Empirical exchange rate models fit: Evidence from the Brazilian economy

Similar documents
Empirical exchange rate models fit: The case of Brazil

Final Exam Answers Exchange Rate Economics

ANSWER ALL QUESTIONS. CHAPTERS 6-9; (Blanchard)

The Relationship between Money Demand and Interest Rates: An Empirical Investigation in Sri Lanka

Section 4 The Exchange Rate in the Long Run

CHAPTER CHAPTER18. Openness in Goods. and Financial Markets. Openness in Goods, and Financial Markets. Openness in Goods,

1 Purpose of the paper

On the Impact of Inflation and Exchange Rate on Conditional Stock Market Volatility: A Re-Assessment

(1 + Nominal Yield) = (1 + Real Yield) (1 + Expected Inflation Rate) (1 + Inflation Risk Premium)

UCLA Department of Economics Fall PhD. Qualifying Exam in Macroeconomic Theory

CURRENCY CHOICES IN VALUATION AND THE INTEREST PARITY AND PURCHASING POWER PARITY THEORIES DR. GUILLERMO L. DUMRAUF

Portfolio investments accounted for the largest outflow of SEK 77.5 billion in the financial account, which gave a net outflow of SEK billion.

What is Driving Exchange Rates? New Evidence from a Panel of U.S. Dollar Bilateral Exchange Rates

Market and Information Economics

FINAL EXAM EC26102: MONEY, BANKING AND FINANCIAL MARKETS MAY 11, 2004

A Note on Missing Data Effects on the Hausman (1978) Simultaneity Test:

Lecture 23: Forward Market Bias & the Carry Trade

2. Quantity and price measures in macroeconomic statistics 2.1. Long-run deflation? As typical price indexes, Figure 2-1 depicts the GDP deflator,

Uncovered Interest Parity and Monetary Policy Freedom in Countries with the Highest Degree of Financial Openness

Exch a nge r at e a nd fu n da m en ta l s:

R e. Y R, X R, u e, and. Use the attached excel spreadsheets to

The macroeconomic effects of fiscal policy in Greece

Output: The Demand for Goods and Services

Macroeconomics II A dynamic approach to short run economic fluctuations. The DAD/DAS model.

OPTIMUM FISCAL AND MONETARY POLICY USING THE MONETARY OVERLAPPING GENERATION MODELS

Empirical analysis on China money multiplier

Estimating Earnings Trend Using Unobserved Components Framework

Spring 2011 Social Sciences 7418 University of Wisconsin-Madison

Revisiting exchange rate puzzles

Subdivided Research on the Inflation-hedging Ability of Residential Property: A Case of Hong Kong

An Alternative Test of Purchasing Power Parity

Balance of Payments. Second quarter 2012

You should turn in (at least) FOUR bluebooks, one (or more, if needed) bluebook(s) for each question.

EVA NOPAT Capital charges ( = WACC * Invested Capital) = EVA [1 P] each

Documentation: Philadelphia Fed's Real-Time Data Set for Macroeconomists First-, Second-, and Third-Release Values

CENTRO DE ESTUDIOS MONETARIOS Y FINANCIEROS T. J. KEHOE MACROECONOMICS I WINTER 2011 PROBLEM SET #6

INSTITUTE OF ACTUARIES OF INDIA

Ch. 10 Measuring FX Exposure. Is Exchange Rate Risk Relevant? MNCs Take on FX Risk

Two ways to we learn the model

Comparison of back-testing results for various VaR estimation methods. Aleš Kresta, ICSP 2013, Bergamo 8 th July, 2013

ECONOMIC GROWTH. Student Assessment. Macroeconomics II. Class 1

General Equilibrium Perception on Twin Deficits Hypothesis: An Empirical Evidence for the U.S.

An Introduction to PAM Based Project Appraisal

VaR and Low Interest Rates

Supplement to Chapter 3

DYNAMIC ECONOMETRIC MODELS Vol. 7 Nicolaus Copernicus University Toruń Krzysztof Jajuga Wrocław University of Economics

Predictive Ability of Three Different Estimates of Cay to Excess Stock Returns A Comparative Study for South Africa and USA

Macroeconomics. Part 3 Macroeconomics of Financial Markets. Lecture 8 Investment: basic concepts

FORECASTING WITH A LINEX LOSS: A MONTE CARLO STUDY

International transmission of shocks:

*Corresponding author Keywords: CNH, Currency Intervention Index, Central Bank Reaction Function, Exchange Rate Intervention.

Problem Set 1 Answers. a. The computer is a final good produced and sold in Hence, 2006 GDP increases by $2,000.

Online Appendix. Using the reduced-form model notation proposed by Doshi, el al. (2013), 1. and Et

MONETARY POLICY IN MEXICO. Monetary Policy in Emerging Markets OECD and CCBS/Bank of England February 28, 2007

Macroeconomics. Typical macro questions (I) Typical macro questions (II) Methodology of macroeconomics. Tasks carried out by macroeconomists

MA Advanced Macro, 2016 (Karl Whelan) 1

Origins of currency swaps

Hedging Performance of Indonesia Exchange Rate

PRESS RELEASE EURO AREA ECONOMIC AND FINANCIAL DEVELOPMENTS BY INSTITUTIONAL SECTOR - FIRST QUARTER August 2012

Purchasing Power Parity (PPP) in the Long-Run: A Cointegration Approach. Md. Nisar Ahmed Shams * S. M. Woahid Murad **

This specification describes the models that are used to forecast

Financial Markets And Empirical Regularities An Introduction to Financial Econometrics

Volume 31, Issue 1. Pitfall of simple permanent income hypothesis model

International Review of Business Research Papers Vol. 4 No.3 June 2008 Pp Understanding Cross-Sectional Stock Returns: What Really Matters?

STATIONERY REQUIREMENTS SPECIAL REQUIREMENTS 20 Page booklet List of statistical formulae New Cambridge Elementary Statistical Tables

External balance assessment:

Watch out for the impact of Scottish independence opinion polls on UK s borrowing costs

Money, Income, Prices, and Causality in Pakistan: A Trivariate Analysis. Fazal Husain & Kalbe Abbas

CHAPTER CHAPTER26. Fiscal Policy: A Summing Up. Prepared by: Fernando Quijano and Yvonn Quijano

International Business And Economics Research Journal Volume 2, Number 10

DOES EVA REALLY HELP LONG TERM STOCK PERFORMANCE?

Reconciling Gross Output TFP Growth with Value Added TFP Growth

a. If Y is 1,000, M is 100, and the growth rate of nominal money is 1 percent, what must i and P be?

The Death of the Phillips Curve?

