Chapter 13 Uruguay: Interventions and Their Effects

Elizabeth Bucacos

Jump to Content Jump to Main Navigation

Author:

Elizabeth Bucacos

Elizabeth Bucacos
Search for other papers by Elizabeth Bucacos in
Current site
Google Scholar
PubMed

Language:: English

Keywords:: BOOK; central bank; intervention; exchange rate; foreign exchange intervention; spot market; intervention policy; dollar sale; foreign exchange; Exchange rates; Foreign exchange intervention; Currency markets; International reserves; Global; Central America

Download PDF (623.8 KB)

Abstract

Foreign exchange intervention is widely used as a policy tool, particularly in emerging markets, but many facets of this tool remain limited, especially in the context of flexible exchange rate regimes. The Latin American experience can be informative because some of its largest countries adopted floating exchange rate regimes and inflation targeting while continuing to intervene in foreign exchange markets. This edited volume reviews detailed accounts from several Latin American countries’ central banks, and it provides insight into how and with what aim many interventions were decided and implemented. This book documents the effectiveness of intervention and pays special attention to the role of foreign exchange intervention policy within inflation-targeting monetary frameworks. The main lesson from Latin America’s foreign exchange interventions, in the context of inflation targeting, is that the region has had a considerable degree of success. Transparency and a clear communication policy have been key. For economies that are not highly dollarized, rules-based intervention helped contain financial instability and build international reserves while preserving inflation targets. The Latin American experience can help other countries in the design and implementation of their policies.

In the wake of the global financial crisis in 2008, the Central Bank of Uruguay faced challenges in global capital flow volatility. It used different tools within the framework of its inflation-targeting regime to cope with the volatility and foreign exchange misalignments. This chapter, in its study of the effects of foreign exchange intervention in Uruguay during 2005–17, analyzes whether responses in the Uruguayan peso/US dollar rate vary with different types of intervention. Relying on a monthly vector autoregressive and a two-step approach using weekly data, the analysis finds robust evidence for a significant but short-term effect of intervention on the nominal exchange rate, depending on the type of intervention. Response seems to be asymmetric depending on whether the intervention involves net purchases or net sales. The evolution of macroeconomic and financial indicators suggests that the combination of foreign exchange intervention with other monetary and macroprudential measures succeeded in Uruguay, so that significant real exchange rate misalignments could be avoided, thus impeding the effects of large capital flow swings and domestic portfolio changes on economic activity and macroeconomic stability.

Introduction

As a small, open economy with free capital movements, a floating exchange rate, and high currency substitution within the framework of an inflation-targeting regime, Uruguay has faced significant challenges during the financial turmoil of recent years. This chapter covers the phases of the economic cycle in the country and the international macro financial cycle during 2005–17.

The global financial crisis that began in the United States in 2008 caused large capital inflows into emerging market economies. Besides “traditional” monetary easing, a series of unconventional policy measures—including long-term asset purchases by the US Federal Reserve, the European Central Bank, the Bank of Japan, and the Bank of England—led to reallocation of funds from advanced economies to the emerging markets (see Kolasa and Wesolowski 2018). This put appreciation pressures on their exchange rates. The macro-financial cycle during that period also exhibited high financial market volatility, capital flow reversals linked to greater uncertainty and risk aversion, and changing expectations of rates of return of currencies and assets. In short, these global and regional shocks posed risks to macroeconomic stability. The sudden and large portfolio changes by domestic and external institution investor markets posed risks of misalignment and macroprudential issues.

In general, the response to such global shocks in financially open economies involves the use of monetary, fiscal, financial, and macroprudential tools. The specific findings of this chapter’s scrutiny of policy responses show that foreign exchange intervention was an effective tool (if not the only one) for dealing with the adverse global capital flow shocks on emerging market economies (for example, see Blanchard, Adler, and de Carvalho Filho 2015; Daude and others 2016).

In Uruguay’s policy response, the countercyclical role of sterilized foreign exchange intervention was complemented by an asset and liability approach to the integrated balance sheet of the public sector and reserve requirements on nonresident investments in public debt in the primary market (Vicente, Malacrida, and Zimet 2017). Taken together, these monetary and financial tools helped temper the adverse effects of large capital flow swings and the related domestic portfolio changes on foreign exchange fundamentals. It also helped control volatility in relative prices, currency markets, and interest rates.

Foreign exchange intervention was a Uruguayan trademark in the fifteen years after the currency was allowed to float, in response to the 2002 crisis.¹ Indeed, Uruguay is among the most active Latin American countries in this regard, and in a region where intervention through varying mechanisms was widespread in the period reviewed.

As such, Uruguay is relevant to the discussion for its significant levels of intervention under its inflation-targeting regime that use monetary aggregates as the instrument, and because it is one of the most highly dollarized countries in the region.

This chapter aims to answer a series of questions. Do interventions affect the level or the volatility of the exchange rate? Are direct interventions (foreign exchange purchases/sales in the spot market) different from indirect interventions (export prefinancing and so on)? Do sterilized interventions have an effect?

Because the chapter focuses on the impact of different types of intervention rather than on effectiveness itself, it concentrates on implementing a well-understood and accepted methodology. To do this, it examines whether one can identify the different exchange-rate responses to the different types of intervention, and it uses two approaches: With monthly data, it estimates a simple vector autoregressive (VAR) in differences between the variables generally used in empirical models of the exchange rate and exchange rate intervention. With monthly and weekly data (not reported here), it applies to Uruguay the two-stage approach proposed by Adler and Tovar (2014).

The results suggest the following:

Interventions affect the level of the exchange rate, but the effect is short-lived;
Although the effect of indirect interventions appears with the expected sign, and is statistically significant, direct interventions get a meaningful response through foreign exchange sales, while purchases “just” prevented the peso from appreciating further;
Sterilized intervention does not seem to affect the level of foreign exchange for more than one week after foreign exchange purchases, explaining why no effect is found when monthly data is used;
Interventions have asymmetric effects on the foreign exchange rate; that is, purchases of foreign exchange (which tend to increase the UY peso/US dollar rate) are more costly in terms of GDP than are sales of foreign exchange (which tend to decrease its cost);
Communication to the public of relevant information regarding the value of the exchange rate seems to play a role in the motives of de facto intervention;
The central bank seems to worry about the level of the appreciation velocity of the foreign exchange, rather than its volatility;
Real exchange rate misalignments seem to have been a reason for intervening only in the case of central bank sales; and
The flip side of sterilized interventions is the increase in the stock of monetary regulation securities, and these excess reserves are a macroprudential buffer with associated costs, because of interest rate differentials between Uruguay and the United States.

The combination of different tools, including foreign exchange intervention, helped avoid foreign exchange misalignments and levels of volatility incompatible with macroeconomic stability. The next section describes the characteristics most relevant to analyzing intervention in Uruguay in the stated period. The chapter describes the VAR methodology and its results, and the results of the two-stage approach.

Uruguayan Intervention and Monetary Policy during 2004–17

Several important characteristics are needed to understand monetary policy and foreign exchange interventions in Uruguay in the period examined, including a short description of the macroeconomic environment, relevant structural facts, and the institutional setup of monetary policy.

The Macroeconomy during 2004–17

Figure 13.1 describes the relevant international environment. Financial conditions had been exceptionally good, though extremely volatile, since the global financial crisis began in 2008, sparking a globally coordinated and expansive monetary policy stance. This resulted in low interest rates and country risk premiums, but these were under a permanent threat of reversion. Commodity prices and terms of trade were high though volatile. Regional demand was also a positive factor through most of the period. Both Argentina and Brazil had internal political problems, but regional output was generally positive. Uruguay’s economy entered a prolonged expansion cycle between 2004 and 2017, albeit with slower growth during 2009 and 2013–17.

These generally favorable results emerged, as noted, despite episodes of high volatility in financial markets, capital flow reversals linked to greater uncertainty and higher risk aversion, and changing expectations on rates of return for currencies and assets during the period.

Structural Facts to Understand Intervention in Uruguay

The US dollar is important to Uruguay’s financial and price systems, and the country remains highly dollarized, even though currency mismatches have recently declined. Deposit dollarization, although down recently, remains close to 80 percent of total deposits. The currency composition of banking credit, however, has changed, with dollarization falling below 50 percent in 2017.

Firms’ financial positions have been transformed since the 2002 crisis, greatly reducing exchange rate exposure on balance sheets, although dollar leverage remains high. At the same time, the government has increased the share of domestic debt and accumulated a large portfolio of international reserve assets. Domestic financial markets remain very shallow, including foreign exchange markets. Domestic markets grew during 2004–17, but their size remains very small by global standards. The wholesale exchange rate market amounts to only 12 percent of GDP and is strongly concentrated on the purchasing side. The stock of exchange rate forwards in Uruguay, meanwhile, amounts only to a little over 3 percent of GDP, and turnover is very small.

The financial system is, likewise, very concentrated, dominated by public banks and pension funds. Only 11 banks are operating, and public banks hold 49 percent of total liabilities, while the four pension funds in the country control a combined asset portfolio (mostly in government bonds) that amounts to 23 percent of GDP.

The US dollar also plays a big role in the price system, as noted. Tradable goods account for close to 40 percent of the consumer price index (CPI) basket, which leads to a high speed of exchange rate pass-through.² Meanwhile, US dollar invoicing of domestic sales is widely extended in the wholesale sector (an internationally unusual practice), as Baron and others (2017) show.

One final important fact is that the government needs to purchase large amounts of foreign exchange to cover its operational flows (Figure 13.2). Because Uruguay is an oil-importing country, and energy-producing utilities are state owned, the government has traditionally needed to purchase foreign currency to pay for those imports. Oil imports amounted to 4 percent of GDP in this period (2004–17), representing 40 percent of energy demand. The change in the net foreign position of the Uruguayan government has reduced the traditional need to purchase currency to service debt.

Intervention Characteristics

This section details features of intervention in the Uruguayan foreign exchange market, mainly regarding agents and types of intervention.

Intervention Goals

As can be seen, the US dollar looms large for prices and financial stability. As the weight of tradable goods in the CPI basket is high, and several nontradables are invoiced in US dollars, the pass-through of the exchange rate to inflation is fast. In addition, because most financial transactions are in US dollars, domestic balance sheets are exposed—despite the recent reduction in currency mismatches—to wealth effects that might affect the asset side more strongly because of quality mismatches between assets and liabilities. These factors justify the strong informational value of the US dollar in the Uruguayan society and the government, and are reasons behind the close watch that the government keeps on exchange rate behavior.

Central bank authorities therefore explain intervention in the foreign exchange market as responding to real exchange rate misalignments and excessive volatility relative to sustainable fundamentals in the market. In particular, the central bank explains its resistance to validating transitory shocks to the appreciation of the domestic currency as concerns over real activity and long-term decisions of the private sector. For example, amid strong capital inflows, in July 2012, Mario Bergara, president of the Central Bank of Uruguay at the time, stated that authorities should be able to differentiate permanent versus transitory shocks: “Zero percent interest rates are not going to last long,” and determine intervention with an aim to reduce the impact of the shock on the real sector.³ On the other hand, authorities have tried to avoid discrete jumps in the exchange rate on financial stability grounds. In September 2015, Bergara claimed that “the Uruguayan (s)ociety is still afraid of what could happen if the exchange rate leaps: that’s why we cannot allow the exchange rate to be a rollercoaster”.⁴

Intervention as a Rule, Not an Exception

Intervention, by Type

Uruguayan authorities use several types of intervention mechanisms in the foreign exchange market. The analysis here concentrates on two classifications:

Sterilized and nonsterilized interventions, and
Direct and indirect interventions

As Uruguay manages monetary aggregates, defining sterilized interventions becomes more problematic than when a monetary authority controls the interest rate. When a central bank controls interest rates, the researcher can assume that interventions are sterilized, as long as the interest rate in the money market remains unaffected, as the monetary effect of the intervention should be zero. When a monetary authority controls monetary aggregates, an intervention is sterilized if the monetary program remains on track, meaning that the expansion generated by the unexpected purchase of foreign currency is compensated by open market operations. Regular purchases of currency by the central bank might not be a good proxy in a dollarized economy, as the central bank might resort to purchasing/selling foreign exchange to sterilize/issue liquidity as part of its regular monetary operations. Since the correct proxy for sterilization is not available, this analysis uses the net of foreign reserves and debt issued by the central bank as the indicator of sterilized intervention.⁵

Figure 13.3 shows that sterilized and nonsterilized interventions are carried out simultaneously, in some months as substitutes and in other months as complements. The interaction of different instruments to intervene in the exchange market accumulated international reserves in the central bank, which significantly increased the stock of monetary regulation securities along with its maintenance cost (see Figure 13.4).

