Publication draft, October 14, 2010
Monetary Policy in Emerging Markets: A Survey
Jeffrey Frankel, Harvard Kennedy School In the Handbook of Monetary Economics,
edited by Benjamin Friedman and Michael Woodford (Elsevier: Amsterdam, 2011). The author would like to thank Olivier Blanchard, Ben Friedman, Oyebola Olabisi, Eswar Prasad and participants at the ECB conference in October 2009, for comments on an earlier draft; and to thank the MacArthur Foundation for support.
Abstract The characteristics that distinguish most developing countries, compared to large industrialized countries, include: greater exposure to supply shocks in general and trade volatility in particular, procyclicality of both domestic fiscal policy and international finance, lower credibility with respect to both price stability and default risk, and other imperfect institutions. These characteristics warrant appropriate models.
Models of dynamic inconsistency in monetary policy and the need for central bank independence and commitment to nominal targets apply even more strongly to developing countries. But because most developing countries are price-takers on world markets, the small open economy model, with nontraded goods, is often more useful than the two-country two-good model. Contractionary effects of devaluation are also far more important for developing countries, particularly the balance sheet effects that arise from currency mismatch. The exchange rate was the favored nominal anchor for monetary policy in inflation stabilizations of the late 1980s and early 1990s. After the currency crises of 1994-2001, the conventional wisdom anointed Inflation Targeting as the preferred monetary regime in place of exchange rate targets. But events associated with the global crisis of 2007-09 have revealed limitations to the choice of CPI for the role of price index.
The participation of emerging markets in global finance is a major reason why they have by now earned their own large body of research, but it also means that they remain highly prone to problems of asymmetric information, illiquidity, default risk, moral hazard and imperfect institutions. Many of the models designed to fit emerging market countries were built around such financial market imperfections; few economists thought this inappropriate. With the global crisis of 2007-09, the tables have turned: economists should now consider drawing on the models of emerging market crises to try to understand the unexpected imperfections and failures of advanced-country financial markets.
JEL numbers: E, E5, F41, O16
Key words: central bank, crises, developing countries, emerging markets, macroeconomics, monetary policy
Why do we need different models for emerging markets?
Monetary Policy in Emerging Market Countries: A Survey Thirty years ago, the topic of Macroeconomics or Monetary Economics for Developing Countries hardly existed1, beyond a few papers regarding devaluation.2 Nor did the term “emerging markets” exist. Certainly it was not appropriate at that time to apply to such countries the models that had been designed for industrialized countries, with their assumption of financial sectors that were highly market-oriented and open to international flows. To the contrary, developing countries typically suffered from “financial repression” under which the only financial intermediaries were uncompetitive banks and the government itself, which kept nominal interest rates artificially low (often well below the inflation rate) and allocated capital administratively rather than by market forces.3 Capital inflows and outflows were heavily discouraged, particularly by capital controls, and were thus largely limited to foreign direct investment and loans from the World Bank and other international financial institutions.
Over time, the financial sectors of most developing countries – at least those known as emerging markets – have gradually become more liberalized and open. The globalization of their finances began in the late 1970s with the syndicated bank loans that recycled petrodollars to oil-importers. Successive waves of capital inflow followed after 1990 and again after 2003. The largest outpouring of economic research was provoked not so much by the capital booms themselves as by the subsequent capital busts: the international debt crisis of 1982-89, the emerging market crises of 1995-2001, and perhaps the global financial crisis of 2008-09.
In any case, the literature on emerging markets now occupies a very large share of the field of international finance and macroeconomics. International capital flows are central to much of the research on macroeconomics in developing countries. This includes both efficient-market models that were originally designed to describe advanced economies and market-imperfection models that have been designed to allow for the realities of default risk, procyclicality, asymmetric information, imperfect property rights and other flawed institutions.
