An Estimation of the Equilibrium Multilateral Real Exchange Rate of Argentina: 1975-2005 María Lorena Garegnani and Guillermo J. Escudé 1
Banco Central de la República Argentina
Abstract This paper contains an empirical analysis of Argentina’s Multilateral Real Exchange Rate (MRER) during 1973-2004 using an equilibrium correction model. It shows that the U.S.A.’s MRER (measured by the Federal Reserve’s Real Broad Dollar Index) and Argentina’s terms of trade are good long and short run determinants of Argentina’s MRER. Whereas MRER misalignments were not signi…cantly corrected at all during the periods of unilateral pegs to the U.S. dollar, during the rest of the sample there was a correction of around 24% during the …rst quarter. Furthermore, the paper shows that the level of the rate of growth of the MRER is signi…cantly a¤ected by whether or not there was a unilateral peg to the dollar. Key words: JEL: C22, E41.
Introduction
Argentina has been an amazing laboratory of exchange regimes over the last 30 years. Its experience has shown that extended periods of pegging to the U.S. dollar, when they coincide with major real appreciations of the U.S. dollar, can have devastating e¤ects on Argentina´s economy. This is precisely what happened at least twice in the last 30 years: …rst during the period of crawling pegs to the U.S. dollar (the “Tablita” experience) in the late 70s and, second, during the more protracted Convertibility experience of the 90s. During both episodes the inability of unilateral pegs to a single currency to 1
Research Department, Central Bank of Argentina. The opinions expressed in this paper are the authors’and do not necessarily re‡ect those of the Central Bank of Argentina. Mailing addresses:
[email protected],
[email protected].
Preprint submitted to Elsevier Science
21 June 2005
facilitate the adjustment of the economy to external shocks (among them the real appreciation of the U.S. dollar), led to the accumulation of major macroeconomic imbalances such as lack of competitiveness due to a signi…cant real exchange rate misalignment, high unemployment and an explosive growth of public debt. Signi…cantly, both episodes ended in severe triple crises (debt, currency and banking), and both during periods of real dollar appreciations (see Escudé (2004a)). This paper delves into the data to look for the main driving forces behind Argentina’s Multilateral Real Exchange Rate (MRER). After presenting a condensed version of the basic model, the paper focuses on the empirical estimation of an equilibrium correction model for Argentina’s MRER during the period 1975:1-2004:1, using quarterly data. This method identi…es a set of fundamental variables that determine internal and external equilibrium and then links them to the multilateral real exchange rate (MRER). The cointegration analysis is made using the system-based procedure from Johansen (1988) and Johansen and Juselius (1990). This method has the advantage of determining the long-run relationship between the real exchange rate and its determinants that is valid in the long-run although there could be large deviations in the short run. We …nd that the MRER of the U.S.A. and Argentina’s Terms of Trade (TT) are highly signi…cant in explaining Argentina’s MRER and can be used in a very parsimonious model that does an excellent job of tracking the path of the latter and prima facie can be used to obtain a measure of real exchange rate misalignment. We …nd that when a unilateral peg to the U.S. dollar does not prevail, around a quarter of the misalignment is corrected in the …rst quarter. However, during the periods with unilateral pegs to the U.S. dollar, there is no signi…cant correction of the misalignment. Furthermore, the level of the rate of growth of the MRER is signi…cantly lower (meaning that the rate of real peso deprecation is lower, or the rate of real peso appreciation higher) when there are such unilateral pegs. We know that much more needs to be done to have a reliable measure of real exchange rate misalignment for Argentina. In particular, the lack of reliable and su¢ ciently long macroeconomic time series and the recurrence of crises, regime changes and structural breaks makes it an awesome task. Hence, we take the results of this paper as a very preliminary but nevertheless promissory step in that direction. The next section brie‡y describes the theoretical model we have in mind. Section 3 presents a description of the data and the econometric results. Section 4 shows graphically how the model tracks the MRER and the resulting misalignment of the MRER from its (preliminary) long run determinants, and section 5 concludes. 2
