Exchange-Rate Swings and Foreign Currency Intervention
Author:
Andrew Filardo 0000000404811396 https://isni.org/isni/0000000404811396 International Monetary Fund

Search for other papers by Andrew Filardo in
Current site
Google Scholar
PubMedClose
,
Mr. R. G Gelos
Search for other papers by Mr. R. G Gelos in
Current site
Google Scholar
PubMedClose
https://orcid.org/0000-0003-4362-676X
, and
Thomas McGregor 0000000404811396 https://isni.org/isni/0000000404811396 International Monetary Fund

Search for other papers by Thomas McGregor in
Current site
Google Scholar
PubMedClose
Publication Date:
29 Jul 2022
eISBN:
9798400215322
ISBN:
9798400215322
Language:
English
Keywords:
intervention proxy; interventions decade; exchange rate misalignment; commodity terms of trade data; FXI IV policy shock
Download PDF (4.1 MB)

This paper develops a new approach for exploring the effectiveness of foreign currency intervention, focusing on real exchange cycles. Using band spectrum regression methods, it examines the role of macroeconomic fundamentals in determining the equilibrium real exchange rate at short-, medium-, and low frequencies. Next, it assesses the effectiveness of FX intervention depending on the degree of cycle-specific misalignments for 26 advanced- and emerging market economies, covering the period 1990–2018, and using different techniques to mitigate endogeneity concerns. Evidence supports the hypothesis that central banks can lean effectively against short-run cyclical misalignments of the real exchange rate. The effects are present in quarterly data—i.e., at policy-relevant horizons. The effectiveness of intervention rises with the size of the misalignment, and with the duration of one-sided interventions. FX sales appear to be somewhat more effective than FX purchases, and intervention is less effective in more liquid FX markets.

Abstract

This paper develops a new approach for exploring the effectiveness of foreign currency intervention, focusing on real exchange cycles. Using band spectrum regression methods, it examines the role of macroeconomic fundamentals in determining the equilibrium real exchange rate at short-, medium-, and low frequencies. Next, it assesses the effectiveness of FX intervention depending on the degree of cycle-specific misalignments for 26 advanced- and emerging market economies, covering the period 1990–2018, and using different techniques to mitigate endogeneity concerns. Evidence supports the hypothesis that central banks can lean effectively against short-run cyclical misalignments of the real exchange rate. The effects are present in quarterly data—i.e., at policy-relevant horizons. The effectiveness of intervention rises with the size of the misalignment, and with the duration of one-sided interventions. FX sales appear to be somewhat more effective than FX purchases, and intervention is less effective in more liquid FX markets.

I. Introduction

Over the last decade, the role of exchange rate intervention has taken on greater prominence in policy discussions. Key questions have centered on the effectiveness of exchange rate intervention as a policy instrument in a central bank’s arsenal—that is for smoothing macroeconomic volatility, preventing the buildup of financial vulnerabilities, and helping attain inflation targets at times when the effectiveness of conventional monetary policy instruments reaches limits.2

In this context, several studies have explored the question of the optimal policy mix for small open economies in response to external shocks. The key insight is that, in the presence of key real and financial frictions, the use of additional policy instruments—such as foreign exchange interventions and capital flow measures—can, in specific settings, improve policy tradeoffs that arise in more standard models of international finance and monetary policy (see Basu et al., 2020, and Adrian et al., 2020, 2021).

The debate has also been informed by country experiences deploying this instrument. Exchange rate intervention remains popular among many emerging market- and small, open advanced economies, whereas the major advanced economies have not intervened much over the past decade. Surveys by the Bank for International Settlements (BIS) (BIS, 2004, 2015, 2019b) and World Bank (2013) of EM central bank motives and objectives for exchange rate intervention show consistent patterns. Most central banks profess to assign high importance to stem volatility, rather than to achieve a particular exchange rate. However, some also name smoothing the impact of commodity price fluctuations and enhancing competitiveness as objectives. These latter objectives remain particularly controversial, both in terms of their desirability and feasibility in practice.3

However, the effectiveness of FX interventions remains subject to debate. Few questions in international finance are as studied as the policy effectiveness of sterilized foreign exchange intervention. Yet, the research debate is far from resolved.4 Survey articles by Edison (1993), Sarno and Taylor (2001), Menkhoff (2013), Villamizar-Villegas and Perez-Reyna (2017) summarize the results from a wide range of empirical studies of interventions. These academic surveys offer a much less favorable assessment of the effectiveness of exchange rate interventions, suggesting that there remains a considerable gap between the policy and academic views about effectiveness.

Several key empirical challenges have continued to hold back progress in assessing accurately the policy effectiveness of exchange rate intervention. The first challenge is a difficulty in identifying reliable econometric ‘instruments’ with which to isolate the independent effect of exchange rate intervention on exchange rate developments. Estimates of effectiveness are known to be plagued by an endogeneity bias. Such biases for at least two reasons. First, authorities may purchase (sell) foreign exchange as the currency appreciates (depreciates) in an attempt to prevent movements in the exchange rate. If such operations are ineffective, central bank foreign asset purchases (sales) aimed at offsetting appreciations (depreciations) would be positively correlated because of other cyclical factors driving the exchange rate. Second, the authorities may purchase (sell) foreign exchange simply to take advantage of the capital inflows (outflows). Once again, such operations would result in a positively (wrongly) signed coefficient in a regression analysis. The academic response to this challenge has been to focus on very short-run intervention impacts, e.g., event studies and the use of high-frequency data (Menkhoff 2010; Fatum and Hutchison, 2003; Fratzscher et al., 2019; Dominguez, 2003) which enable better identification. However, the studies are of limited value policy-wise because they are not able to address the question of whether interventions have a longer-lasting effect that shapes the path of exchange rates over relevant policy horizons.

The second challenge arises from data limitations. The availability of high-quality data on exchange rate intervention has traditionally been very limited, constraining accurate statistical testing. More recently, this constraint has been loosened somewhat as data availability has improved in terms of quality and cross-country coverage.5

The third challenge is the size and variation of exchange rate interventions in the past. The size of exchange rate interventions decades ago pales in comparison to the size of potential interventions today. Foreign reserve buffers have generally grown over time, with levels now often exceeding commonly used foreign exchange reserve adequacy benchmarks that were established in the 1990s. Indeed, the lack of statistical significance of effectiveness in past studies may in part simply reflect the small variations of the interventions rather than an inherent lack of effectiveness. The increased use of foreign currency reserve buffers by a wider range of central banks in the past two decades offers a better environment in which to test the effectiveness hypothesis.6

Despite these challenges, recent empirical research has made further strides in our understanding of intervention effectiveness. Several studies using new cross-country databases have found evidence supporting the effectiveness of exchange rate interventions in reducing exchange rate misalignments. Daude et al. (2016), using quarterly data from 2003–2011, find that interventions are on average effective in influencing the real exchange rate, with a higher effectiveness associated with greater exchange rate misalignment; their measure of misalignment is estimated from an error correction model and captures the tendency of exchange rates to converge over time to long run trends. Adler et al. (2019) document persistent impacts of exchange rate intervention with a half-life of 12–23 months for a broad international set of country experiences. Blanchard et al. (2015) take a different approach focusing on cross country differences in quarterly exchange rate behavior between those countries that intervene heavily and those that do not. They conclude that those countries heavily reliant on intervention experience less exchange rate volatility than non-intervening countries. Consistent with this evidence, Adler and Tovar (2014) report similar results for Latin American countries using weekly data, Hofmann et al. (2019) report persistent intervention impacts over two quarters in the case of Colombia using high-frequency data, and Menkhoff et al. (2020) document a multi-month impact in the case of Japan.

At the same time, new theoretical work casts new light the drivers of exchange rate misalignments and the mechanism through which intervention may be effective. Gabaix and Maggiori (2015) and Maggiori (2022), for example, argue that misalignments (with respect to the exchange rate consistent with macroeconomic fundamentals) stem in large part from imperfect financial intermediation; their research emphasizes the risk-bearing capacity of global financial institutions and its implications on the demand for foreign assets, gross capital flows, and other drivers of persistent disconnects of exchange rates from macroeconomic fundamentals.

Relatedly, in earlier work, Bacchetta and van Wincoop (2006), building on the empirical results of Evans and Lyons (2002), find that private information-based order flows drive short-term exchange rate dynamics.

While these financial frictions seem to be important in the short run, the evidence so far suggests that macroeconomic determinants matter more at longer horizons (Mark, 1995). The notion that the drivers of nominal exchange-rate movements likely differ across different horizons is the so-called exchange determination puzzle (see, e.g., Lyons (2002), Engel et al. (2008)). On the one hand, the weak explanatory power of macro fundamentals over the short run may mean that interventions to counter financial forces may help guide exchange rates toward their fundamental equilibrium levels. On the other hand, exchange rate intervention is likely to be much less effective when leaning against fundamental-driven, longer-term exchange rate movements. However, empirical evidence differentiating intervention effectiveness on these issues is still missing.