Price distortion induced by a flawed stock market index

The Mathematics Of Stock Option Valuation - Part Four Deriving The Black-Scholes Model Via Partial Differential Equations

MODELLING THE US SWAP SPREAD

Forecasting Performance of Alternative Error Correction Models

The Global Factor in Neutral Policy Rates

Problem 1 / 25 Problem 2 / 25 Problem 3 / 11 Problem 4 / 15 Problem 5 / 24 TOTAL / 100

An Analysis of Trend and Sources of Deficit Financing in Nepal

Stylized fact: high cyclical correlation of monetary aggregates and output

Ch. 1 Multinational Financial Mgmt: Overview. International Financial Environment. How Business Disciplines Are Used to Manage the MNC

Exam 1. Econ520. Spring 2017

Proposed changes to the compilation of zone aggregates for monetary aggregates and other variables in the OECD Main Economic Indicators publication

Suggested Template for Rolling Schemes for inclusion in the future price regulation of Dublin Airport

National saving and Fiscal Policy in South Africa: an Empirical Analysis. by Lumengo Bonga-Bonga University of Johannesburg

Capital Strength and Bank Profitability

Financial Econometrics Jeffrey R. Russell Midterm Winter 2011

Linkages and Performance Comparison among Eastern Europe Stock Markets

Banks, Credit Market Frictions, and Business Cycles

Principles of Finance CONTENTS

Measuring the Effects of Exchange Rate Changes on Investment in Australian Manufacturing Industry

The Empirical Study about Introduction of Stock Index Futures on the Volatility of Spot Market

Inventory Investment. Investment Decision and Expected Profit. Lecture 5

Risk Premium and Central Bank Intervention. Pınar Özlü

1. To express the production function in terms of output per worker and capital per worker, divide by N: K f N

Effective factors on velocity of money in Iran

ARE MALAYSIAN EXPORTS AND IMPORTS COINTEGRATED?

BUDGET ECONOMIC AND FISCAL POSITION REPORT

Transcription:

Empirical exchange rae models fi: Evidence from he Brazilian economy Marcelo L. Moura Ibmec São Paulo Rua Quaá 300, São Paulo-SP Brazil CEP:04546-042 Tel.: 4504-2435 marcelom@isp.edu.br Adauo R. S. Lima Ibmec São Paulo Rua Quaá 300, São Paulo-SP Brazil CEP:04546-042 Tel.: 4504-2388 adauorsl@ isp.edu.br THIS VERSION: APRIL 12, 2007

Empirical exchange rae models fi: Evidence from he Brazilian economy 2 ABSTRACT In his aricle, we es he adequacy and forecasing performance of empirical exchange rae models for an emerging commodiy exporer economy wih independenly floaing regime. In paricular, we sudy hose models using daa from he Brazilian economy. The esed economic models include he Flexible Price Moneary Model (FPMM) and is specificaion of he Asse Model, he Sicky Price Moneary Model (SPMM), he Porfolio Balance Model and he Marke Model based on real-ime informaion used in inernaional rade desks. Our main resul is o show ha, opposed o he resuls shown in he classic lieraure, some of our specificaions may forecas moves in he nominal exchange rae o a beer resul han ha of a drifless random walk. In paricular, he bes specificaions include variables ha capure he moneary policy (M1 and ineres rae), counry risk (EMBI) and erms of rade. Key-words: Empirical Models of Exchange Rae Deerminaion, Ou-of-sample forecasing, Emerging Economies. JEL Codes: F31, F41, F47. 2

3 1 Inroducion In he presen sudy, we es he adequacy of he empirical exchange rae models for an emerging commodiy-based economy wih independenly floaing regime 1. Our purpose is o assess he in-sample and ou-of sample fi of hose models. Our sraegy consiss of esimaing he main empirical models of exchange rae deerminaion, including conrols for risk percepion of economic agens, condiions of he exernal marke, inervenions of he Cenral Bank and oher effecs known in he lieraure. Our analysis replicaes for an emerging economy he sudy carried ou in he classic aricle by Meese and Rogoff (1983) bu wih a broader se of economic models. The original Meese and Rogoff work showed ha a simple drifless random walk model would be more effecive for he exchange rae forecasing han he models ha involve macroeconomic fundamenals. Meese and Rogoff s research has generaed an exensive lieraure. Mark (1995) argues ha he moneary fundamenals migh obain some success o explain he behavior of he exchange rae if he saisical ess were given more power. However, a hos of auhors, for example, Kilian (1999) and Giorgianni (2001) remained skepics and suggesed ha he resuls obained by Meese and Rogoff may sill seem robus, even afer all he daa and inense academic invesigaion gahered for over weny years. Some excepions o his skepicism are presen in recen works. Chen (2004) analyzes commodiies producers (Ausralia, Canada and New Zealand) for OCDE counries. The auhor concludes ha for Ausralia and New Zealand he global price of heir respecive expored commodiies is likely o have a meaningful and sable impac on heir respecive currencies. However, in he case of Canada, he evidence was less conclusive. Guo e Savikcas (2006) make use of variables ha reflec he agens expecaion owards he fuure behavior of he economic fundamenals, like he erm srucure of ineres raes, credi risk, and he idiosyncraic risk of he Unied Saes sock marke, among ohers. Their analysis sugges ha he idiosyncraic risk of hose asses forecas he American dollar s behavior facing he G7 s main currencies, and conclude ha he exchange rae does no follow a drifless random walk.. Cheung, Chinn and Pascual (2003) added oher models and elemens o he 1970s radiional specificaions, such as, he correlaion beween he exernal ne asse and he differenial of relaive produciviy in he radable goods secor beween counries (Balassa-Samuelson effec) a he deerminaion of he exchange rae. The auhors concluded ha, in line wih a grea par of he exising lieraure, i is very difficul o find empirical esimaions of srucural models ha may consisenly ouperform a random walk, having he MSE as basis of comparison. On he oher hand, he srucural models provide a beer forecasing for exchange rae movemens han ha provided by he random walk. Specific sudies for Brazil, like Muinhos, Alves and Riella (2003), sae ha he random walk is no he bes hypohesis o explain he behavior of he exchange rae in Brazil. Using daa from May 1999 o December 2001, he auhors conclude ha a model derived from he heory of uncovered ineres rae pariy capures he Brazilian exchange rae s behavior beer. This model akes ino consideraion he sovereign risk premium (in he sudy measured by he C-Bond spread, in relaion o Treasury Bills, as a variable in he specificaion of he uncovered pariy. Summing up, he exising lieraure allows us o draw some imporan conclusions. Firs of all, i is difficul o find empirical economic models ha consisenly ouperform a drifless random walk for he ou-of-sample esimaions. Second, he addiion of specific variables of some 1 This definiion follows he exchange rae arrangemens adoped by he IMF, and available a hp://www.imf.org/exernal/np/mfd/er/index.asp 3