1
A synthesis of a plausible theoretical model
The estimation of exchange rate misalignment is one of the most challenging empirical problems in open-economy macroeconomics (Edwards (1989), MacDonald (1997), Hinkle and Montiel (1999), Ho¤mann and MacDonald (2000), among others). The main di¢ culty is that the equilibrium MRER is not observable. Ba¤es, Elbadawi and O´Connell (1999) perform a single-equation procedure for estimating both the equilibrium real exchange rate and the degree of misalignment, based on a theoretical model by Montiel (1999), and apply it to Côte d’Ivoire and Burkina Fasso. In this paper we use that methodology as a baseline, and adapt it to some of the peculiarities of Argentina’s experience with exchange regimes based on a model by Escudé (2005). The model we have in mind represents a medium run, which abstracts from investment, growth and changes in the multilateral structure of the small open economy’s (SOE) international trade, as we specify below. There are three goods: an exportable good (which is also consumed domestically), imported inputs to production, and non tradables (i.e. goods that are produced domestically and only consumed domestically). Firms in the non tradable sector are monopolistically competitive. In both sectors output is produced with labor (which is mobile between sectors and immobile internationally) and capital (which is bolted down and does not depreciate). The SOE is assumed to be a price taker for exportable and imported goods. Firms in the exportable sector are perfectly competitive and have a (constant) imported input requirement per unit of output. For simplicity, the Law of One Price (LOP) prevails for exports and imports, with full and instantaneous pass-through of nominal depreciations to peso prices. This simpli…cation, however, is ameliorated by our assumption of a multilateral non-commodity trade environment (with the U.S.A. and Europe as trade partners) where in the model’s long run the LOP does not hold for the goods that the U.S.A. and Europe produce and trade, for any of the multiple reasons that have been used to justify this fact, which include pricing to market and local currency pricing (Obstfeld (2002), Engel (2002)). Due to non tradable price stickiness, this makes the U.S.A.’s MRER a key fundamental for the SOE’s MRER, along with its TT. Households are assumed to consume exportable and non-tradable goods (or services), which are also the two categories of goods produced domestically. Hence, it is convenient to de…ne the MRER as the relative price between exportable (X) and non-tradable (N) goods. Given our assumption on the LOP and full and immediate pass-through for the SOE, we can de…ne the MRER as e ( S m )=PN , where S m is the multilateral nominal exchange rate (pesos per a geometrically trade weighted basket of currencies), is the 3
geometrically trade weighted basket of export price indexes, and PN is the peso price of non-tradables. In (at least) two of the subperiods we consider below, Argentina has had a unilaterally …xed (during the Convertibility period) or almost …xed (the crawling peg regime of the Tablita period) of the peso to the U.S. dollar 2 . Let St be the peso/dollar nominal exchange rate, and assume that a signi…cant fraction of trade ( EU ) is done with the Euro area and the rest ( U S = 1EU ) with the U.S.A., and that these coe¢ cients hold both for exports and imports. Furthermore, these coe¢ cients are constant in the model’s long run (which, hence, is actually a medium run). The MRER (e) can be de…ned as a geometrically weighted average of bilateral real exchange rates (…rst equality), or equivalently, as a ratio between the multilateral nominal exchange rate and the non-tradables price index (second equality): e=
SP U S PN
!
US
(S= )P EU PN
!