Figure 13.5 presents the estimated cost of the surplus reserves, calculated as follows:

\begin{matrix} C S R_{t} = \frac{(R_{t} - S R_{t}) (i_{t}^{U_{y}} - i_{t}^{U S A})}{Y_{t}}, & (13.1) \end{matrix}

where R_t is the annual average reserves expressed in millions of US dollars; SR_t is the annual average “comfortable” level of reserves;⁶ $(i_{t}^{U_{y}} - i_{t}^{U S A})$ is the annual average spread between Uruguayan and US bonds, and Y_t is current Uruguayan GDP expressed in millions of US dollars.

Interventions can be either direct or indirect. Indirect interventions refer to several operations: (1) Settling local currency securities in US dollars (this avoids an effect on the foreign exchange market of foreign exchange conducted in the purchase/sale of public paper), (2) exchange forwards settlements, and (3) export prefinancing. Direct interventions refer to sales or purchases in the exchange market, both spot and forward. The main reason for the distinction is the communication role of interventions. Other agents observe direct intervention and internalize information for financial decisions; while indirect interventions are not visible to the market in real time.⁷, ⁸ As Figure 13.4 shows, indirect interventions are much bigger by volume than direct interventions, though the mean monthly interventions are similar (see Annex 13.1).

Monetary Policy Design and Implementation

For background on monetary policy conduct in Uruguay during 2004–17, the analysis details the institutional setup of monetary policy and its history.

Institutional Setup of Monetary Policy

The Central Bank of Uruguay carries out monetary policy with a dual mandate from its charter law’s statement of main goals: “price stability that is consistent with growth and employment” and financial stability. The charter also describes how the central bank interacts with the Ministry of Finance: there is a Macroeconomic Coordination Committee to generate an environment of coordination between the central bank and the executive. If a disagreement arises about the “monetary policy system,” the opinion of the Ministry of Finance prevails.⁹ Practice shows that this committee meets quarterly, just before the meeting of the Monetary Policy Committee. The committee, internal to the central bank, is in charge of deciding monetary policy and is comprised of the members of the board of the bank (the only ones with voting power), the heads of the Monetary Policy and Markets and Economic Advising divisions, and other top officers of the central bank invited in an advisory function. Monetary policy is conducted through central bank bills and other forms of liquidity injection/sterilization, including foreign exchange operations.

Meanwhile, the Debt Committee coordinates placement of central bank and government paper in the market. Traditionally, central bank paper is issued in short maturities and government paper with maturities over two years.¹⁰ In Uruguay, the countercyclical role of sterilized foreign exchange intervention is complemented by the use of an asset and liability approach to the integrated balance sheet of the public sector and reserve requirements on nonresident investments in public debt in the primary market.¹¹

Stages of Monetary Policy since 2002

Uruguay moved to a floating exchange rate system after the July 2002 collapse of the exchange rate crawling band. The period from 2002 to the present can be divided into four stages according to the monetary system implemented:

1. As the country was exiting the turmoil of the crisis, it transitioned toward inflation targeting, a period that lasted until the end of the first semester of 2004.
2. Between the second semester of 2004 and the first of 2007, Uruguay had what can be called an inflation-targeting system managing monetary aggregates as the instrument of monetary policy.
3. From the second semester of 2007 to the first semester of 2013, the country operated an inflation-targeting regime with the one-day nominal rate as the instrument of monetary policy.
4. From the second semester of 2013, the country returned to the management of its inflation-targeting regime with monetary aggregates.

The first (transition) phase was a reorganization phase. Uruguay would not solve fiscal sustainability until the May 2003 debt restructuring was completed. In that phase, the central bank tried to show the public that it could deliver on monetary policy by setting targets for the monetary base, with no reference to an inflation target. In that period, the commitment to inflation was gradually increased, while the commitment to the monetary target (first the base but later M1)¹² was phased out. Inflation as a reference was introduced for 2003. Then, the language of the central bank would start to give greater relevance to inflation, to finally call it an inflation target by the end of the first semester of 2004. Initially, the central bank announced a target for the monetary base of the following year. Later, that reference was transformed into a target for M1 and by the second semester of 2004 there was a reference for M1 in a target that implied no commitment. One of the priorities of the transition was restoring international reserve assets to a comfortable level. This was accomplished by late 2003 according to the safety criteria of the central bank.¹³

In the second semester of 2004, during the second phase, even though the central bank would not officially recognize it, Uruguay was in an inflation-targeting regime. That is, the central bank had a public target for inflation and no commitment whatsoever to monetary aggregates.

The first phase of inflation targeting with monetary aggregates was marked by consistent surprises in money demand growth. Money demand until 2002, because of the incentives that led to the dollarization of the Uruguayan economy (see Licandro and Licandro 2003), trended downward. After the debt restructuring, and the implementation of several regulatory changes to deal with the financial stability threat that posed currency mismatches, the trend of money demand changed, and Uruguay started to experience a period of strong remonetization. The favorable environment described in the previous section was also a factor. As a result, monetary policy was consistently more contractive than forecasted, leading to the piercing of the lower bound of the inflation target in the first semester of 2005, and to increasing public pressure on monetary policy because of the strong appreciation of the currency.

Against this background, the central bank decided to change the instrument of monetary policy to the one-day interest rate in the second semester of 2007 (the third phase). International volatility marked the period. On the negative side, both the Lehman Brothers and the European debt crises occurred. On the positive side, Uruguay regained investment grade status by all rating agencies, and it experienced positive financial market access shocks that enhanced the volatility of capital inflows. Furthermore, Uruguay experienced strong foreign direct investment associated with high commodity prices, the development of pulp paper production, and increasing restrictions on agricultural production in Argentina.

During the Lehman Brothers episode, the Latin American region, mostly with inflation and inflation expectations under control, decided to provide liquidity and lower interest rates; however, Uruguay took a completely different approach. Concern over the effects on financial stability of a jump in the exchange rate only six years after the Uruguayan crisis of 2002, led the central bank to exercise an interest rate defense of the exchange rate. As the liquidity shortage in foreign currency grew, the Uruguayan government decided to assist the portfolio change by offering a repurchase of short-term domestic paper, but in limited amounts. As a result, the interest rate jumped. Once the worst of the international liquidity crunch passed, Uruguay allowed the interest rate to return to 10 percent, a level chosen to control inflation and inflation expectations.

Despite the environment that followed the Lehman Brothers episode, both inflation and inflation expectations remained outside the target set by the Macroeconomic Coordination Committee. In May 2013, two days before the announcement of tapering by the US Federal Reserve, Uruguay announced a return to the management of monetary aggregates. The announcement was made amid concerns about the cost of sterilization and high portfolio capital inflows after Uruguay’s credit rating was raised to investment grade by a second agency, effectively opening it up to institutional investor capital.

The next section explores the effectiveness of government actions on the exchange rate, identifying whether different types of intervention (sterilized or nonsterilized, direct or indirect) had different impacts on the exchange rate.

Var Methodology

The data set for the empirical analysis was obtained from three sources: the Electronic Stock Exchange, the Central Bank of Uruguay, and Bloomberg Finance L.P. It uses monthly data from January 2007 to December 2016 for exchange rates, net foreign exchange purchases of governmental institutions, domestic and foreign interest rates, and risk and volatility measures, such as the VIX index and the Emerging Market Bond Index (EMBI).

The analysis uses a multivariate time series approach to measure the impact of interventions over the parity Uruguayan peso–US dollar (Annex 13.1 presents the main descriptive statistics for the variables used). During the period, the central bank participated heavily in the exchange market, mainly through open market operations sterilizing capital inflows. There is no significant size difference between direct and indirect interventions. Included in the analysis are other variables that are standard in the literature: the exchange rate in Brazil, Uruguayan interest rates, expected inflation and expected depreciation, the VIX Index, and the EMBI for emerging markets.¹⁴^, ¹⁵

Figure 13.6 shows the very close relationship between the Uruguayan peso and the international US dollar value, which leads to thoughts about possible cointegration. That assumption is backed by a Johansen test for cointegration. As net foreign exchange purchases (NFXP) is I(0) and endogenous, we decided to estimate a VAR model in differences for the exchange rates series, treating interventions as an endogenous variable.

The VAR model, expressed in matrix form, is represented as follows:

\begin{matrix} Y = B Z + C X + U, & (13.2) \end{matrix}

where Y denotes the endogenous variables vector, Z is the autoregressive matrix of the endogenous variables, and X represents the matrix of the exogenous variables.

The endogenous variables are: $Δ E_{t}^{B r l}$ , in differences; $Δ E_{t}^{U y}$ domestic exchange rate, in differences; NFXP_t net foreign exchange purchases, in levels; I_COMM_t communications, a discrete variable, nominal interest rate, in levels; $r_{t}^{U I}$ indexed unit interest rate, as a proxy of real interest rate, in levels; $d_{t + 1}^{e}$ expected depreciation, in levels.¹⁶ The exogenous variables are: $i_{t}^{U S}$ nominal interest rate for the United States, in levels; $ρ_{t}^{E M E}$ emerging markets EMBI; VIX_t VIX index, in levels, as a measure of volatility.

Estimation Results

The analysis estimates a general model for all interventions, a model separating direct and indirect interventions in the exchange market, a model distinguishing sterilized and nonsterilized interventions, and a model differentiating the central bank’s and central government’s interventions (see Table 13.1). It also analyzes the response of the exchange rate and the expected depreciation to a change of one standard deviation in net foreign exchange purchases for different types of intervention. It uses Choleski factorization for identification, which is particularly controversial, since the arbitrary order assigned to the endogenous variables could determine the impulse response functions. The order assigned was:

\begin{matrix} Δ E_{t}^{B r l} \to Δ E_{t}^{U y} \to N F X P_{t} \to I_{-} C O M M_{t} \to d_{t + 1}^{e} \to i_{t} \to r_{t}^{U I} & (13.3) \end{matrix}

The logic in assigning this order is that the domestic market observes the exchange rate of its relevant market from a financial economic perspective (Brazil) whose movements are transmitted to the domestic exchange rate, the central bank in the foreign exchange market, said intervention is collected by the agents through the expected depreciation, which, via Fisher’s open parity, affects domestic interest rates.

^{Table 13.1.}

Models Estimated, Monthly Data

Source: Authors’ calculations. Note: Not available are data that discriminates between direct and indirect interventions before March 2011. VAR = vector autoregression. *p < 0.1; **p < 0.05; ***p < 0.01.

