Measuring Real Exchange Rate Misalignment:

A Dynamic OLS Approach

 

 

 

 

 

 

 

 

 

 

Teferi Mequaninte

 

tefmeq@yahoo.com

 

Monetary Policy and Economic Research Directorate

National Bank of Ethiopia

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

May, 2006

 

Abstract

 

The paper tries to develop a model for the real exchange rate of Ethiopia .More specifically, the study attempts to examine whether Ethiopia's real exchange rate is misaligned with respect to its long run equilibrium level and answers such question as: (1) what are the constituent parts of the long run equilibrium RER?  (2) Which variables set the movement of the equilibrium real exchange rate? (3) Based on the findings calculate the degree of misalignment and (4) what policy measures could be taken to realign the real exchange rate with its equilibrium level. The empirical estimation results concludes that, terms of trade (TOT), external aid inflows (ODA), commercial policy stance (CPS), central bank reserve (CBR) and investment to GDP ratio were found to influence the long-run real exchange rate in the case of Ethiopia. However, variables such as nominal devaluation and real money supply found to have no effect on the real exchange rate.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1. Introduction

 

The Ethiopian economy, with support from the World Bank and International Monetary Fund (IMF), has since October, 1992 witnessed the introduction of adjustment program to halt the down turn of the economy and to move the economy on the path of sustained growth and development.  The real exchange rate (RER), by virtue of its impact on the international competitiveness of an economy, assumed an overriding importance among the cohort of policy variables (Haile Kibret, 1994).  Evidences from Latin America, Asia and African countries revealed that the link between real exchange rate behavior and economic performance is strong (Elbadawi and Soto, 1995).  According to the evidences, sustained RER misalignment usually generates sever macroeconomic disequilibrium (Edwards, 1989). The equilibrium exchange rate, according to Elbadawi (1994) is a path upon which an economy maintains both internal and external balance.

 

Real exchange rate misalignment as defined by Edwards (1989) refers to a situation where the real exchange rate diverges from its long-run equilibrium, though the equilibrium rate is not actually observed. Ethiopia's real exchange rate has been noted for being misaligned through out its development history (see Tewodros, 2004: Melesse, 2001 and Equar, 2001).  This misalignment has contributed to the deterioration of the trade balance, domestic inflationary pressure and the weak competitive position of the country (Asmerom Kidane, 1997).

 

The main objective of this paper is therefore, to develop an empirical model for the real exchange rate in Ethiopia.  More specifically, the study attempts to examine whether Ethiopia's real exchange rate is misaligned with respect to its long run equilibrium level and answers such question as: (1) what are the constituent parts of the long run equilibrium RER?  (2) Which variables set the movement of the equilibrium real exchange rate? (3) Based on the findings calculate the degree of misalignment and (4) what policy measures could be taken to realign the real exchange rate with its equilibrium level.

 

Following this introductory part, section two reviews the literature and section three deals with model specification. The data and methodology, Empirical results and the computation of the real exchange misalignment will be done in section four, five and six respectively. Finally section seven gives the concluding remark and policy implications.    

2. Literature Survey

Broadly speaking, there are three competing literatures on the real exchange rate for developing countries; a measure based on purchasing power parity (PPP) (Balassa, 1990), a measure based on using the black market premium (Quirk et al., 1987), and a model based approach (Edwards, 1988, 1994; Elbadawi 1994). The discussion of the Orthodox Purchasing Power Parity - PPP theory defines the real exchange rate as e=E P*/P, where E is the nominal exchange rate, P* and P are foreign and domestic price indices, respectively. This approach assumes an unchanged equilibrium exchange rate throughout the period and calculates the misalignment by deducting the actual real RER from some base year in which case the RER is believed to be in equilibrium. Edwards (1989) has criticized the application of the PPP theory on the ground that it gives inadequate consideration to changes in the equilibrium RER caused by fundamentals.

 

The second approach measures the misalignment using the black market exchange rate. This approach is also criticized by Montiel and Ostry (1994) as the informational content of the parallel market is limited in terms of capturing various shocks along the adjustment path. According to Aron (1994a) the parallel market is seen as a thinly traded market solely used for illicit activities.

