A measure of exchange market pressure (EMP) in Brazil

(1)

Faculty of Economics and Business

Master Thesis

A measure of Exchange Market Pressure (EMP) in Brazil

July 2015

Abstract

This paper constructs an Exchange Market Pressure (EMP) Model for Brazil on the basis of the measure suggested by Klaassen and Jager (2011). The EMP model consists of the exchange rate change, the interest rate differential between the actual and the counterfactual interest rate and the official intervention component of the Banco Central do Brasil. Generally, it was found that the EMP measure works well for the period from 2009 to 2014. Peaks and troughs in Exchange Market Pressure could be underpinned by macroeconomic trends in Brazil. Although Brazil mainly faced excess supply on the forex market for the period under examination, an instance of excess demand after the Global Financial Crisis could be correctly identified. Between 2001 and 2008, it was found that the EMP measure is not a good approximation of exchange market pressure. This can be explained

by a high risk premium and weak monetary

institutions during this period.

Name: Thilo Hamann Student ID: 10089799 Email: Thilo.Hamann@student.uva.nl Supervisor: Franc Klaassen Words: 10296

(2)

2

Statement of Originality

This document is written by Thilo Hamann who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

3

1. I

NTRODUCTION

In a time in which the global economy is more interconnected than ever before, the interdependence between countries and contagion effects became especially apparent during the Global Financial Crisis (GFC) (Rose and Spiegel, 2010). An important role in the GFC is attributed to Emerging Market Economies (EME) which experienced large capital inflows since 2007 as they were seen as a safe harbor for investors (Tsangarides, 2011; Gradziuk, 2012). A drawback of these capital inflows was that some of the incoming capital was of a rather short term nature. Alexandre Tombini, the Brazilian Central Bank governor, stated in the Financial Times that the rising interest rates of the developed world acted as a “vacuum cleaner” for the Brazilian capital market as capital started leaving Brazil again (Rathbone and Wheatley, 2014).

One of the EMEs that was subject to major capital inflows during the GFC was Brazil. Amongst other macroeconomic developments, increased capital led to strong economic growth compared with the average OECD country (Figure 1) but also to an appreciating exchange rate. Looking at the value of the

Brazilian Real per dollar we can clearly see an appreciation of the Real since late 2004 (Figure 5). However, the appreciation is not as big as one would expect given the vast amount of capital inflows. This raises the question whether the Brazilian Central Bank, the Banco Central do Brasil (BCDB), is intervening in the forex market to avoid large changes in the exchange rate. Such fluctuations could have a negative impact on economic growth for instance by increasing the exchange rate risk (Edison, 1993).

When a central bank intervenes in the forex market (i.e. when a country has a fixed or managed exchange rate regime), the actual exchange rate does not represent its true value due to the arbitrary effect of interventions. Even when the central bank does not intervene in the forex market (i.e. when the country operates under a floating exchange rate regime), pressure on the exchange rate exists. The term Exchange Market Pressure (EMP) captures both cases. A general definition of the concept of EMP was pioneered by

100 150 200 250 300 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

Brazil in international comparison

Real GDP at purchasing power parity, 2000=100 BRAZIL Mexico Chile China India South Africa Indonesia Russian Federation OECD

(5)

5 Girton and Roper in 1977. While the measurement models of EMP used in academic literature are rather controversial and not yet complete (see Girton and Roper, 1977; Weymark, 1998; Klaassen and Jager, 2011), the definition of EMP is consistent among researchers. Klaassen and Jager define EMP as the “excess supply on the forex market if policy makers would refrain from actions to offset that excess supply, where this excess supply is expressed in the (relative) depreciation required to remove it” (Klaassen and Jager, 2011). Officially, given the free floating exchange rate of the Brazilian Real, one could immediately say: “EMP is fully reflected in the depreciation of the Real since the BCDB has no intention to intervene in the forex market”. However, the full story is not that simple. One reason for this is that the BCDB is pursuing an inflation targeting strategy for their monetary policy. The exchange rate, among other variables, has a significant effect on inflation. A prime example would be imported deflation. Imported deflation can be stylized in the following way: given a decrease in import prices, the price level in the importing country decreases via the appreciation of the exchange rate in the home country. This ultimately leads to deflation. To counter the effect of imported deflation a central bank can, for example, alter the exchange rate (Banco Central Do Brasil, 2011, p.55). Thereby, the exchange rate becomes an interesting tool for the central bank to influence the price level in their country. Interventions in the forex market to achieve policy goals can either be realized by changing the official interest rate, the selic rate in the Brazilian case, or by buying international reserves. Interestingly, the selic rate has risen constantly since 2013 and international reserves of the BCDB have steadily increased since mid-2005 (Figure 2 and 3). Whereas the selic rate rise can also be attributed to several other reasons aside from exchange market pressure, the rise in international reserves is a clear indication of fending off appreciating pressures on the Brazilian currency, which would have negative repercussions for the domestic economy (The Gulf Times, 2014).

This paper aims at quantifying the trend of increasing international reserves and a rising selic rate in Brazil. Furthermore, it will add to the current academic literature by determining a practical monthly index of EMP for Brazil using the approach suggested by Klaassen and Jager (2011). This approach has been chosen due to its easy implementation as well as its demonstrated success. Thus, the main research question of this paper is: to what extent is Brazil currently facing Exchange Market Pressure?

(6)

6 To analyze EMP in Brazil, monthly data from the period between 2001 to 2014 will be examined. Moreover, intervention periods via changes in international reserves will be determined to identify whether the EMP measure correctly predicts the presence of EMP in Brazil. Furthermore, we will determine whether changes in reserves can serve as a proxy for official interventions of the Brazilian Central Bank. The structure of the thesis is as follows: the next section will give an overview of the economic situation of Brazil and the role of the BCDB. Next, the evolution of EMP models will be outlined, as well as existing literature on EMP with the main focus on Brazil. Section 3 deals with modelling EMP for Brazil using the approach of Klaassen and Jager. This includes the methodology section and a short interpretation of the EMP model. Subsequently, section 4 presents and analyzes the EMP measure for the Brazilian case and shows a proxy for the intervention component of EMP. The paper subsequently finishes with a conclusion in section 5.

0 5 10 15 20 25 30 In %

Selic Rate

Figure 2 Source: DataStream

Figure 3 Source: DataStream

0 50000 100000 150000 200000 250000 300000 350000 400000 In M ill io n US $

International reserve holdings

(7)

7

2. L

ITERATURE REVIEW

The section starts off with a brief description of the economic situation in Brazil including the BCDB’s monetary policy practice and target levels. Subsequently, the current literature on EMP modelling will be presented including past EMP related research done in Brazil.