In this paper, we build on recent empirical and theoretical insights and offer a novel approach for measuring the effectiveness of exchange rate intervention at policy-relevant horizons. In addition to using an expanded cross-country dataset, our approach is based on cycle-specific exchange rate misalignments.7 The intuition behind this strategy is that the relation between exchange rates and fundamentals may be different at different cycle lengths—longer-term swings in exchange rates may be driven by longer-term swings in fundamentals in different ways than short-term exchange rates swings are by shorter-term movements in fundamentals.8 If this is the case, the effectiveness of exchange rate intervention may naturally vary by the nature of the cyclical forces at any point in time driving the cyclical misalignments.

Econometrically, our approach entails a multi-step estimation procedure. After specifying cycle lengths, we use spectral regression methods to estimate the equilibrium real exchange rate at short-, medium-, and long-run cycles for 30 advanced- and emerging market economies.9 We then use these estimates to derive exchange-rate misalignments, cycle by cycle. Third, we examine the effectiveness of exchange rate intervention for each cycle-specific misalignment using a quarterly unbalanced panel estimation. To address endogeneity concerns, we employ FXI “surprises” as deviations from estimated policy rules (Brandao-Marques et al., 2020). In a robustness exercise, we also use capital flows to other countries as an instrument, following Blanchard et al. (2015). We use quarterly data.

Our empirical findings strengthen the case for the effectiveness of exchange rate intervention. However, we find intervention is effective only when leaning against misalignments with respect to short-cycle misalignments, but not with respect to medium-cycle misalignments and long-cycle misalignments. For a short-cycle misalignment of 10 percent, a tenth-percentage-point-of-GDP exchange rate intervention is associated with a statistically significant percent change in the exchange rate ranging from 1.5 to 4.5 percent. Our evidence also provides support for the view that persistent, one-sided interventions increase the effectiveness relative to one-off interventions as do relatively large interventions. We also find that FX sales are generally more effective than purchases. In terms of cross-sectional differences in our sample of economies, effectiveness varies across regions (with intervention in Asian economies generally more effective) and across exchange rate regimes (with intervention by regular interveners more effective than by floaters). Finally, we present some initial evidence that FX intervention is more effective when market depth is low.

That said, we do not rule out the possibility that the estimated intervention effects are even larger than we report because, despite our attempts, we may have not been able to fully overcome the inherent endogeneity bias challenge. As in the case of Daude et al. (2016), our estimates should therefore be seen as a lower bound for the true intervention effects given that the endogeneity bias tends to work against finding intervention effectiveness.10

The rest of the paper is organized as follows. The next section describes our dataset and its sources. Section III outlines our new approach to measuring exchange rate misalignments. Section IV outlines our econometric model and presents the results on the effectiveness of exchange rate intervention. Section VI concludes and draws policy implications.

II. Data Description

Our sample consists of 30 advanced and emerging market economies covering the period 1990–2018. The sample size is constrained by the availability of exchange rate intervention data, which starts only in 1990:Q1, and the availability of data on regressors in our empirical analysis, which reduces the sample size to 26 countries. The full country list and data details can be found in the Appendix A.

Exchange rate data: The real effective exchange rate (REER) index11 and nominal bilateral exchange rate with the US dollar (USNER) data are available at a monthly frequency from the IMF’s International Financial Statistics (IFS) database. In the analysis, we focus on both the trade-weighted real effective exchange rate—which more accurately reflects trade flow relationships—and the real bilateral exchange rate relative to the US dollar—which may be more important from a financial perspective.

Macroeconomic fundamentals: real GDP growth rates, general government fiscal deficit, monetary policy rates, and CPI inflation all come from the IMF’s IFS database and are available at a quarterly frequency for most advanced and emerging market economies. Commodity terms of trade data are available at a monthly frequency from the IFS database. In addition, we use data on GDP per capital in real USD terms from the IFS database to proxy for productivity levels.

Exchange rate interventions (FXI): The exchange rate intervention proxy is defined as the quarterly change in central banks’ Net Foreign Assets (NFA), adjusted for valuation effects due to movements in exchange rates. Appendix B describes the construction of this FXI proxy measure. Considerable variation is evident in Figure 1.12

Figure 1.
Figure 1.

Exchange Rate Intervention (FXI), by Country From 1990–2018

Citation: IMF Working Papers 2022, 158; 10.5089/9798400215322.001.A001

Global and policy variables: Net foreign assets (NFA), exports (X) and imports (M), and gross and net capital inflows (GKI and NKI) are taken from the IMF’s Balance of Payments and International Investment Position (BoPIIP) database. Official reserves come from the IMFs Composition of Official Foreign Exchange Reserves (COFER) database. We proxy for global financial volatility using the S&P 100 Volatility Index (VXO) from the St. Louis Federal Reserve FRED database and for capital account openness using the overall restrictions index from Fernández et al. (2016).

III. New Approach to Measuring Exchange Rate Misalignments

This section describes a new approach to measuring exchange rate misalignments. The conventional approach is to estimate an “equilibrium” exchange rate and define the misalignment as the deviation of this estimate from the exchange rate. Our approach emphasizes cycle-specific misalignments, i.e., misalignments of short-, medium-, and long-cycles in the exchange rate. To do this, we first decompose the equilibrium exchange rate into short-, medium-, and long-cycles and then use spectral regression methods to derive real exchange-rate deviations from their equilibrium levels at short-, medium-, and long-run cycles. This section describes the multi-step procedure.

A. Exchange Rates and Their Macroeconomic Fundamentals at Different Cyclical Frequencies

As is conventional in the literature, we are interested in estimating the relationship between the real exchange rate and macroeconomic fundamentals of the form:

eri,tf=βXi,tf+θif+εi,tf,(1)

where eri,tf is the real effective exchange rate index for country i at time t; Xi,tf is a vector of macroeconomic fundamentals consisting of: log per capita income, net foreign assets (as percent of GDP), log openness (sum of exports and imports as percent of GDP), government consumption (as percent of GDP), and the log of the commodity terms of trade index; and θif are country fixed effects. This choice of the macroeconomic fundamental variables reflects findings in the literature.13

One key difference between our approach and that in the literature is our emphasis on cycle-specific misalignments. Namely, f ∈ {s,m,l} indexes the three frequency bands that corresponding to the short-, medium-, and long-cycles. Operationally, after specifying the frequency bands, Equation 1 can be estimated with band spectrum regression methods (following Engle 1974, Granger and Engle 1983, Phillips 1991, and Corbae, Ouliaris, and Phillips 2002, among others). First, all variables are transformed to the frequency domain using a Fourier transformation.14 The transformed dependent variables can then simply be regressed on the transformed explanatory variables for each selected frequency band. In our case, we transform the exchange rate and our macro fundamentals into their cyclical components at different frequencies, before proceeding with a standard linear regression.

B. Identifying Short-, Medium-, and Long-Cycles

We choose the frequency bands corresponding to the short-, medium-, and long-cycles following the tradition of Burns and Mitchell (1946). In particular, we use a statistical algorithm that identifies peaks and troughs in the real bilateral exchange rate (against the US dollar) as in Bry and Boschan (1971) and Albuquerque et al. (2015).15 Figure C.1 in Appendix C plots the real effective exchange rate for the 22 advanced and emerging market economies for which we have monthly data, along with the identified peaks and troughs.

The cyclical nature of real exchange rates is visually striking for our sample of countries, with clear peaks and troughs. From peak-to-peak, the average length of the estimated cycles is around 10 years. With the duration estimate for exchange rate cycles, we calibrate the decomposition of the exchange rate series into different sub-cycles. The long-run cycle is estimated by filtering out frequencies associated with cycles of duration of 10 years or more; the medium-run cycle is defined as filtering out frequencies associated with cycles of duration greater than 4 years and less than 10-years. The remaining frequencies of interest are associated with short (including business) cycles of 1 to 4 years.16 For less than a year, we treat the relationship between macroeconomic fundamentals and the real exchange rate as being statistically unpredictable.17

C. Relationship of Cycle-Specific Exchange Rates and Macroeconomic Fundamentals

Given these calibrated cycle lengths of interest, Equation 1 can be estimated as a set of cycle-specific regressions. Table 1 presents the results. For each cycle-specific regression, the coefficients on the fundamental variables are highly significant and generally have the expected signs. The coefficients on NFA and trade openness are both quite negative, as expected, and consistent across cycles, while the coefficient on income is positive, as expected. The coefficient on government consumption differs across cycles. The commodity terms of trade are measured such that a decrease is an improvement in the terms of trade (and tends to be associated with an appreciation in the domestic currency).

Table 1.

