Real Exchange Rate Misalignment and Economic Growth : An Empirical Study for the Maghreb Countries

It has long been recognized in academic and policy debates that domestic policies play an important role in explaining economic growth. The paper investigates the role of real exchange rate (RER) misalignment on long-run growth in three countries of the Maghreb countries (Tunisia, Algeria and Morocco) over the period 1980-2008. We first estimate equilibrium RER relying on the Fundamental Equilibrium Exchange Rate (FEER) approach, from which misalignment is derived. Second, we estimate a dynamic panel growth model in which among the traditional determinants of growth, our measure of misalignment is included. The results indicate that the coefficient for RER misalignment is negative, which means that a more depreciated (appreciated) RER helps (harms) long-run growth. As a consequence, an appropriate exchange rate policy would close the gap between RER and its equilibrium level.


Introduction
The assessment of equilibrium values of the real exchange rate (RER) has always been an important issue in international macroeconomics, especially in the current context of global imbalances.Indeed, since the mid of the 1990s -the beginning of a period characterized by the increasing of emerging countries to global imbalances -the accelerating international financial integration process has engendered a growing disconnection between RER variability and growth (Béreau and al, 2009).
Within this context of growing international financial integration and global imbalances, it seems particularly interesting to focus on the impact of currency misalignments on growth sine persistent RER gaps are likely to affect the economic growth of countries.Indeed, the significant and persistent deviation of RER from equilibrium level, i.e., RER misalignment, could have implications on the balance of the economy.There is a vast theatrical and empirical literature that suggests that RER misalignment is one of the key indicators in identifying a country's economic vulnerability.As Kaminsky and al. (1998) underline, an overvaluation of the currencies is often the sign of the inconsistency of the decisions of macroeconomic policies that may lead to an unsustainable current account deficit, increasing external debt and the risk of possible speculative attacks.On the opposite, it is expect that RER undervaluation -which could be attributed to competitive devaluations -may drive the exchange rate to a level that encourages exports and promote growth.Consequently, an important question concerns the measure of misalignment that is the evaluation of equilibrium exchange rate.
The interest of studying the link between currency misalignment and growth is particularly notable for China.Chinese authorities have been frequently accused of maintaining the value of the yuan against major currencies at a very low level to finance China's spectacular growth, through the promotion of its exports.This export-led growth has generated surging Chinese current account surpluses, creating a major source of tension among trading partners who experienced important trade deficits with China (especially the United States and the European Union).The undervalued national currency (excess depreciation with respect to the equilibrium rate) has been the main elements of the recent growth success of China and other emerging Asian currencies.
Our aim in this paper is to investigate the relationship between RER misalignment and economic growth in three countries of the Maghreb countries (Tunisia, Morocco and Algeria) using panel data techniques during the period 1980-2008.One of the main empirical contributions of the paper is to test a model specification for the long-run equilibrium RER and then use these to obtain estimated RER misalignment and assess how robust the results are when they are included as an explanatory variable in the panel growth model.Here, we use the System Generalized Method of Moments (GMM) estimator for dynamic panels (Arellano and Bover, 1995;Blundell and Bond, 1998) to deal with the problems of unobserved country effects and endogenous regressors in a dynamic setting.
We organize the paper in 5 sections.Section 2 reviews the literature on fundamental RER equations and the impact of RER misalignment on growth.Section 3 deals with the empirical estimation of both equilibrium RER and currency misalignment.Section 4 estimates the relationship between economic growth and a set of explanatory variables, by paying a special attention to the impact of RER misalignment.Finally, section 5 concludes.

Review of the Literature
In the present section we briefly describe the empirical literature on two key issues for our analysis: (a) the measurement of RER misalignment, and (b) his impact on economic growth.