economies, such as expor prices and counry-risk premium improve he performance of he economic models. Finally, economic variables ha have forward-looking componens may improve he resuls of he models for he ou-of-sample forecasing. The purpose and conribuion of his work is o carry ou a deailed sudy abou he empirical fi of hese exchange rae models o an emerging economy like he Brazilian economy. The following secion presens he economic models used in his work and he respecive in-sample esimaions. In he analysis of he in-sample resuls, we focus on he saisical significance of he esimaed coefficiens for he macroeconomic variables and wheher heir respecive signs are in line wih he ones expeced by he economic heory. In Secion 3, we analyze he forecasing performance of he esimaed models agains ha of he drifless random walk. We follow he mehodology suggesed by Cheung, Chinn and Pascual (2003), in which he assessmen crierion is he Mean Squared Error (MSE), and he saisic proposed by Diebold and Mariano (1995) measures is saisic significance. The las secion presens he conclusion of he sudy. 4 2 Specificaion of he models and esimaion 2.1 Specificaion of he models The Flexible-Price Moneary Model (FPMM) The Flexible Price Moneary Model perspecive was very represenaive in he 1970s when he floaing exchange raes were adoped by he main indusrialized economies, afer he emergence of he Breon Woods sysem in 1973. The FPMM assumes ha, in each counry, he equalizaion of currency supply and demand deermines he price level in each counry. Furhermore, relaive prices in each counry and exchange raes are relaed by he purchasing power pariy relaionship. In economeric erms, he FPMM o be esimaed could be presened by: ( m m ) + β2( y y ) + β4 ( i i ) s = β 0 + β1 + µ (1) Where s is he exchange rae logarihm (R$/US$), m and m * he M1 logarihms in Brazil and in he Unied Saes, respecively; y and y * he indusrial producion logarihm in boh counries and i and i * he logarihm for he swap ineres raes for a year in Brazil and in he Unied Saes, respecively 2. The variable µ is a random erm. The Sicky-Price Moneary Model (SPMM) Despie he fac ha he FPMM was he dominan approach o deermine he exchange rae in he early 1970s, is weak empirical resuls led o he concepion of models ha ook over fricions in he economy, inducing anoher form of convergence for long-run marke equilibrium. Dornbusch (1976) inroduces he idea of sicky prices in he shor run o he exchange models, which enables jumps in he nominal and/or real exchange rae o beyond is long-run equilibrium value. The exisence in he sysem of variables ha jump, in his specific case, he exchange rae and he ineres rae, would make up for he sickiness in oher variables, ha is, he prices of goods. Thus, he adjusmen velociy in various markes would be differen. Consider π and π * logarihms for he inflaion raes in Brazil and in he Unied Saes, which ry o capure price sickiness in boh economies, he Sicky-Price Moneary Model (SPMM) can be described by he following equaion: 2 (1+ pre swap ineres rae logarihm) was used for he domesic raes and he USA raes. 4

5 s ( m m ) + β 2( y y ) + β 4( i i ) + β5( π π ) ν = β 0 + β1 + (2) Where ν is a random erm. In order o adap he radiional models for an emerging economy in which commodiies play an imporan role in expors, we included oher variables ha are specific o models (1) and (2). In oher words, here will be he addiion of some variables ha capure he counry-risk premium, he evoluion of relevan prices for he rade balance, and a erm ha capures he relaive evoluion of produciviy in he radable goods secor beween Brazil and he USA. Using hese conrol variables 3, we can rewrie he above models: T ( m m ) + β ( y y ) + β ( i i ) + Γ j, β j s = β 0 + β1 2 4 + µ (3) s T ( m m ) + β 2 ( y y ) + β 4 ( i i ) + β 5 ( π π ) + Γ j, β j ν = β 0 + β1 + (4) Where T Γ j, would be he ransposed vecor of he conrol variables. The Porfolio Balance Model, Asse Model and Marke Model The moneary models formerly shown, flexible prices and sicky prices, assume he perfec subsiuion beween home and exernal asses and heir effecs on he exchange rae. However, he exisence of home-bias (home agens preference for home asses), liquidiy difference, solvency risk, ribuary differences and even he currency-exchange risk can affec he presumed equilibrium in he moneary models, which makes he home asses and he exernal asses imperfec subsiues. The Porfolio Balance Model assesses how his flawed subsiuion beween home and exernal asses can affec he agens demand for home and exernal asses. For he Porfolio Balance Model we made use of wo specificaions. The purpose is o capure he effecs of he Porfolio Balance Models o changes in he agens risk percepion, he aleraions in he inernaional marke counry-risk premium credi condiions and he exernal price effecs on he rade balance. In he case of he Brazilian economy, boh he Cenral Bank of Brazil and he Naional Treasury ac upon he exchange markes by means of dollar denominaed domesic deb insrumens or dollar-based derivaives, and ha is why we added he evoluion of he public secor foreign currency ne domesic liabiliies o capure he effec of hose acions on he nominal exchange rae. The wo specificaions differ from he imposiion or no of he PPP. Thus, he specificaion can be expressed as: s ( p p ) + β z + β CRB + β FCL + β ( i i ) + β EM + β DXR + β VIX = β 0 + 1 2 3 4 5 6 7 + η (5) Equaion (5) represens he model ha imposes PPP, while he specificaion ha assumes price sickiness could be expressed as: s ( i i ) + β EM + β DXR + β VIX + β ( π π ) = β + β z + β CRB + β FCL + β + η 0 1 2 3 4 5 6 7 8 (6) Where z is he logarihm of he produciviy differenial for he radable goods secor, CRB is he CRB baske, FCL is he ne foreign currency liabiliies, herein measured by he monhly resuls of he curren accoun, EM is he EMBI+ Brazil, DXR is he logarihm of he public secor dollar denominaed ne domesic liabiliies and VIX is he volailiy indicaor which aims o capure he changes in he inernaional invesors risk-aversion 4. The erm (p p *) indicaes ha PPP is valid a 3 The descripion of he daa is shown in Appendix 1. 4 See he definiion of hose indicaors in he appendix. 5

