Chapter 7 Brazil: Taking Stock of the Past Couple of Decades

Marcos Chamon, David Hofman, Nicolas Magud, and Alejandro Werner
Published Date:
February 2019
  • ShareShare
Show Summary Details
João Barata and R. B. Barroso 

The opinions expressed in this chapter are the sole responsibility of the author and are not necessarily those of the Central Bank of Brazil.

This chapter describes the exchange rate regimes, possible policy rationales, and public operational guidelines for foreign exchange intervention in Brazil over the past couple of decades. Based on an assessment of intervention effectiveness, it proposes a flowchart that approximates the decision-making process and that may guide policymakers facing similar conditions. Historical records show that a managed floating regime often emerges in response to high levels of volatility in global liquidity. International reserve accumulation is the natural response to periods of high global liquidity and a good backing for interventions in periods of high volatility. Brazilian authorities have favored rules-based and preannounced strategies with swap interventions to address hedging demand, and spot or repo interventions to address liquidity pressures. They use discretion regarding the timing of the interventions and the corresponding announcements.


Foreign exchange intervention in Brazil is not for beginners.1 The exchange rate regime has been a managed float since the demise of the crawling peg in the late 1990s, but with evolving perceptions about the desirability of intervention and the pecking order of different intervention policies.

There is wide consensus that international reserve accumulation has served the country well, but there is also an ongoing debate about the level of reserves and its relation to different deployment strategies. Intervention policies during past crises have covered the spectrum of spot, credit, and futures market operations, all targeted and calibrated to very specific concerns of policymakers. While these policies tend to be discretionary regarding the timing of large intervention programs, their inclination has been for rules-based approaches that preclude signaling a preference for specific exchange rate levels, accompanied by a carefully designed communication strategy to avoid bad equilibrium outcomes. There are certainly many lessons to learn from their systematic study of the strategies and results in the past couple of decades.

The views presented in this chapter build heavily on commentary by the policymakers responsible for designing intervention policies over the years (Franco 2000; Fraga 2000; Bevilaqua and Azevedo 2005; Mesquita and Torós 2010). They also build on academic and policy work that explores a range of issues related to capital flows, exchange rates, and official interventions (Barroso 2011, 2014, 2017; Barroso, Pereira da Silva, and Sales 2016).

Given that the information is well documented by these and other sources, the picture that emerges from the exercise is panoramic in nature. By approaching a large set of regimes and intervention episodes, the similarities and basic rationale guiding policy decisions should become clear—this is probably the value added from this chapter relative to others in the literature. Another distinct advantage of the chapter is the cautious, normative motivation. The goal here is not only to summarize facts and opinions but also to extract useful evidence-based policy recommendations.2

The first challenge is to present the main features of the exchange rate regimes and associated intervention policies adopted over the past couple of decades. Linking the regime with the intervention results in a richer descriptive classification than the traditional de facto approaches offered in Ilzetzki, Reinhart, and Rogoff (2017). The analysis begins with the crawling peg of the 1990s and then moves along the managed float regimes that persist to this day. The rich experience with intervention policy implies that there are many significant subperiods to explore. The second section below takes some time to introduce the main features, rationale, and likely effectiveness of the intervention policies adopted in each of the exchange rate regimes.

The first years of floating offered a rare opportunity for experimentation, with progressive free floating giving way to managed floating on the back of a sequence of crises. These crises allowed policymakers to conduct a reality check, so to speak, of their preconceptions regarding the floating regime, and it set the basis for the intervention strategy in the subsequent years. The lessons of such a reality check are summarized in the third section of the chapter. They point to the need for accumulating international reserves in proportion to foreign liabilities—an effective leaning-against-the-wind strategy that hedges aggregate risk, reducing the incidence and severity of crises, and essential to efficient global safety net design.

There are noticeable regularities in the operation of intervention policy during the managed float period, which are also explored in the third section of this chapter. The timing of intervention follows global liquidity conditions, with good times stimulating reserve accumulation, and bad times leading to the official provision of foreign currency liquidity by the central bank. Policymakers define the size of interventions with transaction-level information on currency markets.

The communication strategy favors the preannouncement of the size of derivative interventions to buy time and to smooth balance sheet adjustment. It also favors unlimited interventions in spot and credit markets for as long as necessary. The effectiveness of intervention policy is explored in the fourth section of the text. The effect of intervention on the level and volatility of the exchange rate has the expected sign, according to the academic literature that explores instrumental variable or event study methodologies.

The fourth section of the chapter also presents a flowchart summarizing the decisions behind intervention policy in Brazil. It has both a descriptive dimension, as it fits the historical experience, and a normative dimension, as a blueprint for future policy. This flowchart will help policymakers from other jurisdictions to appreciate the Brazilian foreign exchange framework, and to see the wisdom of such a framework.

Exchange Rate Regimes

A panoramic view of exchange rate regimes in Brazil since the 1990s shows a trend toward a managed float with a large stock of international reserves— although not without experimentation around pegs and free-floating approaches. De facto classification systems, as in Ilzetzki, Reinhart, and Rogoff (2017), capture the broad boundaries of regimes but are probably too coarse to capture the rich experience in the country. It is particularly challenging to capture the different models during the managed-float period, especially during the 2000s.

Table 7.1 classifies recent experience, with a listing of the different periods, the de-facto regime, the economic context, and the broad lines of the intervention rationale. The text that follows offers supporting evidence for the classification. Figure 7.1 helps to gauge exchange rate dynamics in different periods; it is in real terms, so there is no trend associated with different levels of inflation (which is positive for Brazil). An important disclaimer is in order: To keep the focus on exchange rate regimes, this section pays only cursory attention to other aspects of the economic and political context.

