Journal Issue

Republic of Madagascar: Selected Issues and Statistical Appendix

International Monetary Fund
Published Date:
September 2005
  • ShareShare
Show Summary Details

III. Competitiveness in Madagascar: An Assessment of the Real Effective Exchange Rate8

A. Introduction

1. The objective of this chapter is to estimate a fundamental real effective exchange rate (FEER) based on long-run fundamentals and estimate the short-term factors that explain deviations from the real effective exchange rate (REER) from the FEER, allowing us to estimate the behavioral effective exchange rate (BEER). At any point in time and, indeed, for extended periods of time, the actual real effective exchange rate may deviate from its long-run level because of policy variables or short-term shocks. It is therefore important to compare the real exchange rate from the FEER and the BEER.

5. We find that at end-2004, following the sharp nominal depreciation of the currency, the real effective exchange is below its long-run fundamental effective exchange rate (determined by the terms of trade, relative productivity, and the net wealth of the country) and to the BEER (mostly determined by short-term factors or exogenous shocks, fiscal policy, and trade agreements). Controlling for these short-term factors, the scope of the discrepancy between the real exchange rate and the BEER remains significant. However, since end-2004, it is likely that part of the real depreciation has been reversed following the combination of a stable nominal exchange rate and the high level of inflation in the first months of 2005.

6. The section takes a long-term approach in the estimation of the FEER. The FEER is assumed to be driven by productivity relative to partner countries, the net foreign asset position as a measure of wealth of the country, and the terms of trade. Positive long-run productivity and wealth shocks are expected to appreciate the FEER. Productivity shocks will tend to reduce the price of tradable goods and increase demand for nontraded goods, therefore leading to an appreciation in the FEER. Permanent shocks in the nation’s wealth will also tend to appreciate the equilibrium exchange rate by increasing the demand for nontradable goods. The effect of the terms of trade is ambiguous. While the income effect will tend to appreciate the effective exchange, the substitution effect will tend to depreciate it. In developing countries, the literature finds that the income effect dominates.9

7. Potential factors behind deviations of the REER from the FEER are monetary and fiscal policy indicators, productivity, and terms of trade shocks. We also control for the exchange rate policy, Madagascar experienced episodes of fixed and crawling pegs and, more recently, a floating exchange rate. The exchange rate regime can be divided into three periods: from 1982 to early 1991, the authorities operated in a de facto crawling peg regime, from 1991 to the beginning of 1994, the country was under a de facto fixed exchange rate regime, and from mid-1994 to the present, the economy has operated under a floating exchange rate regime. We show that deviations from the FEER depend on the exchange rate regime.

8. In Section II, we will describe the data set and the methodology used in the analysis. In Section III, we will present the main findings.

B. Data and Methodology

Data set

9. We use quarterly data from 1980 to 2003. The variables used in estimating the long-run FEER are relative productivity, net wealth, and terms of trade. For productivity, we construct a relative productivity index based on output per worker in Madagascar and an international productivity index (source: World Economic Outlook). Wealth is measured by the ratio of net foreign assets to nominal GDP. Terms of trade are represented by the price of Madagascar exports divided by the price of imports into Madagascar.

10. A set of other variables is used to explain the deviations of the REER from the FEER. Monetary policy is measured by the ratio of reserve money to GDP, public expenditures (as a share of GDP) is a measure of government consumption. We include two dummies, one to capture the African Growth Opportunity Act (AGOA) and its favorable impact on export competitiveness, another one to capture the large scaled tax exemptions on imported goods. This set of variables is used to determine the short-term dynamic and explain their respective contribution to deviations of the real effective exchange rate from its FEER.


11. We first test all variables to determine their unit root properties. Variables found to be stationary will not be used in the estimation of the long-run relationship of the real effective exchange rate. Second, the long-run REER equation is estimated:

where X is a vector of explanatory variables. The next stage is to use the estimated coefficients α to compute the long-run FEER using the Hodrik-Prescott filtered explanatory variables. The idea is to decomposeX into its long-term trend component X and the short-term shocks components X. FEER is then determined according to

12. Deviation is then defined as the difference between the actual real exchange rate and the FEER.

13. The next stage consists in explaining the main determinants of deviatios, using as explanatory variables shocks to their long-term components X~ and a set of other short-term shocks Y. The following equation will be estimated:

C. Results

Estimating the long-run equilibrium real effective exchange rate (REER)

14. Most variables have a unit root, with the notable exception of the (log of) the ratio of government expenditure to GDP and rreserve money to GDP (see appendix).