^{Table 13.1.}

Models Estimated, Monthly Data

VAR Global Interventions			VAR Direct and Indirect Interventions			VAR Sterilized and Nonsterilized Interventions
	ΔE^UP_t	NFXP_t	ΔE^Uy_t	NFXP_DIR_t	NFXP_ND_t	ΔE^Uy_t	NFXP_STER_t	NFXP_NONSTER_t
ΔE^Brl_t-1	1.290*** (0.500)	33.460*** (73.288)	1.323*** (0.610)	196.317 (191.648)	-A131S (116.816)	1.291*** (0.518)	0.868 (0.676)	-0.945 (1.258)
ΔE^Uy_t-1	1.195*** -0.0896	-29.658*** (13.138)	1.098*** (0.140)	-33.980 (44.132)	-59.206*** (26.900)	1.116*** (0.100)	-0.423*** (0.128)	0.282 -0.239
ΔE^Uy_t-1	-0.404*** (0.084)	21.156* (12.251)	-0.344*** (0.118)	30.756 (37.221)	27.524 (22.688)	-0.342*** (0.091)	0.458*** (0.118)	-0.410* (0.220)
NFXP_t-1	0.001* (0.000)	0.296*** (0.089)
NFXP_t-2	-0.001 (0.001)	0.201*** (0.085)
NFXP_IND_t-1			0.129* (0.073)	-0.296* (0.161)	0.265*** (0.097)
NFXP_EST_t-1						0.001 (0.090)	0.672*** (0.118)	-0.513*** (0.220)
NFXP_NONSTER_t-1						0.104***	0.096 (0.062)	0.152 (0.115)
NFXP_NONSTER_t-2						0.040 (0.048)	0.137*** (0.062)	-0.057 (0.116)
I-COMM_t-1	0.238*** (0.112)	0.918 (16.362)	0.561*** (0.161)	-6.895 (50.625)	-22.424 (30.858)
i^Uy_t-1	0.014 (0.036)	11.070***	-0.053 (0.057)	3.174 (18.000)	35.177*** (10.972)
d^e_t-1	0.264* (0.148)	-27.749 (21.654)				0.246* (0.149)	-0.416*** (0.180)	-0.838*** (0.362)
γ^UI_t	0.142*** (0.065)	6.936 (9.591)	0.011 (0.152)	-11.837 (47.781)	-62.454*** (29.124)	0.133** (0.067)	0.175* (0.109)	-0.531*** (0.202)
ρ^EME_t	0.285*** (0.090)	-25.693** (13.200)	0.437*** (0.174)	-123.026*** (54.542)	-64.724** (33.245)	0.278*** -0.092	-0.004 -0.119	-0.318 -0.222
i^USA_t	0.398*** (0.100)	-38.656*** (14.670)	-0.145 (0.290)	15.563 (90.981)	-96.255* (55.456)	0.417*** (0.101)	-0.149 (0.132)	-0.342 (0.246)
VIX_t	0.016** (0.001)	-0.986 (1.144)				0.015** (0.008)	-0.023*** (0.010)	0.0207 (0.019)
D_1	0.124928 (0.275)	-265.301*** (40.379)	0.680 (0.455)	26.034 (142.987)	-388.197*** (87.155)	0.063 (0.447)	-1.214*** (0.583)	-2.504*** (1.084)
C	-1.092*** (0.330)	161.715*** (48.433)				-1.008*** (0.342)	0.350 (0.446)	1.253* (0.830)
R²	0.945	0.797	0.955	0.577	0.850	0.947	0.846	0.464
Adjusted R²	0.935	0.759	0.937	0.399	0.783	0.935	0.810	0.341
No. of observations	131	131	82	82	82	131	132	133

^{Table 13.1.}

Models Estimated, Monthly Data

VAR Global Interventions			VAR Direct and Indirect Interventions			VAR Sterilized and Nonsterilized Interventions
	ΔE^UP_t	NFXP_t	ΔE^Uy_t	NFXP_DIR_t	NFXP_ND_t	ΔE^Uy_t	NFXP_STER_t	NFXP_NONSTER_t
ΔE^Brl_t-1	1.290*** (0.500)	33.460*** (73.288)	1.323*** (0.610)	196.317 (191.648)	-A131S (116.816)	1.291*** (0.518)	0.868 (0.676)	-0.945 (1.258)
ΔE^Uy_t-1	1.195*** -0.0896	-29.658*** (13.138)	1.098*** (0.140)	-33.980 (44.132)	-59.206*** (26.900)	1.116*** (0.100)	-0.423*** (0.128)	0.282 -0.239
ΔE^Uy_t-1	-0.404*** (0.084)	21.156* (12.251)	-0.344*** (0.118)	30.756 (37.221)	27.524 (22.688)	-0.342*** (0.091)	0.458*** (0.118)	-0.410* (0.220)
NFXP_t-1	0.001* (0.000)	0.296*** (0.089)
NFXP_t-2	-0.001 (0.001)	0.201*** (0.085)
NFXP_IND_t-1			0.129* (0.073)	-0.296* (0.161)	0.265*** (0.097)
NFXP_EST_t-1						0.001 (0.090)	0.672*** (0.118)	-0.513*** (0.220)
NFXP_NONSTER_t-1						0.104***	0.096 (0.062)	0.152 (0.115)
NFXP_NONSTER_t-2						0.040 (0.048)	0.137*** (0.062)	-0.057 (0.116)
I-COMM_t-1	0.238*** (0.112)	0.918 (16.362)	0.561*** (0.161)	-6.895 (50.625)	-22.424 (30.858)
i^Uy_t-1	0.014 (0.036)	11.070***	-0.053 (0.057)	3.174 (18.000)	35.177*** (10.972)
d^e_t-1	0.264* (0.148)	-27.749 (21.654)				0.246* (0.149)	-0.416*** (0.180)	-0.838*** (0.362)
γ^UI_t	0.142*** (0.065)	6.936 (9.591)	0.011 (0.152)	-11.837 (47.781)	-62.454*** (29.124)	0.133** (0.067)	0.175* (0.109)	-0.531*** (0.202)
ρ^EME_t	0.285*** (0.090)	-25.693** (13.200)	0.437*** (0.174)	-123.026*** (54.542)	-64.724** (33.245)	0.278*** -0.092	-0.004 -0.119	-0.318 -0.222
i^USA_t	0.398*** (0.100)	-38.656*** (14.670)	-0.145 (0.290)	15.563 (90.981)	-96.255* (55.456)	0.417*** (0.101)	-0.149 (0.132)	-0.342 (0.246)
VIX_t	0.016** (0.001)	-0.986 (1.144)				0.015** (0.008)	-0.023*** (0.010)	0.0207 (0.019)
D_1	0.124928 (0.275)	-265.301*** (40.379)	0.680 (0.455)	26.034 (142.987)	-388.197*** (87.155)	0.063 (0.447)	-1.214*** (0.583)	-2.504*** (1.084)
C	-1.092*** (0.330)	161.715*** (48.433)				-1.008*** (0.342)	0.350 (0.446)	1.253* (0.830)
R²	0.945	0.797	0.955	0.577	0.850	0.947	0.846	0.464
Adjusted R²	0.935	0.759	0.937	0.399	0.783	0.935	0.810	0.341
No. of observations	131	131	82	82	82	131	132	133

All Foreign Exchange Interventions

As can be seen in Table 13.1, the effect of interventions on the exchange rate level is barely significant and lasts only a short time. The response of the change in the exchange rate to an increase of one standard deviation in the interventions in the exchange market is significant and positive, with a very short duration, one period ahead, as presented in Figure 13.1.1. The expected depreciation (one-year horizon) is negative and significant for the second and third period ahead. This result is consistent with the perception among market players that the effect of intervention is transitory. In a scenario of appreciation of the peso, intervention briefly sustains the domestic price of the US dollar, but after the effect of intervention is over, expected depreciation anticipates an appreciation of the peso.

Evidence suggests that communications are effective in the short term. When the government communicates the concerns behind monetary policy actions and foreign exchange intervention, this communication moves the exchange rate in the desired direction (Table 13.1).

Direct and Indirect Foreign Exchange Interventions

Direct interventions refer to operations done directly in the exchange market. Contrary to expectations of the analysis, indirect interventions have a stronger effect on exchange rate changes than direct interventions do. The effect on the expected depreciation is similar to the previous model for general intervention. Direct interventions do not seem to explain changes in the foreign exchange rate in a statistically meaningful way. Indirect interventions are barely significant in the short term. Figure 13.2.1 presents the impulse response functions of these VAR models. Interventions seems to be incorporated by the agents in their expectations: When the effect of a purchase of foreign exchange over the exchange rate fades, expectations of domestic currency appreciation begin to appear. As Table 13.1 shows, indirect interventions are negatively correlated with direct interventions, so they seem to be used as substitutes in some periods.

Sterilized and Nonsterilized Interventions

Table 13.1 also shows that sterilized interventions do not affect the exchange rate. Nonsterilized interventions are effective in impacting the exchange rate level for a short period. When the effect of the nonsterilized interventions over the exchange rate end in the medium term, expectations of local currency appreciation appears. It seems that sterilized and nonsterilized interventions are used simultaneously in a complementary form, and are thus positively correlated.

Overall, the evidence in this section points to a short-lived effect of nonsteril-ized intervention. Effects of sterilized intervention using this VAR methodology could not be found. Communications do seem to play a role in the short term, as expected by the literature.

The next section delves deeper into the analysis of weekly data using the approach suggested by Adler and Tovar (2014).

Two-Stage Approach

This section describes the two-stage approach to weekly data and presents the estimation results. It analyzes the problem in more detail by expanding the frequency data, and in that way, it is expected to improve the understanding of the effectiveness and duration of foreign exchange intervention.

Econometric Approach

It is quite difficult to discern the direction of causality between foreign exchange intervention and foreign exchange performance, as intervention affects the exchange rate, and the decision to intervene depends on the evolution of the exchange rate. In effect, simple correlation between them would wrongly suggest that an increase in foreign exchange purchases would appreciate the exchange rate (measured as domestic currency per foreign currency). To overcome this endogeneity problem, as is well-known in the foreign exchange intervention literature (Kearns and Rigobon 2005), this analysis follows the two-stage estimation process in Adler and Tovar (2014). In the first stage, a de facto reaction function for the central bank is estimated; in the second stage, predicted values of this reaction function are used as instruments in the estimation of a behavioral equation for the exchange rate.¹⁷

First Stage: Central Bank Reaction Function

The central bank reaction function is modeled as a censored variable and estimated with a Tobit model, with weekly data during January 2005–December 2017.¹⁸^,¹⁹ Formally:

\begin{matrix} I_{t} = \max (0, β_{0} + β_{1} D e_{t - 1} + β_{2} (r e e r_{r} - r e e r_{t}^{e q}) + β_{3} Δ_{t} + β_{4} σ_{t} + β_{5} R_{t}^{M 2} + β_{6} C_{t} + u_{t}), & (13.4) \end{matrix}

where I_t is the amount of interventions (net purchases or net sales as percentage of GDP) in different versions: sterilized versus nonsterilized, direct versus indirect; De _t-1 is the lagged change in the bilateral exchange rate (Uruguayan peso/US dollar) that captures short-run movements; $r e e r_{r} - r e e r_{t}^{e q}$ stands for real effective exchange rate misalignments; Δ_t is a measure of the nominal appreciation/depreciation velocity (over trend (HP)); σ _t is a nominal exchange rate volatility measure (over trend (HP)); $R_{t}^{M 2}$ is a reserve assets over M2 ratio, which captures precautionary motives; C_t reflects how communication of relevant information may affect the decision to intervene and is presented as a dummy variable that takes the value of –1 (the message entails sales by the central bank); 0 (the message is neutral regarding the market intervention); and 1 (the message entails purchases by the central bank).²⁰^,²¹^,²² Finally, u _t is the error term. All variables are weekly and standardized.