 

 In the modern theory the real exchange rate, RER is defined as the relative price of tradable goods (PT) to non-tradable goods (PN) i.e., E PT*/ PN and uses a formal model for determining the RER. Its principal advantage is the capability of incorporating changes in the equilibrium real exchange rate and involves the calculation of the Fundamental Equilibrium Exchange Rate (Williamson 1994). Despite its advantages, however, there is no direct measure for the prices of tradable and non-tradable goods in this approach.  Therefore, whether to use consumer price index (CPI) or world wholesale price index (WPI) to substitute price of non-tradable and tradable respectively or what such choice represents has been unsettled issues. Elbadawi (1994) and Edwards (1989) for instance, argued WPI is a good proxy for the price of tradable and the CPI for that of non-tradable.  Their reasoning is that WPI contains mainly tradable and CPI mainly non-tradable.  This study will also use WPI as a measure of tradable and CPI as a proxy for non-tradable.

Another dilemma in using the modern theory of real exchange rate is that on the choice of nominal exchange rate.  Whether to use bilateral exchange rate with respect to the strong US Dollar or the multilateral exchange rate of the trading partners and what weights and which country's currencies should be included in the multilateral exchange rate is still an unresolved issue.  This study, in congruent with other empirical studies, will use the multilateral real exchange rate and trade weights will be used in the selection of trading partners.

 

Various studies on the determinants of the real exchange rate and the effects of real exchange rate misalignment have been undertaken. Edwards (1989) for example, developed a theoretical model of real exchange rate behavior and devised an empirical equation of how to estimate the real exchange rate dynamics using pooled data for a group of twelve developing countries. According to him, the important fundamentals that determine the real exchange rate are; the terms of trade, level and composition of government consumption, controls on capital flows, exchange and trade controls, technological progress, and capital accumulation. The study found that in the short-run, real exchange rate movements are affected by both real and nominal factors. In the long run however, only real factors affect the sustainable equilibrium real exchange rate. Edwards (1989) further investigated weather there was any link between real exchange misalignment and economic performance. His conclusion was that the countries whose real exchange rates were closer to equilibrium out-performed those with misaligned real exchange rates. Similarly Cottani et al (1990) also argued that in parts of Latin America, unstable real exchange rates inhibited export growth, while in Asia, export expansion was fostered by stable exchange rates. On the other, in Africa, the wide spread poor performance of the agricultural sector and economic growth in general could be attributed to persistently misaligned real exchange rates.

 

Cottani et al's argument was authenticated by other empirical findings. Ghura and Grennes (1993), for example, investigated the impact of real exchange rate misalignment on economic performance using a panel data for sub-Saharan countries. They too found that real exchange rate misalignment negatively affected income growth, export and imports, and investment and savings. In all the above studies, the most common determinants of real exchange rate were found to be terms of trade, openness, capital inflows and nominal devaluation. Other studies employing cointegration analysis in the empirical analysis of the real exchange rate as stated in Mkenda (2001), includes; Baffes et al (1999) for Cote d'Ivoire and Burkina Faso, Elbadawi and Soto (1997) for seven developing countries, Kadenge (1998) for Zimbabwe, Gelband and Nagayasu (1999) for Angola, and Aron et al (1997) for South Africa.

 

With in the context of Ethiopia, empirical studies on the determination of the equilibrium exchange rate have been undertaken.  Tewodros (2004), in his annually based construction of Equilibrium Real Exchange Rate for Ethiopia, for the periods 1970/71-2003/04, noted that both the actual REER and equilibrium exchange rate seemed to appreciate until the end of the Derg regime and depreciate there after.  Melesse (2001), using quarterly data for the period's 1985/86 - 1990/00, also observed the real exchange rate of the Birr had been over valued from 1986/87 - 1990/91 and then fluctuates between under valuation and convergence to the equilibrium after wards.  Equar (2001), in his Master's thesis and using quarterly data for the periods between 1985 and 2000 found that ERER is above the actually observed real exchange rate (i.e., the Birr was overvalued). According to him, after the introduction of the economic reform, however, the misalignment is gradually narrowed.