In the latter part of the 19th century most of the vineyards of Europe were destroyed by the microscopic aphid Phylloxera vastatrix. Eventually a desperate last resort was tried: grafting susceptible European vines onto resistant American root stock. Purist French vintners initially disdained what they considered compromising the refined tastes of their grape varieties. But it saved the European vineyards, and did not impair the quality of the wine. The New World had come to the rescue of the Old. In 2007-08, the global financial system was grievously infected by so-called toxic assets originating in the United States. Many ask what fundamental rethinking will be necessary to save macroeconomic theory. Some answers may lie with models that have been applied to fit the realities of emerging markets, models that are at home with the financial market imperfections that have now unexpectedly turned up in industrialized countries as well. Purists will be reluctant to seek salvation from this direction. But they should not fear. The hardy root stock of emerging market models is incompatible with fine taste.
1. Why do we need different models for emerging markets? At a high enough level of abstraction, it could be argued, one theory should apply for all. Why do we need separate models for developing countries? What makes them different? We begin the chapter by considering the general structural characteristics that tend to differentiate these countries as a group, though it is important also to acknowledge the heterogeneity among them.
Developing countries tend to have less developed institutions (almost by definition), and specifically to have lower central bank credibility, than industrialized countries.4 Lower central bank credibility usually stems from a history of price instability, including hyperinflation in some cases, which in turn is sometimes attributable to past reliance on seignorage as a means of government finance in the absence of a well-developed fiscal system. Another common feature is an uncompetitive banking system, which is again in part attributable to a public finance problem: a traditional reliance on the banks as a source of finance, through a combination of financial repression and controls on capital outflows.
Another structural difference is that the goods markets of small developing countries are often more exposed to international influences than those of, say, Europe or Japan. Although their trade barriers and transport costs have historically tended to exceed those of rich countries, these obstacles to trade have come down over time. Furthermore developing countries tend to be smaller in size and more dependent on exports of agricultural and mineral commodities than are industrialized countries. Even such standard labor-intensive manufactured exports as clothing, textiles, shoes, and basic consumer electronics are often treated on world markets as close substitutes across suppliers. Therefore these countries are typically small enough that they can be regarded as price-takers for tradable goods on world markets. Hence the “small open economy” model.
Developing countries tend to be subject to more volatility than rich countries.5 Volatility comes from both supply shocks and demand shocks. One reason for the greater magnitude of supply shocks is that primary products (agriculture, mining, forestry and fishing) make up a larger share of their economies. These activities are vulnerable both to extreme weather events domestically and to volatile prices on world markets. Droughts, floods, hurricanes, and other weather events tend to have a much larger effect on GDP in developing countries than industrialized ones. When a hurricane hits a Caribbean island, it can virtually wipe out the year’s banana crop and tourist season – thus eliminating the two biggest sectors in some of those tropical economies. Moreover, the terms of trade are notoriously volatile for small developing countries, especially those dependent on agricultural and mineral exports. In large rich countries, the fluctuations in the terms of trade are both smaller and less likely to be exogenous.
Volatility also arises from domestic macroeconomic and political instability. Although most developing countries in the 1990s brought under control the chronic pattern of runaway budget deficits, money creation, and inflation, that they had experienced in the preceding two decades, most have still been subject to monetary and fiscal policy that is procyclical rather than countercyclical. Often income inequality and populist political economy are deep fundamental forces.
Another structural difference is the greater incidence of default risk.6 Even for government officials who sincerely pursue macroeconomic discipline, they may face debt-intolerance: global investors will demand higher interest rates in response to increases in debt that would not worry them coming from a rich country. The explanation may be the reputational effects of a long history of defaulting or inflating away debt.7 The reputation is captured, in part, by agency ratings.8
Additional imperfections in financial markets can sometimes be traced to underdeveloped institutions, such as poor protection of property rights, bank loans made under administrative guidance or connected lending, and even government corruption.9 With each round of financial turbulence, however, it has become harder and harder to attribute crises in emerging markets solely to failings in the macroeconomic policies or financial structures of the countries in question. Theories of multiple equilibrium and contagion reflect that not all the volatility experienced by developing countries arises domestically. Much comes from outside, from global financial markets.