Table 7.1.De Facto Exchange Rate Regimes in Brazil, 1994–2017
PeriodRegime (De Facto)ContextIntervention
1994–99Crawling pegReal planMonetary policy rationale
1999Free fallingCurrency crisisToo costly to peg
1999–2001Managed floatInflation targetingTried rules-based intervention
2002Free fallingPolitical uncertaintyToo costly to manage
2003–08Managed float + reserves accumulationInflation targeting with international reservesExternal adjustment + global liquidity = prudent to accumulate
2008Managed float + crisis interventionDollar liquidity crisis in spot, future, and creditIntervention in all markets followed by low interest rates
2009–12Managed float + reserve accumulationQuantitative easing in advanced economiesExcessive credit + US dollar liquidity = prudent to accumulate
2013–16Managed float + swap programTaper tantrum, first hike, China, commodity glutOfficial hedging supply rules-based intervention
2016–17Managed float + swap program reversedNormalization, benign interregnumPrudent to rebuild buffer, rules-based accumulation
Source: Author.
Source: Author.

Figure 7.1.Real Effective Exchange Rate in Brazil, 1993–2017

(Index, June 1994 = 100)

Source: Central Bank of Brazil.

From 1994 to 1999, the exchange rate regime in Brazil converged to a crawling peg after initial attempts with free floating. Indeed, the early 1990s in Brazil began with the “Real Plan,” the stabilization plan that famously ended hyperinflation in the country. As with many similar programs in emerging and developing economies at the time, the exchange rate peg was at the heart of the strategy. The creative twist was the phase-in of the peg with a unit-of-account-only currency, before actually issuing the sovereign currency. Another twist was that the new currency was actually allowed to float freely in its first few months of existence, which led to considerable appreciation.

After a good start, 1994–98 saw a succession of emerging market crises (the Tequila crisis in 1994, the Asian crisis in 1997, and the Russian crisis in 1998) that put a lot of pressure on the exchange rate. Policymakers chose to adopt a crawling peg in order to accommodate some of the pressure without severe domestic or external adjustments.3

Indeed, as the pressure on the peg mounted, the government entered an IMF program to defend it. In spite of the program, the continuing deterioration of the economy and costly necessary adjustments eventually led the government to abandon the peg in January 1999. The country then entered a free-falling exchange rate regime—per the terminology of Ilzetzki, Reinhart, and Rogoff (2017). The issuance of US dollar–linked government debt during the peg period smoothed the effect on private balance sheets, and thereby on economic activity, a lesson well taken for the future. While devising a new nominal anchor strategy that could accommodate a floating exchange rate, the central bank relied strongly on large interest rate differentials.4 In July 1999, after the renegotiation of the IMF program to accommodate the new policy framework, the government established the inflation-targeting regime with a floating exchange rate.5

From the establishment of inflation targeting in 1999 up to 2001, the exchange rate regime was that of managed floating. Within the spectrum of floating exchange rate regimes, it is fair to say that policymakers preferred free floating to managed floating during the first years of inflation targeting. In fact, exchange rate volatility was actually quite low at first, and little intervention was deemed necessary. Nevertheless, history has a tendency to repeat itself. Once again, after a good start, there was a succession of crises (Argentina’s default in 2001, the US terrorist attack in 2001, and Brazil’s election in 2002) that put great pressure on the currency. During the first two crises, policymakers offered foreign exchange hedges as the main intervention policy. By the end of 2001, the central bank had issued US dollar–linked securities that came close to $20 billion. In 2002, it announced a derivative contracts program to the same effect (the “swap contract” offers the exchange rate variation plus an onshore US dollar interest rate for the domestic interbank rate). Central bank–issued notes and swaps were substitutes for scarce reserves.6

Later in 2002, during the electoral campaign, convertibility risk became a primary concern to market participants: first, because of allegedly risky positions on foreign debt attributed to the election frontrunner, and second, because the public sector was a net debtor in foreign currency, which led to negative feedback between depreciation and the fiscal position. The rules-based swap program was not sufficient anymore, and discretionary spot interventions were the second line of defense. As seen in Figure 7.1, the high costs of defending the currency eventually led to a free-falling regime that endured until the postelection policy framework became clearer and the default risks were dissipated.

After the election, beginning in 2003, the new government acted aggressively to build credibility. A strong external sector adjustment made possible the initial accumulation of nonborrowed international reserves. At the same time, the central bank announced that it would no longer roll over 100 percent of its US dollar notes and swap contracts. It retired $9 billion of securities in 2003 and $26 billion in 2004. Global liquidity conditions in great part determined the pace of retirement, with discretion exercised in the rollover rate. By 2004, international reserve accumulation was an official policy and the net foreign creditor position an explicit goal for the economy. By 2006, the central bank began to issue reverse swaps to cope with one-sided appreciation bets in the futures market and to build a contrarian position to be unwound when conditions turned. The new policies meant frequent interventions in the spot and futures markets. The rules-based intervention for the spot market strove not to signal any preferred level for the exchange rate (as detailed in the next section). As shown in Figure 7.1, in spite of systematic intervention, the exchange rate strongly appreciated from 2003 to 2008. With international reserves increasing from under $50 billion to over $200 billion, the period is classified as a managed float with international reserve accumulation.

The payoff from the strategy became clear during the global financial crisis. The exchange rate depreciation experienced after September 2008 was much more “controlled” than the ones experienced in free-falling episodes of the recent past. The regime could be classified as a managed float with crisis intervention. Indeed, the accumulated buffer in the futures market allowed the central bank to offer ample hedge at low cost, while accumulated reserves allowed it to offer spot dollars to address a relatively small repatriation demand and to substitute for thin international trade finance. Instead of increasing the interest rate, as in the past crisis episodes, the central bank was able to reduce interest rates as soon as the US dollar markets normalized.7,8

An important aspect behind the success of intervention policies during the global financial crisis is that, with the international reserve accumulation policy and retirement of public foreign debt instruments, the country had become a net creditor in foreign currency. Therefore, the initial exchange rate depreciation strengthened the external position and increased the credibility of the intervention policy.

By the end of 2009, the central bank had already zeroed its position in future and credit markets and resumed the accumulation policy. From 2009 to 2012, quantitative easing in the United States was an important driver of capital inflow to emerging markets. Barroso (2017) estimates that around 50 percent of capital inflows to emerging markets was caused by quantitative easing by the US Federal Reserve—from $50 billion to $80 billion in the case of Brazil. The decision to steepen the accumulation of international reserves in that period was associated with an abundance of global liquidity. Reserves increased from around $200 billion to $350 billion, and the regime was once again a managed float with reserve accumulation.