2.1 Brazil’s economic data

After various crises in the 1990s, the Brazilian economy started to boom in the early 2000s. The OECD has made a comparison to show Brazil’s superior growth compared to other OECD countries. Taking 2000 as a base year for all BRIICS (Brazil, Russia India, Indonesia, China and South Africa) countries as well as the average OECD country, it becomes apparent that Brazil, based on real GDP at Purchasing Power Parity, consistently outperformed the average OECD country since 2000. In numbers: The OECD average in 2012 was 121.36 (taking 2000 as the base year), whereas Brazil’s real GDP at PPP was 147.76, indicating superior economic growth of Brazil (Figure 1). This increasing growth of Brazil compared to the average OECD country can mainly be attributed to a more mature labor market leading to more employment and subsequently a decrease in labor inequality (OECD, 2013). Even the GFC, starting in 2008, had no strong repercussions for the economic growth of Brazil. Due to its increase in economic growth, Brazil has moved up the ladder to one of the largest world economies and has thus gained increasing interest by investors (OECD, 2013). Moreover, Brazil has consistently run a current account deficit since 1999, with a few

exceptions of capital account surplus (Figure 4). In the first two months of 2014, the capital account surplus amounted to 21.4 billion US dollars. It is worth mentioning that the biggest increase in the capital account can be attributed to increases in Foreign Direct Investment and Bonds, Notes and Commercial Papers (Banco Central do Brasil, 2014c, p. 56-57). Moreover, Santander Trade states that during the period from 2000 to 2015, most barriers on the stock market were removed and a lot of companies were privatized,

-14000 -12000 -10000 -8000 -6000 -4000 -2000 0 2000 4000 In M ill io n US $

CA Balance

(8)

8 making Brazil more attractive to foreign investors underpinning the large increase in the current account surplus (2015).

To promote stable and sustainable economic growth, the BCDB has followed an inflation targeting policy since the abolition of the pegged exchange rate vis-à-vis the Dollar in 1999. Inflation targeting is pursued mainly via changes in the selic rate. The selic rate target is set at 11 percent while the current inflation target is set by the BCDB at 4.5 percent with a band of ± 2 percent (Banco Central do Brasil, 2014b). As mentioned in the introduction, Brazil officially has a floating exchange rate; nevertheless, there have been various interventions by the BCDB to avoid large appreciations (see, for instance, Banco Central do Brasil, 2011, p. 62). Thus, the exchange rate has stayed relatively constant vis-à-vis the Dollar since 2004, at around 2.2 Brazilian Real for each Dollar (Figure 5). The only exception occurred in the fourth quarter of 2002, when the Brazilian Real per Dollar exchange rate reached a maximum of 3.8945.

Furthermore, the BCDB enormously increased their holdings of international reserves from 2005 onwards (refer to figure 3). When scaling the international reserves by the forex turnover of Brazil, the picture turns out slightly differently but the message stays the same (figure 6). Before 2005, the holdings of international reserves, scaled by forex turnover, increased constantly to $15.94 billion US dollars in January 2005. By January 2014, they had even increased to $20.51 billion US dollars with a big spike in 2008. The interventions of the BCDB indicate that EMP is not solely reflected in the change of the exchange rate, which would have been the case if Brazil had a perfectly free float, but also in the interventions of the central bank which need to be modelled.

To get a better overview of what EMP actually means and how it evolved the next subsection will elaborate on the history of EMP and its definition.

0 0,5 1 1,5 2 2,5 3 3,5 4 4,5 B razi lian R e al p e r US D o llar

Real vis-à-vis Dollar

(9)

9

2.2 Exchange market pressure measures

The Girton-Roper-Model is an early attempt to model EMP (Girton and Roper, 1977) based on the postwar Canadian experience. In their analytical paper, the dependent variable, termed Exchange Market Pressure, measures the volume of interventions necessary to achieve a certain Exchange Rate target. Two main interventions are mentioned. Firstly, the change in reserves, secondly the change in the exchange rate. Both affect the money market and are used to remove an undesirable disequilibrium. Hence, to derive the amount of official interventions required, the money market model is used, making the Girton-Roper model both model- and definition-dependent. An application of the Girton-Roper-Model to the postwar Brazilian case was conducted by Connolly and Da Silveira (1979). The authors found that the model is a good but imperfect measure for EMP in Brazil between 1955 and 1975.

The EMP measure of Girton and Roper is further developed by Weymark (1998), who adds some tweaks to the model-dependent definition to make it model-independent. Her model-independent definition reads as follows: “EMP measures the total excess demand for a currency in international markets as the exchange rate change that would have been required to remove this excess demand in the absence of exchange market intervention, given the expectations generated by the exchange rate policy actually implemented” (Weymark, 1998, p.109). There are two main differences with Girton and Roper. Firstly, Girton and Roper focus on the domestic money market. The EMP measure of Weymark, however, focuses on the excess demand of the currency in the international market. Hence, EMP can also be obtained from models that do not employ the monetary approach to determine the exchange rate. Secondly, Weymark’s definition of EMP takes into account that the actual observed exchange rate is different from the one which would exist under a perfect float with no interventions. The second crucial difference with Girton and Roper

0 5000 10000 15000 20000 25000 In M ill io n US $

International reserve holdings scaled by Forex

Turnover

(10)

10 is that Weymark’s definition measures the actual rather than the expected external imbalance of the economy. Overall, Weymark finds in her empirical studies that EMP differs widely from the estimates of the Girton-Roper-Model. Nevertheless, it is important to note that both Girton and Roper and Weymark use a model-dependent measure of EMP (by imposing the money market model) that is subject to the shortcomings of the underlying model.

Various authors have addressed the model dependency of Girton and Roper and Weymark by introducing the interest rate component into the EMP definition (see for instance Mody and Taylor, 2007, van Horen et al, 2006 or Klaassen and Jager, 2011). The interest rate component is an important policy instrument in the smoothing of exchange rate fluctuations nowadays and should be taken into consideration when determining EMP. The question, however, is how to introduce the interest rate into the EMP measure to make the EMP measure model-independent. Mody and Taylor and van Horen et al. use the change in the interest rate. However, using the change in the interest rate leads to a theoretical problem. Assume the following: under a fixed exchange rate regime, a central bank is confronted with EMP and is fending it off by increasing the interest rate. If the EMP measure is built around the change in the interest rate, the EMP measure would clearly indicate that EMP exists in period 1, but in period 2 (when EMP is still present) the interest rate does not change again. Thus, one could falsely conclude that EMP is present. Hence, using changes in the interest rate as a measure for EMP is wrong.

Klaassen and Jager (2011) address this problem by using the interest rate differential between the actual interest rate and the counterfactual interest rate. The counterfactual interest rate can be defined as the interest rate that would prevail if policy makers would refrain from influencing the currency to remove excess demand. Under the counterfactual interest rate, policy makers do not have an exchange rate objective. The challenge here is that the counterfactual interest rate is not readily observed. Klaassen and Jager tackle this challenge by using the generalized version of the famous and recognized Taylor rule, which shows the optimal interest rate, as a proxy for the counterfactual interest rate. Given their measure of EMP, Klaassen and Jager were able to demonstrate the presence of EMP in time periods for which traditional measures were unable do so (see for instance in the 1992-1993 crisis in the Exchange Rate Mechanism).