EU
=
S= PN
where P U S and P EU are the price indexes of the U.S.A. and Europe, exogenous euro/dollar nominal exchange rate, ( )
EU
(1)
US
( )
(1) is the
EU
is the exogenous trade weighted basket of foreign currencies per dollar nominal exchange rate ("dollar strength"), S= = S
US
(S= )
EU
is the country’s multilateral nominal exchange rate (that we represented as S m previously), and (PtU S ) U S (PtEU ) EU t is the export price index. will also be our TT because, to simplify, we assume that there is no in‡ation in import prices and that the multilateral import price index is normalized to one. Hence, our TT is basically the index of the dollar and euro invoiced prices of the goods that the U.S.A. and Europe, respectively, import from Argentina. Under our de…nition of the MRER, e is the relevant relative price for output decisions as well as consumption decisions. If the consumption sub-utility function has a Cobb-Douglas speci…cation for the consumption of exportable and non-tradable goods, the (dual) Consumer 2
We also evaluate below the inclusion of the recent period in which the nominal exchange rate with the dollar has been ‡uctuating within a small band or 2.9 pesos/dollar with strong daily Central Bank sterilized intervention in the foreign exchange market within the unilaterally …xed (or almost …xed) exchange rate period.
4
Price Index is also a Cobb-Douglas index of these goods’prices: P
( S= ) (PN )1 ;
(2)
where 0< 0; Tc < 0?; T < 0; T > 0; T < 0; Tg < 0? 8
(15)
In this case we have two partial derivatives with ambiguous sign. We shall assume that the direct e¤ects through the demand for exportables predominates over the indirect e¤ect through the product wage in the exportable sector. Under this assumption, inserting T (e; c) and (4) in (10) yields the External Balance equation with a positive slope in the e-c plane, as depicted in Figure 1: T (e; c) =
1
!
1 p
1
!
1= 1+r
(1 + ) +
1
1
r
!
R0 :
We can now easily see the e¤ects of permanent changes of the mean values of exogenous variables on the long run equilibrium values of the MRER and private consumption. For example, a permanent dollar strengthening (increase in ) has the e¤ect of shifting the XB line leftward (which is what the minus sign underneath this variable means in the graph). This makes the steady state MRER increase and private consumption of private goods decline. The reason, of course, is that the strengthening of the dollar makes the trade weighted trade surplus decline in dollar terms. To restore the dollar trade surplus, the MRER must increase, given c, or c must decline, given e. In fact, both e rises and c falls, the relative magnitude of these changes being given by slope of the Internal Balance line.
An increase in the TT ( ) shifts XB to the right due to the increased trade balance and IB to the left because the increased labor demand in the exportable sector makes w increase. This makes e fall, and has an ambiguous e¤ect on c. An opening of the economy based on higher import requirements for the exportable sector has the opposite e¤ects of an increase in the terms 9
of trade: it decreases the trade balance due to increased imports and reduces the demand for labor in the exportable sector. Hence, e increases. An increase in the international dollar interest rate or in the exogenous component of the risk premium increases the cost of the public foreign debt, generating a reduction in the steady state debt level (4) and, hence, in the net interest payments abroad. Hence, the trade balance can fall, shifting XB to the right and making e fall and c increase. An increase in the provision of public goods (increase in public expenditures) increases the demand for non-tradables and hence the demand for labor in the non-tradable sector. This reduces the product wage in this sector w, according to (12). Therefore, labor demand increases while labor supply falls because, given e, the real wage falls. Both e¤ects shift IB to the left. The increase in government demand for exportables reduces the trade balance, thus requiring a higher MRER, given c, or requiring lower private consumption of exportables c, given c. Of course, the e¤ect on welfare is in general ambiguous and depends on the utility of the consumption of public goods relative to the utility of the consumption of private goods. Although we have not formalized it here (see Escudé (2005)) an across the board increase in productivity (zF ) shifts both lines to the right, increasing c with an ambiguous e¤ect on e, and an exogenous increase in the participation rate (or willingness to work zH ) increases e and c by shifting IB to the right. In the empirical part of this paper, below, the MRER of the U.S.A. ( ) is shown to play an important role as an explanatory variable for Argentina’s MRER. Our interpretation hinges around this stylized model. The