Journal Issue
Share
Article

Are African Current Account Deficits Different? Stylized Facts, Transitory Shocks, and Decomposition Analysis

Author(s):
Luisa Zanforlin, César Calderón, and Alberto Chong
Published Date:
January 2001
Share
  • ShareShare
Show Summary Details

I. Introduction

Policymakers view the evolution of the current account balance as a key leading indicator of the health of a country’s economy. Given the macroeconomic fragility of the developing world, especially African countries, arriving at a clear understanding of the factors that affect the current account balance appears to be a sensible strategy for effective policymaking. The contribution of this paper is to help provide clues toward the design of such a strategy in Africa, not only by helping uncover the possible factors that may affect the behavior of the current account but also by using them to estimate whether current account balances in African countries are in line with fundamentals.

Recent studies on current account determination have included Debelle and Faruqee (1996), and Calderón, Chong, and Loayza (1999). The former use a panel of 21 industrial countries over 1971-93 and an expanded cross-sectional data set that includes an additional 34 industrial and developing countries in order to explain long-term variations and short-run dynamics of the current account by specifying cross-section and panel data models, respectively. Calderón, Chong, and Loayza (1999) provide a general characterization of the empirical linkage between current account deficits and a broad set of economic variables for 44 countries for the period 1966-95. Their main findings are that current account deficits are persistent; increases in domestic growth generate larger deficits; transitory increases in public or private savings have positive effects on the current account but their permanent changes have insignificant effects; increases in terms of trade or appreciation of the real exchange rate are linked with larger current account deficits; and finally, either higher growth rates in industrialized economies or higher international interest rates reduce the current account deficit in developing economies.

The first objective of this paper is to identify similar stylized facts for sub-Saharan African countries and to see how they differ from other developing countries. Indeed, there are quite a few characteristics that indicate that what drives the current account balances in Africa may be different. For one, deficits in the current account in Africa and, in particular, in sub-Saharan Africa, have been very large in recent years. During the period 1975-95, sub-Saharan Africa had the world’s highest current account deficit (5.9 percent of the gross national disposable income). Moreover, the rate of growth of the region has been dismal (Easterly and Levine, 1997), the reliance on foreign aid has been very high (World Bank, 1998), public and private savings have been very low, the concentration of exports on single primary products continues to be quite significant, and the distortions in the economy (for instance, as measured by the black market premium) have also been very large. All of these characteristics emphasizing the fact that the determinants of current account balances in Africa are probably different, not only in terms of magnitude but, given the many peculiarities of the region, also in terms of direction.

The second objective of this paper is to use our findings about the determinants of the current account from current account determination in order to calculate whether the current account deficits in Africa are “excessive” with respect to fundamentals, as defined by the difference between the actual and estimated fitted values. This is done for the region and for each African country of our sample by source of disequilibrium: external, internal, or global (the evolution of the world economy).

We take advantage of a large macroeconomic data set that includes information on savings, aid flows, and other relevant national income variables for a panel of 64 countries for the period 1975-1995, of which 30 are African countries. We follow Calderón, et al. (1999) and adopt a reduced-form approach, instead of holding to a particular structural model, so that a comprehensive menu of determinants, identified by the literature, may be included. Finally, given the fact that current account deficits may be persistent (Calderón, et al, 1999), we use a dynamic panel data approach that helps control for problems of joint endogeneity of the explanatory variables and correlated unobserved country-specific effects (Arellano and Bover, 1995; Blundell and Bond, 1997).

This paper is organized as follows. The next section describes the data. Section III presents simple correlation analysis. Section IV very briefly describes the Arellano-Bover-Blundell-Bond dynamic panel data methodology (which is explained in more detail in Appendix I). Section V presents our empirical results, linking them to theoretical work where relevant. In Section VI we calculate the fitted and actual values of the current account deficits for the different regions, as well as for the African countries in our sample; following Easterly Loayza, and Montiel (1997), we estimate whether these values are “excessive” for Africa. Finally, Section VII concludes.

II. Data

Our objective is to provide an extensive analysis of the various determinants of current account deficits in Africa, so as to explain whether such deficits are “excessive” with respect to developing countries as a whole. We rely on previous theoretical and empirical research in order to choose an operational set of variables encompassing the various dimensions of current account determination. From the theoretical literature, we use domestic output growth (Sachs, 1982; Glick and Rogoff, 1995; and Razin, 1995), the real exchange rate (Stockman, 1987), terms of trade fluctuations (Obstfeld, 1982; and Tornell and Lane, 1998), fiscal policy variables (Leiderman and Razin, 1991; and Frenkel and Razin, 1996), capital controls (Mendoza, 1991), and world economy shocks (Glick and Rogoff, 1995; and Razin, 1995).

Data are taken mostly from the International Monetary Fund and the World Bank databases. In addition, two data sets were made available to us: (a) a savings data set from the World Savings Database (Loayza and others 1998),2 which uses data adjusted for capital gains and losses that accrue to the public and private sectors as a result of inflation, that is, for the erosion of the real value of nonindexed public debt; and (b) aid flows, as expressed by the effective development assistance (EDA) provided to a country—an aggregate measure that combines total grants and grant components of all official loans—which are calculated using a new approach based on a loan-by-loan basis, that reflects the financial cost a creditor incurs in making loans on concessional terms (Chang, Fernandez Arias, and Serven, 1998). Aid is a variable of particular importance in the African context (Collier and Gunning, 1999; and World Bank, 1998). For instance, in sub-Saharan Africa, aid represents about 25 percent of the gross national disposable income (Table 2). Detailed definitions and sources are presented in Table 1.