every insan of ime. All he variables are expressed in logarihm. Figure 1 shows he graphs of he variables used in he models. One FPMM varian, he Asse Model, was also esimaed, hrough he specificaion: 6 s ( m m ) + β ( y y ) + β EM + β TT ν = β 0 + β1 2 3 3 + (7) Where TT, is he Brazilian erms of rade. Finally we use a specificaion, called Marke Model, in which real ime variables are used and whose access is immediae o he raders, and ha somehow affec he marke operaors decision when buying and selling currencies. s ( i i ) + β EM + β CRB + β VIX + β HG ν = β 0 + β1 2 3 4 5 + (8) Where, HG is he CS High Yield Index II ha ries o capure changes in he credi scenario in he inernaional markes. Again he variables are in logarihm. 2.2 - Model esimaion In order o avoid possible spurious regressions, he models were esimaed assuming in firs differences he dependen and independen variables 5. A general expression for he relaion wih he exchange rae is: s = Χ Π + ε (9) Where X, is he vecor for he considered economic variables, ε is a random erm, and Π is he vecor for he esimaed coefficiens. The specificaion in firs difference would involve he following regression: s = Χ Π + υ (10) Where υ is a random erm. Since in he exchange models here may be group deerminaion of all variables presen in he equaion, i is jusifiable o make use of insrumenal variables ha would lead o consisency gains in he esimaed parameers. As he exchange rae variaion, as well as he variaion of some oher variables presen in he specificaions, does no show a normal disribuion, we used he Generalized Mehod of Momens (GMM) wih he Time Series Weighing Marix (HAC). The esimaors generaed by he GMM are robus, and opposie o hose obained by Maximum Likelihood, he GMM does no require he exac informaion from he disribuion of errors of he specified models 6. 5 According o Engle and Granger (1987), in case he non-saionary variables in heir levels have a long-run equilibrium relaion, ha is, in case hey co-inegrae, when specifying he models in firs difference, a specificaion misake will be occurring. However, he sample limiaion does no allow he use of he Vecor of Error Correcion (VEC). 6 As insrumens we used level variables wih wo and hree lags for each esimaion. In he Flexible and Sicky Prices Moneary Models, we have also included as insrumens, he shor-run ineres rae differenial beween he counries and he CS high Yield Index II Spread o Wors (HG). We used hose variables in level and wih no lags and wih, one, wo and hree lags. In he Porfolio Balance Model, besides he level variables wih wo and hree lags, we included he shorrun ineres rae differenial beween he counries, using he same lags as in he moneary models. In he Marke Model, he insrumens used were he level variables wih wo and hree lags and he shor-run ineres rae differenial beween counries wih no lags and wih, one, wo and hree lags. 6

Tables 1 hrough 5 conain he resuls from he various specificaions aking ino accoun he daa from he whole sample (March 1999 o December 2005) for he esimaions of he models 7. As a whole, he specific variables seem meaningful o explain exchange rae in-sample changes, and heir non-inclusion would generae specificaion problems of he models, in he case, problems wih omied variables, in line wih Meese (1990). FPMM Resuls Table 1 presens he FPMM resuls. In he original specificaion, equaion (1), moneary expansion and ineres rae are significan while he indusrial producion is no. The inclusion of conrols, specificaions (2) hrough (8) ends no o aler hose resuls. The moneary expansion difference beween he wo counries is no significan in only one of he models being he posiive sign of he esimaed coefficiens in line wih he FPMM heory. An increase in he moneary expansion implies increase in he price levels, which leads o he nominal exchange rae devaluaion since he model assumes ha PPP is valid a every insan of ime. The ineres rae differenial beween Brazil and he Unied Saes is posiive and significan a a 99% confidence inerval in six of he eigh specificaions esed. Despie he fac ha hese resuls are inuiively conradicory an increase in he home ineres rae, relaively o US inernaional raes, would lead o he devaluaion of he Brazilian Real hey are in line wih wha is expeced by he economic heory ha bases he model. In he FPMM, an increase in he home ineres rae would induce o a fall in money demand; mainained he fixed money supply, he price level mus go up o counerbalance he ineres rae increase, given he PPP, he home price increase devaluaes he exchange rae. The inclusion of conrol variables improves he original model fi, according o he Akaike informaion crierion (AIC). The EMBI+ Brazil is srongly significan in all he specificaions. The posiive sign is in line wih he expecaions ha he worsening in he solvency risk would induce o exchange rae devaluaion, reflecing he deerioraion of he economic agens expecaions abou he fuure fundamenals of he economy. The erms of rade variable is saisically significan in one of he wo specificaions in which i is included, and he negaive sign came in line wih wha is expeced. On he oher hand, CRB commodiies index is no saisically significan o he period in quesion. The produciviy differenial (Balassa-Samuelson) in he radable goods secor is significan in wo of he four specificaions. The negaive sign obained is in line wih he heory. SPMM Resuls The SPMM resuls are presened in Table 2. In he original model, specificaion (1), only he ineres rae differenials and hose of inflaion beween Brazil and he US are saisically significan. The addiion of conrols, specificaions (1) hrough (8) alers hese resuls, he inflaion rae loses is saisic significance and he moneary expansion, M1, is posiive and significan. 7 7 In he esimaion of he models we have also considered resric models in which he shor-run ineres rae differenial balanced wih he risk premium is given by βj ( i risk premium i*) and no as βj ( i i*) + βmembi, presened in he shown models. To avoid he problem of consrucing his variable because he EMBI+ Brazil could appear for overone-year duraion, we used he Brazil premium risk for a year. The resriced models, wih ineres differenial balanced by EMBI+ Brazil boh for in-sample and ou-of-sample were also esimaed, and he resuls obained were similar o hose obained by he resriced models using he one-year premium. In boh cases, he resriced models demonsraed resuls largely similar o he models presened in his aricle considering he sign and significance of he variables. Using he Akaike informaion crierion (AIC) o selec he models, we verified ha he same specificaions for he unresriced models generaed beer models han hose obained wih he resric specificaions. The daa are available upon reques. 7

The posiive and significan sign of he ineres rae is no in line wih wha is expeced in his model. Since he prices are sicky in he shor run, an increase in he nominal ineres raes implies he increase in he real ineres rae, which, in urn, aracs exernal capials and appreciaes he exchange rae. The negaive sign obained for he inflaion differenial in specificaion (1) is in line wih wha is anicipaed by he economic heory since he uncovered ineres rae pariy is valid. The differenial in he expansion of moneary demand is significan in mos of he esimaed models (7 ou of 8), wih he posiive sign for he coefficiens, in line wih he heory. However, he elasiciy obained is differen from he uniy, like some sudies sugges. As observed in he FPMM esimaions, he inclusion of conrol variables improves he SPMM specificaion. The counry-risk premium is again posiive and saisically significan in all he specificaions. The erms of rade variable is negaive, as expeced, and saisically significan. Variables ha capure effecs of he CRB index and he Balassa-Samuelson effec are no saisically significan in pracically all he cases. Porfolio Balance, Marke and Asse Models Resuls In relaion o he variables ha measure he risk percepion (VIX and EMBI) and he ineres rae, he resuls in erms of significance and sign are he same as hose from FPMM and SPMM models; herefore, he same inerpreaion is valid here. Likewise, he CRB index and he Balassa- Samuelson effec do no show robus resuls in erms of saisical significance. In he case of he Porfolio Balance Model, i is worh noing ha he PPP hypohesis, specificaion (2), differs lile from he specificaion ha does no assume he PPP, specificaion (1), providing non-conclusive answer wheher one model is in fac beer han he oher. The Brazilian ne exernal liabiliies, herein measured by he curren accoun variaion, are significan and posiive. Such resul is expeced by he heory given ha he changes in he curren accoun are associaed wih a relaive ransfer of wealh beween counries wih effecs on he exchange rae risk premium. The inflaion differenial beween counries is marginally significan and he negaive sign is in line wih wha is expeced by he heory. The dollar denominaed governmen liabiliy has a negaive sign and is saisically significan a 99% confidence inerval which implies ha he hedge operaions provided by he Cenral Bank of Brazil are effecive in he exchange rae conrol 8. In he Marke Model (Table 4), which uses marke real ime variables, he High Yield Spread is no significan in any of he esimaed models, very likely reflecing he high correlaion of his variable wih he EMBI+ Brazil (37%) and wih he VIX (48%), suggesing he presence of mulicolineariy in he specificaions, which makes he individual idenificaion of he effec of he variables difficul. In he Asse Model, which assumes ha oday he exchange rae reflecs he expeced presen discouned value of he economy s fuure fundamenals, only he produc differenial is no saisically significan and he signs esimaed for he oher variables were in line wih wha was anicipaed by he economic heory. 4 Ou-of-sample forecasing Our forecasing exercise used he ou-of-sample esimaion mehodology adoped by Meese and Rogoff (1983a) as a way o assess he performance of he models for he exchange rae forecasing. Iniially, we esimaed he specificaions for he period January 1999 hrough December 2001. Then, for each esimaed model, we made one-, hree-, six- and welve-monh projecions 8 8 These hedge operaions are characerized by he sale of dollar denominaed reasury bonds and swap operaions common policies adoped by he Cenral Bank of Brazil during he period sudied, and mainly a momens of grea urbulence and exchange rae speculaion. 8