15. We estimate Equation.(1), using as explanatory variables relative productivity, net foreign assets, and the terms of trade. Testing for stationarity in the residuals, following Engle-Granger (1987), we accept the hypothesis of a long-run relationship between these variables; that is, the four variables are cointegrated. Results are summarized in Table 2.

16. All coefficients have the expected sign, suggesting that the long-run REER is positively explained by productivity and net foreign assets. The coefficient of the terms of trade variable is negative, suggesting that the substitution effect dominates the revenue effect; that is, in the presence of a positive shock on the terms of trade, consumption will switch away from nontradable goods toward tradable goods. This result is robust to different specifications. Cady (2003) finds that terms of trade are not significant. In a recent study on real exchange rate misalignments and economic performance, Aguirre and Calderon (2005) find a negative relationship between the terms of trade and the real exchange rate for 20 countries (out of sample of 60). It is also possible that inappropriate exchange rate intervention, as well as structural changes in the exchange rate regime, contributes to this negative coefficient. To that effect, in the next subsection we control for structural changes in the conduct of exchange rate policy, and we show that the response of the real exchange rate to terms of trade shocks is dependent on the exchange rate regime.

Chapter II. Table 2.OLS Estimates of a Long-Run Relationship
Dependent variables: log (reerq)
Model 1
Log (produc)0.370***
Log (tot_weo)-0.693***
R-square adj.0.59
Number of observations80
Notes: (***) represents significance at the 1 percent level.

17. Using the estimated coefficients of Model 1, we compute the equilibrium FEER, using Equation (2). That is, we apply the model on the Hodrik-Prescott-filtered explanatory variables. Alternative filtering methods could be applied, but the Hodrik-Prescott is the most appropriate to extract long-run trends in the variable. In Figure 1, we compare the FEER to the actual REER. Assuming the long-term components of productivity, net foreign asstes (NFA) and the terms-of trade do not change in 2004, the REER is about 20 percent below its FEER at end-2004. Based on the long-run fundamentals, the real effective exchange rate is below its fundamental level. We need to determine to what extent this deviation is explained by policy or exogenous factors.

Chapter II. Figure 1.Fundamental Effective Exchange Rate (FEER) and the Real Effective Exchange Rate (REER)

18. It is important to stress again that this large deviation at end-2004 is in terms of the fundamental and long-run factors. The long-run equilibrium rate has been relatively stable since 1993, reflecting declining terms of trade and improved NFA position, compensated by a declining productivity.

19. Deviations of the real effective exchange rate to the long run level would depend on the temporary shocks and on monetary and fiscal policies. These factors could explain a misalignment in terms of both the extent of the deviation and its duration. In the next subsection, we estimate the impacts of policy as well as exogenous factors on deviations.

Estimating determinants of REER misalignments

20. We analyze the factors explaining deviations of REER from the FEER. Using the computed deviations, we can now estimate Equation (3), using shocks on productivity, money supply, and the terms of trade as the main explanatory variables. We also add public expenditures, a dummy capturing the AGO A, given the the strong response of Malagasy exports to it facilitated access to the U.S. market; and a dummy capturing the broad-based tax exemptions on investment and other kind of goods.

21. Our sample period covers episodes of different exchange rate regimes, which can affect the response of the exchange rate to various expenditure or monetary shocks. To this end, we add interaction variables to evaluate the impact of various variables under alternative exchange rate regimes. We include interactive variables, of the type x*float and x*peg, which represent the value of variable x under a floating exchange rate regime and a fixed peg respectively. Madagascar switched to a floating exchange rate regime in 1994; before that, it experienced crawling and fixed pegs. Results are summarized in Table 3.

22. Variables that are found to affect real effective deviations significantly are productivity shocks, monetary shocks, NFA shocks, public expenditure, and the AGOA dummy (model 1). Productivity shocks yield a real appreciation of the currency. Monetary policy acts in two stages: a monetary expansion policy leads to a real appreciation of the currency followed by a real depreciation. This results is puzzling, because one would first expect a depreciation in the nominal exchange rate followed by a real appreciation as inflation picks up. One possible explanation could be that the exchange rate has been heavily managed and the authorities have prevented the nominal exchange rate freely, therefore delaying the expected nominal depreciation. The AGOA leads to a real appreciation of the currency through its positive and significant impacts on export competitiveness in the textile industry. We prefer to use AGAO as a short-term factor rather than a long-term one: even though it is set to expire in 2015, Madagascar benefited mostly from its third party provision (AGOA III), which will expire in 2007.