 

This study adds to the existing works on the real exchange rate for Ethiopia in a unique way.  In all empirical studies above (by Melesse, Equar and Tewodrose, for instance) important variables such as Central Bank reserve and real money supply were skipped from the regression analysis despite their relevance in the real exchange rate specification and hence the mission of the National Bank of Ethiopia (NBE). In addition to this, the present study replaces capital control by Official Development Assistance (ODA) as the later constitutes major capital inflow to the country.

 

3. Model Specification

In the Behavioral equilibrium exchange rate model, the real exchange rate () is defined as the domestic relative price of tradable goods () to non-tradable goods (), that is, compatible with the attainment of internal and external equilibrium, and ℮ is the nominal exchange rate. Internal equilibrium presupposes that the market for non-tradable clears in the current period and is expected to be so in the future. External equilibrium implies that the current account balances both in the current and future periods are compatible with long-run sustainable capital flows (Elbadawi,1994).Thus, using equations below, the hypothesized relationships can be specified.

 

Based on the works of Melesse (2001) and Equar (2001), and as observed by Edwards (1989), the dynamics of the behavior of the real exchange rate are given by equation as follows:

 

      

 Where

  = Deviations of the actual real exchange rate from its equilibrium level

   = Inconsistency in the macroeconomic policy framework

    = Nominal exchange rate devaluation

     = Positive parameters capturing vital aspects of the adjustment   process

 

Equation (2) gives an indication of the main fundamentals that influence the behavior of the equilibrium real exchange rate:

 

                                                                   

 

Where

 

=    The equilibrium real exchange rate

   =     External terms of trade

    =     External aid inflows (defined as real net ODA to Ethiopia)

  =  Government consumption of non-tradable (measured by the share of government       consumption in GDP)

   =       Commercial policy stance (measured by the black market premium)

          =        Growth rate of real GDP (used as a measure of technological progress)

   =        Investment to GDP ratio.

 

Edwards (1989) stressed that this equation of equilibrium RER does not provide an explicit distinction between permanent and temporary movements in the fundamentals. Thus, it would be necessary to breakdown the fundamentals in to these components as the long-run equilibrium exchange rate is determined by the permanent components.

 

Equation (2) puts the equilibrium real exchange rate as a function of real fundamentals. But the actual real exchange rate, as given in equation (1), responds to both real fundamentals and macroeconomic policies represented by. Thus, Aron et al (1997) and Mkenda (2001) included central bank reserve (CBR) and real money supply (M2) to the model of the real exchange rate to capture the role of macroeconomic policies. Finally, some measures of nominal devaluation should also be introduced to real exchange rate model to capture the impact of nominal devaluation.

 

Incorporating all the above, the model for the real exchange rate (RER) that is used for estimation can be formulated as:

 

 

 Stands for nominal devaluation and Ut for the error term.  

 

The expected theoretical impacts of the respective fundamentals are as follows:

 

  (?)     -    Since terms of trade is defined as the relative price of exports to imports, its impact on the RER is theoretically ambiguous and depends on the relative strengths of the direct income effect operating through the demand for non-tradable and the indirect substitution effects that operates through the supply of non-tradable. To illustrate the impact of the direct income effect, let the price of exports increase (improvement in TOT), and the price of imports stay constant. The increases in the price of exports increases income and then raises the demand for both imports and non-tradable domestic goods. Since the price of imports is given, the higher demand would only affect the price of non-tradable goods and hence a real exchange appreciation will occur. If a deterioration in the terms of trade occur, it may lead to the opposite effect (reducing income and the demand for all goods and hence resulting in a depreciation in the RER). Sometimes, the indirect substitution effect may dominate the direct income effect. For example, an improvement in terms of trade may provide sufficient foreign exchange resources to producer of non-tradable goods in the economy. The increased resources may then enable the producers to increase their production of non-tradable goods, hence lowering its price and to a depreciation in the RER. If a deterioration in terms of trade occurred, it may lead to the opposite effect (an appreciation of the RER). In Elbadawi and Soto’s (1997) study of seven developing countries, in three case, an improvement in the term of trade appreciated the real exchange rate, while in four cases, an improvement in the terms of trade depreciated the real exchange rate.