Table 1.Description of Variables
VariableDescriptionSources
Current account deficit (CAD)Current account deficit as apercentage of gross national disposable income (GNDI). GNDI is equal to GNP plus all net unrequited transfers. CAD > 0 (< 0) indicate a deficit (surplus) in the current account balance.Loayza and others (1998)
Gross national savings (GNS)GNS is equal to GNDI minus consumption expenditureLoayza and others (1998)
Private and public savingsThe definition of government we use is the central government (this excludes non-financial public enterprises). We use data adjusted for capital gains and losses due to inflation. Percentage of GNDILoayza and others (1998)
Real effective exchange rate (REER)REER is the multilateral real exchange rate computed using the following formula:
where e is the exchange rate (in unit of local currency per US$) and P is the consumer price index CPI of the domestic country, ek and Pk are exchange rates and CPIs for trading partners respectively. The δ coefficients represent IMF-generated weights based on both bilateral trade shares and export similarity. An increase in REER indicates a real appreciation of the domestic currency.
IMF, - International Financial Statistics (IFS). World Bank -World Development Indicators (WD1).
Terms of trade (TOT)Terms of trade is computed as the ratio of export prices to import prices (base 1987 = 100)World Bank – WDI
Exports (X)We consider the ratio of exports to GNDI as our measure of openness.Loayza and others (1998) World Bank – WDI
Exchange controls (EXCC)Average of three dummy variables that capture restrictions on current and capital account transactions, as well as multiple exchange rate regimes.Grilli and Milesi-Ferreti (1995)
Black market premium (BMP)Black market premium on foreign exchange (BMP) is the ratio of the parallel to the official exchange rate. We use log (1 + BMP).Wood (1988) for period 1960-88. International Currency Analysis Inc. (various years) after 1988.
Aid (EDA)Aid is proxied by the ratio of effective development assistance (EDA) to Gross domestic investment (GDI). EDA is an aggregate measure of aid flows combining total grants and the grant component of all official loansChang and others (1998) World Bank – WDI.
External DebtExternal debt is measured by the ratio of total external debt (EDT) to gross national product (GNP)World Bank – WDI.
Macroeconomic UncertaintyMacroeconomic uncertainty is proxied by the standard deviation of monthly CPI inflation.IMF – IFS.
Industrialised output growthThe growth rate for industrial economies is computed from the (population-weighted) real GDP (in constant US$) of OECD economies.IMF – IFS.

World Bank - WDI.
International real interest rateTo compute this variable, we use the nominal eurodollar London rate (as proxy for the nominal international rate) and the CPI percentage change for industrial countries (as proxy for external inflation).IMF – IFS.
Table 2.Summary Statistics
Current

Account
Domestic

Output
Savings

(% GNDI)1/
ExportsBlack

Market
ExchangeEffect.

Develop.
Deficit

(% GNDI)
Growth

Rate
PrivatePublic(%GNDI)Premium

2/
Controls

3/
Assist.

(%GDI)
Latin America0.0400.0270.1380.0330.2520.2300.5510.035
East Asia0.0200.0630.2320.0460.3510.1430.3040.016
South Asia0.0220.0530.1490.0380.1530.2470.5720.116
North Africa0.0330.0410.1410.0650.2820.3520.7070.037
Egypt0.0730.0620.1170.0530.2070.0691.0000.073
Morocco0.0600.0400.1450.0360.2110.0570.8420.040
Tunisia0.0480.0390.1680.0570.3790.0650.5830.033
sub-Saharan0.0590.0280.0830.0350.2190.3730.6580.244
Africa
Cent. Afr. Rep.0.0440.0070.0170.0440.1940.0250.5780.399
Cameroon0.0790.0340.1080.0550.1770.0240.4390.060
Ethiopia0.0170.0230.0790.0130.0880.9980.7330.535
Ghana0.0200.0160.0420.0340.2180.9650.9330.255
Kenya0.0470.0380.1730.0120.2520.1450.7020.085
Mauritius0.0450.0460.1790.0370.5870.0520.7840.042
Niger0.0270.0130.0880.0320.2430.0270.4810.375
Nigeria0.0360.0350.1410.0370.3270.6170.7860.000
Rwanda0.0150.0350.0820.0340.0570.3600.6670.264
Sudan0.0870.0520.0570.0230.0530.4720.8890.137
Sierra Leone0.109-0.0010.0090.0140.2110.5740.5740.180
Uganda0.0170.0390.0450.0470.0610.6460.6970.365
Zaire0.0470.0000.0730.0140.1920.5910.6890.102
Zimbabwe0.0360.0330.1430.0320.2830.4720.9290.051

Private and public savings are calculated using data for central government.

Calculated as log(l+variable).

Calculated as the simple average of dummies for presence of current or capital controls, and multiple exchange rate regimes.

Private and public savings are calculated using data for central government.

Calculated as log(l+variable).

Calculated as the simple average of dummies for presence of current or capital controls, and multiple exchange rate regimes.

Additionally, in order to ensure a minimum time-series dimension and to allow an adequate implementation of our dynamic panel data methodology, we include countries that have at least six consecutive annual observations. This approach generates a sample with 64 countries and 759 annual observations for the period 1975-95 and a sub-sample for Africa of 30 countries and 302 observations for the same period.

III. Basic Correlation Analysis

We use both cross-country 1975-95 average (“long-run”) and pooled annual cross-country time-series (“short run”) data and correlate current account deficit (as percentage of gross national disposable income) with its possible determinants, by regions.3 Results are shown in Table 3. This yields some basic stylized facts that help us understand the econometric results of the next section. In particular, we make the following findings:

Table 3.Cross-Country and Panel Correlations Between CAD and Regressors 1/2/.
DevelopingLatin AmericaEast AsiaNorth AfricaSouth AsiaSub SahAfrica
Cross-

country
PanelCross-

country
PanelCross-

country
PanelCross-

country
PanelCross-

country
PanelCross-

country
Panel
Domestic
GDP (logs)-0.417

(0.00)
-0.262

(0.00)
-0.437

(0.05)
-0.231

(0.00)
-0.955

(0.00)
-0.298

(0.00)
-0.083

(0.85)
-0.037

(0.68)
-0.334

(0.52)
-0.227

(0.02)
-0.171

(0.42)
-0.099

(0.06)
GDP growth-0.007

(0.96)
-0.063

(0.04)
0.233

(0.32)
-0.043

(0.40)
-0.749

(0.09)
-0.289

(0.00)
0.104

(0.81)
0.085

(0.34)
0.047

(0.93)
0.029

(0.77)
0.428

(0.04)
0.025

(0.64)
Private savings

(% GNDI) 3/
-0.406

(0.00)
-0.416

(0.00)
-0.156

(0.51)
-0.372

(0.00)
-0.738

(0.09)
-0.569

(0.00)
-0.655

(0.08)
-0.490

(0.00)
-0.172

(0.74)
-0.249

(0.01)
-0.130

(0.54)
-0.299

(0.00)
Public Saving

(% GNDI) 3/
-0.176

(0.16)
-0.156

(0.00)
-0.292

(0.21)
-0.207

(0.00)
-0.696

(0.12)
-0.195

(0.05)
-0.192

(0.65)
-0.231

(0.01)
0.763

(0.08)
0.389

(0.00)
-0.066

(0.76)
-0.082

(0.12)
External
Exports

(% GNDI)
0.082

(0.52)
0.034

(0.27)
0.381

(0.10)
0.104

(0.04)
0.439

(0.38)
0.012

(0.90)
-0.120

(0.78)
-0.143

(0.11)
0.596

(0.21)
0.402

(0.00)
0.101

(0.64)
0.089

(0.10)
REER (logs)-0.031

(0.81)
0.087

(0.00)
0.311

(0.18)
0.223

(0.00)
-0.091

(0.86)
-0.010

(0.92)
0.426

(0.29)
0.439

(0.00)
-0.777

(0.07)
-0.263

(0.01)
-0.254

(0.23)
-0.116

(0.03)
Terms of

Trade (logs)
-0.095

(0.46)
-0.128

(0.00)
-0.407

(0.07)
-0.225

(0.00)
0.614

(0.20)
-0.096

(0.33)
0.363

(0.38)
0.069

(0.43)
0.190

(0.72)
-0.024

(0.81)
-0.102

(0.64)
-0.157

(0.00)
BMP 4/0.154

(0.22)
0.030

(0.32)
-0.009

(0.97)
-0.054

(0.29)
-0.441

(0.38)
0.000

(1.00)
-0.357

(0.39)
0.119

(0.18)
0.049

(0.93)
-0.005

(0.96)
0.334

(0.11)
-0.018

(0.74)
BoP cont 5/0.04

(0.75)
0.00

(0.94)
-0.19

(0.42)
-0.10

(0.06)
-0.28

(0-59)
-0.05

(0.62)
0.32

(0.44)
0.38

(0.00)
-0.30

(0.56)
-0.15

(0.13)
0.12

(0.58)
-0.03

(0.55)
Aid (%GDI)0.266

(0.03)
0.108

(0.00)
0.366

(0.11)
0.151

(0.00)
0.813

(0.05)
0.268

(0.01)
0.821

(0.01)
0.322

(0.00)
0.096

(0.86)
0.086

(0.37)
-0.007

(0.97)
-0.077

(0.15)
World Economy:
Indstr. Output

(in logs)
0.076

(0.55)
-0.088

(0.00)
-0.146

(0.54)
-0.170

(0.00)
-0.461

(0.36)
-0.071

(0.47)
0.415

(0.31)
0.048

(0.59)
0.336

(0.51)
-0.032

(0.74)
-0.003

(0.99)
-0.122

(0.02)
Indstr. Output

Growth rate
0.319

(0.01)
-0.150

(0.00)
-0.079

(0.74)
-0.202

(0.00)
0.853

(0.03)
-0.313

(0.00)
0.090

(0.83)
-0.025

(0.78)
0.390

(0.44)
-0.143

(0.14)
0.380

(0.07)
-0.124

(0.02)
Intl. int. rate 4/0.070

(0.58)
0.053

(0.08)
-0.098

(0.68)
0.047

(0.36)
-0.018

(0.97)
0.247

(0.01)
0.306

(0.46)
0.064

(0.47)
0.264

(0.61)
-0.061

(0.53)
-0.005

(0.98)
0.002

(0.96)

Standard errors in parenthesis.

CAD expressed as percentage of GNDI.

Private and public saving are calculated using data for central government.

Calculated as log(l+variable).

Calculated as the simple average of dummies for presence of current or capital controls, and multiple exchange rate regimes.

Standard errors in parenthesis.

CAD expressed as percentage of GNDI.

Private and public saving are calculated using data for central government.

Calculated as log(l+variable).

Calculated as the simple average of dummies for presence of current or capital controls, and multiple exchange rate regimes.

  • Current account deficits are negatively and significantly correlated with real income levels. In absolute values, the cross-country (long-run) correlation is higher than the pooled-data (short-run) correlation. The lowest correlations are registered for North Africa (-0.08 cross-country and -0.04 pooled) and sub-Saharan Africa (-0.17 cross-country and -0.09 pooled).

  • Current account deficits are negatively and significantly linked with domestic growth rates in the pooled-data (short-run) sample, although such correlation is not significant in the cross-country (long-run) sample. However, the correlation for the cross-country sample is positive and significant for sub-Saharan Africa.4 This is consistent with the expected response of the current account to an increase in the growth rate driven by permanent productivity shocks or investment shocks (Glick and Rogoff, 1995).

  • Current account deficits are negatively correlated to private and public savings. Such correlation is higher for private than for public savings (-0.41 vs. -0.18 with cross-country, and -0.42 vs. -0.16 with pooled data). This empirical regularity also holds for Africa.

  • The correlation between current account deficits and real exchange rate appreciation is positive and significant but low (approximately 0.09) for the pooled-data sample. This is consistent with the Mundell-Fleming prediction that a depreciation of the domestic currency (i.e., a decrease in the real exchange rate) helps reduce the current account deficit; again, however, the sign is reversed for sub-Saharan Africa.