ahead for he exchange rae level. A a second momen, we displaced, using he rolling regression mehod, he esimaion of he models one period ahead, keeping he size of he iniial sample. We repeaed he procedure o he exhausion of he sample. This procedure is hen compared wih he forecasing of a model ha assumes ha he exchange rae follows a drifless random walk. The ou-of-sample forecasing analysis followed he mehodology used by Cheung, Chinn e Pascual (2003). Firsly, we calculaed he raio beween he Mean Squared Error 9 (MSE) of each specificaion and he MSE of he random walk. To es he saisic significance of his raio, we used he saisic proposed by Diebold and Mariano (1995), in which, under null hypohesis, here is no difference beween he wo esimaions forecasing performance, ha is, he forecasing generaed by he economic models is as good as he forecasing generaed by a drifless random walk 10. Table 4 presens he raio beween he MSE of he economic models and he esimaions generaed by a drifless random walk for he various ou of sample periods: 1, 3, 6, and 12 monhs ahead. Thus, numbers inferior o one indicae ha he economic models ouperformed a drifless random walk for he ou-of-sample forecasing of he exchange rae n-periods ahead; numbers superior o one indicae ha he economic models underperformed a drifless random walk. Below he MSE raio, we find he p-value of he Diebold Mariano es. Under he null, we es he equaliy beween he loss funcions of he economic and drifless random-walk models, and, under he alernaive hypohesis, we es wheher he loss funcion of he economic model is inferior o ha generaed by a drifless random walk 11. For shor-run forecasing, 1 and 3 monhs ahead, he Asse Model is he only one ha ouperforms a drifless random walk a a forecas horizon of 3 monhs. For long-run forecasing horizons, 6 and 12 monhs ahead, he economic models ouperform he forecass obained by a random walk in 37,5% of he FPMM cases, 56% of he SPMM, 50% of he Marke Model, and 100% of he Asse Model. The Porfolio Model, however, is always ouperformed by he drifless random walk. 5 Conclusions The resuls of his sudy show ha he economic variables may explain he behavior of an independenly floaing exchange rae in an emerging and commodiy exporer economy like he Brazilian. The specificaions herein esimaed generaed resuls consisen wih hose forecased by he heoreical economic models, mainly when we considered moneary policy variables and used conrol variables ha capured he economic agens risk percepion and he erms of rade condiions for he home marke. As a whole, he resuls of he in-sample esimaions indicae ha variables ha measure he economic agens risk percepions, such as EMBI+, VIX, and Ne Exernal Liabiliies are saisically significan and posiive. As expeced, he worsening (improvemen) of he counry risk percepion leads o an exchange rae depreciaion (appreciaion). 9 9 MSE = L(y ) = E[(y^ -y ) 2 ]where y^ is he esimaed value and y is he observed value and E is he expeced value operaor. 10 Tesing wheher he economic models forecasing performance is differen from he forecas generaed by random walk is equivalen o esing wheher he loss series sample average d, is zero, having: d = L (y )-L(r ). The saisic of Diebol-Mariano es is given by: S = d /(LRV d ) 1/2. Where d = 1/T d and LRV is a consisen esimaor of he asympoic variance T 1/2 d. Diebold and Mariano (1995) show ha under null hypohesis boh models have he same forecasing power, S A ~ N(0,1). 11 For convenience, we have oped o presen only a his momen he p-value of Tes l. The complee resuls are available upon reques 9

The ineres rae is saisically significan and presens a posiive sign whenever included, which shows ha an increase in he ineres rae does no arac exernal capial so as o induce he exchange rae appreciaion. A possible inuiion is ha periods of increased ineres raes are associaed wih momens of inernal urbulence, and hus, correlaed wih he exchange rae devaluaion. Tha is precisely wha happened o he Brazilian economy in 1999, 2001 and 2002. Terms of rade, when included in he specificaion, are always saisically significan and show a negaive sign as expeced by he heory, ha is, improving he erms of rade ends o appreciae he exchange rae. However, in he case of he CRB index and he Balassa-Samuelson effec, hey do no show robus resuls, hey are no consisenly significan, and he sign alernaes. In line wih he argumens presened by Meese and Rogoff (1983a), in he sense ha he ouof-sample forecasing would be one of he imporan crieria for he assessmen of he empirical exchange models, we also esimaed he forecasing performance of hose models for shor-run horizons, 1 and 3 monhs, and long-run, 6 and 12 monhs. The specificaions which have he bes forecasing performance for boh 6- and 12-monh horizons are he FPMM (2), SPMM (2), (6), and (7), and finally he Asse Model. Considering only he saisically significan variables, he EMBI+ Brazil variable and M1 differenial are always presen in hese specificaions, and he ineres rae differenial and erms of rade are presen in mos cases. These resuls indicae ha he exchange rae in Brazil is linked wih economic facors and does no follow a random walk, corroboraing he analysis carried ou by Muinhos, Alves and Riella (2003). Our sudy indicaes ha forecasing he fuure behavior of he exchange rae for an emerging commodiy exporer economy wih independenly floaing regime like Brazil mus include macroeconomic fundamenals. In paricular, moneary policy variables, like ineres raes and M1, variables ha measure he risk percepion of he economic agens, like EMBI+ Brazil and VIX, and variables ha measure he exporing marke condiions, like erms of rade. In line wih he analysis carried ou by Obsfeld and Rogoff (1996), he exchange rae as well as he price of any asse reflecs he agens expecaions owards he behavior of oher variables. Fuure sudies should ry o es hese resuls in oher emerging economies. References BALASSA, B. The purchasing power pariy docrine: a reappraisal. Journal of Poliical Economy, v.72, p. 584-596, 1964. BERKOWITZ, J.; GIORGIANNI, L. Long-horizon exchange rae predicabiliy? Review of Economics and Saisics, v. 83, p. 81-91, 2001. BILSON, J.F.O. The moneary approach o he exchange rae: some empirical evidence, Inernaional Moneary Fund Saff Papers, v. 25, p. 48-75, 1978. BUITER, W.H.; MILLER. M. Moneary policy and inernaional compeiiveness: he problems of adjusmen. Oxford Economic Papers, v. 33, Supplemen, p. 143-175, 1981. CHEN, Y. Exchange raes and fundamenals: evidence from commodiy economies. Job Paper, Universiy of Washingon, Nov. 2004. Available a: <hp://faculy.washingon.edu/yuchin/papers/ner.pdf>. Accessed in: March 10 2006. CHEUNG, Y; CHINN, M.D.; PASCUAL, A.G. Empirical exchange rae models of he nineies: are any fi o survive?. Sana Cruz Deparmen of Economics, Working Paper Series, 1333, 2003. 10 10