 

   (-)     - By increasing real incomes and consequently the demand for both traded and non-traded goods, it tends to cause the RER to appreciate. In his study of twelve developing countries, Edwards (1989) found that an increase in capital inflows appreciated the real exchange rate, as expected.

 

  (?)   -  Increases in government expenditure on non-tradable appreciates  the RER, while those on tradable causes the RER to depreciate. Edwards (1989) found that an increase in government consumption appreciated the real exchange rate in four of the equations he estimated for a group of twelve developing countries, while in the other two equations, an increase in government consumption depreciated the real exchange rate.

 

  (+)   - A reduction, for instance, in an import tariff can decrease the domestic price of imports, which are a part of tradables. This can in turn decrease the local currency price of tradables, leading to an appreciation in the real exchange rate. An increase in import tariffs can have the opposite effect. That is, it can raise the domestic price of imports, there by depreciating the real exchange rate. However, the demand for imports and consequently for foreign exchange will increase, leading to a depreciation in the real exchange rate. In their study of Cote d’Ivoire and Burkina Faso, Baffes et al (1999) found results consistent with the theory; reforms that are aimed at liberalizing trade are consistent with a depreciated real exchange rate.

 

 (?)  - Technological progress appreciates the RER if gains emanating from productivity enhancement in the   tradable Sector overrides those in the non-tradable sector. Edwards (1989) found that an increase in technological progress depreciated the real exchange rate in all his regressions. Aron et al (1997), on the other hand, found that an increase in technological progress appreciated South Africa’s real exchange rate.

 

 (?) -   Central Bank reserve intervention indicates the capacity of the Bank to defend the currency (Aron et al, 1997). An increase in reserve has the effect of appreciating the real exchange rate, while a decrease in reserves depreciates the real exchange rate. In their study of the determinants of the real exchange rate for South Africa, Aron et al (1997) found results consistent with the theory; an increase in reserves appreciated the real exchange rate.

 

      (+)    - Nominal devaluation tends to depreciate the RER.

         (-)     -Increase in real money supply depreciates the real exchange rate

 

 (?)   - Its impact on the real exchange rate depends on whether an increase in investment changes the composition of spending on traded and non-traded goods. If an increase in the share of investment in GDP changes the composition of spending towards traded goods, it will lead to a depreciation in the real exchange rate (Baffes et al ; Edwards, 1989). On the other hand, a change towards non-traded goods appreciates the real exchange rate. For example, Baffes et al (1999) found that an increase in the share of investment in GDP depreciated the real exchange rate in Cote d’Ivoire. Edwards (1989) also found that increases in the share of investment in GDP resulted in depreciation in the real exchange rate in his study of twelve developing countries.

Following the definition of the real exchange rate, a negative sign (i.e.,-) represents an appreciation of the real exchange rate. This is because the real exchange rate is inversely related to spending (consumption) on non-tradable goods. This happens so because if we start from a position of internal balance, a rise in spending creates an excess demand for non-tradable goods at the original real exchange rate. In order to restore equilibrium, a real appreciation is required, which would switch supply toward non-tradable goods, and demand toward tradable goods.

 

4. Data and Methodology

All data used in this study relate to the period 1970/71 to 2003/04 and was obtained from the Macro econometric team of the National Bank of Ethiopia (NBE) and the Organization for Economic Cooperation and Development (OECD) website for the ODA. The data used is annual and the variables are in logs.

 

The paper employs the Stock Watson Dynamic OLS estimation procedure in determining the presence of relationship between the real exchange rate and its determinants. DOLS approach has certain advantages over both OLS and the Johansen maximum likelihood procedures. The OLS apriorily categorizes variables as endogenous and exogenous with implication of endogeneity problem and also the error term is not normally distributed (auto-correlation). The Johansen method, being a full information technique is exposed to the problem of misspecification and small data observation. However, the Stock Watson method is a robust single equation approach, which corrects the problem of endogeniety by inclusion of leads and lags of the first difference of the regressors and for serially correlated errors by a GLS procedure. In addition DOLS has the same good property as the OLS for small sample size and it has the same asymptotic optimality property as the Johansen distribution.