Various authors have also included capital controls in measuring EMP. Capital controls are one of the factors that affect the exchange rate. Eichengreen et. al. (1995) recognize that effective capital controls extend the time a currency peg can be sustained (p.257). Capital controls, on the one hand, reduce the probability of speculative attacks since traders are discouraged from attacking a currency when capital controls are in place. On the other hand, they make it difficult for capital to leave the country which subsequently keeps the exchange rate relatively stable. Moreover, capital controls can change the purpose of interest rate differentials and reserve changes since exchange rate stability is pursued via sterilized purchases on the forex market rather than via the interest rate.

(11)

11 To wrap up: After Girton and Roper’s first attempt at defining EMP many different researchers have worked on developing the EMP measure further. Starting from a model-dependent and definition-dependent description of EMP, the literature progressed to a model-free and definition-indefinition-dependent EMP measure which is used in the most recent academic literature.

3. EMP

M

EASURE FOR

B

RAZIL

In section 2, the economic situation of Brazil and the conduct of monetary policy by the BCDB were elaborated upon. Moreover, the history of the EMP measure was briefly laid out. This section aims at giving an overview of the Klaassen and Jager EMP measure, which will be used in this paper to define EMP for Brazil. To provide a clear understanding of the Klaassen and Jager EMP measure, the underlying assumptions as well as an interpretation of the model itself are presented below.

3.1 The EMP Model for Brazil

Firstly, a two-country setting is assumed with a domestic and a foreign country. In the Brazilian case, the domestic country is Brazil. The foreign country is assumed to be the United States. This is a reasonable assumption given the fact that the Brazilian currency was pegged to that of the US before 2000. The BCDB still monitors the Dollar-Real exchange rate closely, even though this is not officially stated by the BCDB. The policy goal of the BCDB is, in some respect, to manage the exchange rate, which they certainly do (see section 2.1). On the other hand, the goal of the Federal Reserve is not to influence the Brazilian exchange rate. Given the Fed’s objective of maximum employment, stable prices, and moderate long-term interest rates as stated in the Federal Reserve Act section 2A – this can fairly be assumed for the American case. Management of the exchange rate by the BCDB implies that EMP is not fully reflected in the change of the exchange rate (Δs_t). Thus, the BCDB’s policy instruments in influencing the exchange rate need to be modelled when defining EMP in order for a complete view. Instead of including all available monetary policy tools, we follow Klaassen and Jager and use a ”more concentrated central bank instrument set” for Brazil. Money market tools, which include the official discount rate, open market operations and bank reserve requirements of the BCDB, are represented in the selic rate (i_t). Furthermore, money market tools must be effective in influencing the exchange rate. Andrade and Kohlscheen (2013) address this issue. They examine the effect of official interventions on the demand for foreign exchange in Brazil. They find that these interventions were effective in affecting the exchange rate, but that they did not alter the amount of foreign supply in the market directly.

Secondly, the counterfactual interest rate (𝑖_𝑡𝑑_{) plays a crucial role in the EMP measure of Klaassen}

and Jager. The counterfactual interest rate can be defined as the interest rate that would prevail if a policy-maker would refrain from any actions to offset excess supply on the forex market. A passive policy-policy-maker

(12)

12 is defined as a central bank that focuses on domestic or external objectives such as inflation, the current account balance or output, rather than on exchange rate objectives. One of the suggested proxies for the counterfactual interest rate is the famous Taylor rule, since it takes into account output and inflation objectives (Taylor, 1993). Using the Taylor rule, the domestic counterfactual short-term interest rate can be derived under the following assumptions: firstly, the counterfactual Brazilian and American central banks follow the same Taylor rule. Secondly, Brazil and the US both are exposed to a highly integrated financial market, which implies that both countries will react to external shocks, even though the magnitude of the reaction can differ. Moreover, we are imposing the standard Taylor values of an equilibrium interest rate of r_t𝑒𝑞= 2% and γ = 0.5, thus assuming no differences in policy preferences. Finally, we arrive at the fully observable counterfactual interest rate as shown below (for a detailed derivation, please see Klaassen and Jager, 2011). π_te_{is the expected Brazilian inflation at time t and 𝑔}

𝑡 is the relevant gap which are here

considered to be inflation gap and output gap as Taylor suggests. Asterisks denote US variables.

i_td_{= i} t ∗_{+ (π} t e_{− π} t e∗_{) + 0.5(𝑔} 𝑡− 𝑔𝑡∗) (1)

It is important to note here that inflation expectations and inflation itself are modelled in the EMP measure for Brazil. We model both to correct for the different inflation expectations in Brazil and the US. In addition, it is worth noting that recent research by Hegerty (2014) found, using a VAR model, that inflation is a key component in determining EMP in Brazil and hence needs to be modelled.

Thirdly, Klaassen and Jager use official interventions by the central bank on the forex market (c_t) in their measure of EMP. These are scaled by a measure of forex market turnover (v_t) to account for a potentially increasing market size. The BCDB interventions channels are fourfold: Forward Interventions, Spot Interventions, repurchase lines of credit and foreign currency loans will all be captured in the ct

component (Banco Central Do Brazil, 2014a). To sum all of the components into one foreign intervention component, we follow the same approach as the Bank of International Settlements, which combines the different foreign exchange components via addition to show the total forex turnover of Brazil. This seems to be a reasonable approach when posing the assumption that all components affect the exchange rate to a similar respect. Moreover, we want to look into the total effect of foreign intervention by the BCDB and not on a component by component basis.

Finally, capital controls are assumed to have no effect on EMP in Brazil and are hence not modelled. This assumption is underpinned by Jinjarak et al. (2013), who analyze the effect of five different capital controls that were implemented by Brazil during the Global Financial Crisis. Using an analysis of counterfactuals, constructing a scenario for Brazil without these controls, they found that none of these tools were effective in limiting capital inflows to Brazil. This is also in line with other research that either assumes free capital mobility or refrains from modelling capital controls due to their complexity.

(13)

13 Based on the above assumption regarding the different EMP components of Klaassen and Jager, the below model-free measure can be derived by imposing a few more assumptions (for a detailed derivation please see Klaassen and Jager, 2011). Firstly, the exchange rate (s_t) is an explicit function of i_t, c_t and x_t, where x_t captures all other potential determinants of the exchange rate that are not correlated to i_t, c_t. Lastly, we assume that the functional form of s_t is constant over time which, yields:

EMP_t= Δs_t+ w_i(i_t− i_td_{) + 𝑤} 𝑐c_𝑣t

𝑡 (2)

Substituting (1) for i_td_{yields the EMP measure below, imposing the Taylor Rule for the counterfactual}

interest rate.

EMP_t= Δs_t+ w_i(it− [it∗+ (πte− πte∗) + 0.5(𝑔𝑡− 𝑔𝑡∗)]) + 𝑤𝑐c_𝑣t_𝑡 (3)

The above equation for EMP hence consists of the exchange rate change (Δs_t), the Brazilian selic rate (i_t), the Brazilian counterfactual interest rate (i_td_{) and the intervention component (c}

t) scaled by the forex

turnover 𝑣_𝑡. w_i and w_c are the weights referring to each EMP component. The following section clarifies the exact scaling of the weight components.