First-Stage Regression Model of Exchange Rate Macro Fundamentals

article image
Notes: Significance levels: 1 percent (***), 5 percent (**), 10 percent (*). Dependent variable is the log of the real effective exchange rate (REER). Fundamentals: log per capita income, net foreign assets (as percent of GDP), log openness (as percent of GDP), log government consumption (as percent of GDP), log of the commodity terms of trade index. Newey-West corrected standard errors for heteroskedastic autocorrelation up to 4 quarter lags.

The results—and in particular the size of the coefficients—confirm that the macroeconomic fundamentals are more powerful in explaining movements in the exchange rate over longer cycles compared to short cycles, with the exception of trade flows, which are more important for the short and medium cycles.18

D. Defining Cycle-Specific Exchange Rate Misalignments

Armed with the estimates from Equation 1, we define the cycle-specific equilibrium real exchange rate eeri,tf at each frequency f ∈ {s,m,l} in the following way (where the hat notation denotes the OLS estimate):

erri,tf=α^+β^Xi,tf+θ^if.(2)

Accordingly, the cycle-specific misalignments missitf at each frequency are defined as,

misi,tf=eri,tf−eeri,rf,(3)

where eri,tf is the real exchange rate for each country at frequency f ∈ {s,m,l}.

These measures of misalignments—shown in Figure 2—differ from those found in the literature in two ways.19 First, these misalignments indicate the extent to which the exchange rate deviates from its fundamental value at different cyclical frequencies. Second, the frequency-domain method estimates are theoretically unbiased (Engle, 1974)) as long as all variables are transformed using the same frequency band in each regression. This unbiasedness property of the estimators is generally not present when misalignments are constructed as the difference between the actual exchange rate and some equilibrium level, as in Daude et al. (2016).

Figure 2.
Figure 2.
Figure 2.

Cyclic-Specific Exchange Rate (REER) Misalignments

Citation: IMF Working Papers 2022, 158; 10.5089/9798400215322.001.A001

IV. Estimates of FXI Effectiveness Using Cycle-specific Misalignments

This section begins by describing the baseline FXI effectiveness model and results. It then addresses various robustness issues and extensions of the baseline to include large and persistent interventions, sales versus purchases of foreign assets, and cross-sectional comparisons among the sample population.

A. Baseline Panel Regression Model

Our baseline regression model to assess the effectiveness of FXI is as follows:

erit=ρeri,t−1+μXi,t+φZi,t+Σfβfmisi,t−1fδFXIi,t+Σfμf(FXIi,t×misi,tf)+θi+εi,t(4)

where eri,t is the log of the real effective exchange rate for country i and time t; Xi,t are the usual exchange rate fundamentals; Zi,t are the a set of policy variables and global factors20; FXIi,t is our interventions proxy; the misi,tf variables are the measures of short-, medium-, and long-run misalignments; and θt are country fixed effects. The coefficients μf measure the extent of the correlation between FXI and cycle-specific misalignments.21 All models are estimated using fixed effects with heteroskedasticity and serial correlation corrected Newey-West standard errors.

B. Baseline Results

The results support the hypothesis that central banks are effective at leaning against short-run misalignments in the real exchange rate. Table 2, column 1 presents the baseline specification in Equation (4). The sample is restricted to 26 countries for whom we have data on policy variables for. The evidence of cycle- specific FXI effectiveness can be read off the cross terms FXI*mis S, FXI*mis M, and FXI*mis L. Theory suggests that these coefficients should be negative, i.e. effective FX purchases (positive) lead to a deprecation (negative) of the exchange rate.

Table 2.

Effectiveness of Leaning Against the Wind

article image
Notes: Significance levels: 1 percent (***), 5 percent (**), 10 percent (*). Estimated with country fixed effects. Dependent variable is the log of the real effective exchange rate (REER). FXI is foreign exchange intervention as a share of GDP. Fundamentals: log per capita income, net foreign assets (as percent of GDP), log openness (as percent of GDP), log government consumption (as percent of GDP), log of the commodity terms of trade index; Policy variables: real policy rate, reserve-to-trade ratio (both in percent), Chin-Ito index for capital account openness; Global variables: interactions between the VXO index and capital account openness and between the real policy rate and capital account openness. Short-, medium-, and long-cycle misalignments are relative to their respective cyclical equilibrium real exchange rate based on Equation 1. Newey-West corrected standard errors for heteroskedastic autocorrelation up to 4 quarter lags.

In terms of magnitude, the estimates indicate that for a short-run misalignment of 10 percent, a tenth-percentage-point-of-GDP of exchange rate intervention (FXI), in this case purchases of foreign exchange, is associated with a statistically significant 1.5 percent depreciation in the exchange rate.22

The coefficients on the medium-cycle and long-cycle misalignments (FXI*mis M and FXI*mis L) are relatively modest.23 Moreover, the medium-cycle misalignment coefficient is not of the correct sign and not statistically different from zero. This is evidence that exchange rate intervention is ineffective in leaning against the forces associated with the medium-cycle misalignments. In contrast, the long-cycle misalignment coefficient appears to be statistically significant but small. Our subsequent robustness checks suggest that this finding is statistically fragile and may be spurious. One explanation for this finding could be that short-run misalignments are often generated by external shocks in the context of shallow or illiquid markets, a setting in which FX intervention may be particularly effective, especially in the presence of financial frictions. Such financial frictions may be less binding over the long-run, perhaps explaining why we find FX interventions less effective at longer horizons. We explore this further in section IV.G.

C. Addressing the Endogeneity Bias Challenge

We address the endogeneity by using policy “surprises” as deviations from estimated policy rules. Specifically, we follow Brandao et al. (2020) in estimating the FXI “surprises” as the residual from the following linear intervention rule:

FXIi,t=βXi,t+δσi,treer+θi+εi,tfxi,(5)

where the vector X denotes the fundamentals; σi,treer is the monthly variance of the real effective exchange rate index within the quarter; θi are country fixed effects, and εi,tfxi is the residual and our measure of the FXI IV policy shock. Figure 3 plots the FXI surprises against FXI.

Figure 3.
Figure 3.

FXI Surprise and Capital Flow IVs

Citation: IMF Working Papers 2022, 158; 10.5089/9798400215322.001.A001

Notes: Figures plots FXI (as % of GDP) on the vertical axis against the FXI policy surprise variable (blue circles) and the fitted FXI instrument using capital flows (red circles) on the horizontal axis respectively; black line is the 45 degree line.

The results from re-running Equation 4 and replacing FXI with FXI surprises are shown in Column 2 of Table 2.24 Several key findings stand out. First, the coefficients on FXI and the FXI cross terms have the expected negative sign. Second, the coefficient on FXI*mis S is threefold larger than in Column 1 and is statistically more significant. Third, the FXI*mis M and FXI*mis L coefficients are of the expected sign, albeit statistically insignificant. All this suggests that use of the FXI surprises helped to attenuate the endogeneity problem and reinforces our baseline finding that exchange interventions are effective in influencing short-cycle misalignments; these results even suggest that exchange rate intervention may have a modest impact on medium- and long-cycle misalignments.

D. Robustness

We report three statistical checks that confirm the robustness of our conclusions.25 Column 1 of Table 3 shows estimates from Equation 4 but replaces the fixed effects baseline specification with a specification in first differences. Column 2 reports the results when replacing the fundamentals with the equilibrium REER—that is the simple fitted REER obtained from eq. 1 without filtering the regressors. Column 3 considers the usefulness of global capital flows as an instrument, as proposed by Blanchard et al. (2015).26

Table 3.

Robustness Checks

article image
Notes: Significance levels: 1 percent (***), 5 percent (**), 10 percent (*). Estimated with country fixed effects. Dependent variable is the log of the real effective exchange rate (REER). FXI is foreign exchange intervention as a share of GDP. Fundamentals: log per capita income, net foreign assets (as percent of GDP), log openness (as percent of GDP), log government consumption (as percent of GDP), log of the commodity terms of trade index; Policy variables: real policy rate, reserve-to-trade ratio (both in percent), Chin-Ito index for capital account openness; Global variables: interactions between the VXO index and capital account openness and between the real policy rate and capital account openness. Short, medium, and long run misalignments are computed relative to their respective cyclical equilibrium real exchange rate level based on the macroeconomic fundamentals model. Newey-West corrected standard errors for heteroskedastic autocorrelation up to 4 quarter lags.

The main result is that central banks are effective at leaning against short-cycle misalignments, as indicated by the negative and statistically significant coefficients on FXI*mis S in all three columns. The other coefficients are fairly robust, even though the signs on the cross terms with mis M and mis L are economically small, statistically insignificant, and tend to be positive.