The next section of this chapter considers goods markets and concludes that the small open economy model is probably most appropriate for lower-income and middle-income countries: prices of traded goods are taken as given on world markets. Two key variants feature roles for nontraded goods and contractionary effects of devaluation. The subsequent three sections focus on monetary policy per se. They explore, respectively, the topics of inflation (including high-inflation episodes, stabilization, and central bank independence), nominal anchors, and exchange rate regimes. The last three sections of the chapter focus on the boom-bust cycle experienced by so many emerging markets. They cover, respectively, procyclicality (especially in the case of commodity exporters), capital flows, and crises.
2. Goods markets, pricing, and devaluation As already noted, because developing countries tend to be smaller economically than major industrialized countries, they are more likely to fit the small open economy model: they can be regarded as price-takers, not just for their import goods, but for their export goods as well. That is, the prices of their tradable goods are generally taken as given on world markets.10 It follows that a devaluation should push up the prices of tradable goods quickly and in proportion.
Traded goods, pass-through and the Law of One Price
The traditional view has long been that developing countries, especially small ones, experience rapid pass-through of exchange rate changes into import prices, and then to the general price level. There is evidence in the pass-through literature that exchange rate changes are indeed reflected in imports more rapidly when the market is a developing country than when it is the United States or another industrialized country.11 The pass-through coefficient tells to what extent a devaluation has been passed through into higher prices of goods sold domestically, say, within the first year. Pass-through has historically been higher and faster for developing countries than for industrialized countries. For simplicity, it is common to assume that pass-through to import prices is complete and instantaneous.
This assumption appears to have become somewhat less valid, especially in the big emerging market devaluations of the 1990s. Pass-through coefficients appear to have declined in developing countries, though they remain well above those of industrialized economies.12
On the export side, agricultural and mineral products, which remain important exports in many developing countries, tend to face prices that are determined on world markets. Because they are homogeneous products, arbitrage is able to keep the price of oil or copper or coffee in line across countries, and few producers have much monopoly power. The situation is less clear, however, regarding the pricing of manufactures and services. Clothing products or call centers in one country may or may not be treated by customers as perfect substitutes for clothing or call centers in another country.
When export prices are sticky
There is good empirical evidence that an increase in the nominal exchange rate defined as the price of foreign currency (that is, a devaluation or depreciation of the domestic currency) causes an increase in the real exchange rate.13 There are two possible approaches to such variation in the real exchange rate. First, it can be interpreted as evidence of stickiness in the nominal prices of traded goods, especially non-commodity export goods, which in turn requires some sort of barriers to international arbitrage, such as tariffs or transportation costs. Second, it could be interpreted as a manifestation that nontraded goods and services, which by definition are not exposed to international competition, play an important role in the price index. Both approaches are fruitful, because both elements are typically at work.14
If prices of exports are treated as sticky in domestic currency, then traditional textbook models of the trade balance are more relevant. Developing countries tend to face higher price-elasticities of demand for their exports than do industrialized countries. Thus it may be easier for an econometrician to find the Marshall-Lerner condition satisfied, though one must allow for the usual lags in quantity response to a devaluation, producing a J-curve pattern in the response of the trade balance.15
The alternative approach is to stick rigorously to the small open economy assumption, that prices of all traded goods are determined on world markets, but to introduce a second category: nontraded goods and services. Define Q to be the real exchange rate:
E ≡ the nominal exchange rate, in units of domestic currency per foreign.
CPI ≡ the domestic Consumer Price Index, and
CPI* ≡ the world Consumer Price Index.
Assume that the price indices, both at home and abroad, are Cobb-Douglas functions of two sectors, tradable goods (TG) and nontradable goods (NTG), and that for simplicity the weight on the nontradable sector, α , is the same at home and abroad:
We observe the real exchange vary, including sometimes in apparent response to variation in the nominal exchange rate. The two possible interpretations, again, are: (1) variation in the relative price of traded goods (EPTG *)/PTG, which is the case considered in the preceding section, or (2) variation in the within-country relative price of nontraded goods (i.e., the price of nontraded goods relative to traded goods). In this section, to focus on the latter, assume that international arbitrage keeps traded goods prices in line: PTG = EPTG*. Then the real exchange depends only on the relative price of nontraded goods.