3.2 wi and wc in the EMP Measure

The weights in the EMP model above pose the biggest challenge in the Klaassen and Jager EMP measure since they are not observable. Throughout the history of the EMP, two different approaches have been used to tackle this challenge.

If model-dependency is assumed as in the measure of Girton and Roper, unit weights are implied. Weymark extends the Girton and Roper unity implication and arrives at weights which are unknown parameters, but that can be estimated. Eventually, the estimated weight then measures the effectiveness of the intervention instrument.

For model-independent measures of EMP, it is quite challenging to derive the exact weight components given that the exchange rate is tied to macroeconomic fundamentals and that one wants to estimate the weights from a structural exchange market model. To date, there is still no full-fledged agreement among researchers (see for instance: Klaassen, 2011 and Penecost et. al., 2001). The approach that generated the largest consensus is the one by Eichengreen et. al. (1994), who estimate the weight component by applying the ratio of two sample standard deviations (𝜎). In practice, this measure is easy to apply. Nevertheless, it does not only reflect the effectiveness of the intervention component but also how frequently this intervention component is used by the authorities. This in turn, leads to an unobserved bias in the estimator.

Despite this bias, the approach of Eichengreen et. al. (1994) is also applied in this paper and the weights are therefore defined as follows:

(14)

14 w_i= 𝜎∆𝑠 𝜎(𝑖𝑡−𝑖𝑡𝑑) 𝑎𝑛𝑑 wc= 𝜎∆𝑠 𝜎(𝑐𝑡 𝑣𝑡) (4)

3.3 Interpretations of the Klaassen – Jager EMP measure

To get a better understanding of the above derived EMP measure as used in this paper, a short interpretation of the measure is outlined below. There are various ways in which EMP can exist. In general, we can say that there are 3 cases for EMP in Brazil: let us examine each case briefly.

Case 1: Δst≠ 0

EMP exists when there is a change in the actual exchange rate between period 1 and period 2. A change in the exchange rate is a clear indicator that the currency is under pressure, regardless of the exchange regime applied by the BCDB. This pressure is immediately reflected in a corresponding appreciation.

Case 2: (i_t− [i_t∗_{+ (π} t e_{− π} t e∗_{) + 0.5(𝑔} 𝑡− 𝑔𝑡∗)]) ≠ 0

The second case of EMP occurs when the selic rate differs from the counterfactual interest rate, visible in the equation above. The selic rate, as assumed, encompasses all open market operations via the money market tools of the BCDB. The counterfactual interest rate signals the optimal level of the interest rate according to the Taylor Rule. A mismatch between these interest rates signals usage of the money market tools by the BCDB. This could be an indicator to fend off pressure of the exchange rate. If the selic rate is higher than the counterfactual interest rate, this signals excess supply on the forex market. There is excess demand on the forex market if the selic rate is lower than the counterfactual one.

Case 3: 𝜎∆𝑠

𝜎(𝑐𝑡

𝑣𝑡)

∗ct

𝑣𝑡≠ 0

Lastly, EMP is present when the BCDB officially intervenes in the foreign exchange market. In this case, the exchange rate does not represent its true value.

4. EMP

IN

B

RAZIL

The previous section laid out the methodology for EMP measure used in this paper. This section will show the practical application of the EMP measure to the Brazilian case. Starting off with the data collection and description, the section finishes with the an analysis of the monthly EMP measure for Brazil

4.1 Data Collection Method and Analysis

For the EMP measure for Klaassen and Jager to be applicable to the Brazilian case, not only the assumptions and methodology must be applicable to the Brazil, but the data must also be available for every component of the EMP measure. For this purpose, a monthly dataset beginning in November 2001 and continuing

(15)

15 through to the end of December 2013 has been gathered. The timespan of this analysis is restricted by the availability of the inflation expectation parameter. After the exchange rate regime was changed in 2000, the BCDB only started gathering yearly inflation expectation data from November 2001 onwards. The frequency of these observations is restricted by the intervention dataset from the BCDB, which is only available in monthly frequencies. Overall, two main databases were used to obtain the relevant data for the Brazilian EMP: Thomas Reuters’ DataStream and the database of the BCDB for an extensive range of Brazil-specific data.

Based on the data obtained, we can analyze the behavior of each EMP component in this section. To provide a solid overview, the EMP measure is split into its three components: Δs_t, w_i(i_t− i_td_{) and 𝑤}

𝑐_𝑣ct_𝑡.

4.1.1 The exchange rate change

In section 2, we introduced Figure 5 showing the actual exchange rate for the Brazilian Real vis-à-vis the US Dollar. The exchange rate data was collected via DataStream, which lists the BCDB as its source. Daily data was collected and the month end value was used since all data in this paper is on a monthly basis. Based on this dataset, the exchange rate change (Δst) is shown in figure 7 below.

Generally, one can say that the exchange rate changes remain within in a range of ±0.1 Real per Dollar per month. Nevertheless, there are a few exceptions; especially in between 2001 to 2003, when exchange rate change fluctuations were quite volatile (positive and negative), reaching a maximum of 0.8726 from August to September 2002. Also in 2008, the exchange rate change was above its average band, reaching its peak at 0.2799 between August and September.

4.1.2 The interest rate component

As has been assumed, all money market tools of the BCDB are captured by the selic rate (it). Daily data has

been collected for the selic rate via DataStream, which again lists the BCDB as its source, and has been

-0,1 -0,05 0 0,05 0,1 0,15 Δ B razi lian R e al p e r US D o llar

Real/Dollar change per month

(16)

16 averaged out to be consistent with monthly data intervals (please refer to figure 8 for a graphical representation).

The counterfactual interest rate (i_td_{) consists of various sub-components, as described in section 3.}

Let us examine each.

4.1.2.1 Foreign Interest Component

For the foreign interest component (i_t∗_{), daily data for the US federal fund rate has been collected via}

DataStream, now listing the Federal Reserve as a source. The daily data has been averaged to fit the monthly intervals of the dataset. The federal fund and the selic rate are shown in figure 8 below.

The federal fund rate has risen constantly since June 2004 and reached its peak in February 2007 at 5.26 percent. After the federal fund rate decreased to 0.15 percent in January 2009, it never recovered from this fall and remained relatively constant around this percentage. When comparing the federal fund rate with the selic rate, it is apparent that the selic rate is a lot more volatile. Moreover, the selic rate is set at a higher percentage than the federal fund rate due to the higher inflation target set for Brazil by the BCDB.

4.1.2.2 Expected inflation gap

The yearly expected inflation gap component (πte− πte∗) consists of the Brazilian and US yearly inflation

expectation component. The Brazilian expected yearly inflation was gathered directly from the BCDB database. The “General Price Index – Internal Supply” (IGP-DI) was deemed the most suitable option here. The data set collected is on a daily basis. To bring it to the standard monthly dataset used in this paper, the average was taken (an acceptable calculation since no distinct outliers were identified in the dataset). For the US, yearly expected inflation data was directly downloaded from DataStream in a monthly form, since daily data was not available. Here, DataStream lists the Federal Reserve Bank of Cleveland as its source. Figure 9 depicts the yearly expected inflation gap by taking the difference between yearly expected inflation for Brazil and the US.