E. Effectiveness of Persistent, Large, and One-Sided Interventions

The results in the previous section suggest that FX intervention can smooth out misaligned exchange rates by one-off interventions. We now shift our attention to the impacts of large and persistent interventions. Are persistent interventions more effective?27 We investigate this possibility using the following model:

erri,t=peri,t−1+μXi,t+φZi,t+Σfβfmisi,t−1f+δ1FXIi,t+Σfδfxif(FXIi,t×misi,tf)+βCFXIi,t+φ1(FXIi,t×CFXIi,t)+Σfφcfxif(FXIi,t×CFXIi,t−1×misi,tf)+θi+εi,t,(6)

where CFXIi,t−1=Σj=14FXIi,t−j is the 4-quarter cumulative sum of our FX intervention variable. Xi,t are the usual exchange rate fundamentals; Zi,t are the policy variables and global factors. The φcfxi coefficients measure the effect of FXI conditional on the previous quarters’ cumulative FXI and on the cyclical misalignment. We use the FXI policy surprises instrument as our FXI variable.

Figure 4.
Figure 4.
Figure 4.

Exchange Rate (REER) Misalignments, by Cycle Type

Citation: IMF Working Papers 2022, 158; 10.5089/9798400215322.001.A001

We find evidence that persistent interventions increase intervention effectiveness. The results are presented in Table 4. Cumulative interventions are found to be effective at leaning against short-run misalignments. For every additional percent of CFXI, the exchange depreciates by a further .26 percent. This suggests that policy makers face an additional “bang for their buck” if they intervene systematically against misalignments over consecutive quarters. Note that these results are for persistent interventions but do not distinguish between persistent purchases or sales; it is important to note that sales may be more effective because markets understand that there is an absolute lower bound at zero reserves and that there may be an implicit lower bound above zero at which rating agencies may downgrade sovereign debt. If a central bank were to sell reserves during a period of stress, it could signal a strong commitment to offset non-fundamental pressures.

Table 4.

Persistent and One-Sided Interventions

article image
Notes: Significance levels: 1 percent (***), 5 percent (**), 10 percent (*). Estimated with country fixed effects. Dependent variable is the log of the real effective exchange rate (REER). FXI is foreign exchange intervention as a share of GDP. CFXI are cumulative FX interventions for the last 4 quarters. Fundamentals: log per capita income, net foreign assets (as percent of GDP), log openness (as percent of GDP), log government consumption (as percent of GDP), log of the commodity terms of trade index; Policy variables: real policy rate, reserve-to-trade ratio (both in percent), Chin-Ito index for capital account openness; Global variables: interactions between the VXO index and capital account openness and between the real policy rate and capital account openness. Short, medium, and long run misalignments are computed relative to their respective cyclical equilibrium real exchange rate level based on the macroeconomic fundamentals model. Newey-West corrected standard errors for heteroskedastic autocorrelation up to 5 quarter lags. Column (5) uses a randomly generated FXI variables drawn from a normal distribution.

Next, to investigate the impact of particularly large interventions, we adjust the above equation as follows:

erri,t=ρeri,t−1+μXi,t+φZi,t+Σfβfmisi,t−1f+δ1FXIi,t+Σfδfxif(FXIi,t×misi,tf)+βP/Si,t+μ1(FXIi,t×P/Si)+ΣfμP/Sf(FXIi,t×P/Si,t×misi,tf)+θi+εi,t,(7)

where P/S denotes either large FX purchases (P) or sales (S), defined as FX interventions in either direction that are greater than 1 percent of GDP. The μP/S coefficients capture the differential effect of FXI conditional on the size of the interventions and the cyclical misalignments. We continue to use our FX policy surprises as our FXI measure. Figure 5 summarizes the data on sales and purchases.

Figure 5.
Figure 5.

FX Sales vs. Purchases

Citation: IMF Working Papers 2022, 158; 10.5089/9798400215322.001.A001

The evidence is consistent with the perspective that policy makers generally lean against the wind, either purchasing foreign assets during periods of capital inflow pressures and exchange rate appreciation. or selling foreign assets during periods of capital outflow pressures and currency depreciation. Table 5 documents that both large FX purchases and sales tend to strengthen the countercyclical effect of intervention with respect to short-cycle misalignments. FXI sales appear to be somewhat more effective in terms of economic size (3.6 percent vs. 3.0 percent).

Table 5.

Large Interventions (Sales vs. Purchases)

article image
Notes: Significance levels: 1 percent (***), 5 percent (**), 10 percent (*). Estimated with country fixed effects. Dependent variable is the log of the real effective exchange rate (REER). FXI is foreign exchange intervention as a share of GDP. Large purchases and sales are dummy variables for FX purchases or sales greater than 1 percent of GDP. Fundamentals: log per capita income, net foreign assets (as percent of GDP), log openness (as percent of GDP), log government consumption (as percent of GDP), log of the commodity terms of trade index; Policy variables: real policy rate, reserve-to-trade ratio (both in percent), Chin-Ito index for capital account openness; Global variables: interactions between the VXO index and capital account openness and between the real policy rate and capital account openness. Short, medium, and long run misalignments are computed relative to their respective cyclical equilibrium real exchange rate level based on the macroeconomic fundamentals model. Newey-West corrected standard errors for heteroskedastic autocorrelation up to 5 quarter lags. Column (5) uses a randomly generated FXI variables drawn from a normal distribution.

F. Delving Into the Cross Section

While the baseline results captured in Table 2 highlights the importance of accounting for cyclical misalignments when gauging the effectiveness of exchange rate intervention for the full sample of economies, this section digs into the cross-sectional differences across our sample of countries. Table 6 reports the results for (i) Latin American and Asian economies, and (ii) managed floaters and interveners.

Table 6.

Country Differences, by Regions and FXI Intensity

article image
See notes to Table 2. The classification of interveners and floaters is based on the IMF’s 2020 https://www.elibrary-areaer.imf.org/Pages/Home.aspx Database.

Differences between Asia and Latin America. Columns (1) and (2) highlight the regional differences in the impact of FXI on the misalignments. Nearly all the coefficients have the expected signs. There is some evidence that Asian interventions are more impactful. For Asia, which relies heavily on the dollar in both trade and finance, the results for both the short- and medium-cycle misalignment measures are statistically significant.

Differences between countries which are managed floaters versus interveners. The distinction between the managed floaters and the interveners is the degree to which a country relies on exchange rate intervention in its respective policy framework. In general, the FXI*mis S coefficients for interveners and floaters are both negative, but the coefficient is statistically significant and more negative for interveners. The other coefficients in the table are statistically insignificant and may reflect low explanatory power of the limited regional datasets. The results do suggest that a credible intervener may be able use its interventions to signal its intentions, and expect markets to react accordingly. This does not suggest that there are not more nuanced differences among the groups of floaters and interveners. Delving into the details of how policy makers can effectively signal its intervention intentions is left for future research.

G. Market Depth and Liquidity

FX intervention is likely to be more effective in countries where market depth is shallow and is subject to bouts of illiquidity. To investigate this issue, we construct a measure of the liquidity of FX markets as follows:

spreadi,t=bidi,t−offeri,teri,t×100(8)

where market liquidity is measured as the difference between the average spread between the bid- and offer-rate within the quarter, normalized by the nominal exchange rate, and measured in percent. More shallow, less liquid FX markets should exhibit a larger spread. We add this variable to our baseline specification as a triple interaction with FXI and our different misalignment terms, controlling for the spread on its own, as well as its interaction with the misalignments.

The results, presented in Table 7 based on our baseline specification, suggest that less liquidity FX markets (as captured by a larger bid-offer spread) are indeed associated with a more effective leaning-against-the wind effect of FX interventions. For a 1-percentage point increase in the bid-offer spread28, a tenth-percentage-point-of-GDP FX purchase is associated with a statistically significant 4-percent larger depreciation in the exchange rate, when starting from a 10 short-run misalignment. The impact associated with medium- and long-run misalignments is not statically significant. When we use our FX policy surprise measure, the differential impact increases to 7 percent.

Table 7.

Market Depth and Liquidity

article image
Notes: Significance levels: 1 percent (***), 5 percent (**), 10 percent (*). Estimated with country fixed effects. Dependent variable is the log of the real effective exchange rate (REER). FXI is foreign exchange intervention as a share of GDP. Fundamentals: log per capita income, net foreign assets (as percent of GDP), log openness (as percent of GDP), log government consumption (as percent of GDP), log of the commodity terms of trade index; Policy variables: real policy rate, reserve-to-trade ratio (both in percent), Chin-Ito index for capital account openness; Global variables: interactions between the VXO index and capital account openness and between the real policy rate and capital account openness. Short-, medium-, and long-cycle misalignments are relative to their respective cyclical equilibrium real exchange rate based on Equation 1. Newey-West corrected standard errors for heteroskedastic autocorrelation up to 4 quarter lags.