If the relative price of nontraded goods goes up in one country, that country’s currency will exhibit a real appreciation. 16
Two sources of variation in the relative price of nontraded goods make this simple equation useful and interesting, particularly for developing countries. They are very different in character: one is best thought of as monetary in origin and short-term in duration, the other as real and long-term.
Begin with the latter: the famous Balassa (1964)-Samuelson (1964) effect. An empirical regularity that shows up robustly in long-term data samples, whether cross-section or time series, is that when a country’s per capita income is higher, its currency is stronger in real terms. This real appreciation can in turn usually be associated with an increase in the relative price of nontraded goods, as per the above equation. The elasticity coefficient is estimated at around .4.17 Balassa and Samuelson identified the causal mechanism as productivity growth that happens to be concentrated in the tradable good sector. (Bergin, et al, 2006, and Ghironi and Melitz, 2005, have shown theoretically why this may be no coincidence.) Regardless of the mechanism, the empirical regularity is well-established.18
Having said that, individual countries can deviate very far from the Balassa-Samuelson line, especially in the short run. There has been an unfortunate tendency, among those papers that invoke the Balassa-Samuelson relationship, to try to assign it responsibility for explaining all variation in the relative price of nontraded goods and therefore in the real exchange rate, even in the short run. A more sensible approach would be to recognize the large temporary departures of the real exchange rate from the Balassa-Samuelson line, and to think about what causes them first to appear and then disappear gradually over time.
Fortunately, we have long had some simple models of how monetary factors can explain large temporary swings in the real exchange rate. A monetary expansion in a country with a currency peg will show up as inflation in nontraded goods prices, and therefore as real appreciation, in the short run. A devaluation will rapidly raise the domestic price of traded goods, thereby reducing the relative price of nontraded goods and showing up as a real depreciation. The Salter-Swan model originally showed these effects, and their implications for internal balance (attaining the desired trade balance) and external balance (attaining the desired point on a tradeoff between output and price level acceleration).19
Dornbusch (1973, 1980) extended the nontraded goods model, in research on the case of pegged countries that was once as well-known as his famous overshooting model for the case of floating countries. The extension was to marry Salter-Swan with the Monetary Approach to the Balance of Payments. No flexible-price assumptions were harmed in the making of this model: The nominal price of nontraded goods was free to adjust. But, in the aftermath of a devaluation or in the aftermath of a domestic credit contraction, the levels of reserves and money supply would lie below their long-run equilibria. Only via a balance of payments surplus could reserves flow in over time, thereby gradually raising the overall money supply and nontraded goods prices in tandem. In the long run all prices and quantities, including the real exchange rate, would be back to their equilibrium values -- but only in the long run. Movements in the relative price of nontraded goods that arise from money factors in the short run and from Balassa-Samuelson in the long run remain a good way to think about real exchange rates in developing countries.
Contractionary effects of devaluation
Devaluation is supposed to be expansionary for the economy, in a “Keynesian approach to the trade balance,” that is, in a model where higher demand for domestic goods, whether coming from domestic residents or foreign, leads to higher output rather than higher prices. Yet, in currency crises that afflict developing countries, devaluation often seems to be associated with recession rather than expansion.