In general, the yearly expected inflation components for Brazil and the US are quite stable at an average 3.9 percent. However, there is one exception: In December 2002, the Brazilian yearly inflation expectation reached an all-time high of 18 percent after the abolition of the exchange rate target with the US. As a consequence, the yearly expected inflation gap increased to 16.05 percent. The main reason for this spike in yearly expected inflation rates, according to the December 2002 Inflation Report of the BCDB, was the off-season agricultural period causing a supply shock to the economy (Banco Central do Brasil, 2002). Nevertheless, this inflation spike is an exception. After 2004, the yearly inflation expectation seems rather stable, but still lies above the target of 4.5 percent set by the BCDB. In July 2008, the yearly inflation expectation skyrocketed to 7.7 percent. This can be attributed to the Global Financial Crisis. However, this

(17)

17 did not affect the yearly expected inflation gap component drastically. It is worth mentioning that the Brazilian yearly inflation expectations consistently lie above the yearly inflation expectations of the US in the period under consideration, resulting in a positive yearly expected inflation gap component.

Overall, yearly inflation expectations for both countries seem relatively stable, with the Brazilian yearly inflation expectations fluctuating a bit more than the US yearly inflation expectations. For both countries, they seem to lie within the bandwidth of their targets as set by the Central Banks; indicating a stable yearly expected inflation gap.

4.1.2.3 Gap component

The other gap component (𝑔_𝑡− 𝑔_𝑡∗_{) in the EMP measure is twofold. Firstly, it consists of the yearly inflation}

gap (π𝑡− π𝑡∗) component. The yearly inflation gap is modelled to correct for the different yearly inflation

targets of the Brazilian and US central banks. Both datasets were gathered via DataStream on a daily basis and are averaged to fit the monthly EMP model. For the Brazil, the data source is the BCDB; for the US, DataStream lists the Bureau of Labor Statistics. Figure 10 depicts the yearly inflation gap by taking the difference between the Brazilian and US yearly inflation.

It becomes clear from figure 10 that the yearly inflation gap follows the same pattern as the yearly expected inflation component. Also, the actual yearly inflation of Brazil consistently lies above that of the US, resulting in a positive inflation gap. The reason for this is the higher yearly inflation target set by the BCDB as compared to the Federal Reserve. Furthermore, it is worth mentioning that the actual yearly inflation for both countries fluctuates a lot more when compared to the yearly inflation expectations (refer to Figure 9).

Secondly, the gap component consists of the actual output gap (y_𝑡− y_𝑡∗_{). Yearly data on the output}

gap for Brazil and the US was gathered from the dataset of the OECD. This poses a challenge for this paper since all other data is collected at monthly intervals. Hence, we apply linear interpolation to obtain a monthly dataset for each country. Given the fact that the data does not vary much over the period examined in this paper, linear interpolation does not pose any problems and is the correct tool to use. Figure 11 below shows the monthly data for the output gap.

The output gap differs widely between the two countries. While Brazil (solid line) started off in November 2001 with a GDP under its potential GDP, it consistently performed above potential GDP after March 2006, reaching a maximum in January 2010 at 4.6 percent above potential GDP. This is in line with Figure 1, which shows that Brazil is currently amongst one of the best performing emerging countries. However, after this all-time high, the percentage has steadily declined ever since.

(18)

18 0 5 10 15 20 25 30 in %

Selic Rate / Federal Fund Rate

Federal Fund Rate Selic Rate 0,00 5,00 10,00 15,00 20,00 in %

Expected Inflation Gap

-5 0 5 10 15 20 In %

Inflation Gap

-6 -4 -2 0 2 4 6 Ou tp u t gap in %

Output Gap

Brazil US Figure 9

Source: BCDB & DataStream

Figure 8 Figure 10 Source: OECD Figure 9 Figure 10 Figure 11 Source: DataStream Source: DataStream Figure 8 Figure 11

(19)

19 For the US, the picture is the opposite (dotted line). While in the period between (late) 2001 and 2008, the US was outperforming potential GDP, the US underperformed after January 2008, reaching a minimum of 4.68 percent below potential GDP in April 2009. Since that all-time minimum, the output gap of the US has declined gradually.

4.1.2.4 The counterfactual interest rate

Based on all the components above, the counterfactual interest rate can be constructed. It is depicted in figure 12 in a direct comparison with the selic rate.

Interestingly, it seems that for the time period between November 2001 and late 2008, the selic rate differs widely from the counterfactual interest rate. The gap is as big as 20.53 percent in January 2005, which seems to be an unreasonable difference between the two rates. However, when looking at the general trend of the selic rate compared to the counterfactual, it becomes apparent that they are rather similar. Upward and downward movements are represented in both rates during the same period. Nevertheless, the gap remains. The reason for the big difference between the selic and counterfactual interest rate can be derived from various sources. Firstly, the risk premium, which is not modelled in the counterfactual scenario, was very high in the period 2001 to 2008. Figure 13 below shows the total equity risk premia for Brazil. The risk premium data was gathered from the free source ‘Damodaran Online’. On this webpage, all annual risk premia from 2001 till 2014 can be found for Brazil. Damodaran uses a 4 step approach to calculate the total equity risk premium for a country. In the first step, he estimates the mature market risk premium. Next, he estimates the default risk of Brazil via a dataset from Moody’s. The default risk is subsequently converted into the country risk premium. Finally, all premia are added to obtain the final total equity risk premia of Brazil. 0 5 10 15 20 25 30 In te re st rate in %

Counterfactual VS Selic Rate

Selic

Counterfactual Figure 12

(20)

20 Figure 13 reveals that the risk premium during the period of 2001 to 2003 increased to an all-time high of 15.76 percent in 2003. Afterwards, the risk premium decreased steadily but still remained remarkably high.

Since 2008, the risk premium has been rather stable at around 8.5 percent. The high average risk premium of 10.79 percent during the period from 2001 to 2008 can explain the big difference between the selic and the counterfactual rate as shown in figure 12, as the counterfactual interest rate as constructed in this paper does not fully account for the Brazilian risk premium with a separate variable.

Secondly, Brazil was hit with various supply shocks between 2001 and 2005. After the introduction of the floating exchange rate, Brazil suffered from the slowdown of the world economy; most notably in 2001 to 2002, as well as an energy shortage in 2001. These were reasons why the inflation target was missing between 2001 and 2003. Nevertheless, Brazil has had a good track record regarding the inflation targets set ever since, which can explain the big gap in the beginning of the 2000s as well as the subsequent closing of the gap (OECD, 2005).