Available a:: <hp://reposiories.cdlib.org/cgi/viewconen.cgi?aricle=1033&conex=ucscecon>. Assessed in: December 15 2005. CLARIDA, R.; TAYLOR, M. The erm srucure of forward exchange premiums and he forecasabiliy of spo exchange raes: correcing he errors. Review of Economics and Saisics, v. 79, p. 353-361, 1997. DIEBOLD, F. X.; MARIANO, M. Comparing Predicive Accuracy. Journal of Business and Economic Saisics, v. 13, p. 253-65, 1995. DORNBUSCH, R. Expecaion and exchange rae dynamics. Journal of Poliical Economy, v. 84, p. 1161-76, 1976.. Exchange rae economics: where do we sand? Brooking Papers on Economic Aciviy, 1, p. 143-85, 1980.. Moneary policy under exchange rae flexibiliy. In: BIGMAN, D.; TAYA, T. (eds.). Floaing exchange raes and he sae of World rade paymens. Cambridge, Massachuses: Harper & Row, Ballinger, 1984, p. 3-31. ENGEL, D.; GRANGER, C.W.J. Co-inegraion and error correcion: represenaion esimaion and esing. Economerica, v.55, p. 251-276, 1987. FRANKEL, J. A.; ENGEL, C.M. Do asse demand funcions opimize over he mean and variance of real reurns? A six-currency es. NBER Working Paper, n. 1051, 1982. FRENKEL, J. A. A Moneary approach o he exchange rae: docrinal aspecs and empirical evidence. Scandinavian Journal of Economics, v. 78, p. 200-224, 1976. GRANGER, C.W.J.; NEWBOLD, P. Spurious regression in economerics. Journal of Economerics, v.2, p. 111-120, July 1974. GUO, H.; SAVICKAS, R. Idiosyncraic volailiy, economic fundamenals, and foreign exchange raes. Federal Reserve Bank of S. Louis Working Paper Series, n. 2005-025B, May 2006. HANSEN, L. P. Large Sample Properies of Generalized Mehod of Momens Esimaors. Economerica, v. 50, p. 1029-54, 1982. HONG, Y.; LEE, T. Inference on predicabiliy of foreign exchange raes via generalized specrum an nonlinear ime series models. Review of Economic and Saisics, v. 85, p. 1048-62, 2003. HOOPER, P.; MORTON, J. Flucuaions in he Dollar: A Model of Nominal and Real Exchange Rae Deerminaion. Journal of Inernaional Money and Finance, v. 1, p. 39-56, 1982. KILIAN, L. Exchange raes and moneary fundamenals: evidence on long-horizon predicabiliy. Journal of Applied Economerics, v. 14, p. 491-510, 1999. LEWIS, K.K. Are foreign exchange inervenion and moneary policy relaed and does i really maer?. Journal of Business, v. 68, p. 109-127, 1995. MACKNINNON, J. G. Numerical Disribuion Funcion for Uni Roo Tess. Journal of Applied Economerics, v. 11, p. 601-618, 1996. 11 MARK, N. C. Exchange raes and fundamenals: evidence on long-horizon predicaabiliy. American Economic Review, v. 85, p. 201-218, 1995. 11

. Inernaional macroeconomics and finance: heory and economeric mehods. Malden, Massachuses: Blackwell Publishers Inc., 2001, 67-70 p. MARTIN, J. P.; MASSON, P. R. Exchange rae and porfolio balance. NBER Working Paper, n. 377, 1979. MEESE, R. Currency flucuaions in he Pos-Breon Woods era. The Journal of Economic Perspecive, v. 4, n. 1, p. 117-134, 1990. MEESE, R.; ROGOFF, K. The ou-of-sample failure of empirical exchange rae models: sampling error or misspecificaion?. In: FRENKEL, J., (ed.), Exchange Raes and Inernaional Macroeconomics. Chicago: Universiy of Chicago Press, 1983b, p. 67-105.. Empirical exchange rae models of he sevenies: do hey fi ou of he sample?. Journal of Finance, v. 43, p. 933-948, 1983.. Was i real? The exchange rae-ineres differenial relaion over he modern floaing-rae period. Journal of Finance, v. 43, p. 933-948, 1988. MEGALE, C. Faores exernos e o risco-país. Rio de Janeiro, 2003. 96 f. Disseração (Mesrado em Economia) PUC-Rio. MUINHOS, M. K.; ALVES, S. A. L.; RIELLA, G. Modelo macroeconômico com seor exerno: endogeneização do prêmio de risco e do câmbio. Pesquisa e Planejameno Econômico, v. 33, iss. 1, p. 61-89, abr. 2003. OBSTFELD, M.; ROGOFF, K. Foundaions of inernaional macroeconomics. Cambridge, Massachuses: MIT Press, 1996, cap. 8, 9, p. 529, 625. ROSSI, J. O modelo moneário de deerminação da axa de câmbio: Teses para o Brasil. Brasília: IPEA, 1995. p. 28, (Working Paper, n. 393). SAMUELSON, P.A. Theoreical Noes on Trade Problems. Review of Economics and Saisics, v. 46 (May), p. 145-146, 1964. SARNO, L.; TAYLOR, M.P. The economics of exchange raes. Cambridge: Cambridge Universiy Press, 2002, cap. 4, p. 104-107, 115-118. 12 12