Faced with high global liquidity and apparently improving balance sheets, the financial and nonfinancial private sectors tapped international capital markets. Policymakers offered some moderation with a mixture of interest rate policy, macroprudential policy, capital flow management, and leaning against the wind. Barroso, Pereira da Silva, and Sales (2016) show that quantitative easing boosted economic activity, credit markets, and asset prices—accounting for the full feedback of these factors on each other, and accounting for the endogenous policy responses in terms of macroprudential policy, monetary policy, and international reserve accumulation. In retrospect, it was hard to strike a balance, even though policymakers seemed confident. The taper tantrum in 2013 put this conviction to a test.

With the end of quantitative easing getting closer, and after announcements to that effect, it became increasingly clear that Brazil had overstretched its foreign borrowing and built external imbalances—despite the countervailing policies adopted in the previous period. Facing a potential “rush to the door,” policymakers adopted the time-honored decision to offer swap contracts to help private participants orderly hedge and retire exposures. This was the beginning of the managed-floating-with-swap-intervention program. In August 2013 the Central Bank of Brazil announced daily sales of $500 million in swap contracts until the end of the year. It extended the program in December of the same year and again in mid-2014—in response to deteriorating external conditions on the back of financial stress in China and hesitating moves by the US Federal Reserve.

By March 2015, the central bank had accumulated swap exposures of $108 billion. Based on the estimates of the quantitative easing-induced capital inflows to Brazil in previous years, it seems the program was more than adequate. With the political turbulence the country would face in the coming years (for example, an impeachment of the president), it proved wise to have had a large program. A crucial difference with analogous programs from the early 2000s is that international reserves now offered a hedge to the central bank, which minimized concerns with convertibility risks and contingent fiscal liabilities. Instead of the free-falling regimes of the past, it allowed for a controlled adjustment of the exchange rate (see the following section on the effectiveness of interventions).

With the worst of the political turbulence in the past, and given the onset of a benign external environment and the continuation of macroeconomic adjustment (in external balances, private balance sheets, and interest rates), some exchange rate appreciation was in order. Even though there had been no speculation of convertibility risk in the immediate past, the onshore US dollar interest rate and market commentary did provide warning signs during peak political turbulence. Policymakers took notice and turned their attention to the concept of international reserves net of the swap position. With this mindset, and assuming the benign external environment was temporary, the central bank began issuing large amounts of reverse swaps to reduce its net swap position—even if implying a slightly slower convergence of inflation to the target. By the end of 2017, with a pause in the second semester of 2016 to reassess the benign external environment, the central bank had reduced its swap position to $24 billion. This is referred to in Table 7.1 as “managed float + swap program reversed.” In both the swap and reverse swap periods, there was no further accumulation of international reserves.

It is interesting that after all the swings in the exchange rate and exchange rate regimes of the past couple of decades, at the end of 2017 the real exchange rate was back to the point where it began in 1994.

Intervention Strategies and Rationale

Reality Check from the First Years of Floating

The good initial performance of the floating exchange rate appeared to corroborate a more benign view of free floating.9 In admittedly simplistic terms, this view is summarized with three propositions: First, the exchange rate mostly reflects fundamentals, although with short-run overshooting a la Dornbusch. Second, depreciation stimulates the economy with foreign demand and is a first line of defense in a crisis. Third, liquidity problems are rare and can be handled within the multilateral safety net system. The next years of the floating regime saw the reemergence of crises and exchange rate volatility, and the ensuing reality check and slow convergence toward a new consensus with a more nuanced view on exchange rate flexibility.

This is the first reality check of the period: Leaning against the wind might be a good policy after all, since volatility and persistent trends are disconnected from fundamentals and from efficient allocations. There is hardly any controversy in the statement that the exchange rate should reflect fundamentals. In principle, this allows it to function as an effective shock absorber and to give correct signals to private agents and policymakers. The problem is that floating exchange rates are much more volatile than fundamentals (Obstfeld and Rogoff 2000). Evans (2012) shows that interest rate differentials cannot explain exchange rate variance for either short or long horizons.

The literature often relies on risk-sharing shocks or uncovered interest parity shocks, sometimes reverse engineered into a preference shock, to account for the stylized facts, but the debate is still open. In principle, if no clearly identified fundamental shocks drive the exchange rate volatility, it seems unreasonable to suppose such volatility would support efficient allocations. In reality, it leads to risk premiums that likely distort allocations across sectors and time and affect the transmission of shocks to inflation through synchronized price setting. The persistence of the shock is also a concern. It is reasonable to expect (and consistent with empirical evidence) that long appreciation trends are followed by sharp depreciations, as it is difficult to shift resources to a previously shrinking tradable sector (Caballero and Lorenzoni 2007).

Stepping back a moment, it is not even clear if depreciation is a net positive for economic activity, in all cases, through its effect on the foreign demand channel. For emerging markets, the common incidence of external credit constraints and private sector reliance on foreign currency debt have wide-ranging implications. Large depreciations interact negatively with foreign currency debt, which leads to a reduction of external credit lines and market access. The resulting reductions in domestic demand might be larger than the increase in foreign demand. This is the case in a relevant class of models (Jeanne and Korinek 2010), where the reduction in domestic demand falls more heavily on nontradables and further depreciates the currency, which creates a negative feedback loop.

This is the second reality check of the period: It seems important to offer a hedge or outright liquidity to the private sector to minimize the negative financial accelerator mechanism associated with depreciations. If convertibility is not an issue, which is the case when large international reserves are available, the hedge seems to be a good first line of defense. The argument depends on assuming that tapping foreign capital markets and allowing private balance sheets to have exposure to the risk will yield welfare gains. It also depends on the foreign exchange markets being sufficiently underdeveloped to raise significantly the incidence of one-sided markets.10

For the multilateral system to work, crises must be sufficiently infrequent and nonsynchronized. The emerging market crisis of the 1990s and the global financial crisis warrant some caution in this regard. In parallel to new injections of resources at the multilateral level, international reserve accumulation accelerated in emerging markets. Barroso (2011) documents that crisis episodes in Latin America, broadly defined, are less frequent, although in a nonlinear way, in the presence of a large ratio of reserves to foreign liabilities. That is the third reality check: Besides smoothing the severity of a crisis, international reserves reduce the incidence of crises and are therefore key ingredients of the global safety net. As already pointed out, emerging market crises are essentially balance sheet crises. The ratio of large international reserves to external liabilities could, in principle, minimize such risks.