Thirdly, let us examine the policy of the BCDB with respect to the Selic rate. The selic rate was established in November of 1979 under the aegis of “Circular 466”. For years, the Brazilian Real was pegged to the US Dollar. As stated in section 2, this peg was resolved in 1999 and Brazil allowed the Real to float freely. A change in the exchange rate regime poses various challenges. The introduction of independent monetary policy also has implications for operation of the Selic rate. No less than 9 out of 17 changes to the selic system occurred from 2002 to 2006. The biggest change occurred in 2002 with the restructuring of the Brazilian payment system, during which the Selic was reformed (Banco Central Do Brasil, 2014d). This is a clear indicator that monetary institutions were not yet perfectly established. In the first OECD Economic Survey of Brazil in 2005, it was also recommended that the institutions for macroeconomic policymaking in Brazil needed to be strengthened further to remain on a sustainable path of economic growth. One of the

0,00% 2,00% 4,00% 6,00% 8,00% 10,00% 12,00% 14,00% 16,00% 18,00% R isk Pr e m iu m in %

Risk Premium Brazil

(21)

21 main reasons was the bad investment climate due to excessively high interest rates. Moreover, it was noted that monetary institutions had not yet set up a complete inflation targeting system, introduced with the establishment of the free float (OECD, 2005).

Apart from the big difference between the selic rate and the counterfactual one in the period between 2001 and 2008, there appears to be a big spike in the counterfactual interest rate in 2002. The spike can mainly be attributed to the skyrocketing yearly expected inflation component as explained above.

In conclusion, the counterfactual interest rate as constructed in this paper is not a good proxy for the selic rate for the period from 2001 to 2008, given the large discrepancies between the selic rate and the counterfactual one. However, since 2009, the counterfactual interest rate and the selic rate seem to be moving a lot closer together. The average difference decreased to a remarkably low 2.15 percent, which is noticeably smaller than before; the average difference from 2001 to 2008 was 9.58 percent. In addition, the counterfactual interest rate seems to follow a similar pattern to the selic rate since 2009; this was not the case prior to 2009. The main reasons for the convergence of the selic rate and the counterfactual one are the decrease of the risk premium, lesser exposure to supply shocks and finally a more mature and credible Central Bank with a lower inflation target.

4.1.3 The intervention component

The last component of the EMP measure used in this paper is the intervention component (ct

𝑣𝑡). Let us look

at the two components separately.

The official intervention component (𝑐_𝑡) was collected from the BCDB webpage directly. In the dataset, the intervention component consists of four variables: Forward interventions, spot interventions, repurchase lines of credit and foreign currency loans which are summed to build the intervention component. The monthly intervention component is modelled below in figure 14.

The official interventions by the BCDB are quite volatile. Until 2007, the BCDB intervened much less in the foreign exchange market than it did in subsequent years. In May 2007, the biggest positive intervention was recorded at 14.626 billion US Dollars and one and a half years later, in October 2008, the biggest sale was recorded at 9.510 billion US dollar. The biggest purchase of US Dollars was triggered by the US subprime mortgage crisis in early 2007, during which the risk perception of banks increased. CDS premiums of major Brazilian banks quadrupled within a period of three months. The increased risk perception led the BCDB to intervene in the market and to switch from a rather restrictive monetary policy to an expansive monetary policy. This led to a large injection of liquidity into the economy, thereby reducing the risk of a systemic collapse of funds and banks (BCDB, 2007). The biggest sale of Real can be attributed to the announcement of the Fed on the 29th_{of October to introduce the Dollar-Swap-Line between the Fed,} the BCDB and other South American Central Banks. A swap line is mainly used for short term lending at a

(22)

22 going exchange rate between two Central Banks. The introduction of these facilities improved liquidity conditions in the Global Financial Market and was designed to mitigate the difficulties in obtaining Dollar funding. (Federal Reserve, 2008).

After 2006, the volatility and frequency of official interventions also increased. Overall, the official interventions of the BCDB are of a rather positive nature, which implies more buying than selling of international reserves. This is another indication that Brazil is currently faced with excess supply on the forex market, positive EMP, and that the BCDB tries to avoid large appreciations of the exchange rate by intervening in the market.

Nevertheless, the increased amount of official interventions can also be a natural phenomenon since the forex market has grown over time. As a consequence, we must scale the intervention component by the size of the forex market (𝑣𝑡). Data on the size of the forex market has been collected from the Bank of

International Settlements (BIS). The challenge with the dataset of the BIS is its long-term nature. The BIS only has data points for each third year, which poses a challenge for the monthly dataset used in this paper. Since it seems that the forex market is increasing exponentially, a cubic spline was used to interpolate monthly data. The outcome is depicted in figure 15 below.

As can be seen from figure 15, the forex market increased by a factor of approximately 3 from 2001 to 2013. Looking at all the instruments, the size of the forex market was estimated to be 6 billion US Dollars in November 2001. By 2013, this had increased to 17 billion US dollars. When looking at the composition of the instruments used, the BIS differentiates between spot, outright forwards, foreign exchange swaps, currency swaps and options. As mentioned earlier, the biggest part of the Brazilian forex market consists of spot interventions. This is also reflected in the dataset of the BIS. Interestingly, the amount of outright forwards only started to increase from 2010 onwards. In 2007, outright forwards made up a mere 6.8 percent of the whole Brazilian forex market. This percentage increased to 43.4 percent by 2010. Moreover, the amount of foreign exchange swaps, currency swaps and options is rather limited.

In figure 16, we scale the official interventions of the BCDB (Figure 14) by the size of the Brazilian forex market. As expected, it becomes clear that the volatility of the intervention compared to figure 14 has decreased. However, the picture remains the same.

(23)

23 -15 -10 -5 0 5 10 15 20 In B ill io n US $

BCDB Intervention Component

Figure 14 0 2 4 6 8 10 12 14 16 18 20 In B ill io n US $

Size of the Brazilian Forex Market

Source: BIS Figure 15 Figure 14 Source: BCDB -1,5 -1 -0,5 0 0,5 1 1,5 2 2,5 In B ill io n US $

Scaled Interventions

Figure 16 Source: BCDB

Figure 15 Source: BIS

(24)

24

4.1.4 The weight component

For the correct weighting of the components in the EMP measure, we use the approach of Eichengreen et al. (please refer to section 3.1.1). The relevant standard deviations of the exchange rate change, the differential between the selic rate and the counterfactual one and the intervention component are depicted in the table below. The table also shows the corresponding weights.

𝜎∆𝑠 0.137218 𝜎(𝑖𝑡− 𝑖𝑡𝑑) 5.63517 𝜎(𝑐𝑡 𝑣𝑡) 0.616348 𝐰𝐢= 𝝈∆𝒔 𝝈(𝒊𝒕− 𝒊𝒕𝒅) 0.02435 𝐰𝐜= 𝝈∆𝒔 𝝈(𝒄𝒕 𝒗𝒕) 0.222632

From the table above, it becomes apparent that the standard deviation of the interest rate differential is quite high. This can be attributed to the big gaps between the selic rate and the counterfactual interest rate between 2001 and 2009. The weight component for the interest rate differential is quite low, at 0.024, as compared to the component of the official interventions, which is 0.223.

4.2 The EMP measure for Brazil

Based on the above monthly dataset, the EMP measure for Brazil can be constructed. To provide a good overview of the collected data, figure 17 illustrates all 3 components of the EMP measure, multiplied with their corresponding weights so that EMP is the unweighted sum.