13 TABLES Table 1 In-sample resuls: exchange rae changes Flexible Price Moneary Model and conrols (1) (2) (3) (4) (5) (6) (7) (8) Consan 0,010*** 0,006** 0,005* 0,007** 0,011*** 0,004 0,001 0,004 Produc 0,039 0,023 0,089 0,038 0,023-0,025-0,063-0,030 M1 0,246** 0,291*** 0,383*** 0,274*** 0,170 0,238** 0,239** 0,208** Ineres Raes 1,726*** 0,961*** 1,097*** 1,081*** 1,671*** -0,166-0,380 1,053*** EMBI+ Brazil.. 0,223*** 0,150*** 0,215***.. 0,302*** 0,295*** 0,249*** Terms of Trade.... -1,167***...... -0,118.. Commodiies Index - CRB...... 0,081...... 0,233 Balassa Samuelson Effec........ 0,212-0,457-0,631** -0,605*** Saisic J (1) 0,105 0,088 0,063 0,091 0,112 0,094 0,107 0,098 AIC -6.187-6.589-6.139-6.622-6.189-6.355-6.229-6.521 Noes: ***, ** and * denoe significance levels a 1%, 5% e 10%, respecively. (1) H0: he over idenificaion of he insrumens is saisfied. 13

14 Table 2 In-sample resuls: exchange rae changes Sicky Price Moneary Model and conrols (1) (2) (3) (4) (5) (6) (7) (8) Consan 0,000 0,004 0,005* 0,005 0,014*** 0,004 0,010* 0,004 Inflaion -1,209** -0,564-0,180-0,525 0,551-0,272 0,531-0,006 Produc -0,013 0,052 0,079 0,032 0,084 0,069 0,121 0,054 M1 0,202 0,312*** 0,363*** 0,296*** 0,259** 0,349*** 0,430*** 0,366*** Ineres Raes 1,486*** 1,028*** 1,077*** 1,018*** 1,616*** 0,846*** 1,257*** 0,913*** EMBI+ Brazil.. 0,219*** 0,153*** 0,215***.. 0,250*** 0,121** 0,248*** Terms of Trade.... -1,038***...... -1,096***.. Commodiies Index - CRB...... 0,010...... 0,096 Balassa Samuelson Effec........ 1,059*** 0,255 0,355 0,203 Saisic J (1) 0,129 0,105 0,083 0,104 0,113 0,137 0,082 0,129 AIC -6.146-6.591-6.190-6.562-6.034-6.512-6.051-6.477 Noes: ***, ** and * denoe significance levels a 1%, 5% e 10%, respecively. (1) H0: he over idenificaion of he insrumens is saisfied. 14

15 Table 3 In-sample resuls: exchange rae changes Composie Models Porfolio Model (1) Porfolio Model (2) Marke Model Asse Model Consan -0,001-0,007* 0,013*** -0.001 M1 0.444*** Produc 0.071 Balassa-Samuelson Effec -0,746*** 0,001 Commodiies Index - CRB 0,119 0,170-0,357* Terms of Trade -0.677* Ineres Raes 1,491*** 1,234*** 0,899*** Inflaion -1,257**.. Dollar Denominaed Governmen Liabiliy Ne Exernal Liabiliies (Curren Accoun adjused) -0,064** -0,081*** 1,246** 1,945*** VIX 0,060* 0,107*** 0,191*** EMBI+ Brazil 0,123*** 0,166*** 0,147** 0.228*** High Yield Spread -0,016 PPP.. 1 Saisic J (1) 0,103 0,148 0,082 0,104 AIC -6.711-6.585-6,090-6,103 Noes: ***, ** and * denoe significance levels a 1%, 5% e 10%, respecively. (1) (1) H0: he over idenificaion of he insrumens is saisfied. 15

16 Table 4 Ou-of-sample forecass. Models / Periods ahead (1) 1 monh 3 monhs 6 monhs 12 monhs FPMM original specificaion 1,089 0,646 0,987 0,485 0,944 0,434 1,127 0,700 + EMBI+ Brazil 1,115 0,679 0,863 0,374 0,733 0,168 0,705 0,053 + Terms of Trade 1,032 0,540 0,654 0,170 0,678 0,099 0,648 0,021 + CRB commodiies index 0,972 0,476 0,741 0,309 0,838 0,369 1,086 0,591 + Balassa-Samuelson effec 1,273 0,903 0,963 0,458 1,025 0,530 1,216 0,846 + Balassa + EMBI 0,895 0,408 0,719 0,284 0,647 0,144 0,601 0,008 + Balassa + EMBI + TT 0,756 0,254 0,724 0,285 0,805 0,292 0,667 0,023 + Balassa + EMBI + CRB 0,889 0,403 0,506 0,125 0,567 0,068 0,722 0,142 SPMM original specificaion 2,220 0,941 1,160 0,734 0,925 0,398 0,964 0,443 + EMBI+ Brazil 0,937 0,393 0,764 0,244 0,642 0,096 0,675 0,019 + Terms of Trade 0,968 0,452 0,612 0,120 0,670 0,124 0,721 0,041 + CRB commodiies index 0,870 0,366 0,718 0,261 0,630 0,081 0,978 0,453 + Balassa-Samuelson effec 1,050 0,588 0,850 0,326 0,909 0,384 1,191 0,697 + Balassa + EMBI 1,137 0,672 0,762 0,269 0,566 0,049 0,580 0,050 + Balassa + EMBI + TT 0,782 0,261 0,616 0,197 0,486 0,045 0,612 0,076 + Balassa + EMBI + CRB 1,045 0,540 0,843 0,379 0,563 0,076 0,804 0,276 16

17 Table 4 Coninuaion Ou-of-sample forecass. Models / Periods ahead (1) 1 monh 3 monhs 6 monhs 12 monhs Porfolio Model wihou PPP 1,315 0,700 1,523 0,778 2,429 0,962 6,129 0,993 Porfolio Model wih PPP 1,639 0,861 1,569 0,826 2,032 0,943 4,390 0,995 Marke Model 1,019 0,526 0,855 0,336 0,679 0,124 0,784 0,062 Asse Model 0.807 0,298 0.529 0,083 0.525 0,037 0.531 0,001 Noes: The able presens he ou-of-sample MSE raio beween he economic models forecas and he drifless random walk forecass. Values below uniy indicae ha he economic models had a beer forecas performance han he drifless random walk. The gray numbers below he raios indicae H0 p-values for he Diebold-Mariano es (DIEBOLD; MARIANO, 1995, p. 4). Shadowed boxes indicae ha he economic model has ouperformed he drifless random walk a a confidence inerval level of 90% or higher. (1) H0: MSE of he economic model = MSE of a drifless random-walk; H1: MSE of he economic model < MSE of a drifless random-walk. 17