There we have the main elements of the “new consensus” emerging out of the first years of floating. The central bank should lean against the wind in the exchange rate market to improve allocative efficiency and to accumulate reserves in the case of appreciation trends. It should be ready to offer hedge and liquidity to private participants in the case of one-sided markets. International reserves should be commensurable with external debt to reduce crisis severity and incidence, complementing the global financial safety net. The next sections detail the intervention strategy adopted in subsequent years and the rationale for the operational framework.

International Reserve Accumulation Policy

Interventions in the spot market in Brazil from 2004 to 2012, excluding interventions in 2008 because of the global financial crisis, have followed an official policy of international reserves accumulation. In fact, this was a de facto rather than official policy until the global financial crisis, after which it conspicuously appeared in official speeches and documents. For example, in the first edition of the “International Reserve Management Report,” published in June 2009, the central bank states that the increase in reserves is the “result of the reserve accumulation policy started in 2004.” The motivation for the policy (as explained in the previous section), was to build a sufficiently large buffer to the economy while leaning against appreciation trends. The policy has an operational framework, which covers daily intervention strategy, timing, rationale, and tentative reserve adequacy metrics. As for the timing, it is associated with a set of indicators, including appreciation trends, capital inflows, and growth in foreign debt liabilities that occur in the context of global liquidity—such as the ones prevalent in the two major periods of international reserve accumulation.

The basic operational guideline is that the central bank should “buy net order flow.” This is a rules-based definition of the size of daily interventions. As explored in the foreign exchange market microstructure literature (Evans and Lyons 2002; Vitale 2007), net order flow is the main proximate driver of exchange rate dynamics. The rationale for this result is that interdealer trade based on private customer order flow reveals the aggregate order flow to the market, and this public information is fully priced by the market at the end of the day. If the central bank buys from dealers based on central bank observations of aggregate order flow, it affects interdealer trade but conveys no additional information to the market. The same result follows if the central bank buys order flow less than proportionally or up to an error. This flexibility is relevant for the actual operation of daily interventions, given its interaction with onshore dollar liquidity.

Onshore interest rates tend to respond to dollar liquidity. As a result, systematically buying in excess of order flow tends to attract even more order flow. To avoid these negative feedback dynamics, the operational rule could react less than proportionally to the order flow and include a random component to the decision. This operational rule is a good description of actual interventions during accumulation episodes.11 The rule reduces the effectiveness of the leaning-against-the-wind policy and the pace of accumulation (see below for the link between the size of the intervention and the effect on the exchange rate). In addition, the proportion between intervention and order flow might be modulated in the presence of policy trade-offs, such as appreciation trends that help control inflation through their effect on tradable goods, while still offering some support to activity through favorable investment conditions.

The central bank communicated its rules-based intervention strategy to convey that it was not targeting any specific level for the exchange rate or any specific target for the rate of change of the exchange rate. “Buy net order flow” is a clever compromise between leaning against the wind and the floating regime.

The implementation of the “buy net order flow” rule relies on the central bank having an informational advantage in the spot market. This is the case in Brazil, where policymakers have complete information on the net order flow, based on the electronic records of private transactions reported by financial institutions. Each participating institution in the over-the-counter spot market collects partial information through its own order flow, but not the aggregate total. As a result, even though the central bank is following a rules-based approach, market participants cannot anticipate the size of the intervention on any single day and must incorporate these transactions in the interdealer market.

There is no official reserve adequacy metric. However, the reality check from the initial years of floating made clear the importance of the reserves to foreign debt ratio. The excess of debt relative to reserves prompts negative feedback loops in depreciation episodes. As seen in Figure 7.2, from 2004 to 2008, intervention raised international reserves to the same order of magnitude as the sum of public and private debt. This was presented sufficiently many times in official central bank communication to make it a plausible de facto adequacy metric. Other metrics, such as the ratio to short-term debt or imports, were also mentioned with some frequency, but reserves were much larger than what would be indicated by such metrics. During the second accumulation period, from 2009 to 2012, reserves increased almost in tandem with external debt, so again the ratio looked like a good adequacy metric candidate, although with a large safety margin. To the extent that it signaled a fully developed hedging market, domestic institutional investors’ holdings of foreign assets could be another factor to be considered in the future.12

Figure 7.2.International Reserves and Net Public and Private Debt, 1993–2017

(Billions of US dollars)

Sources: Central Bank of Brazil; and author.

One missing piece of the argument is the cost of holding international reserves. In a rare public assessment by Central Bank of Brazil staff, based on precautionary models, Silva (2011) discusses a range of scenarios for international reserves cost and the expected output loss during a crisis and finds that if a crisis occurred only once every decade or so, the avoided losses would compensate the cost. Such exercises are viewed with skepticism in policy circles, where self insurance is not usually seen as the exclusive rationale for accumulating reserves. For example, it is often mentioned in international policy circles that reserves might have a role in reducing external credit constraints, improving financial stability, or stabilizing net foreign asset positions. It is also mentioned that reserves might be at whatever level results from leaning-against-the-wind policies of the past, given the reputational cost of reducing them afterward. Even in the context of very simple models that are based on self insurance, one must recognize that there is a high level of uncertainty regarding the effect of international reserves on crisis incidence and severity. High uncertainty aversion typically means that holding excessive reserves is much better than holding just the exact amount of reserves.