On the Measurement of RER Misalignment
In the present paper, the measurement of RER misalignment relies on the notion of the fundamental equilibrium exchange rate (FEER).This (called "equilibrium RER") was defined by Nurkse (1945) as the relative price that helps attain internal and external equilibrium simultaneously.Edwards and Savastano (1999) survey the literature on the measure of RER misalignment and they found that most empirical efforts can be classified: (a) single equation models and (b) general equilibrium simulation models.In both approaches the RER is defined as the relative price of traded and non-traded goods that achieves simultaneously external and internal equilibrium (Note 1).
The single-equation approach have usually derived reduced forms for the equilibrium RER from a wide variety of theoretical models and most of these efforts have been based on Edwards (1989) and Obstfeld andRogoff (1995, 1996).The general empirical approach is to relate the actual RER to that exchange rate that would be consistent with the medium term fundamentals driving the equilibrium exchange rate, such as fiscal policy and the terms of trade, government spending, trade policy, among other factors.Here, misalignment occurs when RER deviation from the equilibrium path is persistent.Misalignment could arise -among other factors -due to inadequate macroeconomic, trade and exchange rate policies.
We follow the single-equation approach in this paper.This approach consists of (a) estimating an equilibrium relationship between the RER and a set of fundamentals, (b) then using the coefficients and the medium-term values of the fundamentals to compute the equilibrium exchange rate, and (c) finally computing the exchange rate misalignment as the difference between the actual exchange rate and the equilibrium value.
The RER is a broad summary measure of the prices of one country to the price of another country or group of countries.It can generally be expressed as: RER = EP/P*, where P is the domestic price index, P* is the foreign price index and E is the nominal exchange rate (units of foreign currency per domestic currency).Note that our definition of RER implies that an increase (decrease) in RER denotes a real appreciation (depreciation) of the local currency.
We also use the annual real effective exchange rate (REER) defined as the annual index of domestic prices (consumer price index) for a country (i) toward the annual index of the prices of main trading partners, multiplied by the nominal exchange rate of the country (i).
We thus estimate the equilibrium REER equation from the model specified in Berg and Miao (2010): where subscripts i and t represent country and time indexes, respectively, and i  and t i,  are country-specific intercepts and disturbance terms.
prod , stands for the relative productivity in the traded-goods sector (relative to the non-traded goods one), t i govc , is the government consumption (as a share of GDP), t i invest , is the investment (as a share of GDP) and t i open , is the trade openness (Note 2).All variables are expressed in logarithm terms.
Equation (1) represents our fundamental long-run REER equation -the baseline equation for our estimation of the equilibrium REER -and has several testable predictions.First, according to the Balassa-Samuelson, if productivity in the tradables sector grows faster than in the non-tradables sector, the resulting higher wages in the tradables sector will put upward pressure on wages in the non-tradables sector, resulting in a higher relative price of non-tradables (i.e., a real appreciation).As productivity data by sectors are not available for a sufficient number of countries, we follow Coudert and Couharde (2008) in using a proxy given by the real GDP per capita.Second, an increase in openness should cause REER depreciation.Trade liberalization reduces the domestic prices of tradables causing a demand shift away from nontraded goods.Under some fairly reasonable cross price elasticities assumptions, nontradable prices should fall, producing a real depreciation.Following Tera and Valladares (2010), openness is proxied by the sum of exports and imports over GDP.Third, the expected signs on government consumption and investment are ambiguous, depending on the share of tradable goods in the relevant spending baskets.For example, if government spends relatively more non-tradable goods, an increase in government consumption should lead to an REER appreciation.
One of the reasons for finding the determinants of the REER is to be able to estimate his degree of misalignment.The misalignment in the REER corresponds to the difference between the observed and the equilibrium REER.However, computing equilibrium REER is not straightforward.Indeed, as mentioned by Arberola (2003), finding a long-run cointegration relationship between the REER and it determinants would yield an estimate of the equilibrium rate if we were able to observe the equilibrium level of the determinants.Therefore, to calculate the long-run equilibrium REER we need to isolate the permanent values of the macro fundamentals from their short-run fluctuations.
There are several procedures to filter or decompose macroeconomic time series.Here, we use the Hodrick and Prescott (HP) framework (1997) to obtain the permanent (equilibrium) components of the fundamental variables (Note 3).Indeed, the equilibrium REER is obtained by feeding the estimated model with the permanent components of the fundamentals (estimated with the HP filter) These permanent components are characterized as sustainable levels and are therefore consistent with the concept of equilibrium.The equilibrium REER is normalized (through the country-specific intercept) so that the long-run misalignment for each country is set equal to zero.
At each point in time, the RER misalignment is calculated as the difference between the observed REER and its predicted equilibrium value, that is, we compute: where it q ˆis the predicted REER value from equation (1).If the difference is positive (negative), we observed over (under) valuation of local currency.