18 Appendix Daa Descripion The daa cover he period from January 1999 o December 2005. The following series for he Brazilian price indexes were used: he IPCA, calculaed by IBGE was used as consumer inflaion rae measure, he IPA-DI, esimaed by FGV, as radable inflaion rae indicaor, and he IPCA non-radable inflaion rae series (IPCAn), calculaed by MCM (a Brazilian consuling firm), as non-radable inflaion rae proxy for Brazil. For he Unied Saes, he Consumer Price Index was used as he consumer price index, he Service CPI Less Energy Services (CPIn), as non-radable inflaion rae measure, and he Producer Price Index (PPI), as radable goods inflaion rae measure. The Bureau of Labor Saisics calculaed he US series. In all he cases, we used he original series wihou seasonal balance. The inflaion rae for boh counries was calculaed by he 12-monh logdifference in he consumer price index (IPCA and CPI). As produc proxy, given he absence of GDP monhly series in boh counries, he indusrial producion original series for Brazil and he Unied Saes, calculaed by IBGE and by he Bureau of Labor Saisics were respecively, used. The exchange rae (R$/US$) used refers o he las marke price a he end of he monh closing rae, obained a Bloomberg. The SELIC Rae and he FED Fund Rae were used as shor-run ineres raes for Brazil and he Unied Saes, respecively. However, given he differen basis for calculus beween raes in Brazil and in he Unied Saes, he following ransformaion was made in he SELIC rae series so ha i would be made linear and comparable o he FED Fund: ( ) ( 252 [ 1+ /100 ) 1] 1/ 252 SELICn = SELIC As long-run rae, he 1-year Pre-DI swap rae and he US Swap semi 30/360 1Y were used boh series obained a Bloomberg 13. The same procedure used for SELIC was replicaed a he Pre-DI swap so ha i would become linear. The Gross Governmen Deb daa abou he Brazilian GDP are provided by he Cenral Bank of Brazil (BCB); for he Unied Saes, he daa were obained a Bloomberg 12. The risk premium used was EMBI+ Brazil (Emerging Marke Bond Index Brazil) calculaed by JP Morgan, which measures he risk spread of he Brazilian sovereign exernal deb over a general risk-free bond, in he case, he Unied Saes Treasury. The CS High Yield Index II Spread o Wors (HG) was used and calculaed by Bank Credi Suisse 13. The HG includes corporae bonds considered below invesmen grade and reflecs he risk percepion of he marke credi. The VIX measures he implici volailiy of he prices in a baske of opions in he S&P 100 index, and shows he marke expecaions for he 30-day volailiy. I is 12 Bloomberg Code:.DBT%GDP Index 13 Bloomberg Code: DLJHSTW Index 18

19 published by he Chicago Board Opions Exchange (CBOE), and widely used as marke risk measure 14. The ne exernal liabiliies is no published monhly by he Cenral Bank of Brazil; hus, a June-2005 ne exernal liabiliies based series was buil and updaed wih June-2005 monhly curren accoun liquidiy, which is also provided by he BCB. The domesic dollar denominaed governmen deb daa, which also include he posiion in USD-SELIC swap of he Cenral Bank, are released by he Cenral Bank of Brazil and by he Naional Treasury. I s worh noing ha despie being measured in dollars, he dollar-dominaed bonds are issued and seled in naional currency, in his case, he Brazilian Real. The USD-SELIC swap is seled by he differenial of reurns beween he dollar Pax 19 and he SELIC Rae. Insead of using direcly he CRB15 as reference for he commodiies prices in he inernaional marke, a series was buil from he CRB segmened indexes, using differeniaed consideraions in order o approach his series o he Brazilian Expor baske. The following series were used: CRB Energy (10%), CRB Meal (30%), CRB Grains (40%), CRB Raw Indusrial (10%) e CRB Indusrial (10%)16. Daa relaed o erms of rade, consruced and released by FUNCEX were also used. For daa concerning he SELIC rae and FED Fund, Brazil 1-year Pre-DI swap rae and he US Swap, CRB and is segmens, CS High Yield Index II e EMBI+ Brazil, we used marke quoaion a he end of period (monhly). Given he inexisence of monhly produciviy daa boh in Brazil and in he Unied Saes, he following produciviy proxy relaed o he radable goods secor beween he wo counries was applied (Z)17: 14 The VIX index can be obained a Bloomberg hrough he code VIX Index. Furher deails on VIX can be obained a hp://www.srickne.com/vix.hm 15 The CRB is calculaed daily by he Commodiy Research Bureau, in USA, and comprises he price of 22 commodiies. CRB daa is available a hp://www.crbrader.com/crbindex/spo_background.asp. 16 The Bloomberg codes are respecively CRBFENRG Index, CRB METL Index, CRBFGRNS Index, CRBFINDU Index e CRB RIND Index. 17 We assume ha he radable goods secor produciviy index can be expressed by invering he radable goods price index in each counry ((1/IPA-DI e 1/PPI); he same procedure is used o obain he secor produciviy index of non-radables (1/IPCAn e 1/CPIn). Denominaing Zbr and Zus he relaive produciviy beween he radable goods and he non-radable goods in Brazil and in he Unied Saes, respecively, we have: Zbr = ( IPCAn / IPA DI) Zus = ( CPIn / PPI) Balassa-Samuelson assume similar produciviy for he non-radable goods for he secor, which leads o: Zbr / Zus = ( IPCAn / IPA DI) / CPIn / PPI ( ) in logarihm, he above expression can be rewrien as: Zbr / Zus = log IPCAn / IPA DI log CPIn / PPI = ( ) ( ) Z 19

20 Figure 1 Daa 2800 EMBI+ Brazil, basis poins 4.0 Exchange Rae BRL / US$ 2400 3.6 2000 3.2 1600 1200 2.8 800 2.4 400 2.0 0 1.6 25 Ineres raes, 1 year, differenial Brazil - US 4.8 Log of M1, Brazil/US 20 4.6 4.4 15 4.2 10 4.0 3.8 5 3.6 Log of Indusrial Producion Index, Brazil/US Log of Inflaion, Brazil/US.32.14.28.24.20.12.10.16.08.12.06.08.04.00.04.02 -.04.00 20

21 Figure 2 Daa 360 CRB Commodiies Index 98 Brazilian Terms of Trade, index number -0.5 Balassa-Samuelson, log 320 280 96 94-0.6-0.7-0.8 240 200 160 92 90 88 86-0.9-1.0-1.1-1.2-1.3 Log of Ne Exernal Liabiliies, Brazil 12.75 12.70 12.65 12.60 12.55 12.50 12.45 Dollar Denominaed Public Domesic Deb, US$ millions 90000 80000 70000 60000 50000 40000 30000 20000 10000 0 CS High Yield Index II Spread o Wors, basis poins 1100 1000 900 800 700 600 500 400 40 35 30 25 20 15 VIX, % p.y. 300 10 21