Intervention during the Global Financial Crisis

The global financial crisis was very much a US dollar liquidity crisis. During the crisis, the Central Bank of Brazil effectively acted as a foreign currency liquidity provider of last resort in spot, credit, and futures markets. The spot market US dollar shortage came from foreign investors’ demand to repatriate funds. The credit market shortage came from exporters that lost credit lines abroad. The futures market shortage came from exporters that “overhedged” during the past appreciation trend, as well as from private participants, including banks, who had to roll over short-term hedges of medium- or long-term US dollar liabilities. The interventions were successful, judging by the normalization of US dollar liquidity captured, for example, by the spread between onshore and offshore US dollar interest rates (in January 2009, these measures were back to the levels of August 2008). Interventions were also short-lived in comparison with the protracted consequences of the crisis in global financial markets. They also benefited from the extraordinary swap arrangements by the US Federal Reserve on a global scale.

Communication was a key element of the policies. The Central Bank of Brazil preannounced a ceiling to the entire volume of swap interventions to reassure private participants, and it announced it would offer as many US dollars as necessary in the spot and credit markets. Stone, Walker, and Yasui (2009) provide an interesting empirical assessment of the effects of interventions by the central bank during the financial crisis in 2008. In general, announcements are found to have a greater impact on the level of the exchange rate than the interventions themselves. This result signals two things: First, there was sufficient central bank credibility; and second, there were sufficient reserves to support the claims. Another communication strategy was that the central bank continued to report its trade credit assets as qualified international reserves.

Calibrating the size of the interventions captured in Figure 7.3 was key to their success. For the spot market, the central bank had complete information from the order flow, disaggregated to the transaction level, and so could make a good diagnosis of the source of the liquidity demand. The bank also had good estimates of the net international liability position at a disaggregated level, particularly the debt instruments. The data indicated that repatriation was manageable, given that the central bank would also reduce hedging costs with swap interventions. In total, the central bank sold $14.5 billion in spot auctions, or 7 percent of pre-crisis reserves.

Figure 7.3.International Reserves, Swap Position, and Repo Position, 2003–17

(Billions of US dollars)

Source: Central Bank of Brazil; and author.

The central bank also had full information on trade credit operations down to the transaction level, which is not large in Brazil, relative to the size of the economy, and there is a stable relationship between trade credit and trade activity. Based on this assessment, authorities committed to offer as much US dollar credit as necessary, until credit markets resumed normal operations. In total, the central bank loaned $24.5 billion in repo auctions, or 12 percent of precrisis reserves.

As already noted, preannouncing a high ceiling for swap operations was essential to the intervention communication strategy. The central bank has access to registers at the central derivative clearing of the country, which provide a good basis for estimating the total hedging needs. Matching this with debt information registered with the central bank, one has a good idea of the rollover needs, in case there is a mismatch in maturity of debt and hedges. The coarse granularity of the market, with enough big players and public comprehensive balance sheet information, also helped in the assessment. Informed by this exercise, the $50 billion initial ceiling looked like a sufficiently high number. Considering actual interventions, the central bank sold $33.0 billion in swap auctions, or 16 percent of precrisis reserves, which gives a rough estimate of the spot market pressure one would face without swap intervention. Because the swap is a nondeliverable forward contract with domestic currency settlement, there was essentially no pressure on international reserves.

It is worth commenting on the interaction of foreign exchange intervention policy and monetary policy during the crisis. As noted, for the first time Brazil was able to use countercyclical monetary policy during a crisis. However, the monetary easing did not come immediately after the crisis. The strategy at the time, communicated by central bank authorities, was first to normalize the transmission channels and then begin monetary policy normalization. Accordingly, the central bank first announced its intervention package, including foreign exchange intervention, and then made sure the liquidity squeeze subsided before beginning a new easing cycle. In this context of the global financial crisis, it is fair to assume that an aggressive easing would interact adversely with the US dollar liquidity crisis.

The Swap Program after the Taper Tantrum

The liquidity squeeze in spot and futures US dollar markets was not as large as during the crisis, but private balance sheets were more fragile. The main concern was accumulated foreign debt from the quantitative easing period. There was a lot of anxiety in global markets with policy normalization. From the taper tantrum in May 2013 to liftoff in December 2015, and the second hike in federal funds in November 2016, the normalization process advanced by a sequence of bold moves intercalated with setbacks and financial turbulence. The motivation for the intervention policy was to avoid panic and allow rational calculations to dominate the adjustment of debt profiles and foreign currency exposures. Spot interventions could have the opposite effect and induce market participants to liquidate positions as soon as possible. Swap interventions create conditions for adjustment, and by the nature of nondeliverable forwards, preserve international reserves for use in case of panic.

The key communication strategy to obtain such an effect is to preannounce a “sufficiently large” intervention. Unlike the crisis intervention that set a total ceiling and left some freedom for daily operations, this time around, both the total amount and the daily interventions were set in advance (see the previous section for the exact timing and content of the announcements). The very strict rules-based intervention was designed, in part, to avoid signaling any preference for the level of the exchange rate.

To set the size of the swap intervention, policymakers must consider a large number relative to foreign liabilities, based on the consolidated information from the central bank and central clearings datasets. The level of corporate sector debt in the country around the tapering of quantitative easing was close to the average emerging market, just below 50 percent of GDP, with 65 percent in local currency versus 35 percent in foreign currency. Of the 35 percent in foreign currency, 12 percent referred to exporters, 6 percent to nonexporters with local hedge, 5 percent to nonexporters without local hedge but with foreign headquarters, 5 percent to nonexporters without local hedge but with foreign assets, and 6 percent to unhedged exposure. This last part alone totaled $70 billion; and there were residual unhedged positions in all debt categories. It is difficult, however, to distinguish what is the normal level of risk taking from what is excessive and therefore susceptible to instability. One alternative is to estimate what the debt levels would have been without quantitative easing. Estimates known to policy-makers at the time point to a figure of the same order of magnitude as unhedged plus some imperfectly unhedged positions (Barroso 2017; Barroso, Pereira da Silva, and Sales 2016). By this metric, it was clear that conditions warranted a sizable intervention.