Evidence on RER Misalignment and Economic Growth
The RER misalignment is a key macroeconomic policy variable, particularly in the case of developing countries, being used to predict future exchange rate shifts among floaters and to evaluate the need to adjust the exchange rate among countries with less flexible regimes.On the one hand, sustained exchange rate overvaluation could constitute a warning sign of adjustment of relative prices and a possible decline in the aggregate growth rate of the economy.On the other hand, since the RER fluctuations determine production and consumption choices between domestic and foreign goods, the RER misalignment could be used as a tool to influence the actual state of the economy.Thus, there were countries which had tried to maintain their currencies undervalued in order to stimulate growth through the channel of exports.
The literature on equilibrium RER goes back to the 1960s (Balassa, 1964) and the second half of the first decade of the new century has shown an increase in the number of empirical studies on RER misalignment and growth (Note 4).The literature on exchange rate misalignment has not reached a consensus in terms of how misalignment is measured, since part of the literature is based on deviations from PPP while other studies focus on the deviation of the RER from some equilibrium level.Another issue that is frequently examined in the literature on RER misalignment is the notion that overvaluation processes that last for a significant period of time are good indicators of possible currency crises (Frankel and Rose, 1996) and ultimately have an impact on relative price adjustment and create a negative correlation with growth.Razin and Collins (1997) investigate the relation between economic growth and RER misalignment considering that there are two possible channels through which RER misalignment might influence growth.First, it could influence domestic and foreign investment, by influencing the capital accumulation process which is a well known engine of growth.Second, a RER that is out of line could affect the tradables sector, and the competitiveness of this sector in respect of the rest of the world.In exploring the relationship between RER misalignment and economic growth, they found that while very high overvaluation appears to be associated with slower growth, moderate to high (but not very high) undervaluation appears to stimulate growth.In light of the above discussion, it can be argued that RER misalignment can distort price signals, result in misallocation of resources across sectors, and generate a negative impact on growth.
Rodrik ( 2008) is one of the recent studies on RER misalignment and growth, with estimation results for a set of 184 countries and time series data from 1950 to 2004.The author develops an index to measure the degree of RER undervaluation adjusted for the Balassa-Samuelson effect using real per capita GDP data.The main empirical result is that overvaluation hurts growth, undervaluation facilitates it.For most countries, high growth periods are associated with undervalued currencies.In fact, there is a little evidence of non-linearity in the relationship between a country's RER and its economic growth.An increase in undervaluation simulates economic growth just as well as a decrease in overvaluation.The magnitude and statistical significance of the estimated coefficient for RER undervaluation is higher for developing countries due to the fact that such countries are often characterized by institutional fragility and market failures (Note 5).
Berg and Miao ( 2010) develop an empirical investigation on RER misalignment and growth in order to compare the results with Rodrik ( 2008) and what they call the Washington Consensus (WC) view, which is based on a fundamental equilibrium exchange rate model (Note 6).Their main result is that WC and the Rodrik views of the role of misalignment in growth are observationally equivalent for the main growth regressions but there are some identification problems since the determinants of RER misalignment are also likely to be explanatory variables in the growth regression.The empirical findings support those from Rodrik (2008) in the senses that not only are overvaluations bad but undervaluations are also good for growth, a result that it is not consistent with the WC view.
Eichengreen ( 2008) develops a historical review of the literature on RER and growth, focusing attention on possible mechanism through which a competitive RER fosters growth.Avoiding real overvaluation may simply encourage the optimally balanced growth of traded -and nontraded -goods producing sectors.Alternatively, there may be nonpecuniary externalities associated with the production of exportables (learning by doing effects external to the firm) that do not exist to the same degree in other activities -meaning that market forces, left to their own devices, may produce a RER that is too high.The main policy recommendation therefore is for such countries is to keep their RER at a competitive level and with lower volatility since they are mainly useful for jump-starting growth based on development experiences, such as the high growth East Asian economies (Note 7).
The work developed by Aguirre and Calderon (2005) is among those using a measure of RER misalignments as deviations of actual exchange rates from their equilibrium for 60 countries over 1965-2003 using panel and time series cointegration methods.Using dynamic panel data techniques they find that RER misalignments hinder growth but the effect is non-linear: growth declines are larger, the larger the size of the misalignments.Although large undervaluations hurt growth, small to moderate undervaluations enhance growth.These results are robust when controlling for movements in the equilibrium RER.Hausmann and al. (2005) also recognize potential non-linearities in the relationship between growth and RER misalignments for eighty episodes when growth accelerates by at least two percentage points and that acceleration lasts for at least eight years.Their main empirical finding is that RER depreciation is one of the factors associated with the occurrence of such growth accelerating episodes.Gala and Lucinda (2006) developed a dynamic panel data analysis using Difference and System GMM techniques, for a set of 58 countries from 1960 to 1999, with a measure of RER misalignment incorporating the Balassa-Samuelson effect and other control variables for the growth regression such as physical and human capital, institutional environment, inflation, the output gap and terms of trade shocks.The main empirical evidence supports the argument that a real depreciated (appreciated) exchange rate is associated to higher (lower) growth rates.
One of the main contributions of our empirical estimates in the next section is to extend the determinants of RER including not only differences in per capita income but also the government consumption, openness and investment.In order to measure RER misalignment we then subtract the actual RER from its estimated value.The main purpose of this transformation is to investigate the role of RER misalignment in our growth model, based on the System GMM estimation.