Political costs of devaluation
In a widely quoted statistic, Cooper (1971) found that political leaders often lose office in the year following devaluation. Frankel (2005) updated the estimate and verified statistical significance: A political leader in a developing country is almost twice as likely to lose office in the six months following a currency crash as otherwise. Finance ministers and central bank governors are even more vulnerable. The political unpopularity of devaluations in developing countries helps explain why policy makers often postpone devaluations until after elections.20
Why are devaluations so unpopular? It is often thought that they have adverse distributional effects. But the urban elites that are most important to the political process in most developing countries are more likely to benefit from the relative price effects of devaluation (an increase in the price of agricultural products relative to services) than is the rural population. One possibility is that devaluations act as a proxy for unpopular IMF austerity programs or other broad reform packages. IMF-associated austerity programs have often resulted in popular unrest. 21 I also tested the proposition that devaluations are acting as a statistical proxy for unpopular IMF austerity programs by conditioning the previous calculation on the adoption of IMF programs. The IMF program variable does not seem to raise the frequency of leader job loss, relative to devaluations that did not involve an IMF program.22 There is more support for the hypothesis that finance ministers and central bankers are likely to lose their jobs when a devaluation is perceived as violating previous public assurances to the contrary; but this only explains part of the effect. The dominant reason appears to be that devaluations are indeed contractionary.
Edwards (1986) and Acar (2000) found that devaluation in developing countries is contractionary in the first year, but then expansionary when exports and imports have had time to react to the enhanced price competiveness. (In the very long run, devaluation is presumed neutral, as prices adjust and all real effects disappear.) Bahmani-Oskooee (2006) also found some evidence of contractionary effects. For the countries hit by the East Asia crisis of 1997-98, Upadhyaya (1999) found that devaluation was at best neutral in the long run, while Chou and Chao (2001) found a contractionary effect in the short run. Ahmed, et al. (2002) find that contractionary devaluations are a property of developing countries in particular. Rajan and Shen (2006) find that devaluations are only contractionary in crisis situations, which they attribute to debt composition issues.
Connolly (1983) and Kamin (1988) did not find contractionary effects. Nunnenkamp and Schweickert (1990) rejected the hypothesis of contraction on a sample of 48 countries, except during the first year in the case of manufactured exports (as opposed to agricultural). Some who find no negative correlation attribute the findings of those who do to the influence of third factors such as contemporaneous expenditure-reducing policies.
Confirming a new phenomenon, Calvo and Reinhart (2001) found that exports do not increase at all after a devaluation, but rather fall for the first eight months. Perhaps firms in emerging market crises lose access to working capital and trade credit even when they are in the export business.
Effects via price pass-through
Through what channels might devaluation have contractionary effects? Several of the most important contractionary effects of an increase in the exchange rate are hypothesized to work through a corresponding increase in the domestic price of imports, or of some larger set of goods. Indeed, rapid pass-through of exchange rate changes to the prices of traded goods is the defining assumption of the “small open economy model,” which has always been thought to apply fairly well to emerging market countries. The contractionary effect would then follow in any of several ways. The higher prices of traded goods could, for example, reduce real incomes of workers and therefore real consumption.23 Or they could increase costs to producers in the non-traded goods sector, coming from either higher costs of imported inputs such as oil or higher labor costs if wages are indexed to the cost of living.24 Krugman and Taylor (1978) added increased tariff revenue to the list of ways in which devaluation might be contractionary.25 The higher price level could also be contractionary through the “real balance effect,” which is a decline in the real money supply. The tightening of real monetary conditions, which typically shows up as an increase in the interest rate, could then exert its contractionary effect either via the demand side or via the supply side.26
These mechanisms were not much in evidence in the currency crashes of the 1990s. The reason is that the devaluations were not rapidly passed through to higher prices for imports, for domestic competing goods, or to the CPI in the way that the small open economy model had led us to believe. The failure of high inflation to materialize in East Asia after the 1997-98 devaluations, or even in Argentina after the 2001 devaluation, was good news. But it called for greater scrutiny of the assumption that developing countries were subject to complete and instantaneous pass-through.27 Balance sheet effect from currency mismatch