Second Stage: Exchange Rate Equation

The aim of the second stage of this estimation process is to link movements in the exchange rate to central bank intervention, using the forecast values of the reaction function previously estimated as an instrument. In that way, the analysis uses a variable that is highly correlated with foreign exchange intervention but relatively free of the endogeneity problem reported in the literature. It also includes other variables as controls: interest rate differentials, sovereign spreads, commodity price shocks, and the US trade-weighted exchange rate. Formally,

\begin{matrix} e_{t} = \max {0, f [(i_{t} - i_{t}^{*}), S_{t}, P_{t}^{E}, P_{t}^{S}, P_{t}^{B}, {\hat{I}}_{t}, D U S_{t}],} & (13.5) \end{matrix}

where e_t stands for the log of the nominal bilateral exchange rate (Uruguayan pesos against the US dollar); i_t is the domestic 30-day nominal interest rate for the peso-nominated yield curve; $i_{t}^{*}$ is the effective federal funds rate, and the difference between them shows the interest rate spread; S_t is the EMBI differential for Uruguayan sovereign bonds; $P_{t}^{E}, P_{t}^{S}, a n d P_{t}^{B}$ are the logs of the international price indexes for energy, soybean, and beef (basic products for Uruguay that may influence the evolution of the exchange rate); Î_t is the amount of forecast intervention from the first stage of the estimation procedure; and DUS_t is the trade-weighted nominal exchange rate index for the United States, which has a direct correlation with capital flows in emerging market economies.²³,²⁴,²⁵ All variables are weekly and standardized for January 2005 to December 2017. Optimal lags are selected using information criteria.

The specific model approach used to estimate the exchange rate equation depends on the order of integration of the forecast intervention variable. Because only Î_t is stationary, the analysis has to deal with I(0) and I(1) variables in the same system. Assuming that the intervention shock has transitory effects on the other variables, the analysis pursues a VAR in differences for the I(1) series and the forecast intervention series.²⁶

The analysis of the order of integration of the variables involved at this stage is presented in Table 13.2.3 in the annex.

Estimation Results

The estimations for the first stage, displayed in Table 13.2, suggest that interventions respond de facto to several motives, and that the willingness to intervene depends not only on the actual sign of the intervention—whether net purchases or net sales—but also on the type of it.²⁷ In effect, short-term movements of the exchange rate, nominal appreciation velocity, and precautionary concerns are the most common a priori reasons for intervention. The communication variable does have a role in the level of interventions, particularly when there is currency appreciation, and the objective is to raise the exchange rate; when sterilized interventions are considered, the communication variable is always significant both when the final objective is either to raise or decrease the exchange rate. In line with Adler and Tovar’s previous findings, it seems that real exchange rate misalignments were a reason for intervening only in the case of central bank sales. Intra-week volatility appears to be a driving force in the reaction function for indirect intervention through net purchases in the foreign exchange market.

^{Table 13.2.}

Reaction Function Estimates of Types of Intervention, January 2005 to December 2017

Source: Authors’ calculations. Note: Estimations were made using weekly normalized data. Standard errors that are robust to autocorrelation and heteroskedasticity appear in parentheses. *p < .1; ** p < .05; *** p < .01.

^{Table 13.2.}

Reaction Function Estimates of Types of Intervention, January 2005 to December 2017

	BCU	Direct	Indirect	Sterilized	Nonsterilized
Purchases
Constant	-0.0358 (0.5769)	-0.0106 (0.8639)	-0.0240 (0.6981)	-0.0202 (0.7810)	0.0028 (0.9957)
D(e_t-1)	-1.6432*** (0.0048)	-0.2124*** (0.0000)	—	-0.0857* (0.0606)	—
${r e e r}_{t} - {r e e r}_{t}^{e q}$	—	—	—	—	—
Δ_t	0.1899*** (0.0016)	0.2425*** (0.0000)	0.1520*** (0.0003)	—	0.1553*** (0.0058)
σ_t	—		-0.0995** (0.0281)	—	—
$R_{t}^{M 2}$	-0.1723** (0.0144)	-0.1821*** (0.0079)	—	—	0.1197** (0.0403)
C_t	4.1587*** (0.0000)	—	3.1419** (0.0156)	2.9591*** (0.0929)	—
R²	0.1295	0.1352	0.0562	0.0268	0.0409
Adjusted R²	0.1243	0.1313	0.0520	0.0239	0.0380
Durbin Watson	1.1640	1.1883	1.3840	0.4245	0.7986
No. of observations	674	674	674	674	674
Sales
Constant	-0.0081 (0.9057)	-0.0228 (0.5415)	-0.0073 (0.9089)	-0.0245 (0.5052)	-0.0063 (0.9110)
D(e_t-1)	-0.0754** (0.0200)	-0.1604*** (0.0000)	—	-0.1414*** (0.0001)	—
${r e e r}_{t} - {r e e r}_{t}^{e q}$	—	0.1077** (0.0218)	—	—	—
Δ_t	0.2514*** (0.0054)	0.2534*** (0.0000)	0.2320*** (0.0043)	—	0.1194** (0.0299)
σ_t	—	—	—	—	—
$R_{t}^{M 2}$	—	-0.0857** (0.0243)	—	-0.2483*** (0.0000)	—
C_t	—	1.8975** (0.0200)	—	2.6746*** (0.0009)	—
R²	0.0666	0.0635	0.0525	0.1138	0.0139
Adjusted R²	0.06379	0.0601	0.0511	0.1086	0.0125
Durbin Watson	1.2582	1.1011	1.4300	0.4419	1.3338
No. of observations	674	674	674	674	674

^{Table 13.2.}

Reaction Function Estimates of Types of Intervention, January 2005 to December 2017

	BCU	Direct	Indirect	Sterilized	Nonsterilized
Purchases
Constant	-0.0358 (0.5769)	-0.0106 (0.8639)	-0.0240 (0.6981)	-0.0202 (0.7810)	0.0028 (0.9957)
D(e_t-1)	-1.6432*** (0.0048)	-0.2124*** (0.0000)	—	-0.0857* (0.0606)	—
${r e e r}_{t} - {r e e r}_{t}^{e q}$	—	—	—	—	—
Δ_t	0.1899*** (0.0016)	0.2425*** (0.0000)	0.1520*** (0.0003)	—	0.1553*** (0.0058)
σ_t	—		-0.0995** (0.0281)	—	—
$R_{t}^{M 2}$	-0.1723** (0.0144)	-0.1821*** (0.0079)	—	—	0.1197** (0.0403)
C_t	4.1587*** (0.0000)	—	3.1419** (0.0156)	2.9591*** (0.0929)	—
R²	0.1295	0.1352	0.0562	0.0268	0.0409
Adjusted R²	0.1243	0.1313	0.0520	0.0239	0.0380
Durbin Watson	1.1640	1.1883	1.3840	0.4245	0.7986
No. of observations	674	674	674	674	674
Sales
Constant	-0.0081 (0.9057)	-0.0228 (0.5415)	-0.0073 (0.9089)	-0.0245 (0.5052)	-0.0063 (0.9110)
D(e_t-1)	-0.0754** (0.0200)	-0.1604*** (0.0000)	—	-0.1414*** (0.0001)	—
${r e e r}_{t} - {r e e r}_{t}^{e q}$	—	0.1077** (0.0218)	—	—	—
Δ_t	0.2514*** (0.0054)	0.2534*** (0.0000)	0.2320*** (0.0043)	—	0.1194** (0.0299)
σ_t	—	—	—	—	—
$R_{t}^{M 2}$	—	-0.0857** (0.0243)	—	-0.2483*** (0.0000)	—
C_t	—	1.8975** (0.0200)	—	2.6746*** (0.0009)	—
R²	0.0666	0.0635	0.0525	0.1138	0.0139
Adjusted R²	0.06379	0.0601	0.0511	0.1086	0.0125
Durbin Watson	1.2582	1.1011	1.4300	0.4419	1.3338
No. of observations	674	674	674	674	674

One way to analyze the effectiveness of interventions in changing the level of the Uruguayan peso/US dollar rate is to calculate impulse-response functions. To recover the structural intervention shock, this study applies Choleski factorization, which implies a specific ordering of the variables, ranging from the most to the least exogenous. The trade-weighted US dollar index is determined once international commodity prices are set; in addition, international prices of energy and beef Granger dictate the US dollar index in the United States. The nominal bilateral exchange rate Uruguayan peso/US dollar is contemporaneously affected by the trade-weighted US dollar index but not the other way around; Uruguay is a small open economy and a price-taker from global markets. The interest rate spread is contemporaneously affected by both international conditions and the value of the exchange rate, while Uruguayan country risk is determined after them. Interventions respond to all the previous variables and the direction of the relation is known, for central banks tend to react to dampen movements in the exchange rate rather than the opposite.²⁸ As a result, the ordering of the variables is:

\begin{matrix} {[P_{t}^{E}, P_{t}^{S}, P_{t}^{B}, D U S_{t}, e_{t}, (i_{t} - i_{t}^{*}), S_{t}, {\hat{I}}_{t}]} & (13.6) \end{matrix}

The estimations for the second stage reveal either a short-lived and small effect of interventions (purchases) on the nominal exchange rate or no effect at all, in line with the results previously obtained with monthly data. For example, when only sterilized foreign exchange purchases are considered, additional intervention of 1 percent of GDP will increase the nominal exchange rate by 0.5 percent maximum by the first week following the change.²⁹^,³⁰ Then the effect vanishes.

Similar results are obtained when the central bank is the agent in charge of the intervention. In effect, an increase of 1 percent of GDP in central bank nonsterilized foreign exchange purchases will increase the nominal exchange rate by 1.4 percent the following week. After that, the effect disappears.

Indirect intervention, through export prefinancing and local currency securities in US dollars, or exchange forwards settlements, among other operations, has a smaller (0.7 percentage) although longer impact (12 weeks) on the nominal Urguayan peso/US dollar rate, while direct purchases in the spot market and sterilized sales do not have a statistically significant effect (see Table 13.3 and Annex 13.1 for the impulse-response functions (IRFs)).

^{Table 13.3.}

Types of Main Effects of Interventions, 2005–17

Source: Authors’ calculations. Note: NSSc = nonstatistically significant results, with correct sign.

^¹“Amount” shows the percentage points of nominal Uruguayan pesos/US dollar change.

^²This shows the change of nominal Uruguayan pesos/1 US dollars in percentage points of GDP.

^{Table 13.3.}

Types of Main Effects of Interventions, 2005–17

			First Effect after Impulse		Max Effect after Impulse		Last Effect after Impulse
Increase in		Impulse (% GDP)	One Week¹	Amount²	One Week¹	Amount²	One Week¹	Amount²
Purchases		1	1	1.4	1	1.5	1	1.5
Sales		1	1	-32.5	4	-42.2	4	-42.2
BCU
	Purchases	1	1	1.4	1	1.4	1	1.4
	Sales	1	1	-16.8	1	-16.6		-3.2
Direct
	Purchases	1	—	NSSc	—	NSSc	—	NSSc
	Sales	1	1	-14.9	1	-14.7	12	0.1
Indirect
	Purchases	1	1	0.7	2	0.8	12	0.2
	Sales	1	2	-31.5	3	-38.9	21	-12.9
Sterilized
	Purchases	1	1	0.5	—	—	—	—
	Sales	1	—	NSSc		NSSc	—	NSSc
Nonsterilized
	Purchases	1	6	0.7	19	0.9	111	0.5
	Sales	1	1	-30.4	3	-37.2	20	-14.5

Source: Authors’ calculations. Note: NSSc = nonstatistically significant results, with correct sign.