One explanation for this result is that short-run misalignments are typically generated by external financial shocks in the presence of financial frictions and shallow FX markets. In such a setting, FX intervention is likely to be particularly effective at serving as a liquidity buffer which helps prevent adverse, self-reinforcing liquidity spirals. Misalignments over longer horizons may be driven more by the divergence of exchange rates from macroeconomic fundamentals, which would explain the result that FX interventions are less effective with respect to longer cycle misalignments.

V. Conclusions

The policy debate on the usefulness and effectiveness of exchange rate intervention has been reinvigorated in recent years. One challenge to forging a consensus have been the huge gaps in our understanding of when intervention is likely to work and when it is not.

It is in this context that the findings of our analysis are helpful. Leaning against short-cycle misalignments is found to be effective, both in terms of the economic size and statistical significance; but not for leaning against medium- and long-cycle misalignments. Persistent, one-sided interventions appear to offer additional benefits compared to one-off interventions. The effects are measurable at policy-relevant horizons. That said, there are considerable differences—across regions, and between those economies which actively intervene and those which do not—which deserve further attention.

Our cycle-specific misalignment approach also offers a new perspective on how to think about exchange rate misalignments. This approach emphasizes the short-, medium-, and long-cycle factors driving exchange rates. The findings in this paper suggest that intervening to reduce exchange rate misalignments arising from long-run macroeconomic factors is not likely to be effective. However, in line with the theoretical literature, intervening to reduce exchange rate misalignments arising from financial frictions is more likely to yield results.

Econometrically, this paper demonstrates the usefulness of band spectrum estimation techniques in understanding dynamics of exchange rate misalignments. Certainly, there is more to be learned from our approach. Potential fruitful lines of future research include a more nuanced investigation into the conditions determining when exchange rate intervention is likely to be most effective. This will likely involve delving into the underlying mechanisms driving the misalignments, such as the risk-bearing capacity of global financial institutions. We also leave for future research extensions that incorporate intervention strategies based on market conditions such as market depth and liquidity, intervention tactics with respect to the most effective sizes and types of foreign asset purchases and sales, and the role of policy communication.

VI. References

  • Adler, G., C. Casas, L. Cubeddu, and G. Gopinath, N. Li, S. Meleshchuk, C. Osorio Buitron, D. Puy, and Y. Timmer, 2020, “Dominant Currencies and External Adjustment”, IMF Staff Discussion Note, no 20/05, July. (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

  • Adler, G., and C. Tovar, 2014, “Foreign Exchange Interventions and Their Impact on Exchange Rate Levels”, Monetaria, January-June, pp. 1–48.

    • Search Google Scholar
    • Export Citation

  • Adler, G., N. Lisack, and R. Mano, 2019, “Unveiling the Effects of Foreign Exchange Intervention: A Panel Approach”, Emerging Markets Review, Vol. 40(C), pp. 1–29.

    • Search Google Scholar
    • Export Citation

  • Adrian, T., C. Erceg, J. Lindé, P. Zabczyk, J. Zhou, 2020, “A Quantitative Model for the Integrated Policy Framework,” IMF Working Paper 20/122 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

  • Adrian, T., C. Erceg, M. Kolasa, J. Lindé, and P. Zabczyk, 2021, “A Quantitative Microfounded Model for the Integrated Policy Framework,” IMF Working Paper 21/292 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

  • Albuquerque, R., M. Eichenbaum, D. Papanikolaou, and S. Rebelo, 2015, “Long-Run Bulls and Bears”, Journal of Monetary Economics, vol 76(S), pp. 21–36.

    • Search Google Scholar
    • Export Citation

  • Bacchetta, P., and E. van Wincoop, 2006, “Can Information Heterogeneity Explain the Exchange Rate Determination Puzzle?,” American Economic Review, vol 96(03), pp. 552–576.

    • Search Google Scholar
    • Export Citation

  • Bank for International Settlements, 2005, “Foreign Exchange Market Intervention in Emerging Markets: Motives, Techniques, and Implications”, BIS Papers No. 24 (Basel: Bank for International Settlements).

    • Search Google Scholar
    • Export Citation

  • Bank for International Settlements, 2013, “Market Volatility and Foreign Exchange Intervention in EMEs: What Has Changed?”, BIS Papers No. 73, (Basel: Bank for International Settlements).

    • Search Google Scholar
    • Export Citation

  • Bank for International Settlements, 2019a. Annual Report, June. (Basel: Bank for International Settlements).

  • Bank for International Settlements, 2019b. Reserve Management and FX Intervention, BIS Papers, no 104. (Basel: Bank for International Settlements).

    • Search Google Scholar
    • Export Citation

  • Basu, S., E. Boz, G. Gopinath, F. Roch, and F. Unsal, 2020, “A Conceptual Model for the Integrated Policy Framework,” IMF Working Paper 20/121. (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

  • Blanchard, O., G. Adler, and I. de Carvalho Filho, 2015, “Can Foreign Exchange Intervention Stem Exchange Rate Pressures from Global Capital Flow Shocks”, IMF Working Paper, WP/15/159, (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

  • Bordo, M., 2019, “The Operation and Demise of the Bretton Woods System, 1958–1971”, in (Naomi Lamoreaux, Ian Shapiro, eds.) The Bretton Woods Agreements: Together with Scholarly Commentaries and Essential Historical Documents, Yale University Press (New Haven).

    • Search Google Scholar
    • Export Citation

  • Brandao Marques, L., G. Gelos, M. Narita, and E. Nier, 2020, “Leaning Against the Wind: An Cost-Benefit Analysis for an Integrated Policy Framework”, IMF Working Paper, WP/20/123, (Washington: International Monetary Fund)..

    • Search Google Scholar
    • Export Citation

  • Burns, A., and W. Mitchell, 1946. Measuring Business Cycles, NBER. (Cambridge, Massachusetts: MIT Press).

  • Busetti, F., and M. Caivano, 2017, “Low Frequency Drivers of the Real Interest Rate: A Band Spectrum Regression Approach”, Banca d’Italia Working Paper 1132, September 2017 (Rome: Banca d’Italia)

    • Search Google Scholar
    • Export Citation

  • Bry, G., and C. Boschan, 1971, “Programmed Selection of Cyclical Turning Points” in Cyclical Analysis of Time Series: Selected Procedures and Computer Programs, NBER. (Cambridge, Massachusetts: MIT Press).

    • Search Google Scholar
    • Export Citation

  • Calvo, G., and C. Reinhart, 2002, “Fear of Floating”, The Quarterly Journal of Economics, vol 117(2), pp. 379–408.

  • Canales-Kriljenko, J., 2003, “Foreign Exchange Intervention in Developing and Transition Economies: Results of a Survey”, IMF Working Paper 03/95 (Washington: International Monetary Fund)..

    • Search Google Scholar
    • Export Citation

  • Canzoneri, M., R. Cumby, and B. Diba, 1999, “Relative labor productivity and the real exchange rate in the long run: evidence for a panel of OECD countries,” Journal of International Economics, Vol. 47 (2), pp. 245–266.

    • Search Google Scholar
    • Export Citation

  • Cavallino, P., and N. Patel, 2019, “FX Intervention: Goals, Strategies and Tactics”, BIS Papers. No. 104b, October, (Basel: Bank for International Settlements).

    • Search Google Scholar
    • Export Citation

  • Chamon, M., D. Hofman, N. Magud, and A. Werner (Eds.), 2019, Foreign Exchange Intervention in Inflation Targeters in Latin America, International Monetary Fund, Washington, D.C., pp. 43–62

    • Search Google Scholar
    • Export Citation

  • Chen, S.-L., and H.-Y. Lee, (2006). “Why Use Markov-Switching Models in Exchange Rate Prediction?”, Economic Modelling, vol 23(4), pp.662–668

    • Search Google Scholar
    • Export Citation

  • Corbae, P. D. S. Ouliaris, and P. Phillips, 2002, “Band Spectral Regression with Trending Data,” Econometrica, 2002, Vol. 70 (3), pp. 1067–1109.

    • Search Google Scholar
    • Export Citation

  • Cubeddu, L., S. Krogstrup, G. Adler, P. Rabanal, M. Dao, S. Hannan, L. Juvenal, N. Li, C. Buitron, C. Rebillard, D. Garcia-Macia, C. Jones, J. Rodriguez, K. Chang, D. Gautam, and Z. Wang, 2019, “The External Balance Assessment Methodology: 2018 Update”, IMF Working Paper 19/65, (Washington: International Monetary Fund)

    • Search Google Scholar
    • Export Citation

  • Daude, C., E. Levy Yeyati, and A. Nagengast, 2016, “On the Effectiveness of Exchange Rate Interventions in Emerging Markets”, Journal of International Money and Finance, vol 64(C), pp. 239–261.

    • Search Google Scholar
    • Export Citation

  • Dominguez, K, 2003, “The Market Microstructure of Central Bank Intervention” Journal of International Economics, vol 59(1), pp. 25–45.