  • The correlation between current account deficits and terms of trade fluctuations is negative and significant in the pooled-data sample, but it is not significant in the cross-country sample. The short-run correlation is consistent with the Harberger-Laursen-Metzler effect, by which transitory improvements in the terms of trade help reduce current account deficits (Obstfeld, 1982; Svensson and Razin, 1983; and Mendoza, 1995). The long-run, correlation is consistent with the findings of Razin (1995) and Mendoza (1995) that permanent shocks to the terms of trade do not affect current account balance.

  • There is no significant relationship between current account deficits and distortions (black market premium, exchange controls) in the external sector for the full sample, but there is a positive and significant relationship for the sub-Saharan Africa panel.5

  • Current account deficits are significantly and positively correlated with aid inflows for both samples. The correlation is higher in the cross-country than in the pooled-data sample, thus implying that stronger current account effects might be expected with more persistent the EDA inflows. Interestingly, not only does the sign change for Sub-Saharan Africa, but this correlation also becomes nonsignificant for this region.

  • The correlation between current account deficits and the growth rate for industrial economies is positive and significant with pooled data (0.32), while negative and significant (-0.15) for the cross-country sample. This appears to be consistent with the hypothesis that growth rates are driven by both permanent and transitory global shocks to the economy.

  • The correlation between current account deficits and international real interest rate is not significant in either of the two samples.

To summarize, the simple correlations between the current account deficit and its possible components appear to indicate that the forces that drive the deficits in Africa and, in particular, sub-Saharan Africa, may be different both qualitatively and quantitatively than those other regions, even when compared with the developing world as a whole.

IV. Methodology

The equations specified for the empirical analysis include some explanatory variables that could be jointly endogenous, that is, correlated with the error term. In addition, unobserved country-specific factors may be correlated with the explanatory variables. Given these potential weaknesses in the empirical approach, our preferred method of estimation is the generalized method of moments (GMM) estimator for dynamic models of panel data, introduced by Arellano and Bover (1995) and Blundell and Bond (1997). This so-called GMM system estimator combines in a single system the regression equation in both changes and levels, each with its specific set of instrumental variables. Because of this capability, our model is designed to handle both pooled cross-country and time-series data. It is dynamic, since it allows for independent effects from the lagged current account deficit. Also, this approach allows for a “weak exogeneity” assumption with respect to the explanatory variables while preserving the estimator’s properties.

The consistency of the GMM estimator depends on whether lagged values of the explanatory variables are valid instruments in the current account deficit regression. We address this issue by considering two specification tests suggested by Arellano and Bond (1991) and Arellano and Bover (1995). The first is a Sargan test of overidentifying restrictions, which tests the overall validity of the instruments by analyzing the sample analogue of the moment conditions used in the estimation process. Failure to reject the null hypothesis gives support to the model. The second test examines the hypothesis that the error term is not serially correlated. We test whether the differenced error term (that is, the residual of the regression in differences) is first-, second-, or third-order serially correlated. First-order serial correlation of this error term is expected even if the original error term (in levels) is uncorrelated, unless the latter follows a random walk. Second-order serial correlation of the differenced residual indicates that the original error term is serially correlated and follows a moving average process at least of order one. If the test fails to reject the null hypothesis of absence of second-order serial correlation, we conclude that the original error term is serially uncorrelated and use the corresponding moment conditions. Appendix I presents a detailed explanation of our approach.

V. Regression Results

In this section, we analyze the determinants of current account deficits in Africa for the period 1975-95 using annual data. We will compare the regression results across two samples: a sample of developing countries and a sample of African countries. As mentioned above, we use the Arellano-Bond-Bover-Blundell dynamic panel data techniques to control for joint endogeneity and correlated country-specific effects. The dependent variable is the current account deficit as a percentage of gross national disposable income, or GNDI. Somewhat similar to Calderón, et al. (1999), the determinants considered have been classified in three groups:6 (a) domestic economy conditions, (b) external conditions, and (c) evolution of the world economy. In particular, the set of explanatory variables include the lagged value of the current account deficit, CAD, the domestic output growth rate, private and public savings (both expressed as a percentage of GNDI), exports (as a percentage of GNDI), the real effective exchange rate, the terms of trade, the extent of balance of payments controls, the black market premium, aid flows (effective development assistance), the output growth rate of industrial economies, and the international real interest rate. Given the frequency of the data used, our analysis mainly focuses on the transitory impact of changes in these variables on the current account deficit. This helps us to understand the short-run dynamics of the current account adjustment, as well as some medium-run implications for its sustainability in the context of African economies.7 By using a GMM-IV system estimator (Arellano and Bover, 1995), we allow the explanatory variables to be jointly (weakly) endogenous, except for the terms of trade, the industrialized output growth rate, and the international real interest rate variables that in both our developing and African country samples are most likely to be exogenous. Results are presented in Table 4 (benchmark regressions are [1] and [1a]) and are discussed below.

Table 4.Determinants of CAD in Africa and in All Developing Countries, 1975-95 1/
[1][1a][2][2a][3][3a][4][4a]
LDCsAfricaLDCsAfricaLDCsAfricaLDCsAfrica
Constant0.030