Estimating Equilibrium RER and its Misalignment
The present section attempts to describe the econometric methods used to estimate the equilibrium REER and its misalignment for 3 countries of the Maghreb countries (Tunisia, Algeria and Morocco).To estimate equation ( 1), we use annual data over the period 1980-2008.
The econometric methodology used in this paper is based on panel unit root and cointegration tests.First, we test for unit root in various series.Second, we test for cointegration between the real effective exchange rate and the underlying macroeconomic fundamentals.Finally, we estimate the long-run parameters that we later use for computing the real equilibrium exchange rate and the corresponding misalignment.

Panel unit root tests
To test for the presence of unit roots on panel data, we use the Im, Pesaran and Shin ( 2003) -IPS thereafter-.IPS using the likelihood framework, suggest a new more flexible and computationally simple unit root testing procedure for panels (which is referred as bar t  statistic), that allows for simultaneous stationary and non-stationary series.Moreover, this test allows for residual serial correlation and heterogeneity of the dynamics and error variances across groups.The IPS test is based on the estimation of the following equation: where T is the number of observations over time, N denotes the number of individual members in the panel and t m d , contains deterministic variables.The null hypothesis is defined as 0 : and the alternative hypothesis is 0 : that allows for some (but not all) of individual series to have unit roots.
IPS ( 2003) compute separate unit root test for the N cross-section units and define their bar t  statistic as a simple average of the individual ADF statistics, iT t , for the null as: As indicated in table1, the tests of panel unit root of according to IPS (2003) confirm that all variables are nonstationary in levels but stationary in first differences.We now test for the existence of a long-run relationship between the real effective exchange rate and its determinants.

Cointegration tests
Pedroni (1999,2004) proposes a residual-based test for the null of cointegration for dynamic panels with multiple regressors in which the short-run dynamics and the long-run slope coefficients are permitted to be heterogeneous across individuals.The test allows for individual heterogeneous fixed effects and trend terms and no exogeneity requirements are imposed on the regressors on the cointegrating regressions.
Specially, the tests ask for the residuals estimation from static cointegrating long-run relation for a time series panel of observables it y : where as usual T is the number of observations over time and N is the number of units in the panel.It is possible to interpret the model (3) as N different equations, each of which has K regressors.The variables it y and it x are assumed to be I(1), for each member i of the panel, and under the null of no cointegration the residual it e will also be I(1).i  and i  are scalars denoting fixed effects and unit-specific linear trend parameters, respectively and i  are the cointegration slopes; note that all this coefficients are permitted to vary across individuals, so that considerable heterogeneity is allowed by this specification.
Pedroni considers the use of seven residual-based panel cointegration statistics, four based on pooling the data along the within-dimension (denoted 'panel cointegration statistics') and three based on pooling along the between-dimension (denoted 'group mean cointegration statistics').
Another distinction between the two sets of test is based on the alternative hypothesis specification.In fact, even if both sets of test verify the null hypothesis of no cointegration: 1 :  is the autoregressive coefficient of estimated residuals under the alternative hypothesis ( ), alternative hypothesis specification is different: -the panel cointegration statistics impose a common coefficient under the alternative hypothesis which results:

H
, i  -the group mean cointegration statistics allow for heterogeneous coefficients under the alternative hypothesis and it results: It is straightforward to observe that the first category of four statistics includes a type of non -parametric variance ratio statistic, a panel version of a non-parametric Phillips and Perron (1988)  -statistic, a non-parametric form of the average of the Phillips and Perron t -statistic and an ADF type t -statistic.
The second category of panel cointegration statistics is based on a group mean approach and includes a Phillips and Perron type  -statistic, a Phillips and Perron type t -statistic and an ADF type t -statistic.The comparative advantage of each of these statistics will depend on the underlying data-generating process.
After the calculation of the panel cointegration test statistics the appropriate mean and variance adjustment terms are applied, so that the test statistics are asymptotically standard normally distributed:  2 in Pedroni (1999) (Note 8).Pedroni (2004) explored finite sample performances of the seven statistics.He showed that in terms of power all the proposed statistics do fairly well for T > 100.Moreover Pedroni's (1997) simulations showed that for small time span (T < 20), the between dimension (group t-statistic) is the most powerful.Given our relatively short time span (T = 29), we will pay a particular attention to the group parametric-t statistic ( stat ADF  ) when testing for cointegration.The result of panel cointegration tests are displayed in table 2.
Since simulations made by Pedroni (2004) show that, in small samples, the group-mean parametric-test is more powerful than the other tests, we can conclude that the null hypothesis of no cointegration is rejected in our study, and now turn to the estimation of the long run relationship between the REER and its determinants.

Equilibrium RER and misalignment
As revealed from panel unit root and cointegration tests, our series are integrated of order 1 and cointegrated.It is thus possible to proceed to the estimation of the long-run relationship (1).To this end, we rely on the Fully-Modified Ordinary Least Squares (FMOLS) methodology pioneered by Pedroni (1999Pedroni ( , 2004)).In this sense, the advantage of the FMOLS estimation procedure over other techniques such as the Pooled Mean Group (PMG) method proposed by Pesaran and al. (1999) and the Dynamic Ordinary Least Squares (DOLS) method developed by Kao and Chiang (2000) is that, while slope homogeneity is imposed, short-run heterogeneity is allowed for each member of the panel.The cointegration vector obtained is displayed in table 3.
The results from the panel cointegration estimation (Table 3) appear consistent with the theoretical and empirical literature (Note 9).All long-run coefficient estimates are highly significant (at 1% significance level), displaying expected signs according to theory.
In addition to the statistical significance of our parameters, we are interested in their economic impact, especially when thinking about the effects of alternative policies on the REER.In particular, these results show that the productivity differential contributes to long term REER variations in the Maghreb region.Indeed, a 10% in the domestic productivity of tradables relative to non-tradables (relative to the corresponding variable for trading partner countries) tends to appreciate a country's equilibrium REER by about 4.4%.The government consumption coefficient is positive and statistically significant.Indeed, a positive shock on public consumption engenders a long-term REER appreciation that confirms our expectation that a rise of global demand of non-tradable goods leads to increase in prices.An increase in government consumption is associated with an appreciation of the REER.A 10% increase in government spending to GDP ratio will appreciate the REER by 4.4%.An increase in the investment (as % of GDP) of 10% is associated with an appreciation of the equilibrium REER of more than 11%.Negative coefficient corresponding to the variable of trade opening indicates that commercial liberalization will cause an REER depreciation of 3%.
Using our estimates, we compute the measure of misalignment ( it MIS ) as the deviations of the REER from its equilibrium level, where the latter is obtained by feeding the estimated model with the permanent components of the fundamentals (estimated with the Hodrick-Prescott (1997) filter).These permanent components are characterized as sustainable levels and are therefore consistent with the concept of equilibrium.
Figure 1 presents the evolution of RER misalignment in the 3 selected countries of the region.We observe an alternation between the episodes of overvaluation and undervaluation during the period of study.The determination of RER misalignment by our model confirms this evolution for the panel countries.There are persevering and recurring episodes of misalignments.Thus, persistent misalignment in the RER sometimes can be considered as an indicator of potential crisis, with disastrous consequences on the economy.For this reason, development strategies should include efforts to preserve as long as possible the REER at a near the equilibrium regardless of the exchange rate regime.
Given these misalignments series, let us now investigate their impact on the economic performance of the different countries.