    • Search Google Scholar
    • Export Citation

  • Dominguez, K., 2012, “Foreign Reserve Management During the Global Financial Crisis”, Journal of International Money and Finance, vol 31(8), pp. 2017–2037.

    • Search Google Scholar
    • Export Citation

  • Dooley, M., D. Folkerts-Landau, and P. Garber, 2019, “Bretton Woods II Still Defines the International Monetary System”, Pacific Economic Review, vol 14(3), pp. 297–311.

    • Search Google Scholar
    • Export Citation

  • Edison, H., 1993, “The Effectiveness of Central-Bank Intervention: a Survey of the Literature After 1982,” International Finance Section, Department of Economics, Princeton University, Princeton, N.J.

    • Search Google Scholar
    • Export Citation

  • Eichengreen, B., and R. Hausmann, 1999, “Exchange Rates and Financial Fragility”, in Proceedings: Economic Policy Symposium, Federal Reserve Bank of Kansas City, pp. 329–368.

    • Search Google Scholar
    • Export Citation

  • Engel, C., and J. Hamilton, 1990, “Long Swings in the Dollar: Are They in the Data and Do Markets Know It?”, American Economic Review, vol 80(4), pp. 689–713.

    • Search Google Scholar
    • Export Citation

  • Engel, C. and West, K. D., 2005, “Exchange rates and fundamentals”. Journal of Political Economy, 113(3), pp.485–517.

  • Engel, C., N. Mark, and K. West, 2008, “Exchange Rate Models Are Not as Bad as You Think”, NBER Chapters, in: NBER Macroeconomics Annual 2007, vol 22, pp. 381–441.

    • Search Google Scholar
    • Export Citation

  • Engle, R., 1974, “Band Spectrum Regression”, International Economic Review, vol 15(1), pp. 1–11.

  • Evans, M., and K. Lewis, 1995, “Do Long-Term Swings in the Dollar Affect Estimates of the Risk Premia?”, Review of Financial Studies, vol 8(3), pp. 709–742.

    • Search Google Scholar
    • Export Citation

  • Evans, M. and R. K. Lyons, 2002, “Order Flow and Exchange Rate Dynamics,” Journal of Political Economy, University of Chicago Press, vol 110(1), pp. 170–180.

    • Search Google Scholar
    • Export Citation

  • Fatum, R., and M. Hutchison, 2003, “Effectiveness of Official Daily Foreign Exchange Market Intervention Operations in Japan,” Journal of International Money and Finance, vol 25(2): 199–219

    • Search Google Scholar
    • Export Citation

  • Fernández, A., M. Klein, A. Rebucci, M. Schindler, and M. Uribe, 2016, “Capital Control Measures: A New Dataset”, IMF Economic Review, vol 64(3), pp. 548–574.

    • Search Google Scholar
    • Export Citation

  • Fratzscher, M., O. Gloede, L. Menkhoff, L. Sarno, and T. Stöhr, 2019, “When Is Foreign Exchange Intervention Effective? Evidence from 33 Countries,” American Economic Journal: Macroeconomics, vol 11(1), pp. 132–56

    • Search Google Scholar
    • Export Citation

  • Gabaix, X., and M. Maggiori (2015). “International Liquidity and Exchange Rate Dynamics”, Quarterly Journal of Economics, vol (3), pp. 1369–1420.

    • Search Google Scholar
    • Export Citation

  • Granger, C.W.J. and R. Engle, 1983, “Applications of Spectral Analysis in Econometrics.” Handbook of Statistics vol 3, 1983, pp. 93-109.

    • Search Google Scholar
    • Export Citation

  • Gruss, B., and S. Kebhaj, 2019. Commodity Terms of Trade: A New Database. IMF Working Paper 19/21, (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

  • Hamilton, J., 2018, “Why You Should Never Use the Hodrick-Prescott Filter”, Review of Economics and Statistics, vol 100(5), pp. 831–843.

    • Search Google Scholar
    • Export Citation

  • Hofmann, B., H. Shin, and M. Villamizar-Villegas, 2019, “FX Intervention and Domestic Credit: Evidence from High Frequency Micro Data”, BIS Working Papers, No. 774, (Basel: Bank for International Settlements).

    • Search Google Scholar
    • Export Citation

  • Imbens, G., and J. Wooldridge, 2009, “Recent Developments in the Econometrics of Program Evaluation”, Journal of Economic Literature, vol 47(1), pp. 5–86.

    • Search Google Scholar
    • Export Citation

  • Klaassen, F., 2005, “Long Swings in Exchange Rates”, Journal of Business and Economic Statistics, vol 23(1), pp. 87–95.

  • Levy Yeyati, E., 2010, “What Drives Reserve Accumulation (and at What Cost)?”, voxeu.org, September 30; www.voxeu.org/article/what-drives-reserve-accumulation-and-whatcost.

    • Search Google Scholar
    • Export Citation

  • Lipton, D., 2019, “Foreign Exchange Intervention in Latin American Countries with Inflation Targets”, speech in Mexico City, March 8.

    • Search Google Scholar
    • Export Citation

  • Lyons, R.K., 2002 “New Perspective on FX Markets: Order-Flow Analysis”, International Finance, vol 4(2), pp. 303-320.

  • Maggiori, M., 2022, “International Macroeconomics When Financial Markets Are Imperfect”, Handbook of International Economics, North Holland (Amsterdam).

    • Search Google Scholar
    • Export Citation

  • Mark, N., 1995, “Exchange rates and fundamentals: Evidence on Long-Horizon Predictability”, The American Economic Review, pp.201–218.

    • Search Google Scholar
    • Export Citation

  • Mark,N. and D. Sul, 2001. Nominal Exchange Rates and Monetary Fundamentals: Evidence from a Small Post-Bretton Woods Panel, Journal of International Economics, vol 53(1), pp. 29–52

    • Search Google Scholar
    • Export Citation

  • Menkhoff, L., 2010, “High-Frequency Analysis of Foreign Exchange Interventions: What Do We Learn?”, Journal of Economic Surveys, vol 24(1), pp. 85–112.

    • Search Google Scholar
    • Export Citation

  • Menkhoff, L., 2013, “Foreign Exchange Intervention in Emerging Markets: A Survey of Empirical Studies”, The World Economy, vol 36(9), pp. 1187–1208.

    • Search Google Scholar
    • Export Citation

  • Menkhoff, L., L., M. Rieth, and T. Stöhr, 2020, “The Dynamic Impact of FX Interventions on Financial Markets”, DIW Discussion Paper no 1854, March (Berlin: DIW).

    • Search Google Scholar
    • Export Citation

  • Mussa, M., 1981, “The Role of Official Intervention”, Group of 30 Occasional Papers, No. 6, (Washington: Group of 30).

  • Neely, C., 2008, “Central Bank Authorities’ Beliefs about Foreign Exchange Intervention”, Journal of International Money and Finance, vol 27(1), pp. 1–25.

    • Search Google Scholar
    • Export Citation

  • Phillips, P.C.B., 1991, “Spectral Regression for Co-integrated Time Series,” in Nonparamnetic and Semiparametric Methods in Economics and Statistics, ed. by W Barnett, J. Powell, and G. Tauchen (Cambridge: Cambridge University Press).

    • Search Google Scholar
    • Export Citation

  • Ricci, L., G. Milesi-Ferretti, and J. Lee, 2013, “Real Exchange Rates and Fundamentals: A Cross-Country Perspective”, Journal of Money, Credit, and Banking, vol 45(5), pp. 845–865.

    • Search Google Scholar
    • Export Citation

  • Sarno, L., and M. Taylor, 2001, “Official Intervention in the Foreign Exchange Market: Is It Effective and, If So, How Does It Work?”, Journal of Economic Literature, vol 39(3), pp. 839–868.

    • Search Google Scholar
    • Export Citation

  • Taylor, M., 2004, “Is Official Exchange Rate Intervention Effective?”, Economica, vol 71(281), pp. 1–11.

  • Villamizar-Villegas, M, and D. Perez-Reyna, 2017, “A Theoretical Approach to Sterilized Foreign Exchange Intervention”, Journal of Economic Surveys, vol 31(1), pp. 343–365.

    • Search Google Scholar
    • Export Citation

  • Williamson, J., 1989, “What Washington Means by Policy Reform”, in J. Williamson (ed) Latin American Readjustment: How Much has Happened, Washington: Peterson Institute for International Economics.

    • Search Google Scholar
    • Export Citation

VII. Appendixes

Appendix A. Data Details

Table A.1.

Data Sources

article image
Table A.2.

Country Coverage

article image
* Hong Kong SAR is a territory of the People’s Republic of China with its own currency.
Table A.3.

Summary Statistics

article image
Notes: Table reports the mean of each variable over the period 1990–2018. REER index; DSP per capita in real USD; government consumption, NFA, openness, as percent of GDP; reserves in percent of quarterly imports; FXI as FX purchases in percent of GDP; capital account openness index from Chinn-Ito; real policy rate in percentage points; commodity terms of trade as index. * Hong Kong SAR is a territory of the People’s Republic of China with its own currency.