(0.059)
-0.038

(0.068)
-0.018

(0.065)
-0.060

(0.075)
-0.012

(0.074)
0.383

(0.139)
-0.050

(0.055)
-0.043

(0.076)
CAD lagged0.519

(0.061)
2/0.109

(0.073)
3/0.482

(0.045)
2/0.158

(0.062)
2/0.476

(0.054)
2/0.021

(0.048)
0.501

(0.059)
2/0.199

(0.062)
2/
Domestic
Domestic

output growth
0.107

(0.068)
3/-0.192

(0.115)
3/0.037

(0.075)
-0.276

(0.096)
2/0.086

(0.071)
0.116

(0.086)
0.082

(0.078)
-0.181

(0.083)
2/
Rate
Private savings-0.143

(0.066)
2/-0.368

(0.086)
2/-0.132

(0.052)
2/-0.440

(0.115)
2/-0.278

(0.073)
2/-0.618

(0.091)
2/-0.117

(0.071)
3/-0.170

(0.045)
2/
Public savings-0.202

(0.060)
2/-0.409

(0.099)
2/-0.218

(0.063)
2/-0.374

(0.060)
2/-0.240

(0.068)
2/-0.531

(0.052)
2/2/
Gov exp/GNDl0.368

(0.103)
2/0.408

(0.106)
Macroeconomic

Uncertainty
-0.000

(0.000)
2/-0.000

(0.000)
External
Exports/GNDI0.003

(0.022)
0.006

(0.047)
0.012

(0.022)
0.029

(0.062)
0.073

(0.022)
2/0.046

(0.064)
-0.018

(0.029)
0.028

(0.052)
REER0.006

(0.012)
0.036

(0.016)
2/0.015

(0.013)
0.038

(0.016)
2/0.017

(0.014)
-0.045

(0.031)
0.010

(0.011)
0.007

(0.021)
Terms of trade-0.030

-0.005
2/-0.024

-0.009
2/-0.032

-0.006
2/-0.011

-0.016
-0.029

-0.007
2/-0.027

-0.012
2/-0.038

-0.006
2/-0.066

-0.014
2/
Black market

premium
0.010

(0.007)
3/-0.034

(0.013)
2/0.003

(0.008)
-0.031

0.008
2/0.009

(0.008)
0.014

(0.021)
0.003

(0.008)
0.002

(0.015)
BoP Controls-0.009

(0.006)
3/-0.005

(0.011)
-0.002

0,007
0.004

(0.010)
0.000

(0.006)
0.012

(0.008)
0.000

(0.007)
-0.001

(0.011)
Eff. dev.

assistance
-0.024

(0.012)
2/-0.035

(0.010)
2/-0.023

0.010
2/-0.041

(0.014)
2/-0.033

(0.011)
2/-0.010

(0.012)
-0.021

(0.013)
3/-0.007

(0.009)
Debt/GNP

Lagged
-0.011

(0.010)
-0.100

(0.014)
2/
World economy
Industrial output

growth rate
-0.346

(0.066)
2/-0.201

(0.139)
3/-0.363

(0.072)
3/0.007

(0.250)
-0.311

(0.072)
3/0.152

(0.176)
-0.357

(0.072)
2/0.053

(0.286)
International real

interest rate
-0.143

(0.089)
3/-0.408

(0.225)
3/-0.170

(0.089)
2/-0.344

(0.189)
3/-0.289

(0.080)
3/-0.294

(0.195)
3/-0.030

(0.100)
-0.040

(0.286)
Countries6430612963306430
Observations759302710282739302742301
Sargan test0.2660.5690.2050.5100.2070.5560.2500.472
Serial
Correlation test
1st Order0.0000.0890.0000.0580.0000.0190.0000.006
2ndOrder0.9180.7440.8870.7440.8830.3800.4200.655
3rdOrder0.7940.2730.5800.2360.3820.5220.9260.202

Standard errors in parentheses.

Significant at a 5 percent confidence interval.

Significant at a 10 percent confidence interval.

Standard errors in parentheses.

Significant at a 5 percent confidence interval.

Significant at a 10 percent confidence interval.

Persistence. The lagged value of the current account deficit (CAD) is positive and significant (at 10 percent) but not very persistent (coefficient is 0.11).8 Notice that the CAD for all developing countries (0.52) shows a much higher persistence than the process for African countries. This may be attributed to the greater likelihood of current account reversals in Africa (Milesi-Ferretti and Razin, 1998). The higher levels of external debt in the region (relative to other developing countries) may be linked to policies that, in aiming to speed up the adjustment of the external gap, generate current account reversals. Since the current account determines the accumulation in foreign debt, there will be a critical level of accumulation above which the expected growth of a country will not be perceived by foreign lenders to be sufficient to generate the necessary repayment flows; the deficits will have to revert.9

A. Domestic Economy

Domestic output growth. This variable has a different impact in the two samples. For developing countries, we find a positive and significant coefficient (see regression [1] of Table 4). This is consistent with a situation in which domestic absorption is increasing more than exports. The income elasticity of foreign goods may be low.10 In contrast to the result for the sample of developing countries, the coefficient on domestic growth is negative and significant at the 10 percent level for the sample of African countries. A 1 percentage point rise in the GDP growth rate leads to decrease of about 0.22 percentage points in the current account deficit. This result is consistent with the predictions of the intertemporal approach to the current account, where domestic growth driven by transitory, country-specific productivity shocks might increase savings and, without investment effects, might help to reduce the current account deficit in Africa (Obstfeld and Rogoff, 1995).

Public and private savings. We find that the coefficients for both variables are negative and significant (at the 5 percent level) in both samples. For private savings, we find that the (absolute value of the) coefficient for African countries is almost two-and-a-half times larger than the coefficient for developing countries (-0.37 vs. -0.14, from regressions [1a] and [1] of Table 4). A transitory increase in private savings of 1 percent of GNDI may reduce the current account deficit by 0.41 percent. The larger impact generated by higher private savings is consistent with the idea that increases in consumption in Africa may be directly financed with a much higher proportion of foreign inflows than for the average developing country. However, the coefficient of public savings for African countries is twice as large as relative the one for the sample of developing countries (-0.41 vs. -0.20, from regressions [1a] and [1] of Table 4). The impact of a transitory increase in public savings of 1 percent of GNDI generates a reduction of CAD by 0.46 percentage point for African countries. The strong impact of increases in public savings on the current account suggests that IMF-designed programs, which usually employ fiscal consolidation as one of the tools to correct external imbalances, may be more effective in achieving the desired outcome in African countries.11

To further study the role of public savings, an additional experiment is conducted. We use central government expenditure (as a percentage of GNDI) instead of public savings in order to test whether a fiscal expansion (consolidation) is generated through a surge (decline) of government expenditure (regressions [4] and [4a] in Table 4). We find a positive and significant relationship between government expenditure and current account deficits. Although the “instantaneous-impact” coefficient is slightly larger for African economies (0.41 vs. 0.37), the total effect is larger for the sample of all developing countries. An increase of the government expenditure of 1 percent of GNDI leads to an increase of the current account deficit of 0.51 percentage point for African economies (vs. an increase of 0.74 percentage point for all developing countries). Although an increase in public savings is twice as effective in correcting imbalances in Africa, government expenditures have an impact only marginally greater on the current account than in the full sample of developing countries. Therefore, a reduction in public consumption may not cause a decline in interest rates, leaving investment unchanged and the full effect of the increase in savings may be passed through to the current account.