However, large interventions are not without their problems. First, lower cost of hedging could induce more debt-taking abroad to build arbitrage positions, which is self-defeating, for the stated purpose of the policy. Second, market participants must believe there will be no convertibility problems in the future to keep hedging positions with nondeliverable forwards, so that a large enough swap position might lead the market to self-fulfilling concerns with convertibility, and again the policy is self-defeating. Garcia and Volpon (2014) provide an insightful discussion of the problems, which only increased with the accumulation of swaps in the balance sheet of the central bank. At some point, market participants began focusing on international reserves net of swaps, which was probably an early sign that the intervention was reaching its limit. Policymakers recognized these issues. In particular, the link between swap interventions and US dollar liquidity was clear enough to lead policymakers to conduct regular repo line auctions, either as a feature of the program or conditionally on the onshore US dollar rate (an indicator of liquidity demand).

The policy would start to be reversed once domestic political volatility reduced, and global liquidity conditions improved—at first by discretionary rollovers and then by the auction of reverse swaps to imprint a fast-paced reduction of the swap position. There was no official statement on the goal of reducing swaps to zero, and so there was considerable flexibility in the conduct of the intervention policy. Official communication depicted this movement as an accumulation of “international reserves net of swaps.” It seems this last concept conveys accurately the size of the buffer available to cope with a crisis.

Assessing Intervention Policy

Intervention Effect on the Exchange Rate

The reality check from the first years of floating maintains that leaning against the wind is a sensible strategy. However, this requires sterilized intervention (or swap intervention) to have an actual effect on the exchange rate, which is a controversial proposition. Empirically, it is usually difficult to estimate the effect of intervention due to the simultaneity problem. If the central bank sells foreign currency when the domestic currency depreciates, a naive empiricist might say it caused the depreciation. Indeed, this is the sign of the correlation and the sign of coefficients in ordinary regressions. The traditional econometric solution to get the right causal effect is to find an instrumental variable, that is, a random variable related to the depreciation only through its impact on the central bank. The literature uses instrumental variables related to news, market expectations, and the reaction function of the central bank with mixed results (Dominguez and Frankel 1993; Galati, Melick, and Micu 2005; Kearns and Rigobon 2002; Tapia and Tokman 2004).

For the case of Brazil, one may highlight the results from Barroso (2014), who adopted a method that is both general and distinct from the literature. It begins by showing that realized volatility (that is, the intraday sample volatility) fits the formal requirements of an instrumental variable by pure deductive reasoning. A large class of models makes realized volatility depend on volatility—such as offered in Hansen, Huang, and Shek (2011). Simple algebra then shows lagged realized volatility is orthogonal to the innovations. The correlation with intervention policy is ensured by central bank behavior, which at least in Brazil, is usually cautious with exchange rate volatility. The method allows weighted estimation, which proved to be an important feature. If the endogeneity problem is particularly severe in high volatility periods, with the intervention failing to completely reverse foreign exchange shocks, then it makes sense to reduce their weight.

Across several specifications, including the nonparametric instrumental variable, results are similar. The average effect of a $1 billion sell or buy intervention is a depreciation or appreciation of around 0.50 percent, respectively. In the alternative specification that includes swap interventions as a control, the estimate for spot is a bit lower (0.30 percent), which suggests intervention policies are complementary. In this alternative specification, the effects from swap operations are around 0.25 percent, with the correct sign, but are not statistically significant in the sample (2007–11). This last result might be related to the preannouncement of an intervention ceiling during the global financial crisis, which tends to generate some effects on impact and to confound the effect of ensuing swap auctions. All the results in this paragraph are weighted estimators, which are generally deemed more reasonable by central bank staff in foreign exchange desks and familiar with the method.

Several papers investigate the effects of spot interventions on the level of the real against the US dollar using different methodologies. Wu (2010) studies spot interventions with a structural vector autoregression motivated on microstructure models. Kohlscheen and Andrade (2013) study swap interventions with an event study method using intraday data and a dozen events. It is interesting that both studies look into high-frequency and microstructure features as solutions to the simultaneity problem, not unlike the realized volatility approach emphasized here. Although instrumental variable identification is not generally better, it is less demanding on the identifying assumptions. In the case of realized volatility, the assumptions are actually algebraic necessities, which make them particularly attractive.

Assessing the effects of swap interventions is potentially more complicated than spot interventions. This is because of the preannouncement of such policies. The announcement of a new program is usually unexpected, while its implementation is predictable.13 The interventions from 2013–16 are particularly hard to assess, because daily interventions were essentially deterministic. The appropriate methodology for this case is to build an empirical counterfactual and compare it to the actual exchange rate behavior. For example, the event study method assesses abnormal behavior in this manner. Chamon, Garcia, and Souza (2017) use a synthetic control approach that assumes exchange rates from peers provide a good basis to build a counterfactual for the Brazilian exchange rate without swap operations. The method assumes stability of the correlation with peers, which might not be the case when the market is reassessing relative fragilities. Other than that, it seems like a sound empirical strategy. Their results point to significant exchange rate effects only in the first swap announcement. Nevertheless, one must consider that the stated purpose of the intervention was not to lean against the wind of the exchange rate, but rather to smooth the effect of external turbulence on private balance sheets.

A Flowchart of Intervention Strategies

At the risk of oversimplification, this section attempts to compress the received wisdom from the past couple of decades into a simple flowchart (Figure 7.4). The figure represents a feedback system, so that after running through the decisions, one goes back to the initial position, represented by the open circle. It would be naive to expect that by following this exercise, another central bank could replicate the performance of the Brazilian experience. There is considerable judgment in every branch of the decision tree and considerable reliance on specific features of the domestic markets and the domestic surveillance system.

Figure 7.4.Flowchart of Intervention Policy, Based on the Brazilian Experience

Source: Author.

1 A dollar liquidity problem implies a one-sided market for spot dollar, dollar hedge, or dollar credit.

2 To determine whether the reserves are sufficient, estimate the demand size with data from central clearings, order flow, and market participants.

3 After estimating the size of hedging demand, one should preannounce swap interventions up to this size and implement spot and repo interventions as required to reduce onshore dollar rates.