Appendix B. Foreign Exchange Intervention Proxy

The FXI data used in this paper have been adjusted for valuation effects. The methodology for doing so is found in Brandao-Marques et al. (2020) and Adler et al. (2019). They define the FXI proxy as,

FXIj,t=ΔNFAj,t−ΔValSecj,t−ΔValCurDepj,t.

This measure approximates FXI with the change in the central bank’s Net Foreign Assets (NFA), adjusted for valuation changes and the currency composition. ΔNFAj,t denotes the change in NFA for country) and time t. The breakdown to foreign securities (Secj,t,c), and foreign currency and deposits (CurDepj,t,c), where c stands for currency, is given by the IMF’s Data Template on International Reserves and Foreign Currency Liquidity. The currency composition of each asset class is assumed to be the same and to follow what is published by the IMF’s Currency Composition of Official Foreign Exchange Reserves (COFER) database. These currencies include: US dollar, Australian dollar, Canadian dollar, British pound, Japanese yen, Swiss Franc and Euro.

We follow Dominguez (2012) and assume securities to be all government securities and cash is assumed to earn zero returns beyond the exchange rate return. The valuation adjustment involves: subtracting from the change in NFA a COFER-weighted foreign currency total return index29 (using the US dollar as the numeraire); and subtracting the total return of the government bond indexes of each reserve currency, multiplied by the weight of each currency and share of securities in the central bank’s NFA.30:

ΔValSecj,t=Σc∈CSecj,t,cRj,t−1,cSecΔValCurDepj,t=Σc∈CCurDepj,t,cRj,t−1,cCur

We scaled both the actual FXI and proxy series using the respective country’s annual GDP and correlated the actual data with our proxy measure. The correlation is high for emerging markets in particular, where we lack actual intervention data and use proxy values.

Appendix C. Supplementary Exchange Rate Figures

Figure C.1.
Figure C.1.
Figure C.1.

Peaks and Troughs in Real Exchange Rates

Citation: IMF Working Papers 2022, 158; 10.5089/9798400215322.001.A001

Figure C.2:
Figure C.2:
Figure C.2:

Real Exchange Rate Spectral Decomposition

Citation: IMF Working Papers 2022, 158; 10.5089/9798400215322.001.A001

Appendix D. Historical Overview—Evolving Views on FX Intervention, Post-Bretton Woods

Since the end of the post-Bretton Woods system, academic and policymaker views on the appropriate role of foreign exchange intervention have shifted significantly. This section presents a highly selective synopsis of the evolving zeitgeist on the issue, arguing that in many respects the zeitgeist has come full circle. Equally remarkable, the gap between policy and academic views remains quite wide. However, it should be noted that this synopsis does not do justice to the rich historical record as it glosses over many intense debates, varied international experiences, and dueling academic studies. But it is a useful way to set the stage for the discussion of our empirical results and policy implications.

Our synopsis starts at the end of the Bretton Woods era’s adjustable-peg exchange rate system. By March 1973, the Bretton Woods system had run its course. There were a number of well-known economic and political economy issues which precipitated its demise; these have been documented elsewhere (e.g. Bordo (2019)). A new system had to be devised. With due consideration to the hotly debated pros and cons, the international community chose an international monetary system that in principle would rely less on government intervention and more on markets. This choice ushered in a new era of private-market-oriented exchange rate determination and held out promise for less active official foreign exchange intervention strategies.

Mussa (1981) provides an early yet insightful progress report on the post-Bretton Woods international monetary system. He highlights some notable developments. Exchange rates – both real and nominal – were quite volatile during the 1970s, as might have been expected given the significant break from past practices. Somewhat surprisingly, however, the use and holding of foreign reserves did not drop off sharply during this period. Governments appeared hesitant to relinquish control over exchange rates and acted as if exchange rate determination was too important to be left solely to the market. Mussa notes that “[t]o the extent that influencing the behavior of exchange rates is an important objective of government policy (as it appears in many countries), it is appropriate to include exchange behavior as one of the “targets” of monetary policy.” Indeed, Mussa points out that the ‘leaning against the wind’ monetary policies favored by ‘most central banks’ reflected concerns about excessive swings in exchange rates: ‘Experience over the past seven years indicates that the behavior of exchange rates influences the conduct of monetary policy, but only after exchange rates have moved substantially away from what authorities regard as their appropriate or desired values.” In sum, the experiences of the 1970s underscored the caution policymakers felt about the new international monetary system of market-based exchange rates.

By the end of the 1980s, the zeitgeist evolved significantly, culminating in the so-called Washington Consensus (WC). The WC, a term coined by John Williamson (1989), reflected the collective wisdom of various influential international bodies at the time. Skepticism about the effectiveness of central bank foreign exchange intervention policies became grew during the decade, especially after unsuccessful interventions associated with the Plaza and Louvre Accords and less-than-successful attempts at ‘dirty’ floating regimes in Latin America and at the ERM target zones approach. Moreover, academic research in the 1980s amassed considerable evidence questioning the assumption that foreign exchange interventions were effective (Sarno and Taylor (2001)). With no appetite for returning to a Bretton Woods-type system, the WC set out aspirational policy aims to improve the efficiency of the international monetary system. It emphasized competitive market-based exchange rate regimes and institutional reforms which would strengthen the foundations of free markets. While the initial set of specific aims did not speak directly to the issue of foreign exchange intervention, the WC was seen by many inside and outside Washington as a call for a binary international monetary system of either pure float or pure fixed. From the very beginning, the universal attractiveness of the approach was far from clear.

Indeed, experiences in the 1990s began chipping away support and exposing serious fault lines in the international community. The binary system – pure fixed or pure floating – certainly had its adherents. Various advanced economies (e.g. US, Canada, New Zealand) rejected the active use of exchange rate manipulation via exchange rate intervention in their monetary policy frameworks and adopted domestically focused inflation targeting regimes. The simultaneous arrival of the Great Moderation was interpreted by some as confirming the logic of the WC that flexible exchange rates act, on net, as effective shock absorbers. At the other extreme, continental European countries were growing weary of disruptive attempts at managing exchange rates among themselves and set out on a path to adopt a common European currency, the euro. Moreover, emerging market economies with fixed, quasi-fixed, or ‘dollarized’ exchange rates (e.g. Hong Kong SAR31 and Argentina) saw some initial successes. But various crises, especially the Tequila Crisis and the Asian Financial Crisis, were stark reminders of the risks of both unsuccessful pegs and premature efforts to float. The experience of the 1990s led Eichengreen and Hausmann (1999) to conclude that: “Time will tell which of these options [floating or dollarization] proves more appealing to emerging markets. Let us hope that it will not take too many more devastating crises to see the issues more clearly.”

One dominant theme at the end of the 1990s that proved prophetic about the future of the WC was the notion of “fear of floating” (Calvo and Reinhart (1999)). In the wake of the Asian Financial Crisis, for example, Asian central banks gravitated away from the flexible exchange rate ideal and dollarization. Instead they chose a third way – heavily managed exchange rate regimes. One nuanced difference with the past was rhetoric associated with the interventions. It generally focused on disorderly market conditions and periods of exchange rate swings that deviated significantly from a level that was deemed appropriate or desirable. Detractors of this approach voiced concerns that the one-sided nature of the interventions constituted a form of neo-mercantilist exchange rate policy and could eventually destabilize the international monetary system.32 These developments, among others, exacerbated the fault lines with respect to the Washington Consensus. At the same time, academic research was starting to overturn some of the findings from the decade earlier, reporting evidence that exchange rate interventions were effective. Most of the studies focused on the immediate impact of the interventions of relatively short time horizons. Sarno and Taylor (2001) argue that academic studies in the 1990s, using more high-quality data than in the past, contradicted the exchange rate intervention ineffectiveness hypothesis built up in the 1980s.

The Great Financial Crisis in the late 2000s added more stress to the fault lines. The introduction of unconventional monetary policies among the major advanced economies was widely seen as a clear and present danger in many emerging and small, open advanced economies. Wide swings in exchange rate pressures which did not reflect macroeconomic fundamentals elicited sharp responses. The phrase “currency wars” increasingly appeared in policy debates. In terms of actions, emerging markets responded to volatile capital flows with large, sustained foreign exchange interventions to resist the exchange rate pressures. Small, open advanced economies also experienced persistent one-sided exchange rate pressures which resulted in the adoption of floors (Czech Republic, Switzerland), extensive intervention (e.g., Israel, South Korea), new authority to intervene (Sweden), a more sympathetic view toward those central banks accumulating foreign reserves well above previously considered conventional benchmarks, and greater prominence of exchange rates in monetary policy deliberations. Exchange rate intervention strategies are being debated not in terms of being inconsistent with a well-functioning international monetary system but in terms of how to best use them to promote stability.

