currency board arrangement, preannounced horizontal band narrower than or equal to ±2%, and de facto peg—as fixed exchange rate regimes.11 We look for cases where countries moved away from these regimes; there are 119 such departures in the sample.12Of course, there exist a number of alternative classifications of exchange rate regimes. The most prominent, published by the International Monetary Fund (IMF), has been based on government's stated de jureexchange rate policy.13 But what is of interest here is what officials responsible for exchange rate policy actually do as opposed to what they say. Similarly, we choose not to use Shambaugh's (2004) classification, since this relies exclusively on de jure exchange rate data and provides only a coarse classification (peg/no-peg) at an annual frequency. Levy-Yeyati and Sturzenegger's (2003) categorization incorporates information on movements in both exchange rates and international reserves, but their data set is annual, begins only in 1974, ends in 2004, and has a number of missing and inconclusive observations.14We look for cases where countries moved away from these regimes and then experienced either exchange rate appreciation or (at most) minor depreciation.15 Thus, we rule out cases of devaluation and substantial depreciation—where the change in exchange rate regime was forced by market pressures. These cases of currency crisis and step devaluation have been studied before, as noted earlier. Constructing this sample requires answering three more questions: how long a subsequent period, how big an exchange rate change, and exchange rate movements against what? We examine exchange rate changes over the three months following the change in regime. This is long enough for the exchange rate change to be unaffected by high-frequency considerations—transient financial shocks, for example—while retaining our focus on the aftermath of regime changes. Lengthening and shortening the subsequent period does not notably change the results. We examine currencies that appreciated by any amount or depreciated by less than 5%; the latter figure is again arbitrary but reasonable. Again, imposing slightly larger and smaller values of this threshold do not affect the results. Finally, we consider official exchange rates against both the US dollar and the Special Drawing Rights (SDRs).16We review the observations one by one and exclude cases where the RR data set indicates a regime shift but there was no subsequent change in the official SDR exchange rate. (Virtually without exception, there was also no change in the official dollar exchange rate in these cases.) We also exclude a number of questionable observations.17We are then left with 51 cases. These are tabulated in table 1 along with the three-month rates of change of the SDR, official dollar rate, and parallel market rate (where available). We see there some prominent cases (Canada in 1970, Germany in 1973) along with a number of more obscure ones. The earliest flex is Paraguay in 1960, the most recent Malaysia in 2005.18 Most of these cases have only small changes in the exchange rates we consider.19 We use two variants of our default definition of flexes to check for sensitivity. Our first variant drops (twelve) observations without any appreciation in any exchange rate, while our second drops (four) observations with large (double-digit) appreciations. Both variants are marked in Table 1.Table 1 . 51 Flexes with Three-Month Exchange Rate ChangesCountryYearSDR$Parallel $Australia19741%–2%–3%Botswana1980–2%–1% Canada1970–2%–2%–1%Costa Ricaa19630%0%–1%Costa Rica19710%0%1%Finland19734%–7%–6%France19710%–8%–2%Germanyb19730%–10%–10%Germanya19690%0%0%Greece19660%0%–4%Haitia19852%0%0%Hong Kong1972–1%–1%–2%Irana19740%0%0%Iraq1982–2%0%–4%Ireland19790%–2%0%Israel19700%0%–2%Italya19733%3%0%Jamaicaa19830%0%0%Japan1973–2%–2%–1%Kuwait1975–2%2%1%Liberia1998–2%–5%0%Libya19710%0%–11%Lithuania2003–1%–2% Malawi1973–2%–2%–14%Malaysia2005–2%0% Malaysiaa19751%0%0%Malta19723%3% Mauritania1974–1%–1%0%Mexico19761%1%–7%Moroccoa19732%2%1%Mozambique2004–5%–5% Nepal19783%0%–21%Netherlands1971–3%–3%–3%New Zealand1973–9%–9%–5%Nicaraguaa19934%1%4%Paraguaya19600%0%0%Perua19670%0%0%Philippines19701%1%–4%Portugalb1973–12%–12%–12%Singapore1973–1%–1%–6%South Africab19720%–10%–8%Spain1974–3%–3%–3%Sri Lankaa19680%0%1%Sri Lanka19904%1%–2%Suriname1974–1%0%–2%Swedenb1973–10%–10%–11%Switzerland1973–1%–1%–1%Tunisia19743%3%–3%Turkey19610%0%–4%Turkey19720%0%–1%United Kingdom19725%5%–2%Notes: The observations tabulated are for countries that have exited a fixed exchange rate regime (RR < 5) to a more flexible exchange rate regime and have subsequently either (a) appreciated or (b) depreciated less than 5% over the next three months. Subsequent SDR, dollar, and parallel dollar depreciation rates are tabulated on the right; further description is available in the text. Exchange rates are quoted as domestic price of foreign exchange, so that negative values indicate exchange rate appreciation. a "Variant 1" observations without any appreciation.b "Variant 2" observations without appreciations > 10%.View Table Image: 1 | 2As previously noted, a substantial fraction of our cases (20 out of 51) are clustered around the collapse of the Bretton Woods System, when other currencies began floating against the dollar (although there were also flexes in the 1960s, 1980s, 1990s, and 2000s).20 At some level this is not especially disturbing. The greater flexibility of, inter alia, the German deutschemark and the Japanese yen in the period after the Bretton Woods System dissolved is one of the main precedents to which observers point when imagining the consequences of a Chinese transition to greater flexibility (and the dissolution of the so-called Bretton Woods II system).21 Thus it is not inappropriate that our sample should be weighted toward this episode.III. Selectivity Below we will ask how macroeconomic and financial variables behave in the wake of a decision to abandon a currency peg in favor of greater flexibility and allow the currency to appreciate, or at least not depreciate significantly. A logically prior question, however, is whether the post-exit behavior of our variables is affected by their exceptional behavior in the immediately preceding period. Are our findings for the post-exit period contaminated by selectivity, in other words? There is good reason to think that countries choosing to move to greater exchange rate flexibility do not do so randomly. An obvious source of selection is country size. Very small open economies tend not to have floating exchange rates.22 Those with pegged exchange rates are correspondingly less likely to abandon them for greater flexibility. Conditioning on country size when undertaking the kind of analysis conducted here is doubly important, moreover, insofar as the macroeconomic impact of a change in China's exchange rate regime is an obvious subtext of our study and China is located at an extreme of the country-size distribution.While country size is an obvious source of selectivity, other potential sources are less obvious. For example, it is not obvious that fast growing countries might deliberately decide to flex as opposed to remaining in a fixed exchange rate regime in the hope that the good times associated with that regime might continue to roll. As a Chinese policymaker well might ask, why mess with success? We probe further for selectivity bias by examining whether countries that flex differ systematically from other countries in our sample. Since the decision to abandon a pegged exchange rate regime is an event with potentially important medium-term consequences and because the exact timing of the regime change is unimportant in understanding its determinants, we convert our annual data to three-year averages. (Using five-year and other similar averages makes little difference for the results.) We then construct a binary variable where a value of unity signifies a flexing during the period and all other observations take on values of zero. We use this as the dependent variable in an encompassing set of probit regressions as a way of examining whether any of the usual suspects register significantly on the right-hand side of the equation. Since we already know that country size should vary with the exchange rate regime, we include it (along with country and time effects) as a conditioning variable. We are interested in whether there is evidence that other variables of interest affect the likelihood of flexing. If the rate of GDP growth, for example, has a significant effect in this probit model, that would be prima facie evidence of selection bias—that relatively slow or, more plausibly, fast-growing countries (depending on the sign of the coefficients) are more likely to flex. The answer is in Table 2. The top panel displays a set of coefficients from bivariate probit regressions. Our baseline estimates on the left include both fixed time- and random country-specific effects. Almost without exception the variables of interest (GDP growth, export growth, investment growth, consumption growth, credit growth) do not enter the probit regressions significantly; the same is true of the real effective exchange rate and different measures of capital mobility. We interpret this as little evidence of the presence of selection bias in the relevant sense. While size as proxied by population is systematically associated with the probability of flexing in the bivariate regressions, this variable is very slowly moving and as such is not the focus of our analysis.23 We also provide extensive sensitivity analysis in other columns of the table, dropping time effects, and also using two different variants of our measure of exchange rate flexing. However, our essential (non) results seem robust to these perturbations of our basic setup. The results are also in the results tabulated in the panel of table 2. Table . of Flexes of change of GDP credit real effective exchange measure of capital measure of capital fixed random measure of capital for that coefficients from panel probit regressions 1 if flexing occurred during three period, are three-year averages of variables in left but not a separate at the at the 5% at the Table conclude that when their it reasonable to the determinants of flexing than country use an event study approach to examine the behavior of our