23. Short-term terms of trade shocks have no impact on the misalignment. One explanation, as mentioned in the introduction, might be related to the important structural changes the economy experienced, including the move to a floating exchange rate regime in 1994. We capture these structural changes in models 2–3, and we conclude that the real effective exchange rate response to terms of trade shocks depends on the exchange rate regime.

24. Public expenditures have a negative but not very significant impact on the real exchange rate, while the prior would be that consumption of nontradables would increase, as would their prices. More research is needed to determine the composition of public expenditure. It is possible that in developing countries there is no fundamental difference between public expenditures and monetary policy response because fiscal expansions are often associated with an accommodating monetary policy. 10 This result could also reflect that increases in public expenditure concentrated more on imported goods than on nontradable goods. In support of this view, the ratio of the wage bill to GDP has been well contained and even declined throughout the period.

25. Model 2 decomposes the terms of trade shocks into three variables, capturing the interaction of terms of trade changes with the exchange rate regimes. The impacts of terms of trade shocks depend on the nature of the exchange rate regime. Under a fixed peg, the terms of trade tend to have a positive impact on the deviation; that is, it appreciates the REER, while under the floating exchange rate regime, the contribution of terms of trade shocks to the deviation is negative.

26. In Model 3, we see that the response to monetary policy is invariant across exchange rate regimes, suggesting that deviations from the long-run real effective exchange rate respond in the same way. In the short run, both under a fixed and a flexible peg, expansionary monetary policy yields to a real appreciation of the currency.

Chapter II. Table 3.OLS Estimates of deviations from the long-run REER equilibriumDependent variables : deviation from the long-run REER
Model 1Model 2Model 3Model 4Model 5Model 6
Productivity shock0.176 **0.378 **0.355 *0.179 **0.387 **0.368 *
Terms of trade shock-0.086-0.086
Terms of trade* peg1.7991.8571.8591.886
Terms of trade* crawl-0.142-0.130-0.141-0.132
Terms of trade* float-0.186 *-0.186 *-0.193 *-0.193 *
Monetary shock0.127**0.169 **0.129 **0.174 **
Monetary shock* peg0.173 **0.181 **
Monetary shock* crawl0.1660.180
Monetary shock* float0.166 **0.176 **
Monetary shock (-2)-0.171 ***-0.217 ***-0.217 ***-0.177 ***-0.226 ***-0.224 ***
NFA shock (-1)0.003 ***0.002 ***0.002 ***0.003 ***0.002 ***0.002 ***
NFA shock (-1)*peg
NFA shock (-1)*float
Gov. expenditure-0.098 *-0.127 *-0.129-0.086 *-0.110 *-0.103
AGOA0.091 ***0.110 ***0.113 ***0.099 ***0.123 ***0.125 ***
Tax exemption-0.049-0.070-0.069
Deviation (-1)0.581 ***0.426 ***0.435 ***0.582 ***0.423 ***0.428 ***
Constant0.355 **0.448 **0.4600.327 **0.409 **0.375
R-square adj.0.750.790.780.750.790.79
Number of observations797979797979
Notes:(1) ***, **, and * denotes the significance at the 1, 5, and 10 percent level respectively.(2) Peg and Float are dummies to account for the nature of the exchange rate regime, pegged will take a value of 1 during periods the country is not under a floating exchange rate regime.

27. In models 4 to 6, we add a dummy capturing the tax exemptions granted on imported investment (and some consumer) goods. The tax exemption led to a large increase in imports and to the depreciation of the currency up to a point where demand and supply for foreign exchange are equal. As expected, the sign is negative although not significant.

28. To sum up, in the short term, misalignments are mainly driven by productivity shocks, short-term monetary expansions, or shocks on the net wealth of the country. However, the dynamics depend strongly on the nature of the exchange rate regime. It should be noted that our exchange rate regime dummies are probably capturing other elements, such as the liberalization of the economy and the move toward a more market-oriented economy, which coincided with the move toward a flexible exchange rate regime. The AGOA variable is strongly significant and suggests that the removal of the access would trigger a sharp depreciation of the currency in the short term, away from its fundamental level. Contributions of policy variables, such as the ratio of reserve money to GDP and public expenditures, are significant.