In sum, the economic record over the past five decades indicates that the exchange rate intervention zeitgeist has come full circle. We are still confronting some of the same key practical monetary policy challenges that Mussa (1981) long ago described: how should central banks respond to wide swings in exchange rates when the rates deviate far from what is considered appropriate or desired? Indeed, surveys of intervention practices since then have indicated continued use and reliance on intervention by central banks (e.g., BIS (2005, 2013, 2019b), Canales-Kriljenko (2003), IMF (2019), and Neely (2008)). In this respect our paper addresses this policy issue with the benefit of longer time-series, a wider cross-section of countries than in the past, and an innovative empirical design. In addition, the findings contribute to the active discussions on alternative monetary policy frameworks for the future (BIS (2019a), Lipton (2019)).
2

Chamon et al. (2019) review the Latin American experience with intervention and conclude that the countries have achieve a “considerable degree of success.”

3

Appendix D briefly summarizes the evolving views on the policy role of exchange rate intervention. The differences between the policy and academic views in the post-Bretton Woods period has waxed and waned over time. Arguably the gap is growing again.

4

Availability of cross-country data on financial factors emphasized in new theories of exchange rate intervention remains limited and incomplete, especially comparable balance sheet data and risk-taking capacity of global financial institutions (Gabaix and Maggiori 2015). This limits the ability to distinguish between financial factors and macroeconomic fundamentals as drivers of exchange rate swings movements.

5

At the same time, the size (in terms of daily activity) of the exchange rate market has also increased. The relative importance vis-à-vis the effectiveness of exchange rate intervention is ultimately an empirical question.

6

Our perspective is related to the extensive literature on long-term swings in real exchange rates (e.g., Engel and Hamilton (1990), Evans and Lewis (1995), Klaassen (2005), and Chen and Lee (2006)). These swings in part were seen as persistent changes in macroeconomic fundamentals and global financial conditions.

7

The cyclical nature of the misalignments—i.e., mean reverting—is consistent with the literature that exchange rates and long-run fundamentals are cointegrated (Mark and Sul, 2001) and that PPP holds in the long run (Canzoneri, et al., 1999).

8

Long time series are available for real and nominal exchange rates. However, in the main analysis of the paper, we are constrained to 1990–2018 due to availability of data on macroeconomic fundamentals.

9

Note that our findings are consistent with the two dominant theoretical mechanisms linking intervention to exchange rate dynamics: the portfolio balance and signaling channels. The portfolio balance channel assumes imperfect substitutability between domestic and foreign assets. The signaling channel emphasizes the information that interventions provide about changes to future policy; interventions are also seen to reflect changes in central bank intentions about the exchange rate and willingness to intervene (e.g., Cavallino and Patel (2019)). Assessing the contributions of these mechanisms is left for future research.

10

We use the REER as constructed by the IMF IFS using consumer price indices for all countries except Hong Kong SAR, where we use the REER constructed using unit labor cost indices.

11

Note, our data do not distinguish between exchange rate interventions that are sterilized and those that are not. However, during most of period covered by our regression analysis, central banks in our sample pursued interest rate operating frameworks which allow us to treat the interventions as effectively being sterilized.

13

In practice this is achieved using a Butterworth high-pass filter without a trend. We filter out stochastic frequencies above a pre-specified frequency corresponding to the short-, medium-, and long-run frequencies. The benefit of such a filter is that it is “maximally flat” in that the gain function is as close as possible to a flat line at 0 for the unwanted periods and 1 after.

14

Essentially, we calculate a 3-year moving average, identify turning points within a window of ±2 years, find the nearest max/min of the actual exchange rate that corresponds to this turning point.

15

The filtering results, using a Butterworth spectral filter, are presented in Appendix C, Figure C.2.

16

In our empirical implementation, we abstract from high frequency movements in the exchange rate, which can be driven by a whole host of factors including arbitrage, market sentiment, and liquidity.

17

Busetti and Caivano (2017) follow a similar approach for real interest rate determinants.

18

Note that the sample size is reduced from 30 to 26 countries due to data availability.

19

Policy variables comprise the real policy rate (nominal policy crate deflated using the current inflation rate), the reserve-to-imports ratio, and the Chin-Ito index for capital account openness; the global variables comprise of interactions between the VXO index and capital account openness, and between the real policy rate and capital account openness.

20

Note that the use of contemporaneous control variables attributes variation in the real exchange rate to them rather than FXI. In this sense, the specification may underestimate the contribution of FXI to exchange rate variation and hence yield more conservative estimates of the FXI impact.

21

The dependent variable is the natural logarithm of the real exchange rate index while the FXI variable is measured in levels of foreign exchange purchases as a percent of GDP, so a value of 1 is equivalent to an exchange rate intervention of 1 percent of GDP. The misalignments are defined in terms of the difference of natural logarithms and are mean zero by construction. The coefficients when multiplied by 100 give an approximate relationship as the percent change in the exchange rate.

22

The contemporaneous FXI coefficient is very small (.0005) despite its statistical significance; it is also of unexpected sign. This suggests FXI on its own is not quantitatively important if controls for the cycle-specific misalignments are included in the estimation. The “wrong” sign also suggests that the endogeneity bias may be present.

23

An alternative approach would have been to strip these FXI policy surprises out of the construction of exchange rate misalignment, by including them in the fundamentals model in eq 1. The challenge here is that doing so would restrict the sample size of the fundamentals model. We see a benefit in estimating cycles over as long a horizon as possible.

24

In addition to the robustness checks discussed here, we perform two additional exercises which are not discussed in detail in the interest of conciseness. We replace the FXI variables in our baseline model with a randomly generated FXI variables, with the same mean and standard deviation of actual interventions for each country. In this case, all coefficients are insignificant, suggesting that our results are not due to spuriously correlation between FXI and the misalignment measures. We also explore the possibility that the US bilateral exchange rate may be more sensitive to FXI owing to the prominent role USD reserves play at central banks. We re-estimate the model with the bilateral USD real exchange rate as the dependent variable. The results confirm that our conclusions from the REER baseline results continue to hold but provide weaker evidence on FXI effectiveness when leaning against USRER misalignments.

25

Blanchard et al. (2015) argue that global capital inflows in the rest of the world, i.e., excluding those to the home country, can be used as a valid instrument to orthogonalize exchange rate interventions with respect to domestic fundamentals. See Figure 3 for a comparison of this instrument relative to the FXI surprise instrument. The capital flow instrument implicitly assumes that global capital flows are largely independent of individual country domestic fundamentals. In this case, we regress FXI on a gross and net global capital inflows and use the fitted values as the FXI instrument. The effectiveness estimate is small and of the correct size. In many respects, the measure of global capital flows as a proxy for important global financial intermediation imperfections may be far from perfect but nonetheless points to the need for better financial proxies that capture global financial frictions.

26

For example, persistent or large interventions may have a stronger signaling effect than isolated, smaller ones. Large interventions are more likely to alter financial conditions associated with financial frictions.

27

The average spread in our dataset is 10bps (0.1 percent) with a standard deviation of 20bps.

28

Rj,t−1,cCur is the 3-month interbank rate, from Haver Analytics

29

Rj,t−1,cSec is the treasury’s total return index for the past three months, from Thomson Reuters Datastream

30

Hong Kong SAR is a territory of the People’s Republic of China with its own currency.

31

The stability of the system received considerable attention in the 2000s by those supporting the so-called Bretton Woods II view (see e.g., Dooley et al. (2019)).

Same Series

Other IMF Content

Other Publishers

  • Collapse
  • Expand
Cover IMF Working Papers
Exchange-Rate Swings and Foreign Currency Intervention
Author:
Andrew Filardo
,
Mr. R. G Gelos
, and
Thomas McGregor
Volume/Issue:
Volume 2022: Issue 158
Publisher:
International Monetary Fund
ISBN:
9798400215322
ISSN:
1018-5941
Pages:
41
DOI:
https://doi.org/10.5089/9798400215322.001
Figure 1.

Exchange Rate Intervention (FXI), by Country From 1990–2018
Figure 2.

Cyclic-Specific Exchange Rate (REER) Misalignments
Figure 2.

Cyclic-Specific Exchange Rate (REER) Misalignments
Figure 3.

FXI Surprise and Capital Flow IVs
Figure 4.

Exchange Rate (REER) Misalignments, by Cycle Type
Figure 4.

Exchange Rate (REER) Misalignments, by Cycle Type
Figure 5.

FX Sales vs. Purchases
Figure C.1.

Peaks and Troughs in Real Exchange Rates
Figure C.1.

Peaks and Troughs in Real Exchange Rates
Figure C.2:

Real Exchange Rate Spectral Decomposition
Figure C.2:

Real Exchange Rate Spectral Decomposition