Macroeconomic uncertainty. This variable is proxied by the standard deviation of monthly consumer price inflation over the year (regressions [2] and [2a] in Table 4). The effect of uncertainty on the current account is negative for both samples but significant only for the sample of all developing countries, and its economic impact is quantitatively small. In general, this result is consistent with Ghosh and Ostry (1995), who argue that macroeconomic uncertainty lowers both investment and savings (via a precautionary savings motive) thus leading to a decline in the current account deficit. The full effect of uncertainty might already have been picked up by the savings variable itself.

B. External Sector

Openness, (exports as percentage of GNDI). We find that the coefficients are not statistically significant for either of the two samples. They remain non significant even when excluding variables that are expected to be correlated, such as the real exchange rate, terms of trade, and growth rate of industrialized economies).12 The effects of increased openness or increased foreign demand may be sufficiently represented by the appropriate variables in the regressions.

Black market premium. This variable is statistically significant for both samples. The black market premium may help reduce current account deficits for African countries. For instance, foreign exchange controls that increase the black market premium from 0 percent to 20 percent might lead to a reduction of current account deficits for African countries by 0.76 percentage point of GNDI13 The strong impact of the black market premium suggests that capital and exchange controls are effective in isolating African countries from real shocks in the external sector and, by allowing domestic interest rate to diverge, may thus stimulate automatic-adjustment-type mechanisms.

Balance of payments controls. This variable does not yield statistically significant transitory effects at the 5 percent level for either of the two samples (similar to Debelle and Faruqee, 1996; and Calderón, Chong, and Loayza, 1999).14 It appears that the effect is most likely to work through imperfect exchange rate flexibility. The impact of key external sector relative prices is proxied by the current account effects of both real exchange rate and terms of trade fluctuations.

Terms of trade. We find that the relationship between terms of trade and current account deficits is negative and statistically significant for both samples.15 Although the instantaneous impact coefficient is almost similar for the developing and African country sample (-0.029 and -0.024, respectively), the total effect is larger for the developing country sample, where a 10 percent increase in the terms of trade generates a reduction of the current account deficit by 0.62 percentage point (vs. 0.27 percentage point for the African country sample). A positive change in the terms of trade is linked with an improvement in the current account, provided that the Marshall-Lerner conditions hold. This finding is also consistent with the well-known Harberger-Laursen-Metzler effect, which states that adverse transitory terms of trade shocks generate a decline in the ratio of current to permanent income and a deterioration of the savings and current account positions (Obstfeld, 1982; Svensson and Razin, 1983; and Mendoza, 1995).

Real effective exchange rate. We find a significant relationship between this variable and the current account deficit for African countries, in line with the predictions of standard open economy (Mundell-Fleming) models, whereas we find the predicted positive relationship although statistically insignificant for the full sample of countries. For African countries, a 10 percent temporary decrease in the real effective exchange rate (i.e., a transitory depreciation of the domestic currency) reduces the current account deficit by 0.41 percentage point (vs. a reduction of 0.12 percentage point for the full sample). An appreciation of the real exchange rate may have adverse consequences for the current account deficit, and unlike in the full sample of developing countries, the effect in Africa may be stronger that that of the terms of trade. Real exchange rate movements may have a stronger impact on the current account deficit of African countries because exports of African countries appear to be more price sensitive than developing countries as a whole. This is consistent with the fact that African countries tend to export primary commodities and thus face relatively elastic international demand. While in developing countries as a whole exchange rate policy is not statistically significant, as the only effect on the CAD may be through shocks in the terms of trade, in Africa the strong impact of the real exchange rate, which is additional to the impact of the terms of trade, may indicate that there may be some room for exchange rate policy to influence current account behavior, at least in the short term.

International aid. We find that aid flows (as proxied by the ratio of effective development assistance to GDI) might help close the external gap. That is, the coefficient of aid flows is negative and significant at the 5 percent level for both samples. This result is consistent whether we control for savings or not. Thus, it is consistent with the findings of Boone (1996) that all aid flows are devoted to increasing consumption patterns. Given that the savings effect of aid inflows is controlled for in the regression, aid inflows to developing economies increase the disposable income, possibly generating a trade-off with declining investment levels. In addition, this increase in disposable income would put upward pressure on domestic interest rates, thus, encouraging domestic savings and discouraging investment (and establishing a negative relationship with current account deficits).

External debt. To test the impact of external indebtedness, we include an additional regressor to our benchmark regression in both samples. We add the ratio of total external debt to GNP lagged one period (see regressions [3] and [3a] in Table 4). The idea is to explore whether high levels of indebtedness today will generate an adjustment in the external gap in the future (i.e., we expect a negative coefficient). With the exception of the coefficients of public savings, private savings, and terms of trade, the other regressors lose significance and even change signs. Collinearity of this variable with the other regressors appears to be a problem. The high debt levels in Africa may therefore be a reflection of the need for adjustment.