4 Also factor in trade-offs with business or credit cycles.

The first diamond of the flowchart poses a question about global liquidity. In case of US dollar liquidity problems (as captured by the monitoring of gross order flow and market prices), one should enter in a liquidity smoothing mode. Liquidity is broadly understood to capture spot, future, and credit market pressures. If there are sufficient international reserves for massive intervention (as assessed per the available information from granular debt, derivative, and order flow data at the transaction level), it should be undertaken. Unless there are signs of spot market liquidity pressures (typically high onshore US dollar rates) or convertibility concerns, swap interventions should be the main policy instrument. If reserves do not appear to be sufficient, external credit lines and swap agreements should be called, along with swap and eventual spot interventions, to buy time, but the economy should be ready for free floating in volatile conditions.

In case of abundant US dollar liquidity, one should take the opportunity to exit as much as possible from previous swap interventions. If reserves are less than adequate (typically assessed by its proportion to foreign currency debt of the private sector, which is ideally close to but higher than unity), policymakers should take the opportunity to accumulate liquid US dollar assets. The pace of interventions should consider trade-offs with monetary policy, particularly if appreciation is desirable from that perspective. As in every other step in the decision process, this too requires subjective input from policymakers.

In terms of governance, the authority for intervention policy decision-making lies with the board, with strong leadership from the governor and the deputy governor who have a mandate to oversee market operations, including monetary policy and foreign exchange operations. During periods with acute liquidity problems, other members of the board get closely involved as well. Other than that, the decision-making process of the board is fluid and case-specific. It is not clear whether a more richly detailed flowchart of intervention policy will be warranted in the future.


This chapter explores exchange rate regimes, policy rationales, and operational guidelines for foreign exchange intervention in Brazil over the past couple of decades.

The chapter culminates with an assessment of intervention effectiveness that includes a flowchart that approximates the decision-making process and guides policymakers in implementing exchange rate policy. As such, it has the received wisdom from the Brazilian experience. Policymakers from other jurisdictions should view these lessons with the understanding that they rely on specific features of Brazilian financial markets—such as the large nondeliverable forward and the low level of participation of domestic institutional investors in foreign asset markets and domestic hedging markets.

Much of this wisdom is the result of experimentation in the face of wildly shifting global liquidity conditions, international crises, and domestic uncertainty. The managed float solution emerges almost as a necessity to cope with the high levels of volatility and an inflation-targeting regime. The main lesson from the experimentation is that accumulating international reserves in amounts commensurate to foreign debt liabilities and keeping an otherwise stable macro-environment is a good strategy. It allows leaning against the wind when fragilities are being accumulated, smoothing the adjustment of private balance sheets during less tranquil periods, and reducing crisis frequency and severity. The operational guidelines favor rules-based and preannounced intervention as much as possible, to both preclude signaling central bank preferences on the exchange rate level and reassure markets of normal liquidity conditions going forward. In terms of instruments, swaps address a demand for hedging before it transforms into a demand for liquidity, while direct spot market intervention addresses liquidity pressures. It also preserves sufficient discretion in the timing of intervention programs, which minimizes possible moral hazard concerns and fosters risk-sharing with market participants.

Based on history, the framework is likely to evolve in the next couple of decades, once policymakers and the international community discuss and implement novel solutions and conceptual frameworks to deal with global volatility. It is not yet clear if domestic reforms will inspire large institutional investors with considerable investment abroad to hedge external debt in periods of short dollar liquidity. Nor is it clear how the now-frequent arrangements among groups of countries to share international reserves will evolve in the future. Coupled with the requirement that part of the reserve sharing has to be linked to lending programs by international organizations, it is clear that multilateral arrangements continue to be important. At the same time, several jurisdictions have accumulated large amounts of international reserves and have gathered evidence of their effectiveness in reducing exchange rate volatility.


    BarrosoJoãoBarata. 2011. “Optimal Capital Flow Taxes in Latin America.” Central Bank of Brazil Working Paper Series 268Central Bank of BrazilBrasília.

    • Search Google Scholar
    • Export Citation

    BarrosoJoãoBarata. 2014. “Realized Volatility as an Instrument to Official Intervention.” Central Bank of Brazil Working Paper Series 363Central Bank of BrazilBrasília.

    • Search Google Scholar
    • Export Citation

    BarrosoJoãoBarata. 2017. “Quantitative Easing and United States Investor Portfolio Rebalancing towards Foreign Assets.” In International Spillovers of Monetary Policy edited by ÁngelEstrada García andAlbertoOrtiz Bolaños. Mexico City, Mexico: Center for Latin American Monetary Studies22585.

    • Search Google Scholar
    • Export Citation

    BarrosoJoãoBarataLuiz Pereirada Silva andAdrianaSales. 2016. “Quantitative Easing and Related Capital Flows into Brazil: Measuring Its Effects and Transmission Channels through a Rigorous Counterfactual Evaluation.” Journal of International Money and Finance67(C): 10222.

    • Search Google Scholar
    • Export Citation

    BarrosoJoãoBarataRodrigoBarbone Gonzalez andBernardusVan Doornik. 2016. “Credit Supply Responses to Reserve Requirement: Loan-level Evidence of Macroprudential Policy.” BIS Working Paper 674Bank for International SettlementsBasel.

    • Search Google Scholar
    • Export Citation

    BevilaquaAfonso andRodrigoAzevedo. 2005. “Provision of FX Hedge by the Public Sector: The Brazilian Experience.” BIS Paper No. 24Bank for International SettlementsBasel.

    • Search Google Scholar
    • Export Citation

    BursteinArielMartinEichenbaum andSergioRebelo. 2005. “Large Devaluations and the Real Exchange Rate.” Journal of Political Economy113 (4): 74284.

    • Search Google Scholar
    • Export Citation

    CaballeroRicardo andGuidoLorenzoni. 2007. “Persistent Appreciations and Overshooting: A Normative Analysis.” NBER Working Paper 13077National Bureau of Economic ResearchCambridge, MA.

    • Search Google Scholar
    • Export Citation

    CarvalhoCarlos Vianna andAndréVilela. 2015. “What If Brazil Hadn’t Floated the Real in 1999?Brazilian Review of Econometrics35: 265.

    • Search Google Scholar
    • Export Citation

    ChamonMarcosMárcioGarcia andLauraSouza. 2017. “FX Interventions in Brazil: A Synthetic Control Approach.” Journal of International Economics108: 15768.