Growth and RER Misalignment
Having introduced macroeconomic fundamentals for calculating RER misalignment, we are now in position to investigate the impact of RER misalignment on the economic growth by adopting the System GMM dynamic panel estimation method.

Econometric Methodology
To investigate the impact of RER misalignment on economic growth, we add misalignment among explanatory variables in our growth regression.Following Berg and Miao (2010), we estimate the following variation of the standard growth regression: where t i y , is the log of real GDP per capita, t i X , is a vector of contemporaneous and lagged values of growth determinants expressed in logarithm terms, t i MIS , denotes RER misalignment, i  represents unobserved country-specific factors and t  is a period specific effect.The time-specific effect, t  , allows to control for international conditions that change over time and affect the growth performance of countries in the sample, while i  accounts for unobserved country-specific factors that both drive growth and are potentially correlated with the explanatory variables.
Following Berg and Miao (2010) we retain various usual determinants.According to the neoclassical growth theory, the economic growth rate is a function of the initial position of the economy.The conditional convergence hypothesis states that, other things being equal, countries with lower GDP per capita are expected to grow more due to higher marginal returns on capital stock.We account for the initial position of the economy through the initial level of real GDP per capita to control for conditional convergence (see Barro and Sala-i Martin (1996) among others).Relying on some developments of the endogenous growth theory, we include determinants reflecting trade policies, macroeconomic stabilization policies and institutions.Among those potential determinants, we consider the following variables: (i) trade openness (in percentage of GDP), (ii) government consumption (in percentage of GDP), used as an indicator of fiscal policy, (iii) investment (in percentage of GDP) and (iv) the terms of trade.Finally, to these usual determinants, we add RER misalignment in order to investigate the impact of exchange rate overvaluation and undervaluation on economic growth.

Estimation technique
Our estimation technique addresses issues of endogeneity and unobserved country characteristics.Therefore, to account for endogeneity and country-specific unobserved characteristics, we use the System GMM dynamic panel estimation method.The option to use System GMM is based on the argument that the existence of weak instruments implies asymptotically that the variance of the coefficient increases and in small samples the coefficients can be biased.
To reduce the potential bias and inaccuracy associated with the use of Difference GMM (Arellano and Bond, 1991), Arellano and Bover (1995) and Blundell and Bond (1998) develop a system of regressions in differences and levels.
The instruments for the regression in differences are the lagged levels of the explanatory variables and the instruments for the regression in levels are the lagged differences of explanatory variables.These are considered as appropriate instruments under the assumption that although there may be correlation between the levels of explanatory variables and the country specific effect, there is no correlation between those variables in differences and the country specific effect.
The consistency of the System GMM estimator is assessed by two specification tests.The Sargan test of over identifying restrictions tests the overall validity of the instruments.Failure to reject the null hypothesis gives support to the model.The second test examines the null hypothesis that the error term is not serially correlated.Again, failure to reject the null hypothesis gives support to the model.

Empirical results
In Table 4 we report our regression estimates using the System GMM estimation technique.Before we describe our results, we should mention that the specification tests -both the Sargan test of over-identifying restrictions and the test for higher order correlation -validate our regressions for inference.That is, our instruments are not correlated with the error term and the latter does not display higher order serial correlation.
Let us first comment the results relating to the control variables.All the explanatory variables have the expected sign, whatever the sign and the size of the misalignment.The initial GDP per capita coefficient is negative, meaning that the conditional convergence hypothesis is evidenced: holding constant other growth determinants, countries with lower GDP per capita tend to grow faster.The initial position of the economy is thus a significant determinant of growth, as recognised by the neoclassical theory.The investment variable has also the right sign since there exists a positive relationship between capital accumulation and growth.Trade openness also positively affects growth.Thus, the more countries are outward-oriented the more this contributes favorably to economic growth.These results are in line with those found by Cottani and al. (1990), Aguire and Calderon (2005) and Dufrénot and al. (2009), and, more generally with the neoclassical approach according to which the positive impact of trade on growth is explained by comparative advantages, be they in resource endowment or differences in technology (see Béreau and al. 2009).The terms of trade (Note 10), which capture both changes in international demand for a country's export and the cost of production, are positive and statistically insignificant over the period 1980-2008.Government consumption enters negatively and none significantly, although, as underlined by Toulaboe (2006), there seems to be a consensus that consistent and increasing government balance can hinder economic growth.
Turning now to our main variable of interest, we find that there is a negative and significant relationship between growth and RER misalignment.This result implies that growth would decline in response to increases in the RER misalignment.On the other hand, a similar increase in the REER overvaluation (say, 10 %) would imply a growth decline of approximately 0.4 percentage points.This result is consistent with those of Rodrik ( 2008), Berg and Miao (2010), Aguirre and Calderon (2005), Gala and Lucinda (2006) and Eichengreen (2008) in the sense that an undervalued REER is beneficial for long-run growth, while the opposite is true for an overvalued REER.
The crucial policy recommendation to stem from our work, which is especially relevant for Maghreb countries, is that such countries should avoid periods of long lasting REER appreciation and instead adopt economic policies that are able to keep the REER at a competitive level, which most of the time should be associated with a more depreciated REER relative to its equilibrium level.