^¹“Amount” shows the percentage points of nominal Uruguayan pesos/US dollar change.

^²This shows the change of nominal Uruguayan pesos/1 US dollars in percentage points of GDP.

^{Table 13.3.}

Types of Main Effects of Interventions, 2005–17

			First Effect after Impulse		Max Effect after Impulse		Last Effect after Impulse
Increase in		Impulse (% GDP)	One Week¹	Amount²	One Week¹	Amount²	One Week¹	Amount²
Purchases		1	1	1.4	1	1.5	1	1.5
Sales		1	1	-32.5	4	-42.2	4	-42.2
BCU
	Purchases	1	1	1.4	1	1.4	1	1.4
	Sales	1	1	-16.8	1	-16.6		-3.2
Direct
	Purchases	1	—	NSSc	—	NSSc	—	NSSc
	Sales	1	1	-14.9	1	-14.7	12	0.1
Indirect
	Purchases	1	1	0.7	2	0.8	12	0.2
	Sales	1	2	-31.5	3	-38.9	21	-12.9
Sterilized
	Purchases	1	1	0.5	—	—	—	—
	Sales	1	—	NSSc		NSSc	—	NSSc
Nonsterilized
	Purchases	1	6	0.7	19	0.9	111	0.5
	Sales	1	1	-30.4	3	-37.2	20	-14.5

Source: Authors’ calculations. Note: NSSc = nonstatistically significant results, with correct sign.

^¹“Amount” shows the percentage points of nominal Uruguayan pesos/US dollar change.

^²This shows the change of nominal Uruguayan pesos/1 US dollars in percentage points of GDP.

The results suggest a nonlinear relation between interventions and exchange rate performance. In effect, a purchase in the foreign exchange market of 1 percent of GDP would increase the exchange rate by 1.4 percent the following week, while a sale in the foreign exchange market of 1 percent of GDP would decrease the exchange rate by 32.5 percent.³¹ In other words, the effort needed in terms of GDP to get nominal foreign exchange depreciation through foreign exchange intervention is several times that required to appreciate it. Nevertheless, what really matters are the effects after open market operations are done. Sterilized interventions show that a purchase of 1 percent of GDP would increase the nominal exchange rate by 0.5 percent only in the week following the intervention, while there could be a small effect on the exchange rate because of foreign exchange sales (it is not statistically significant). Almost all of the impact of an intervention occurs during the week it is conducted, which confirms the idea that central banks typically lean against the wind.

In a nutshell, it can be said that central bank foreign exchange interventions have a short-lived impact, whereas indirect interventions have smaller but long-lasting effects; for a very short time that can only be seen on a weekly basis, sterilized foreign exchange intervention barely affects the exchange rate, while the remaining nonsterilized foreign exchange interventions are the ones responsible for the higher, although still small effects on the exchange rate that last longer. Obviously, if no sterilization were done, domestic currency depreciation/appreciation induced by foreign currency purchases/sales would be greater.³²

Conclusion

In this chapter’s study of the effects of foreign exchange interventions in Uruguay during 2005–17, the exchange rate sampled arguably reflects the highest flexibility in the country’s economic history. Yet intervention has been the rule rather than the exception, largely because of undue influence in an open economy like Uruguay’s of unprecedented volatility in the international environment.

Authorities have been very vocal about the need to dampen the impact of this volatility in Uruguay and have been trying to avoid both price or financial instability. In doing so, they have used several avenues to intervene in the foreign exchange market. The main interest in this chapter therefore was to identify how the exchange rate responds to different types of intervention, in addition to the effectiveness itself.

Taken together, the results of this study seem to suggest that this expressed concern of the authorities with exchange rate volatility can be rationalized as a concern over the impact of exchange rate short-term movements, nominal exchange rate appreciation/depreciation velocity, and real exchange misalignment adjustments. Further testing of this hypothesis will be a topic of the research agenda.

The results also suggest that the use of foreign exchange intervention, together with other monetary and financial tools, helped dampen the adverse effect of large swings in capital flows and related domestic portfolio changes in terms of the economic fundamentals, and excessive volatility in relative prices, currency markets, and interest rates.³³

In reaching its conclusion, the analysis applied different methodologies to different datasets, both monthly and weekly.³⁴ This allowed more detailed analysis of foreign exchange movements induced by the interventions and more precisely determined the duration of the effect, if any. The analysis also distinguished between purchases and sales of foreign exchange (on a weekly basis) to search for different reasons to intervene in each situation.

In addition, the endogeneity of exchange rates and interventions has been critical in the investigation. As Kearns and Rigobon (2005) point out: “…failing to account for the endogeneity, when central banks lean against the wind and trade strategically, will likely result in a large downward bias to the coefficient on contemporaneous intervention explaining the negative coefficient frequently obtained.” Here, the analysis adopted two strategies to deal with this problem, depending on the dataset frequency. For monthly data, the order of the variables in the VAR was determined following a specified theoretical framework. For weekly data, the analysis followed the Adler and Tovar (2014) two-stage estimation process. In the first stage, a de facto reaction function for the central bank was estimated; in the second stage, predicted values of this reaction function were used as instruments in the estimation of a behavioral equation for the exchange rate.³⁵

More specifically, the results suggest several conclusions, as outlined in the introduction. It is worthwhile to reiterate the results:

1. Interventions affect the level of the exchange rate, but the effect is short-lived.
2. While the effect of indirect interventions appears with the expected sign, and is statistically significant, direct interventions get a meaningful response through foreign exchange sales, while purchases “just” prevented the peso from further appreciation.
3. Sterilized intervention does not seem to have an effect on the level of foreign exchange longer than one week after foreign exchange purchases, explaining why no effect is found when monthly data is used.
4. Interventions have asymmetric effects on the foreign exchange; that is, purchases of foreign exchange (that tend to increase the exchange rate) are more costly in terms of GDP than sales of foreign exchange (that tend to decrease the foreign exchange).
5. Communication to the public of relevant information regarding the value of the exchange rate seems to play a role in the motives of de facto intervention.
6. The central bank seems to worry about the level of the foreign exchange and about the appreciation velocity of the foreign exchange, rather than its volatility.
7. Real exchange rate misalignments seem to have been a reason for intervening only in the case of central bank sales.
8. And last but not least, as the flip side of sterilized interventions is the increase in the stock of monetary regulation securities, these excess reserves are a mac-roprudential buffer that has associated costs because of interest rate differentials between Uruguay and the United States.

ANNEX 13.1. MONTHLY DATA

Table 13.1.1.

Descriptive Statistics

Source: Authors’ calculations.

Table 13.1.1.

Descriptive Statistics

	E^Uy_ty	NFXP_t	NFXP_STD_t	NFXP_DIR_t	NFXP_IND_t	NFXP_ STER_t	NFXP NON_STER_t
Mean	23.272	82.553	-0.001	18.368	17.961	50.816	31.922
Median	22.257	112.160	0.127	0.000	0.000	90.850	13.855
Maximum	32.133	595.430	2.217	384.520	534.560	568.330	669.080
Minimum	18.428	-969.460	-4.550	-378.800	-903.530	-882.870	-373.100
Standard deviation	3.795	235.119	1.017	83.802	174.228	244.035	169.456
Skewness	0.695	-1.696	-1.696	0.597	-0.920	-0.982	0.392
Kurtosis	2.228	8.846	8.846	10.173	12.679	4.687	4.112
Jarque-Bera	13.811	249.335	249.335	288.651	591.838	36.870	10.178
Probability	0.001	0.000	0.000	0.000	0.000	0.000	0.006
No. of observations	131	131	131	131	131	132	132

Source: Authors’ calculations.

Table 13.1.1.

Descriptive Statistics

	E^Uy_ty	NFXP_t	NFXP_STD_t	NFXP_DIR_t	NFXP_IND_t	NFXP_ STER_t	NFXP NON_STER_t
Mean	23.272	82.553	-0.001	18.368	17.961	50.816	31.922
Median	22.257	112.160	0.127	0.000	0.000	90.850	13.855
Maximum	32.133	595.430	2.217	384.520	534.560	568.330	669.080
Minimum	18.428	-969.460	-4.550	-378.800	-903.530	-882.870	-373.100
Standard deviation	3.795	235.119	1.017	83.802	174.228	244.035	169.456
Skewness	0.695	-1.696	-1.696	0.597	-0.920	-0.982	0.392
Kurtosis	2.228	8.846	8.846	10.173	12.679	4.687	4.112
Jarque-Bera	13.811	249.335	249.335	288.651	591.838	36.870	10.178
Probability	0.001	0.000	0.000	0.000	0.000	0.000	0.006
No. of observations	131	131	131	131	131	132	132

Source: Authors’ calculations.

Table 13.1.2.

Unit Root Analysis

Source: Authors’ calculations. Note: For E^Uy_t the 1 percent, 5 percent, and 10 percent critical values are -3.501, -2.888, and -2.578, respectively. *p < .10; **p < .05; ***p < .01.

Table 13.1.2.

Unit Root Analysis

	Augmented Dickey-Fuller Test for Unit Root
Variable	MacKinnon p-value	Test Statistic
E^Uy_t	0.765	-0.967
E^USA_t	0.631	-1.296
E^Uy_t	0.866	-0.621
NFXP_t	0.017	-3.261**
π^e_t	0.011	-3.390**
i^Uy_t	0.043	-2.922**
i^USA_t	0.003	-3.839**
d^e_t	0.019	-3.218**
VIX_t	0.098	-2.576*
π^EME_t	0.012	-3.376**
π^Uy_t	0.064	-2.781*
No. of observations	130

Source: Authors’ calculations. Note: For E^Uy_t the 1 percent, 5 percent, and 10 percent critical values are -3.501, -2.888, and -2.578, respectively. *p < .10; **p < .05; ***p < .01.

Table 13.1.2.

Unit Root Analysis

	Augmented Dickey-Fuller Test for Unit Root
Variable	MacKinnon p-value	Test Statistic
E^Uy_t	0.765	-0.967
E^USA_t	0.631	-1.296
E^Uy_t	0.866	-0.621
NFXP_t	0.017	-3.261**
π^e_t	0.011	-3.390**
i^Uy_t	0.043	-2.922**
i^USA_t	0.003	-3.839**
d^e_t	0.019	-3.218**
VIX_t	0.098	-2.576*
π^EME_t	0.012	-3.376**
π^Uy_t	0.064	-2.781*
No. of observations	130

Source: Authors’ calculations. Note: For E^Uy_t the 1 percent, 5 percent, and 10 percent critical values are -3.501, -2.888, and -2.578, respectively. *p < .10; **p < .05; ***p < .01.

Table 13.1.3.

Cointegration Analysis

Source: Authors’ calculations. Note: The trace test indicates one cointegrating equation at the 5 percent level, and the maximum eigenvalue test indicates one cointegrating equation at the 5 percent level.

^¹“None” denotes rejection of the hypothesis at the 5 percent level.

^²See MacKinnon-Haug-Michelis (1999) for the p-values.

Table 13.1.3.

Cointegration Analysis

Unrestricted Cointegration Rank Test (Trace)
Hypothesized Cointegration Equations	Eigenvalue	Trace Statistic	5% Critical Value	Probability²
None¹	0.235170	56.63615	42.91525	0.0013
At most one	0.124209	22.58728	25.87211	0.1215
At most two	0.044217	5.743543	12.51798	0.4935
Unrestricted Cointegration Rank Test (Maximum Eigenvalue)
Hypothesized Cointegration Equations	Eigenvalue	Maximum Eigenvalue Statistic	5% Critical Value	Probability²
None¹	0.235170	34.04888	25.82321	0.0033
At most one	0.124209	16.84373	19.38704	0.1127
At most two	0.044217	5.743543	12.51798	0.4935

^¹“None” denotes rejection of the hypothesis at the 5 percent level.