C. Evolution of World Economy

Finally, we test the effect of the evolution of the world economy on the current account. First, we find that a temporary increase in the growth rate of industrial economies leads to a decline in the current account deficit for both samples; however, the impact is larger for the sample of all developing economies, where an increase of 1 percent in the industrial country growth rate lowers the CAD by 0.2 percentage point.16 However, once we include ancillary variables ([2a]-[4a] in Table 4) the sign of the coefficient on growth rate of industrial economies loses significance (but becomes positive). Second, we find a negative and significant relationship (10 percent) between the international real interest rate and current account deficits for both samples (regressions [1] and [1a] in Table 4). This result is consistent with the impact of international interest rate fluctuations on current account deficits for net foreign debtors (as most African economies are). We find that the total impact of international interest rate fluctuations on African economies is approximately one-andone-half times as high as the impact on all developing countries, perhaps a reflection of the high levels of debt in the region. A temporary rise in the international real interest rates of 1 percentage point might lead to a decline of the current account deficit of approximately 0.46 percentage point for African countries (vs. a 0.30 percentage point reduction for the whole sample).

Also in Table 4 notice that the specification tests support the GMM system panel estimator for all the regressions. First, the Sargan test of overidentifying restrictions cannot reject the null hypothesis that the instruments are not correlated with the error term. Second, serial correlation tests clearly reject the null hypothesis that the differenced error term is not first-order serially correlated, whereas they could not reject the presence of second- or third-order serial correlation. In short, both specification tests support the use of appropriate lags of the explanatory variables as instruments for estimation.

VI. Is The Current Account Deficit “Excessivein Africa?

In order to evaluate the contribution of the determinants to current account deficits in Africa, we calculate fitted values for the current account deficits, using both the sample of developing countries and the sample of African countries (i.e., using the coefficient estimates from equations [1] and [1a] in Table 4. These fitted values and the contribution to current account fluctuations from external, internal, and world economy conditions are reported in Table 5. This information is summarized in Table 6. Using the above, we investigate in this section whether the current account deficits are “excessive,” defined as the difference between the actual and the estimated fitted values. Although we do such a calculation for both the full developing country sample (equation [1]) and the sample of African countries (equation [1a]), we present and analyze in the text the estimates of the first only, as they give a better idea of the evolution of the fundamentals for the developing economies.17 Although the coefficients of equation [1a] give a better grasp of the impact of policy changes on current account deficits in African countries, estimates based on those coefficients would give misleading (that is, biased) information regarding the equilibrium levels of the current account deficit for the spectrum of developing economies.

Table 5.Decomposition of CAD Fluctuations in Africa, 1975-95 1/
Sources of CAD Fluctuations (Using Eq.(1) In Table 4)Sources of CAD Fluctuations [Using eq.(1a) in Table 4]
PeriodDom.Ext.WorldCountryCAD (% GNDI)Dom.Ext.WorldCountryCAD (% GNDI)
econ.eff.econeffectPredict.ActualDiff.econ.eff.econeffectPredictActualDiff.
North Africa
Egypt75-94-0.010.00-0.010.160.110.07-0.0402-0.030.03-0.010.010.080.07-0.0058
Morocco75-93-0.020.01-0.010.090.060.06-0.0030-0.030.03-0.01-0.030.060.06-0.0018
Tunisia83-94-0.020.01-0.010.120.080.05-0.0276-0.030.03-0.01-0.010.050.05-0.0037
Sub-Saharan Africa
Burkina Faso77-94-0.010.01-0.010.010.000.030.0315-0.020.03-0.01-0.070.030.030.0030
Burundi85-93-0.010.00-0.010.050.020.040.0248-0.020.03-0.01-0.050.040.040.0006
Cameroon75-93-0.020.00-0.010.160.110.08-0.0358-0.030.03-0.010.010.080.08-0.0056
Cent Afr Rep80-94-0.010.00-0.01-0.01-0.030.040.0721-0.010.03-0.01-0.080.040.040.0041
Cote d’lvoire75-93-0.010.01-0.010.120.110.10-0.0074-0.010.03-0.01-0.010.100.10-0.0033
Ethiopia85-94-0.010.00-0.010.060.040.02-0.0241-0.010.02-0.01-0.040.020.020.0001
Gabon75-92-0.050.01-0.010.200.060.00-0.0632-0.060.03-0.010.030.010.00-0.0081
Gambia76-93-0.010.00-0.010.130.090.09-0.0062-0.020.03-0.01-0.010.090.09-0.0037
Ghana75-94-0.010.01-0.01-0.05-0.040,020.0625-0.010.03-0.01-0.100.010.020.0064
Kenya75-93-0.020.01-0.010.090.050.05-0.0017-0.030.03-0.01-0.030.050.05-0.0015
Madagascar76-940.000.01-0.01-0.010.000.070.0763-0.010.03-0.01-0.080.070.070.0042
Malawi80-94-0.010.00-0.010.140.100.08-0,0248-0.020.02-0.010.000.080.08-0.0046
Mali88-93-0.020.01-0.010.150.120.04-0.0798-0.030.03-0.010.010.090.04-0.0533
Mauritania85-94-0.010.01-0.010.230.200.10-0.1015-0.020.02-0.010.050.110.10-0.0101
Mauritius77-93-0.020.01-0.010.100.060.04-0.0200-0.030.03-0.01-0.020.050.04-0.0023
Mozambique85-91-0.010.01-0.020.540.540.21-0.3274-0.030.02-0.010.210.240.21-0.0280
Nigeria75-93-0.020.01-0.010.060.030.040.0045-0.030.03-0.01-0.040.040.040.0002
Nigeria76-93-0.010.01-0.01-0.03-0.040.030.0708-0.020.03-0.01-0.090.020.030.0055
Rwanda76-93-0.010.00-0.010.04-0.010.010.0255-0.020.02-0.01-0.060.010.010.0015
Senegal75-930.000.01-0.010.030.040.070.0374-0.010.03-0.01-0.060.070.070.0017
Sierra Leone76-930.000.01-0.010.090.110.110.00360.000.03-0.01-0.030.110.11-0.0017
Sudan75-800.000.00-0.010.030.020.090.0629-0.020.020.00-0.060.020.090.0718
Togo77-94-0.020.01-0.010.100.050.050.0008-0.030.03-0.01-0.020.050.05-0.0025
Uganda84-94-0.010.01-0.010.010.000.020.0206-0.020.03-0.01-0.070.010.020.0032
Zaire75-89-0.010.01-0.01-0.04-0.030.050.0801-0.010.03-0.01-0.100.040.050.0062
Zambia85-94-0.010.01-0.010.110.110.07-0.0346-0.01-0.02-0.01-0,020.080.07-0.0030
Zimbabwe80-93-0.020.01-0.010.090.060.04-0.0187-0.030.03-0.01-0.030.040.04-0.0014

Authors’ calculations based on Table 4.