29. Based on the estimation of models 1 to 3, we can estimate the behavioral real effective exchange rate (BEER)—that is, the real effective exchange rate explained by the FEER and the factors described above. We derive the BEER as

30. We can determine the contribution of the variables in the deviation equation in explaining the difference between the FEER and the REER. Figure 2 compares the FEER, BEER and the actual REER, with data expanded to include 2004 preliminary estimates.

Chapter II. Figure 2.Behavioral and Fundamental Real Effective Exchange Rate

31. Misalignment can be defined as the difference between the REER and the BEER; that is this is the deviation of the real effective exchange rate not accounted for by the explanatory variables in the deviation equations. At end-2004, the real exchange rate is around 27 percent below the BEER. Controlling for the tax exemption, the discrepancy is reduced to below 20 percent (Figure 3).

Chapter II. Figure 3.Behavioral and Fundamental Real Effective Exchange Rate

(with tax exemption)

32. An interesting exercise it to estimate the BEER by simulating the expiration of AGOA in 2004. Doing so, we find that the REER is very close to its BEER (Figure 4). This suggests that, in the near future, the expiration of AGOA III, which benefited Madagascar most, carries with it the risk of a possible depreciation of the real exchange rate.

Chapter II. Figure 4.BEER and the AGOA Simulation of an expiry of AGOA in 2004

33. To sum up, the depreciation of the real exchange rate was rapid and of a sizable amount. The study indicates that the current real effective exchange rate is below its long-term equilibrium. Several factors explain this discrepancy: monetary policy, tax exemption, and terms of trade shock. Other factors tend to mitigate the extent of the discrepancy. Since end-2004, part of the real depreciation has been reversed following the combination of a stable nominal exchange rate and the high level of inflation in the first months of 2005. This suggests that the level of real effective exchange rate is broadly appropriate.


Unit Root Tests

It is essential to determine the existence of a unit root test before estimating a long-run equation for the REER. If a variable is stationary, no long term co-movement can exist among the variables. Different types of tests were used to examine the unit root property. In addition to the standard ADF test, which has the unit root process as the null hypothesis and trend stationarity as the alternative hypothesis, the data was also subjected to (1) the Philips-Perron test, and (2) a unit root test that has trend stationarity as the null and the unit root process as the alternative, developed by Kwiatkowski, Phillips, Schmidt, and Shin (referred to as “KPSS”).

Results are summarized in table 1 in which the three tests are applied on the series in level and first difference.

Chapter II. Table 1.Unit Root Tests
LevelFirst Difference
VariablesTestH0Test statisticCritical values (5%)Test statisticCritical values (5%)
Log (reerq)ADF (0)I (1)-1.530-2.892-9.180-2.893
PPI (1)-1.553-2.892-9.180-2.893
KPSSI (0)0.8180.4630.1690.463
Log (produc_r)ADF (4)I (1)-2.478-2.901-6.366-2.899
PPI (1)-2.289-2.899-6.483-2.899
KPSSI (0)1.1760.4630.4140.463
Log (tot-weo)ADF (8)I (1)-0.557-2.895-5.399-2.895
PPI (1)-2.196-2.892-8.109-2.893
KPSSI (0)0.5600.4630.1720.463
Log (gexpgdp)ADF (1)I (1)-3.341-2.893-7.404-2.893
PPI (1)-3.311-2.892-7.486-2.893
KPSSI (0)0.3140.4630.4000.463
Log (resmon)ADF (1)I (1)-3.161-2.894-4.255-2.894
PPI (1)-2.664-2.892-10.448-2.893
KPSSI (0)0.2640.4630.1200.463
nfagdpADF (0)I (1)-1.103-2.892-10.530-2.893
PPI (1)-1.021-2.892-10.536-2.893
KPSSI (0)0.6300.4630.1690.463
Summary note: all variables are found to have a unit root, except for government expenditure and reserve money, which are found to be stationary. The PP test found that reservemoney has a unit root, while ADF rejects the unit root hypothesis and the KPSS accepts the I (0) hypothesis.

The real effective exchange rate (REER) exhibits a nonstationary behavior (Figure 1), confirmed by all three tests. Shaded areas in Figure 1 represent period under a de facto fixed peg, periods where the real effective exchange rate was appreciating substantially. Between the two episodes of fixed peg, the authorities implemented a crawling peg. In 1994, Madagascar switched to a floating exchange rate regime.


Prepared by Samir Jahjah.

Sebastian Edwards: “Exchange Rates in Developing Countries”.

Edwards (op.cit,).

Other Resources Citing This Publication