^²See MacKinnon-Haug-Michelis (1999) for the p-values.

Table 13.1.3.

Cointegration Analysis

Unrestricted Cointegration Rank Test (Trace)
Hypothesized Cointegration Equations	Eigenvalue	Trace Statistic	5% Critical Value	Probability²
None¹	0.235170	56.63615	42.91525	0.0013
At most one	0.124209	22.58728	25.87211	0.1215
At most two	0.044217	5.743543	12.51798	0.4935
Unrestricted Cointegration Rank Test (Maximum Eigenvalue)
Hypothesized Cointegration Equations	Eigenvalue	Maximum Eigenvalue Statistic	5% Critical Value	Probability²
None¹	0.235170	34.04888	25.82321	0.0033
At most one	0.124209	16.84373	19.38704	0.1127
At most two	0.044217	5.743543	12.51798	0.4935

^¹“None” denotes rejection of the hypothesis at the 5 percent level.

^²See MacKinnon-Haug-Michelis (1999) for the p-values.

ANNEX 13.2. WEEKLY DATA

Table 13.2.1.

Data Set

Source: Authors’ calculations. Note: BCU = Banco Central del Uruguay; EMBI = Emerging Markets Bond Index; FRED = Federal Reserve; República AFAP = Uruguayan state-owned pension fund company.

^¹This is measured as a percentage of GDP.

Table 13.2.1.

Data Set

Mnemonic	Description	Source	Log	Dif
p^s	Soybean international price	IMF	Y	Y
P^E	Energy international price	IMF	Y	Y
DUS	Trade-weighted US dollar index	FRED	Y	Y
e	Uruguayan pesos/US dollar exchange rate	BCU	Y	Y
i – i*	30-day ITLUP (Uruguayan peso-nominated yield curve) over effective FFR	https://web.bevsa.com.uy/ and FRED	N	Y
S	EMBI Uruguay	República AFAP	Y	Y
DES_REER	REER misalignment	Authors’ calculation	Y	N
Δ	Uruguayan pesos/US dollar appreciation/ depreciation velocity	Authors’ calculation	N	N
σ	Nominal exchange rate volatility	Authors’ calculation on BCU data	N	N
C	Communication dummy	Authors’ calculation	N	N
R^M2	Reserves to M2 ratio	Authors’ calculation on BCU data	N	N
BCU_PUR	Intervention (purchases) done by BCU¹	BCU	N	N
BCU_SALES	Intervention (sales) done by BCU¹	BCU	N	N
DIR_PUR	Direct intervention (purchases)¹	BCU	N	N
DIR_SALES	Direct intervention (sales)¹	BCU	N	N
INDIR_PUR	Indirect intervention (purchases)¹	BCU	N	N
INDIR_SALES	Indirect intervention (sales)¹	BCU	N	N
EST_PUR	Sterilized intervention (purchases)¹	BCU	N	N
EST_SALES	Sterilized intervention (sales)¹	BCU	N	N
NON_EST_PUR	Nonsterilized intervention (purchases)¹	BCU	N	N
NON_EST_SALES	Nonsterilized intervention (sales)¹	BCU	N	N

Source: Authors’ calculations. Note: BCU = Banco Central del Uruguay; EMBI = Emerging Markets Bond Index; FRED = Federal Reserve; República AFAP = Uruguayan state-owned pension fund company.

^¹This is measured as a percentage of GDP.

Table 13.2.1.

Data Set

Mnemonic	Description	Source	Log	Dif
p^s	Soybean international price	IMF	Y	Y
P^E	Energy international price	IMF	Y	Y
DUS	Trade-weighted US dollar index	FRED	Y	Y
e	Uruguayan pesos/US dollar exchange rate	BCU	Y	Y
i – i*	30-day ITLUP (Uruguayan peso-nominated yield curve) over effective FFR	https://web.bevsa.com.uy/ and FRED	N	Y
S	EMBI Uruguay	República AFAP	Y	Y
DES_REER	REER misalignment	Authors’ calculation	Y	N
Δ	Uruguayan pesos/US dollar appreciation/ depreciation velocity	Authors’ calculation	N	N
σ	Nominal exchange rate volatility	Authors’ calculation on BCU data	N	N
C	Communication dummy	Authors’ calculation	N	N
R^M2	Reserves to M2 ratio	Authors’ calculation on BCU data	N	N
BCU_PUR	Intervention (purchases) done by BCU¹	BCU	N	N
BCU_SALES	Intervention (sales) done by BCU¹	BCU	N	N
DIR_PUR	Direct intervention (purchases)¹	BCU	N	N
DIR_SALES	Direct intervention (sales)¹	BCU	N	N
INDIR_PUR	Indirect intervention (purchases)¹	BCU	N	N
INDIR_SALES	Indirect intervention (sales)¹	BCU	N	N
EST_PUR	Sterilized intervention (purchases)¹	BCU	N	N
EST_SALES	Sterilized intervention (sales)¹	BCU	N	N
NON_EST_PUR	Nonsterilized intervention (purchases)¹	BCU	N	N
NON_EST_SALES	Nonsterilized intervention (sales)¹	BCU	N	N

Source: Authors’ calculations. Note: BCU = Banco Central del Uruguay; EMBI = Emerging Markets Bond Index; FRED = Federal Reserve; República AFAP = Uruguayan state-owned pension fund company.

^¹This is measured as a percentage of GDP.

Table 13.2.2

Descriptive Statistics

Source: Authors’ calculations. Note: For each variable, there were 674 observations.

Table 13.2.2

Descriptive Statistics

	P^S	P^E	DUS	e	i – i*	S
Mean	390.2652	141.1646	107.1886	23.41697	1.808031	252.1682
Median	375.2612	132.4080	104.3829	23.39800	2.842216	221.2000
Maximum	622.9135	249.6072	128.4466	32.31200	3.285013	865.2000
Minimum	197.5892	60.64390	94.07466	18.38200	2.130693	112.2000
Standard deviation	103.8388	44.79428	8.676582	3.488442	1.822798	112.2770
Skewness	-0.006670	0.246069	0.676758	0.633226	-1.233232	2.485051
Kurtosis	2.218970	1.776925	2.363292	2.485153	2.855085	10.54785
Jarque-Bera	17.13604	48.81201	62.83378	52.48684	171.4332	2293.623
Probability	0.000190	0.000000	0.000000	0.000000	0.000000	0.000000
Sum	263038.7	95144.93	72245.09	15783.04	1218.613	169961.4
Sum of squares deviation	7256621.	1350393.	50665.51	8189.892	2236.105	8483916.
	DES_REER	DES_C	R^M2
Mean	8.32E-05	4.38E-06	0.006345	0.007206	-0.024768
Median	-0.001664	4.57E-05	0.002456	0.000000	-0.041530
Maximum	0.098689	0.000364	0.055120	0.285714	1.925887
Minimum	-0.095291	-0.000373	5.70E-06	-0.285714	-1.887973
Standard deviation	0.035362	0.000159	0.009316	0.046166	1.008014
	BCU_PUR	BCU_SALES	DIR_PUR	DIR_SALES
Mean	0.044255	-0.017752	0.045964	-0.007136
Median	0.009560	0.000000	0.024794	0.000000
Maximum	0.535780	0.000000	0.491070	0.000000
Minimum	0.000000	-1.640539	0.000000	-0.334517
Standard deviation	0.072237	0.084930	0.063377	0.032109
Skewness	2.899842	-12.19151	2.485731	-6.030426
Kurtosis	15.15568	208.5092	13.25778	45.57295
Jarque-Bera	5094.228	1202768.	3649.076	54984.92
Probability	0.000000	0.000000	0.000000	0.000000
Sum	29.82779	-11.96464	30.97964	-4.809911
Sum of squares deviation	3.511824	4.854446	2.703165	0.693839
	INDIR_PUR	INDIR_ SALES	EST_PUR	EST_SALES	NON_EST_ PUR	NON_EST_ SALES
Mean	0.028257	–0.020006	0.079723	-0.052296	0.070790	-0.052012
Median	0.000000	0.000000	0.041668	0.000000	0.016631	0.000000
Maximum	0.958648	0.000000	0.449949	0.000000	0.690899	0.000000
Minimum	0.000000	-1.640539	0.000000	-0.641600	0.000000	-1.220810
Standard deviation	0.067219	0.094017	0.093340	0.118820	0.118310	0.103809
Skewness	6.252966	-11.50790	1.327974	-2.608135	2.394146	-4.594779
Kurtosis	67.03102	175.0000	4.541375	9.817313	9.084095	39.92522
Jarque-Bera	119533.1	845693.6	264.8230	2069.327	1683.426	40662.41
Probability	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000
Sum	19.04547	-13.48371	53.73317	-35.24724	47.71254	-35.05593
Sum of squares deviation	3.040911	5.948737	5.863377	9.501601	9.420226	7.252500

Source: Authors’ calculations. Note: For each variable, there were 674 observations.

Table 13.2.2

Descriptive Statistics

	P^S	P^E	DUS	e	i – i*	S
Mean	390.2652	141.1646	107.1886	23.41697	1.808031	252.1682
Median	375.2612	132.4080	104.3829	23.39800	2.842216	221.2000
Maximum	622.9135	249.6072	128.4466	32.31200	3.285013	865.2000
Minimum	197.5892	60.64390	94.07466	18.38200	2.130693	112.2000
Standard deviation	103.8388	44.79428	8.676582	3.488442	1.822798	112.2770
Skewness	-0.006670	0.246069	0.676758	0.633226	-1.233232	2.485051
Kurtosis	2.218970	1.776925	2.363292	2.485153	2.855085	10.54785
Jarque-Bera	17.13604	48.81201	62.83378	52.48684	171.4332	2293.623
Probability	0.000190	0.000000	0.000000	0.000000	0.000000	0.000000
Sum	263038.7	95144.93	72245.09	15783.04	1218.613	169961.4
Sum of squares deviation	7256621.	1350393.	50665.51	8189.892	2236.105	8483916.
	DES_REER	DES_C	R^M2
Mean	8.32E-05	4.38E-06	0.006345	0.007206	-0.024768
Median	-0.001664	4.57E-05	0.002456	0.000000	-0.041530
Maximum	0.098689	0.000364	0.055120	0.285714	1.925887
Minimum	-0.095291	-0.000373	5.70E-06	-0.285714	-1.887973
Standard deviation	0.035362	0.000159	0.009316	0.046166	1.008014
	BCU_PUR	BCU_SALES	DIR_PUR	DIR_SALES
Mean	0.044255	-0.017752	0.045964	-0.007136
Median	0.009560	0.000000	0.024794	0.000000
Maximum	0.535780	0.000000	0.491070	0.000000
Minimum	0.000000	-1.640539	0.000000	-0.334517
Standard deviation	0.072237	0.084930	0.063377	0.032109
Skewness	2.899842	-12.19151	2.485731	-6.030426
Kurtosis	15.15568	208.5092	13.25778	45.57295
Jarque-Bera	5094.228	1202768.	3649.076	54984.92
Probability	0.000000	0.000000	0.000000	0.000000
Sum	29.82779	-11.96464	30.97964	-4.809911
Sum of squares deviation	3.511824	4.854446	2.703165	0.693839
	INDIR_PUR	INDIR_ SALES	EST_PUR	EST_SALES	NON_EST_ PUR	NON_EST_ SALES
Mean	0.028257	–0.020006	0.079723	-0.052296	0.070790	-0.052012
Median	0.000000	0.000000	0.041668	0.000000	0.016631	0.000000
Maximum	0.958648	0.000000	0.449949	0.000000	0.690899	0.000000
Minimum	0.000000	-1.640539	0.000000	-0.641600	0.000000	-1.220810
Standard deviation	0.067219	0.094017	0.093340	0.118820	0.118310	0.103809
Skewness	6.252966	-11.50790	1.327974	-2.608135	2.394146	-4.594779
Kurtosis	67.03102	175.0000	4.541375	9.817313	9.084095	39.92522
Jarque-Bera	119533.1	845693.6	264.8230	2069.327	1683.426	40662.41
Probability	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000
Sum	19.04547	-13.48371	53.73317	-35.24724	47.71254	-35.05593
Sum of squares deviation	3.040911	5.948737	5.863377	9.501601	9.420226	7.252500

Source: Authors’ calculations. Note: For each variable, there were 674 observations.