 

5. The Stock-Watson (DOLS) Empirical Results

The Stock-Watson DOLS estimates for the equilibrium exchange rate in Ethiopia appears in Table 1 below. The behavioral exchange rate model is estimated including up to j=±2 leads and lags; insignificant lags and leads were dropped. By the rule of DOLS since the short-run model is the adjustment period where the leads and lags net out their effects, its analysis and interpretation is not included (Stock et al, 1993). Given the fact that a substantial amount of government consumption contains foreign aid and that there is no disaggregated data for government consumption of non-tradable, GCN is excluded from the empirical estimation. Similarly, since technological progress can be captured by investment to GDP ratio (see Jing Xu, 2003) and our economy is mainly agrarian, GRGDP has been also excluded from the empirical model. By the same logic as net foreign asset (NFA) is the component of broad money (M2) in Central Bank’s asset specification; CBR is out of the empirical estimation. Finally dummy war (DWAR) is included in the empirical estimation.  

 

Table 1: The Stock-Watson (DOLS) Empirical Results

 

Dependent Variable: LOG(REER)

Method: Least Squares

Date: 02/16/06   Time: 20:23

Sample(adjusted): 1972 2002

Included observations: 31 after adjusting endpoints

White Heteroskedasticity-Consistent Standard Errors & Covariance

Variable

Coefficient

Std. Error

t-Statistic

Prob. 

C

7.540841

0.937051

8.047414

0.0000

LOG(TOT)

-0.250354

0.097467

-2.568600

0.0199

LOG(AID)

-0.356436

0.082791

-4.305254

0.0005

LOG(CPS)

0.636917

0.098133

6.490342

0.0000

LOG(INVGDP)

0.558694

0.270346

2.066587

0.0544

DLOG(TOT)

0.189792

0.114774

1.653621

0.1166

DLOG(AID)

0.251486

0.111842

2.248581

0.0381

DLOG(M2)

-0.031646

0.361526

-0.087536

0.9313

DLOG(TOT(-1))

0.175415

0.107297

1.634849

0.1205

DLOG(AID(-1))

0.234128

0.153819

1.522103

0.1464

DLOG(CPS(+1))

0.169086

0.131477

1.286057

0.2157

DLOG(INVGDP(+1))

0.597379

0.251350

2.376687

0.0295

D(NOMDEVAL(1))

-0.260052

0.142871

-1.820192

0.0864

DWAR

0.149712

0.098011

1.527501

0.1450

R-squared

0.878032

    Mean dependent var

4.945367

Adjusted R-squared

0.784762

    S.D. dependent var

0.290335

S.E. of regression

0.134697

    Akaike info criterion

-0.869122

Sum squared resid

0.308437

    Schwarz criterion

-0.221514

Log likelihood

27.47138

    F-statistic

9.413903

Durbin-Watson stat

1.313360

    Prob(F-statistic)

0.000023

 

Variables qualified for the long-run model are TOT, AID, CPS and INVGDP and their respective coefficients are    -0.250, -0.356, 0.637, and 0.559. Except for M2 and Nominal devaluation, all the variables are significant and the results show that taken together, these fundamentals explain 78 percent of the variation in the real exchange rate. The negative parameters on terms of trade, foreign aid and central bank reserve variables imply a tendency towards real exchange rate appreciation. However, commercial policy stance and investment to GDP ratio variables exhibit positive coefficients and, therefore, tend to depreciate the real exchange rate. Given the fact that macroeconomic policies are intended to have only a temporary impact on the economy, the insignificant short-run negative coefficients on the money supply and nominal devaluation also cohorts special attention.

 

The negative and significant effect of the terms of trade on the real exchange rate implies that the indirect substitution effect dominates the direct income effect in the case of Ethiopia. The substitution effect may have been on the supply side, in which case deterioration in terms of trade (adverse terms of trade and drought) may not provide sufficient foreign exchange resources to producers of non-tradable goods in the economy. The decreased resources may then not enable the producers to increase their production of non-tradable goods, hence increasing its price and to appreciation of the real exchange rate. 

 

The coefficient on external aid is also negative and significant. This result suggests that a large percentage of foreign aid may have probably been invested in the non-tradable goods and services such as wages, services and recurrent expenditure. As has been evidenced in Yohans (1996) "the growth in Ethiop