U.S. multilateral real exchange rate typically presents long phases of appreciation and depreciation. Hence, when the strong dollar phase begins it is probable that the appreciation will get gradually more pronounced and that this will persist during a number of years. Producers in a country that pegs its currency to the dollar but only has a small fraction of its trade with the U.S.A. hence …nd it increasingly di¢ cult to compete domestically with imported goods or in foreign markets unless their increase in productivity is su¢ ciently fast to compensate for the real appreciation (or trade and/or …scal policy is/are especially geared to compensate for this). But the speed and persistence of real dollar appreciations are too high to make compensating increases in productivity (or changes in trade or …scal policy) a realistic possibility. The importance of “dollar strength”is highlighted by the Argentine experience during the last two periods in which it pegged to the dollar: the “Tablita” episode of 1979-81 and the extended Convertibility period (1991-2001). In both periods, international dollar appreciations combined with a domestic predetermined exchange rate regime that pegged the peso to the dollar, and a 10
vulnerable process of …nancial liberalization or expansion, ended in an abrupt triple crisis (debt, currency and banking). In the above model, a real dollar appreciation has the impact e¤ect of reducing Argentina’s MRER. In a world with considerable price and wage stickiness, a predetermined (and unilateral) nominal exchange rate that is pegged to the dollar has the e¤ect of preventing a timely correction of the resulting real exchange rate misalignment. This is also true for the e¤ects of other external shocks, such as those that arise from the TT, the international interest rate, or the risk premium that international investors charge for foreign debts. As we show below, more ‡exible exchange regimes are better suited for preventing the persistence of misalignments.
2
Empirical results
The traditional approach for verifying the existence of long-run relationships among a group of variables is based on cointegration techniques. These techniques require the study of the time-series properties of the data. Our data set begins in the …rst quarter of 1975 and ends in the …rst quarter of 2005. The MRER (denominated rexchrate below), corresponds to the trade weighted exchange rate with the three main trade partners of Argentina: the U.S.A., Brazil and the E.U. The macro variables that have been considered as potential fundamental determinants of the MRER are: (i) the Terms of Trade, (tt), which is the ratio between the foreign currency price of exports and the foreign currency price of imports; (ii) di¤erential productivity, (prod), which is the di¤erence in labor productivity between Argentina and the U.S.A.; (iii) the MRER of the U.S.A., as measured by the Federal Reserve’s Real Broad Dollar index (rbdi); (iv) the government expenditure/Gross Domestic Product (GDP) ratio; (v) industrial wages; (vi) the trade balance/GDP ratio and (vii) the imports/GDP ratio. The starting point of the econometric study is the analysis of the time-series properties of the variables. We analyze the order of integration of the data using the Augmented Dickey Fuller (ADF) test, a standard unit root test. These test have a null hypothesis of non-stationarity against the alternative of stationarity. All variables are integrated of order 1, since we can’t reject the null hypothesis of the unit-root tests at traditional signi…cance levels (see Appendix 1). Given this, the cointegration analysis is made using the systembased procedure from Johansen (1988) and Johansen and Juselius (1990). In order to validate a conditional model of the MRER on the macro variables of the system, “weak exogeneity” is also evaluated by means of the Likelihood Ratio (LR) tests. 11
Although the system included all the variables previously described 5 , only three of them were found to have a long-run relationship in the sample period. After simpli…cations, the empirical results are presented in Table 1. Inspecting the eigenvalues and their associated statistics (Maximum and Trace) a rank equal to zero is rejected in favor of a rank equal to one. That is, there is one cointegration vector. The MRER, the TT and the Real Broad Dollar Index have a long-run (cointegration) relationship with coe¢ cients (1, 1.3, 3.9). Di¤erential productivity, the industrial wage, the trade balance/GDP ratio and the imports/GDP ratio were also included but no long-run relationship with them was found. 6 Table 1 The Cointegration Analysis of the rexchrate, tt and rbdi System
Cointegration analysis 1976 (2) to 2005 (1)(6 lags and d822;d82;d844;d881;d892;d911;d021 and constant unrestricted). eigenvalue 0.264313 0.0963049 0.0306444 Ho:rank=p p == 0 p