Balance sheet effects have become easily the most important of the various possible contractionary effects of devaluation. Banks and other firms in emerging markets often incur debt denominated in foreign currency, even while much of their revenues are in domestic currency. This situation is known as currency mismatch. When currency mismatch is combined with a major devaluation, otherwise-solvent firms have trouble servicing their debts. They may have to lay off workers and close plants, or go bankrupt altogether. Such weak balance sheets have increasingly been fingered in many models, not only as the major contractionary effect in a devaluation, but also as a fundamental cause of currency crises in the first place.28
A number of empirical studies have documented the balance sheet effect, in particular the finding that the combination of foreign-currency debt plus devaluation is indeed contractionary. Cavallo, Kisselev, Perri and Roubini (2004) find that the magnitude of recession is related to the product of dollar debt and percentage devaluation. Bebczuk, Galindo and Panizza (2006) find that devaluation is only contractionary for the one-fifth of developing countries with a ratio of external dollar debt to GDP in excess of 84%; it is expansionary for the rest. 29
Why do debtor countries develop weak balance sheets in the first place? What is the origin of the currency mismatch? There are four theories. First, so-called “Original sin:” Investors in high-income countries are unwilling to acquire exposure in the currencies of developing countries.30 Second, adjustable currency pegs: an apparently fixed exchange rate lulls borrowers into a false sense of security and so into incurring excessive unhedged dollar liabilities.31 Third, moral hazard: borrowing in dollars is a way for well-connected locals to put the risk of a bad state of nature onto the government, to the extent the authorities have reserves or other claims to foreign exchange.32
Fourth, procrastination of adjustment: when the balance of payments turns negative, shifting to short-term and dollar-denominated debt are ways the government can retain the affections of foreign investors and thus postpone adjustment.33 These mechanisms, along with running down reserves and staking ministerial credibility on holding a currency peg, are part of a strategy that is sometimes called “gambling for resurrection.” What they have in common, beyond achieving the desired delay, is helping to make the crisis worse when it does come, if it comes.34 It is harder to restore confidence after a devaluation if reserves are near zero and the ministers have lost personal credibility. Further, if the composition of the debt has shifted toward the short term, in maturity, and toward the dollar, in denomination, then restoring external balance is likely to wreak havoc with private balance sheets regardless the combination of increases in interest rate versus increases in exchange rate.
We return to these issues when considering emerging market financial crises in section 8 of this survey.
High Inflation Episodes
Hyperinflation is defined by a threshold in the rate of increase in prices of 50% per month by one definition, 1000% per year by another.35 The first two clusters of hyperinflationary episodes in the 20th century came at the ends of World War I and World War II, respectively. The third could be said to have come at the end the Cold War: in Latin America, Central Africa, and Eastern Europe.36
Receiving more scholarly attention, however, have been the numerous episodes of inflation that, while quite high, did not qualify as hyperinflation. As Fischer, Sahay and Vegh (2002) wrote:
“Since 1947, hyperinflations … in market economies have been rare. Much more common have been longer inflationary processes with inflation rates above 100 percent per annum. Based on a sample of 133 countries, and using the 100 percent threshold as the basis for a definition of very high inflation episodes, … we find that (i) close to 20 percent of countries have experienced inflation above 100 percent per annum; (ii) higher inflation tends to be more unstable; (iii) in high inflation countries, the relationship between the fiscal balance and seigniorage is strong… (iv) inflation inertia decreases as average inflation rises; (v) high inflation is associated with poor macroeconomic performance; and (vi) stabilizations from high inflation that rely on the exchange rate as the nominal anchor are expansionary.”
Dornbusch and Fischer (1993), after the distinction between hyperinflation and high inflation, also made a distinction between high inflation episodes and moderate inflation episodes. The dividing line between moderate and high inflation is drawn at 40%. The traditional hypothesis is of course that monetary expansion and inflation elicit higher output and employment – provided the expansion is an acceleration from the past or a departure from expectations. In any case, at high rates of inflation this relationship breaks down, and the detrimental effects of price instability on growth dominate, perhaps via a disruption of the usefulness of price signals for the allocation of output.37 Bruno and Easterly (1998) found that periods during which inflation is above the 40% threshold tend to be associated with significantly lower real growth .