(25)

25 Comparing the different EMP components, it becomes apparent that the exchange rate change and the intervention component are the most volatile, whereas the interest rate differential seems less volatile. Furthermore, there is a small negative correlation of 4 percent between the exchange rate change and the intervention component. This seems logical: if the BCDB sells (negative intervention) international reserves, the exchange rate should appreciate (positive change) accordingly, which seems to be the case. The effect on the exchange rate, however, seems rather small given the monthly dataset being used in this paper.

Additionally, the interest rate component dominates all other EMP components between 2001 and 2008. The reason for this is the big difference between the selic rate and the counterfactual interest rate as depicted in figure 12, caused mainly by the high risk premium and weak monetary institutions during this period. Also, interventions during this period seem to have been rather minor. Hence, the EMP measure will be biased upward for this period. For the period 2009 to 2014, no EMP component is particularly dominant.

Let us now translate the components of figure 17 into the monthly EMP measure for Brazil. Figure 18 shows the results for the period from November 2001 to December 2013.

-0,08 -0,06 -0,04 -0,02 0 0,02 0,04 0,06 0,08 0,1 0,12 0,14

EMP by component

Exchange Rate Change Interest Differential Intervention Figure 17

(26)

26 The EMP measure for Brazil clearly indicates that exchange market pressure for the Brazilian Real is present. Starting with the beginning of the dataset, we can see that Brazil faced severe EMP in the early 2000s. Indeed, EMP soared in August 2002. This, however, calmed down towards the end of 2002 and turned into the strongest excess demand in the forex market in March 2003. The highly volatile EMP in this period can be attributed to the largest inflation ever recorded in Brazil between 2002 and 2003. This period is followed by a long span of positive EMP from 2003 until mid-2008. As already stated above, the positive EMP in this period can mainly be attributed to the interest rate differential between the actual selic rate and the counterfactual one. This biases the EMP measure upwards via the high risk premium and weak monetary institutions in Brazil during this period. Hence, we should take the EMP measure during this period with a grain of salt.

From 2009 onwards, the EMP measure becomes more accurate. This can be attributed to two reasons. Firstly, the differential between the selic rate and the counterfactual rate decreased drastically after 2008. Secondly, official interventions on the forex market by the BCDB became a lot more frequent and extensive during this period. This put the exchange rate and EMP on a stable path.

The year 2009 shows an additional exception to the constant excess supply in the forex market. Between March and August, the EMP signals excess demand on the forex market. When looking at the inflation report of September 2009, published by the BCDB, various reasons can be identified; all boiling down to the contagion effects of the Global Financial Crisis. For instance, net inflows from the US in the period of March to August resulted in 6.9 billion US Dollars - 7.5 billion US dollars short compared to the same period in 2008. Moreover, the financial segment also fell short by 17.4 billion US Dollars. This can mainly be attributed to a decreased in purchases and foreign exchange sales. To counter the effects of less capital inflows on the exchange rate and to keep its Balance of Payments stable, the BCDB purchased a

-0,15 -0,1 -0,05 0 0,05 0,1 0,15 0,2

EMP for Brazil

(27)

27 total of 7.3 billion US Dollars on the spot market in the first 8 month to strengthen international reserves. This strategy eventually led to a switch in the balance sheets of banks from a 1 billion long position in US dollars in the end of December 2008 to a 1.1 billion short position in August 2009. Nevertheless, excess demand on the forex market could not be avoided fully. The purchase of Dollars eventually came to a hold once US and EU income growth as well as consumption revived. A subsequent switch back to excess supply on the forex market resulted. Furthermore, the direct investments made abroad in 2008, which totaled 20.5 billion US Dollars, only resulted in a net return of 6 billion US Dollars. Also, the direct investment abroad within the first nine month of 2009 was 7.5 billion US Dollars lower than in 2008 (Banco Central do Brasil, 2009).

The most noticeable event in the period after 2009 is the spike in August 2011. The spike can be attributed to the big exchange rate change caused by a record flow of FDI to Brazil of 44.1 billion US Dollars between January and August of 2011. Compared to 2011, the amount increased by a remarkable 157 percent. The main FDI components identified were equity participation in companies of 34.8 billion US Dollars and intercompany loans of 9.2 billion US Dollars. To limit the effects of heavy FDI on the exchange rate and hence strong excess supply in the forex market, the BCDB further increased its intervention on the spot market (Banco Central do Brasil, 2011a).

Overall, it becomes apparent from the EMP measure constructed in this paper that in recent years, Brazil has mainly faced excess supply rather than excess demand on the forex market. This implies a capital inflow into the country, which is can be observed from the increase in FDI to Brazil in recent years. Also, the amount of positive interventions by the BCDB increased to avoid large appreciations of the exchange rate. This further attracted FDI and strengthened the exports of goods and services, which increased consistently from 133 billion US Dollars in 2005 to 281 billion US Dollars in 2013.

4.3 Proxy for the intervention EMP

In the above EMP measure, the intervention component of the BCDB is readily observable from the dataset gathered from the BCDB. However, this is not always the case. Hence, various researches have looked at different options to approximate the intervention component. A general proxy for the intervention component by the BCDB is the change in reserves, used by various other researches in the past (e.g. Fiess and Shankar, 2009).

This section verifies whether the change in reserves can be used as a good proxy for the intervention component for Brazil. Data for the monthly reserve balance of the BCDB was collected directly from the BCDB dataset, the same dataset that was used for the direct intervention component. The same weights as for the EMP measure above were employed, as well as the same scale component. Figure 19 shows the result of applying the proxy for the intervention component to the Brazilian case, ceteris paribus.

(28)

28 Figure 19 confirms that the change in reserves serves as a good proxy for the EMP measure as constructed in this paper. Most of the time, the lines of the proxy and the actual EMP measure coincide. The only period for which the proxy does not fully coincide with the actual EMP measure is from 2005 to 2008. Nevertheless, it largely gives the same picture. In 2008, there is a big discrepancy between the actual EMP and the Proxy, resulting from rather minor changes in the reserves but big interventions. The reserves in the beginning of 2008 were marked at 187 billion US Dollars and at the end of the year, they hit 193 billion; a rather minor change. Nevertheless, the interventions resulted in a total of 5 billion in sales in 2008 which does not match the change in reserves. The reason for this is that the interventions of the BCDB during this time were of a rather unconventional nature (see, for instance, the swap agreement with the US).

Overall, changes in reserves can be seen as a good proxy for the intervention component of the EMP measure as they reflect the same picture as compared to using the accurate measure of actual interventions. Since data on changes in reserves are easier to collect than the actual interventions of central banks, it is a convenient proxy to use when constructing an EMP measure.

-0,15 -0,1 -0,05 0 0,05 0,1 0,15 0,2

EMP and Proxy EMP

Actual EMP Proxy Figure 19

(29)

29

5. C

ONCLUSION

The aim of this paper was to determine a practical monthly index of EMP for Brazil. For this purpose, Klaassen and Jager’s (2011) measure for EMP was applied for the period from 2001 to 2014. Klaassen and Jager define EMP as the “excess supply on the forex market if policy makers would refrain from actions to offset that excess supply, where this excess supply is expressed in the (relative) depreciation required to remove it”. To determine the monthly EMP for Brazil we have shortly outlined the economic situation of Brazil. Moreover, past literature on EMP was elaborated upon and the methodology for the EMP measure used in this paper was defined. Finally, a monthly dataset was collected for each component of the EMP measure and then aggregated to obtain the unweighted EMP measure for Brazil.