    • Search Google Scholar
    • Export Citation

    DominguezKathryn andJeffreyFrankel. 1993. “Does Foreign-Exchange Intervention Matter? The Portfolio Effect.” American Economic Review83 (5): 135669.

    • Search Google Scholar
    • Export Citation

    EvansMartin. 2012. “Exchange-Rate Dark Matter.” IMF Working Paper 12/66International Monetary FundWashington, DC.

    EvansMartin andRichardLyons. 2002. “Order flow and exchange rate dynamics.” Journal of Political Economy110 (1): 17080.

    FragaArminio. 2000. “Monetary Policy during the Transition to a Floating Exchange Rate: Brazil’s Recent Experience.” Finance & Development37 (1): 1618.

    • Search Google Scholar
    • Export Citation

    FrancoGustavo. 2000. The Real Plan and the Exchange Rate: Essays in International Finance. Princeton, NJ: Princeton University.

    GalatiGabrieleWillianMelick andMarianMicu. 2005. “Foreign Exchange Market Intervention and Expectations: The Yen/Dollar Exchange Rate.” Journal of International Money and Finance24 (6): 9821011.

    • Search Google Scholar
    • Export Citation

    GarciaMárcio andTonyVolpon. 2014. “DNDFs: A More Efficient Way to Intervene in FX Markets?Discussion Paper 621Pontifical Catholic University of Rio de JaneiroRio de Janeiro.

    • Search Google Scholar
    • Export Citation

    HansenPeterZhuoHuang andHowardShek. 2011. “Realized GARCH: A Joint Model for Returns and Realized Measures of Volatility.” Journal of Applied Econometrics27 (6): 877906.

    • Search Google Scholar
    • Export Citation

    IlzetzkiEthanCarmen M.Reinhart andKenneth S.Rogoff. 2017. “Exchange Arrangements Entering the 21st Century: Which Anchor Will Hold?NBER Working Paper 23134National Bureau of Economic ResearchCambridge, MA.

    • Search Google Scholar
    • Export Citation

    JeanneOlivier andAntonKorinek. 2010. “Excessive Volatility in Capital Flows: A Pigouvian Taxation Approach.” American Economic Review100 (2): 40307.

    • Search Google Scholar
    • Export Citation

    KearnsJonathan andRobertoRigobon. 2002. “Identifying the Efficacy of Central Bank Interventions: The Australian Case.” NBER Working Paper 9062National Bureau of Economic ResearchCambridge, MA.

    • Search Google Scholar
    • Export Citation

    KohlscheenEmanuel andSandroAndrade. 2013. “Official FX Interventions through Derivatives.” Journal of International Money and Finance47: 20216.

    • Search Google Scholar
    • Export Citation

    MesquitaMário andMárioTorós. 2010. “Brazil and the 2008 Panic.” BIS Paper 54Bank for International SettlementsBasel.

    ObstfeldMaurice andKenethRogoff. 2000. “The Six Major Puzzles in International Macroeonomics: Is There a Common Cause?NBER Macroeconomics Annual15: 33990.

    • Search Google Scholar
    • Export Citation

    SilvaJr.Antonio. 2011. “The Self-Insurance Role of International Reserves and the 2008–2010 Crisis.” Central Bank of Brazil Working Paper 256Central Bank of BrazilBrasília.

    • Search Google Scholar
    • Export Citation

    StoneMarkChristopher W.Walker andYosukeYasui. 2009. “From Lombard Street to Avenida Paulista.” IMF Working Paper 09/259International Monetary FundWashington, DC.

    • Search Google Scholar
    • Export Citation

    TapiaMatías andAndreaTokman. 2004. “Effects of Foreign Exchange Intervention under Public Information: The Chilean Case.” Economia4: 142.

    • Search Google Scholar
    • Export Citation

    VitalePaolo. 2007. “An Assessment of Some Open Issues in the Analysis of Foreign Exchange Intervention.” International Journal of Finance & Economics12 (2): 15570.

    • Search Google Scholar
    • Export Citation

    WuThomas. 2010. “Order Flow in the South: Anatomy of the Brazilian FX Market.” Mimeo University of California Santa Cruz Santa Cruz.

    • Search Google Scholar
    • Export Citation
1This line borrows from Tom Jobin, a Brazilian composer, who once said, “Brazil is not for beginners.”
2Of course, the future may be sufficiently different from the past to render these conclusions useless; with this caveat, it seems a good idea to rationalize past decisions in search of guidance for the future.
3See Franco (2000) for a firsthand account and the next section for the economic rationale for minimizing exchange rate volatility in general.
4See Fraga (2000) for a firsthand account of the brave few months of free floating.
5See Carvalho and Vilela (2015) for a counterfactual evaluation of the costs associated with maintaining the peg versus implementing the inflation-targeting regime with floating.
6See Bevilaqua and Azevedo (2005) for an overview of the period by policymakers.
7Monetary policy also waited for interbank market normalization obtained by the reduction of reserve requirements. See Barroso, Gonzalez, and Doornik (2016) for a review and evaluation.
8See Mesquita and Torós (2010) for the official view from policymakers.
9This is, in part, because of the small pass-through of the large initial depreciation. Burstein, Eichenbaum, and Rebelo (2005) show that this results from the fact that nontradable goods demand does not respond strongly to depreciation in emerging markets, and therefore does not pressure wages, and that it is costly to unilaterally reset prices.
10The Brazilian foreign exchange market is deep and liquid, but arguably underdeveloped, in part due to low levels of foreign asset holdings by domestic private investors with long horizons.
11The regression coefficient of weekly interventions on weekly order flow from 2007 to 2013, in periods of positive spot interventions, is around 0.26, which explains 32 percent of the variation in interventions.
12As of June 2018, the ratio of domestic and foreign cross holdings of equities, debt, and credit was 12, 6, and 9 percent, respectively, which relates to the relevance of international reserves to this day.
13Historically, the exit of the policies has also been less predictable, and hence, amenable to the instrumental variable approach.

    Other Resources Citing This Publication