Table 13.2.3

Unit Root Analysis

Source: Authors’ calculations. ** p < .05; *** p < .01.

^¹The variable seems to be I(0) with at least one break, but it was treated as I(1) for operational reasons.

^²The Elliott, Rothenberg, and Stock DF-GLS test was applied.

^³The Phillips-Perron unit root test was applied.

Table 13.2.3

Unit Root Analysis

	ADF Test Statistic		MacKinnon (1996) One-Sided Critical Values
Variable	t-Statistic	p-value	1% Level	5% Level	10% Level	Order of Integration
P^S	-2.0476	0.2665	-3.4403	-2.8658	-2.5691	1
P^E	-2.2133	0.2018				1
DUS	-1.2035	0.6747	-3.4397	-2.8656	-2.5690	1
e	-0.7404	0.8342				1
i – i*	-1.4121	0.5773				1¹
S	-2.8551	0.0513				1¹
D(P^S)	-7.2636***	0.0000	-3.4403	-2.8658	-2.5691	0
D(P^E)	-6.1387***	0.0000				0
D(DUS)	-19.6386***	0.0000	-3.4397	-2.8656	-2.5690	0
De	-19.5813***	0.0000				0
D(i – i*)	-4.7781***	0.0001	-3.4398	-2.8656	-2.5690	0
D(S)	-10.2883***	0.0000				0
I_PUR²	2.2320**	0.0259	-2.5684	-1.9413	-1.6164	0
I_SALES²	-2.2320**	0.0259				0
I_BCU_PUR	-4.2558***	0.0006	-3.4399	-2.8556	-2.5690	0
I_BCU_SALES²	-2.9789**	0.0374				0
I_DIR_PUR	-3.1345**	0.0246				0
I_DIR_SALES	-3.5069***	0.0081				0
I_IND_PUR	-4.5221***	0.0002				0
I_IND_SALES³	-4.0392***	0.0013				0
I_EST_PUR	-15.1229**	0.0000				0
I_EST_SALES	-3.8515**	0.0026				0
I_NON EST_PUR²	-2.3126**	0.0210	-2.5684	-1.9413	-1.6164	0
I_NON EST_SALES	-3.1914**	0.0209	-3.4400	-2.8657	-2.5690	0

Source: Authors’ calculations. ** p < .05; *** p < .01.

^¹The variable seems to be I(0) with at least one break, but it was treated as I(1) for operational reasons.

^²The Elliott, Rothenberg, and Stock DF-GLS test was applied.

^³The Phillips-Perron unit root test was applied.

Table 13.2.3

Unit Root Analysis

	ADF Test Statistic		MacKinnon (1996) One-Sided Critical Values
Variable	t-Statistic	p-value	1% Level	5% Level	10% Level	Order of Integration
P^S	-2.0476	0.2665	-3.4403	-2.8658	-2.5691	1
P^E	-2.2133	0.2018				1
DUS	-1.2035	0.6747	-3.4397	-2.8656	-2.5690	1
e	-0.7404	0.8342				1
i – i*	-1.4121	0.5773				1¹
S	-2.8551	0.0513				1¹
D(P^S)	-7.2636***	0.0000	-3.4403	-2.8658	-2.5691	0
D(P^E)	-6.1387***	0.0000				0
D(DUS)	-19.6386***	0.0000	-3.4397	-2.8656	-2.5690	0
De	-19.5813***	0.0000				0
D(i – i*)	-4.7781***	0.0001	-3.4398	-2.8656	-2.5690	0
D(S)	-10.2883***	0.0000				0
I_PUR²	2.2320**	0.0259	-2.5684	-1.9413	-1.6164	0
I_SALES²	-2.2320**	0.0259				0
I_BCU_PUR	-4.2558***	0.0006	-3.4399	-2.8556	-2.5690	0
I_BCU_SALES²	-2.9789**	0.0374				0
I_DIR_PUR	-3.1345**	0.0246				0
I_DIR_SALES	-3.5069***	0.0081				0
I_IND_PUR	-4.5221***	0.0002				0
I_IND_SALES³	-4.0392***	0.0013				0
I_EST_PUR	-15.1229**	0.0000				0
I_EST_SALES	-3.8515**	0.0026				0
I_NON EST_PUR²	-2.3126**	0.0210	-2.5684	-1.9413	-1.6164	0
I_NON EST_SALES	-3.1914**	0.0209	-3.4400	-2.8657	-2.5690	0

Source: Authors’ calculations. ** p < .05; *** p < .01.

^¹The variable seems to be I(0) with at least one break, but it was treated as I(1) for operational reasons.

^²The Elliott, Rothenberg, and Stock DF-GLS test was applied.

^³The Phillips-Perron unit root test was applied.

Table 13.2.4

Real Exchange Rate Model

Source: Authors’ calculations. Note: The dependent variable is L_REER; the method is a least squares sample (adjusted): January 5, 2005, to October 4, 2017; and the number of included observations are 666 after adjustments.

Table 13.2.4

Real Exchange Rate Model

Variable	Coefficient	Standard Error	t-Statistic	Probability
Constant	5.124394	0.048779	105.0533	0.0000
Net interest rate	-0.015596	0.001352	-11.53773	0.0000
Government consumption-	-0.526984	0.447669	-1.177174	0.2395
over-GDP ratio
Linear trend	-0.000978	8.19E-06	-119.4008	0.0000
R²	0.968463	Mean dependent variable		4.579273
Adjusted R²	0.968320	Standard deviation dependent variable		0.196787
Standard error of regression	0.035026	Akaike information criterion		-3.859467
Sum of squared residuals	0.812153	Schwarz criterion		-3.832432
Log likelihood	1289.203	Hannan-Quinn criterion		-3.848993
F-statistic	6776.326	Durbin-Watson statistic		0.015912
Probability (F-statistic)	0.000000

Table 13.2.4

Real Exchange Rate Model

Variable	Coefficient	Standard Error	t-Statistic	Probability
Constant	5.124394	0.048779	105.0533	0.0000
Net interest rate	-0.015596	0.001352	-11.53773	0.0000
Government consumption-	-0.526984	0.447669	-1.177174	0.2395
over-GDP ratio
Linear trend	-0.000978	8.19E-06	-119.4008	0.0000
R²	0.968463	Mean dependent variable		4.579273
Adjusted R²	0.968320	Standard deviation dependent variable		0.196787
Standard error of regression	0.035026	Akaike information criterion		-3.859467
Sum of squared residuals	0.812153	Schwarz criterion		-3.832432
Log likelihood	1289.203	Hannan-Quinn criterion		-3.848993
F-statistic	6776.326	Durbin-Watson statistic		0.015912
Probability (F-statistic)	0.000000

Annex 13.3. Interventions in Uruguay

Table 13.3.1.

Interventions in Uruguay

Table 13.3.1.

Interventions in Uruguay

Institution	Type of Intervention	Instrument	Comment about the Instrument
Central Bank	Direct/indirect Sterilized/nonsterilized	Purchases/sales in the spot market	—
		Monetary Regulation Bills (Letras de Regulación Monetaria)	Most common instrument to sterilize purchases
		Certificates of deposit	Mainly used during interest rate target period
		Integration in US dollars Derivatives (SWAPS or financial contracts, futures)	From 2008 to 2017 intermittently Notably with public energy enterprises
		Bank reserve requirements	Increased and reduced (in 2016 and 2017, for example)
		Capital controls	Started in August 2012 and lasted until May 2015 (there were modifications in the meantime)
Ministry of Economy	Direct/indirect Nonsterilized	T-bills
		Bonds
		Purchases/sales to central bank
		Exports prefinancing	Handled by the central bank
Bank of the Republic (state-owned)	Direct	Purchases/sales in the spot market

Table 13.3.1.

Interventions in Uruguay

Institution	Type of Intervention	Instrument	Comment about the Instrument
Central Bank	Direct/indirect Sterilized/nonsterilized	Purchases/sales in the spot market	—
		Monetary Regulation Bills (Letras de Regulación Monetaria)	Most common instrument to sterilize purchases
		Certificates of deposit	Mainly used during interest rate target period
		Integration in US dollars Derivatives (SWAPS or financial contracts, futures)	From 2008 to 2017 intermittently Notably with public energy enterprises
		Bank reserve requirements	Increased and reduced (in 2016 and 2017, for example)
		Capital controls	Started in August 2012 and lasted until May 2015 (there were modifications in the meantime)
Ministry of Economy	Direct/indirect Nonsterilized	T-bills
		Bonds
		Purchases/sales to central bank
		Exports prefinancing	Handled by the central bank
Bank of the Republic (state-owned)	Direct	Purchases/sales in the spot market