For the period from 2001 to 2008, the EMP measure used in this paper has an upward bias. The main reason for the upward bias is the big discrepancy between the selic rate and the counterfactual interest rate. This discrepancy, in turn, is caused by the high risk premium and weak monetary institutions in Brazil during this period. For the period between 2009 and 2014, the EMP measure does not contain this upward bias caused by the interest rate differential. We can conclude that Brazil mainly faces excess supply on in the forex market (positive EMP). The largest positive spike in EMP occurred in August 2011 as a result of the largest increase in FDI thus far recorded. One of the main reason for the excess supply in the forex market is the active intervention by the BCDB to counter further appreciation of the Real. An exception to the positive EMP can be seen during the first 8 months of 2009 and can mainly be attributed to the contagion effects of the Global Financial Crisis. Moreover, a proxy for the EMP that uses the change in reserves for the direct interventions was provided. The proxy yields the same results as the actual EMP measure, with the exception of a period in 2008. This discrepancy can be attributed to the rather unconventional interventions of the BCDB.

There are various ways to further investigate the EMP measure for Brazil. Firstly, one could extend the EMP measure by including the risk premium as an additional component – this would help better explain EMP between 2001 and 2008. Secondly, given the size of the oil industry in Brazil, it would be interesting to see whether there is a relationship between stock prices of major companies and EMP. Thirdly, once available, a daily dataset could be used to construct the EMP measure for Brazil to yield further, more detailed insights into Exchange Market Pressure. Lastly, the EMP measure of Klaassen and Jager could be applied to more emerging markets to explore whether there are common trends in EMP among emerging market economies.

(30)

30

6. B

IBLIOGRAPHY

Andrade, C.S., Kohlscheen, E. (2013). Official Interventions through Derivatives: affecting the demand for foreign exchange. Working paper series Brasilia No. 317. 2013. 1-47.

Banco Central Do Brasil (2002). Inflation report of December 2002. 1-181

Banco Central Do Brasil (2007). Financial stability report of November 2007. 1-156 Banco Central Do Brasil (2009). Inflation report of Spetember 2009. 1-147

Banco Central Do Brasil (2011). Inflation report of March 2011. 1-139. Bance Central Do Brasil (2011a). Relatorio de Inflacao Setembro 2011. 1-136

Banco Central Do Brasil (2014a). Statement of International Reserve growth. Retrieved From: http://www.bcb.gov.br/?serietemp. 01. April 2014.

Banco Central Do Brasil (2014b). Homepage. Retrieved From: http://www.bcb.gov.br/?ENGLISH. 01. April 2014.

Banco Central Do Brasil (2014c). Relatorio de Inflacao Marco 2014. 1-115 Banco Central do Brasil (2014d). Introdução ao Selic. Retrieved From:

http://www.bcb.gov.br/?SELICINTRO. 16. May 2015.

Connolly, M., & Da Silveira, J. D. (1979). Exchange market pressure in postwar Brazil: An application of the Girton-Roper monetary model. The American Economic Review. 448-454.

Edison, Hali J. (1993). The effectiveness of central-bank intervention: a survey of the literature after 1982. International Finance Section, Department of Economics, Princeton University 18.

Eichengreen, B., Rose, A.K., Wyplosz, C. (1995). Exchange Market Mayhem: The Antecedents and Aftermath of speculative attacks. Economic Policy, 10(1), 249-312.

Eichengreen, B., Rose, A.K., Wyplosz, C. (1994). Speculative attacks on pegged exchange rates: an empirical exploration with special reference to the European Monetary System, NBER Working Paper Series. Number 4898.

Federal Reserve (2008). 2008 Monetary Press Releases. Retrieved from:

http://www.federalreserve.gov/newsevents/press/monetary/20081029b.htm. 16. May 2015 Fiess, N., Shankar, R. (2009). Determinants of exchange rate regime switching. Journal of International

Money and Finance 28, 68–98.

Gradziuk, A. (2012). The Eurozone Crisis and Emerging Market Economies. The Polish Quarterly of International Affairs 2. 65-81.

Girton, L., Roper, D. (1977). A monetary model of exchange market pressure applied to the postwar Canadian experience. The American Economic Review. 537-548.

Gulf Times (2014). Central banks under pressure to raise rates. 6. Febuary.

Hegerty, S. W. (2014). Exchange market pressure, commodity prices, and contagion in Latin America. The Journal of International Trade & Economic Development 23(1), 56-77.

Jinjarak, Y., Noy, I., Zheng, H. (2013). Capital controls in Brazil – Stemming a tide with a signal?. Journal of Banking and Finance 37. 2938-2952.

Klaassen, F. (2011). Identifying the Weights in Exchange Market Pressure. Tinbergen Institute Discussion Papers 11-030/2. Tinbergen Institute

Klaassen, F., Jager, H. (2011). Definition-consistent measurement of exchange market pressure. Journal of International Money and Finance 30. 74-95

Mody, A., Taylor, M.P. (2007). Regional vulnerability: The case of East Asia. Journal of International Money and Finance 26, 1292–1310.

(31)

31 OECD (2005). OECD Economic Survey Brazil 2005, OECD Publishing.

OECD (2013). OECD Economic Surveys: Brazil 2013, OECD Publishing.

Pentecost, E.J., Van Hooydonk, C., Van Poeck, A. (2001). Measuring and estimating exchange market pressure in the EU. Journal of International Money and Finance 20, 401-418.

Rathbone, J.P., Wheatley, J. (2014). Pressure mounts on EM nations to lift rates. Financial Times. 27. January.

Rose, A., Spiegel, M. (2010). Cross-country causes and consequences of the 2008 crisis:

International linkages and American exposure. Pacific Economic Review 15(3). 340-363. Santander Trade (2015). Brazil: Foreign Investment. Retrieved from:

https://en.santandertrade.com/establish-overseas/brazil/foreign-

investment?&actualiser_id_banque=oui&id_banque=0&memoriser_choix=memoriser. 16. May 2015

Taylor, J. B. (1993). Discretion versus policy rules in practice. Carnegie-Rochester conference series on public policy 39. 195-214.

Tsangarides, C. G. (2012). Crisis and recovery: Role of the exchange rate regime in emerging market Economies. Journal of Macroeconomics 34. 470-488.

Van Horen, N., Jager, H., Klaassen, F. (2006). Foreign Exchange Market contagion in the Asian crisis: A Regression-based approach. Review of World Economics (Weltwirtschaftliches Archiv) 142, 374– 401.

Weymark, D. N. (1998). A general approach to measuring exchange market pressure. Oxford Economic Papers, 50(1), 106-121.

A measure of exchange market pressure (EMP) in Brazil

Faculty of Economics and Business

Master Thesis