Conclusion
This paper explores the relationship between RER misalignment and economic growth in three countries of the Maghreb region (Tunisia, Algeria and Morocco) over the period 1980-2008.As RER misalignment is not observable, equilibrium exchange rate have been estimated relying on the FEER methodology.Misalignment series are then obtained by the deviation of the observed REER from its equilibrium level.We have then assessed their impact on economic growth using dynamic panel data techniques in order to address both the issue of unobserved country-specific effects and the possibility of endogenous regressors.Our empirical estimation of the System GMM panel growth model has shown that estimated coefficient for RER misalignment is negative and statistically significant, which means that a more real depreciated exchange rate helps real GDP growth while the opposite is true for a REER appreciation.The estimated coefficient of RER misalignment suggests that a 10% increase (appreciation) in RER misalignment can reduce annual per capita GDP growth by 0.4%.This result highlights that countries that pursue major and appropriate exchange rate reforms to reduce RER misalignment are very likely to record gains in real per capita GDP.In other words, it should be relevant for countries, especially Maghreb countries, to maintain their REER at its appropriate level.

Notes
Note 1. Edwards (1989) defines internal equilibrium as the sustainable equilibrium in the market of non-traded goods -which is compatible with the unemployment rate at its natural level.External equilibrium occurs when the current account position can be financed with sustainable capital flows -that is, when the intertemporal budget constraint is satisfied.
Note 2. See appendix for data definitions and sources.
Note 3. In general, time series are viewed as the sum of transitory and permanent components, and the HP filter captures the smooth path of the trend component by minimizing the sun of squares of its second difference.
Note 5. Rodrik (2008) incorporates other variables in the growth models (panel and cross-section regressions), including: lagged growth, initial income level (convergence), institutions (Rule of Law), government consumption, terms of trade, inflation, gross domestic saving, years of education, time and country dummies.
Note 6.The first measure of RER misalignment ( ppp it  ) is the same as in Rodrik (2008), using real per capita GDP to capture the Balassa-Samuelson effect, while the second measure ( FEER it


) is based on the FEER view and incorporates additional variables (terms of trade, openness, investment and government consumption).Note 7. See Aghion and al. (2006) on RER volatility and factor productivity, which is different from the impact on factor accumulation (growth).The authors found that countries with a significant degree of RER variability experience slower productivity growth and the magnitude of such is negatively associated with the degree of financial development.
the seven statistics of Pedroni,  and v are the functions of moments of the underlying Brownian motion functionals.The appropriate mean and variance adjustment terms for different number of regressors and different panel cointegration test statistics are given in Table statistic, the values of the mean and the variance have been computed via Monte Carlo methods for different values of T and i p 's and tabulated byIPS (2003).The results of each one of our five variables are reported in table1, where all the tests have a unit root under the null hypothesis.
Evolution of RER misalignments in Maghreb countries (1980-2008)Appendix.Definitions and Sources of Variables Used in Regression Analyses Gross fixed capital formation as a % of GDP Sum of exports and imports of goods and services as a % of GDP Ratio of export to import prices (2000=100) Log difference of real GDP per capita Initial value of ratio of total real GDP to total population Difference between real effective exchange rate and its estimated equilibrium value