References

Adler, Gustavo, and Camilo Ernesto Tovar. 2014. “Foreign Exchange Interventions and Their Impact on Exchange Rate Levels.” Monetaria 1: 1–48.
- Search Google Scholar
- Export Citation
Barón, Andrea, Gerardo Licandro, Miguel Mello, and Pablo Picardo. 2017. “Moneda de fac-turación de las empresas uruguayas.” Paper presented at the XXXII Jornadas Anuales de Economía, Banco Central del Uruguay Montevideo, Uruguay.
- Search Google Scholar
- Export Citation
Bank for International Settlements. 2018. “Moving Forward with Macroprudential Framework.” In Chapter IV, BIS Annual Economic Report 2018.
- Search Google Scholar
- Export Citation
Blanchard, Olivier, Gustavo Adler, and Irineu de Carvalho Filho. 2015. “Can Foreign Exchange Intervention Stem Exchange Rate Pressures from Global Capital Flow Shocks?” IMF, Working Paper 15/159, International Monetary Fund, Washington, DC.
- Search Google Scholar
- Export Citation
Bucacos, E. 2017. “Financial Conditions and Monetary Policy in Uruguay: An MS-VAR Approach.” IDB Working Paper Series WP-796, Inter-American Development Bank, Washington, DC.
- Search Google Scholar
- Export Citation
Contreras, Gabriela, Alfredo Pistelli, and Camila Sáez. 2013. “Efecto de las intervenciones cambiarias recientes en economias emergentes.” Economía Chilena 16 (1): 122–37.
- Search Google Scholar
- Export Citation
Daude, Christian, Eduardo Levy Yeyati, and Arne J. Nagengast. 2016. “On the Effectiveness of Exchange Rate Interventions in Emerging Markets.” Journal of International Money and Finance 64 (C): 239–61.
- Search Google Scholar
- Export Citation
Druck, Pablo, Nicolás Magud, and Rodrigo Mariscal. 2018. “Collateral Damage: Dollar Strength and Emerging Markets’ Growth.” North American Journal of Economics and Finance 43 (C): 97–117.
- Search Google Scholar
- Export Citation
Ibarra, Ana Maria, Daniel Dominioni, Gerardo Licandro, and Umberto Della Mea. 2011. “Un enfoque de Acceso-en-Riesgo para los Activos de Reserva. Documento de Trabajo Número 2011015, Banco Central del Uruguay.
- Search Google Scholar
- Export Citation
Fisher, Lance A., Syeon-Seung Huh, and Adrian Pagan. 2013. “Econometric Issues When Modelling with a Mixture of I(1) and I(0) Variables.” Working Paper Series 97, NCER Working Paper Series, National Centre for Econometric Research, Brisbane, Australia.
- Search Google Scholar
- Export Citation
Ganón Garayalde, Elena. 2013. “Tipo de cambio, intervenciones y política monetaria en Uruguay en el Corto Plazo.” Documento de Trabajo, Banco Central del Uruguay.
- Search Google Scholar
- Export Citation
Ilzetzki, Ethan, Carmen M. Reinhart, and Kenneth S. Rogoff. 2017a. “Exchange Arrangements Entering the 21st Century: Which Anchor Will Hold?” NBER Working Papers 23134, National Bureau of Economic Research, Cambridge, MA.
- Search Google Scholar
- Export Citation
Ilzetzki, Ethan, Carmen M. Reinhart, and Kenneth S. Rogoff. 2017b. “The Country Chronologies to Exchange Rate Arrangements into the 21st Century: Will the Anchor Currency Hold?” NBER Working Papers 23135, National Bureau of Economic Research, Cambridge, MA.
- Search Google Scholar
- Export Citation
Kearns, Jonathan, and Roberto Rigobon. 2005. “Identifying the Efficacy of Central Bank Interventions: Evidence from Australia and Japan.” Journal of International Economics 66 (1): 31–48.
- Search Google Scholar
- Export Citation
Kolasa, Marcin, and Grzegorz Wesolowski. 2018. “International Spillovers of Quantitative Easing.” European Central Bank, Working Paper Series No. 2172, Frankfurt, Germany.
- Search Google Scholar
- Export Citation
Licandro, Gerardo, and José Antonio Licandro. 2003. “Building the Dedollarization Agenda: Lessons from the Uruguayan Case.” Money Affairs XVI (2): 193–218.
- Search Google Scholar
- Export Citation
Licandro, Gerardo, and Miguel Mello. 2018. “Dolarizacion cultural y financiera de los hogares uruguayos.” Investigación Conjunta-Joint Research. In Decisiones financieras de los hogares e inclusión financiera: evidencia para América Latina y el Caribe, edited by María José Roa García and Diana Mejía, 365–403.
- Search Google Scholar
- Export Citation
Licandro, José Antonio. 1999. “Una evaluacion de las reservas internacionales del Banco Central del Uruguay.” Economica XLV (4): 167–95.
- Search Google Scholar
- Export Citation
Vicente, Leonardo, Fabio Malacrida, and Fernando Zimet. 2017. “The Contribution of an ALM Approach to Monetary and Fiscal Policy: The Case of Uruguay.” In HSBC Reserve Management Trends 2017, edited by Robert Pringle and Nick Carver, 37–55.
- Search Google Scholar
- Export Citation

In 2002, Uruguay experienced triple crises: balance of payments, banking, and fiscal. An initial exogenous contagion from Argentina to the Uruguayan financial sector magnified the inherent weaknesses of the Uruguayan economy and its banking sector. Increasing withdrawals diminished the level of available international reserves, which were eventually insufficient to both service the external debt and continue backing the large proportion of foreign currency–denominated deposits still present within the system. Finally, the Uruguayan authorities had to let the peso freely float– which immediately depreciated by 27 percent—and declared a five-day bank holiday on July 30, 2002. See Bucacos (2017) for a detailed description.

Licandro and Mello (2018) show that the high degree of cultural dollarization in Uruguay is, in part, due to the dollarization of the price system, particularly housing prices.

Mario Bergara in the Ofcial Spanish Trade, Industry and Sailing Chamber. Published in Busqueda, #1670, July 12–18, 2012.

Busqueda #1835, October 1–7, 2015.

Uruguay uses its own debt in open market operations.

Uruguay has a safety criterion for reserve adequacy. It consists of ensuring the level of reserves that would allow the central bank to fulfill all mandates set by the law in 99 percent of cases. The basic criterion is set forth in Ibarra and others (2011).

Information on indirect intervention operations is available at the close of day.

Because of the size and intensity of central bank intervention, financial agents can easily identify when the central bank is intervening.

There is no record that this mechanism was used in the period.

The Macroeconomic Coordination Committee was created by the 2007 Charter Law of the Central Bank.

See Malacrida, Vicente, and Zimet (2017).

M1 is the sum of money in the hands of the public plus domestic-currency transactional deposits of the private sector in the banking system.

See Licandro (1999) and Della Mea and others (2011).

It presents the 1-month rate for the Uruguayan peso-nominated yield curve, the 1-year US dollar-nominated yield curve, and the 1-year node for indexed unit yield curve. The indexed unit is an accounting unit that indexes the Uruguayan peso by inflation.

The analysis does not use EMBI Uruguay, to avoid collinearity with interest rates.

The communications variable takes the value of 1 if the government signals concern over appreciation of the currency, –1 when it expresses concern over the depreciation of the currency, and 0 otherwise. It is constructed using public statements after monetary policy meetings.

Twelve de facto reaction functions are estimated; they depend on the type of intervention and the type of agent.

Adler and Tovar (2014) focused on purchases of foreign exchange only; this part of the document treats net purchases and net sales of foreign exchange separately to investigate the existence of different motives to intervene according to an expected appreciation or depreciation effect.

The dependent variable is truncated, for it takes either zero or a specific value that can be positive (net purchases) or negative (net sales). In this chapter, once the dependent variable is standardized, although still truncated, it becomes a continuous variable.

See the real exchange rate model in Annex 13.2.

Adler and Tovar (2014) also include reserve assets over short-run debt to capture precautionary motives; that data was unavailable for Uruguay.

Those values correspond to monthly data; see Annex 13.2.

The interest rate was not the monetary policy instrument in Uruguay for the whole period.

The correlation between the interest rate spread and the EMBI differential is 0.0378 for the sample period.

Tovar and others (2014); on the other hand, Druck and others (2018) show a high correlation between US dollar real effective exchange rate and commodity prices.

This study does not apply a structural analysis. For a more detailed explanation of the econometric issues when modeling with a mixture of I(1) and I(0) variables, see Fischer and others (2013).

Aggregate purchases and aggregate sales results are not presented here.

Recall that expected de facto intervention is used as an instrument for actual intervention in order to identify interventions properly.

As in Adler and Tovar (2014).

If we consider global intervention, the effect is similar but rather smaller.

This result also reflects the strong appreciation of the US dollar in global markets in the sample.

Results are sensitive to US dollar evolution in global markets during the time of analysis.

This is an asset and liability approach to the integrated balance sheet of the public sector and reserve requirements on nonresident investments in public debt in the primary market.

Weekly datasets are from January 2005 to December 2017, while monthly datasets cover March 2007 to December 2017, since the one-day interest rate was the monetary policy instrument.

Twelve de facto reaction functions are estimated, depending on the type of intervention and the type of agent.

Contributor Notes

The authors are grateful to participants at the Banco Central del Uruguay seminar for their suggestions, and to Pablo Picardo and Laura Zacheo for excellent research assistance. Special thanks are due to Fernando Zimet for his patience in answering the authors’ questions. The views expressed in this chapter are those of the authors and do not necessarily represent the views of the Central Bank of Uruguay.

Save
Cite
Email this content

Share Link

Copy this link, or click below to email it to a friend
Email this content
or copy the link directly:

https://www.elibrary.imf.org/display/book/9781484375686/ch013.xml
The link was not copied. Your current browser may not support copying via this button.

Link copied successfully

Collapse
Expand

Foreign Exchange Intervention in Inflation Targeters in Latin America

Author:

Mr. Marcos d Chamon

Mr. David J Hofman

Mr. Nicolas E Magud

, and

Alejandro M. Werner

ISBN:: 9781484375686

Publication Date:: 28 Feb 2019

Publisher:: International Monetary Fund

DOI:: https://doi.org/10.5089/9781484375686.071

Pages:: 320

View raw image

^{Figure 13.1.}

International Indicators for the Uruguayan Economy, 2004–18

View raw image

^{Figure 13.2.}

Balance of Financial Account: Regional Comparison, 2003–16

(Percent of GDP)

View raw image

^{Figure 13.3.}

Foreign Exchange Net Purchases and Exchange Rate, 2007–17

(Normalized data)

View raw image

^{Figure 13.4.}

Sterilized and Nonsterilized Interventions, 2007–17

(Millions of dollars)

View raw image

^{Figure 13.5.}

Estimated Cost of Surplus Reserves, 2010–17

View raw image

^{Figure 13.6.}

Brazilian Real, Uruguayan Peso, and US Dollar Comovements, 2007–17

View raw image

^{Figure 13.1.1.}

Impulse Response Functions: Response of Exchange Rate and Expected Depreciation to Interventions

(Standard deviations of the corresponding impulse)

View raw image

^{Figure 13.2.1.}

Impulse Response Functions: Response of Exchange Rate to Interventions using Choleski Identification

View raw image

^{Figure 13.2.1.}

Impulse Response Functions: Response of Exchange Rate to Interventions using Choleski Identification

View raw image

^{Figure 13.2.1.}

Impulse Response Functions: Response of Exchange Rate to Interventions using Choleski Identification

Foresee

Table 13.3.1.

Interventions in Uruguay

Institution	Type of Intervention	Instrument	Comment about the Instrument
Central Bank	Direct/indirect Sterilized/nonsterilized	Purchases/sales in the spot market	—
		Monetary Regulation Bills (Letras de Regulación Monetaria)	Most common instrument to sterilize purchases
		Certificates of deposit	Mainly used during interest rate target period
		Integration in US dollars Derivatives (SWAPS or financial contracts, futures)	From 2008 to 2017 intermittently Notably with public energy enterprises
		Bank reserve requirements	Increased and reduced (in 2016 and 2017, for example)
		Capital controls	Started in August 2012 and lasted until May 2015 (there were modifications in the meantime)
Ministry of Economy	Direct/indirect Nonsterilized	T-bills
		Bonds
		Purchases/sales to central bank
		Exports prefinancing	Handled by the central bank
Bank of the Republic (state-owned)	Direct	Purchases/sales in the spot market

Topics

Countries

Series

About

Resources

Abstract

Introduction

Uruguayan Intervention and Monetary Policy during 2004–17

The Macroeconomy during 2004–17

Structural Facts to Understand Intervention in Uruguay

Intervention Characteristics

Intervention Goals

Intervention as a Rule, Not an Exception

Intervention, by Type

Monetary Policy Design and Implementation

Institutional Setup of Monetary Policy

Stages of Monetary Policy since 2002

Var Methodology

Estimation Results

All Foreign Exchange Interventions

Direct and Indirect Foreign Exchange Interventions

Sterilized and Nonsterilized Interventions

Two-Stage Approach

Econometric Approach

First Stage: Central Bank Reaction Function

Second Stage: Exchange Rate Equation

Estimation Results

Conclusion

ANNEX 13.1. MONTHLY DATA

ANNEX 13.2. WEEKLY DATA

Annex 13.3. Interventions in Uruguay

References

Within Same Series

Other IMF Content

Other Publishers Content

Asian Development Bank

Inter-American Development Bank

Nordic Council of Ministers

The World Bank

Contributor Notes

Share Link

Table of Contents

Sections

Figures

Metrics

Foresee