Authors’ calculations based on Table 4.

Table 6.Current Account Imbalances in Africa, 1975-95
RegionUsing Eq. (1) in Table 4Using eq. [1a] in Table 4
FittedActualImbalance 1/FittedActualImbalance 1/
Africa3.685.882.204.885.881.00
North Africa8.416.03-2.386.416.03-0.38
Sub-Saharan2.955.862.914.655.861.21
Source: Table 5.

The imbalance is the difference between the actual and the fitted value of the current account deficit.

Source: Table 5.

The imbalance is the difference between the actual and the fitted value of the current account deficit.

In Table 6, we find that current account deficits are excessive in Africa. They are more than 2 percent of GNDI over the “equilibrium” level (as calculated using the developing country sample). However, the behavior of current account deficits differs across regions. The excessiveness of African current account deficits is primarily driven by the external imbalances of sub-Saharan Africa, where the current account deficits are almost 3 percent of GNDI above the equilibrium level. Note that, regardless of the regression equation used, the same qualitative pattern of excessive current account deficits can be seen but the deviations from equilibrium are smaller when we use equation [1a]. Finally, even though the actual level of the current account deficits for North African countries is lower than the average for all developing countries, it is still significant (with the actual value being equal to 6 percent of GNDI). Across all the countries of north Africa, the behavior of the current account deficits is similar. Although these countries registered high levels of CAD (between 5 and 7 percent of GNDI), these deficits are below their equilibrium levels. However, the behavior of CAD across sub-Saharan African countries is more heterogeneous. Of the 27 countries reported in Table 5, 14 reported positive imbalances (“excessive” deficits; see the difference between predicted and actual values column when using equation [1] in Table 4). Among the countries with the highest (positive) imbalance are Zaire (8.01), Madagascar (7.63), Central African Republic (7.21), Sudan (6.29), and Ghana (6.25). In contrast, countries such as Cameroon, Côte d’lvoire, Gambia, Malawi, Sierra Leone and Zambia registered high current account deficits (above 7 percent of GNDI), but their gaps relative to the equilibrium level is small. Finally, from the decomposition of the current account fluctuations (using equation [1] in Table 4), we find that both internal conditions and the evolution of the world economy have helped close the external gap, while the external conditions (as an overall component) have broadly kept the deficit invariant. However, much of the deficit should be attributed to idiosyncratic factors specific to the African region (i.e. country effects). Note that this situation changes drastically if we conduct our analysis using equation [1a] in Table 4 as our benchmark of decompositions.

VII. Summary and Conclusions

In this paper we had two objectives. First, to understand the key determinants of current account deficits in Africa and how they compare with those of the developing world, and, second, to assess whether such current account deficits are “excessive,” defined as the difference between the actual and estimated fitted values obtained. On the first front, we show that for several key determinants Africa is, indeed, different. Specifically, we find that with respect to the African countries, (a) there is not as much persistence in current account deficits as in the full sample of developing countries, probably because the chances of reversals are greater; (b) unlike in the full sample of developing countries, domestic output growth is positively linked with current account deficits, possibly as a result of differences in income elasticity; (c) the impact of private savings on current account deficits is larger than in other regions, which is consistent with the idea that increases in consumption in Africa may be financed by foreign inflows to a larger extent; (d) fiscal consolidation of IMF-designed programs may be more effective than elsewhere as the impact of public savings on the current account deficit is larger in Africa than in the average developing country; the larger the debt level, the stronger the fiscal impact; (e) macroeconomic uncertainty, openness, and balance of payment controls, are not statistically significant; (f) the high level of debt appears to signal the need for adjustment; (g) aid flows help close the external gap; (h) unlike in the average developing country, a depreciation of the currency reduces the current account deficit, at least in the short run; and (i) the impact of the terms of trade is consistent with the Harberger-Laursen-Metzler effect.

Based on the findings of the first section, we are able to calculate whether the current account deficit in Africa is “excessive” with respect to the fundamentals, as estimated from the full sample of developing countries. In fact, we conclude that the current account deficit is almost 3 percent of the gross national disposable income above the equilibrium level. Both, internal conditions and global changes in the world economy have helped reduce the external gap, but it appears that much of the current account deficit imbalance with respect to equilibrium is country specific. Our findings above open at least one possible path for future research. Since we have used annual data for the period 1975-95, we have mostly captured the transitory “short-run” effects of changes in policy variables. We cannot say much about permanent, long-run effects that deal with current account sustainability. The obvious question is to ask whether Africa is also different with respect to permanent effects, not just transitory shocks. If so, a key issue is to investigate whether the long-run current account deficit is “excessive” or not.

References

    Alonso-BorregoCesar and ManuelArellano1999Symetrically Normalized Instrumental-Variable Estimation Using Panel DataJournal of Business and Economic StatisticsVol. 17 (January1999) pp. 36-49.

    ArellanoManuel and StephenBond1991Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment EquationsReview of Economic StudiesVol 58 (April): pp. 277-97.

    ArellanoManuel and OlympiaBover1995Another look at the Instrumental Variable Estimation of Error-Component ModelsJournal of Econometrics 68: pp. 29-51.

    BlundellRichard and StephenBond1997Initial Conditions and Moment Restrictions in Dynamic Panel Data ModelsDiscussion Papers in Economics n. 97-07 (London: University College).

    BoonePeter(1996) “First Aid for the Aid BudgetEconomic Affairs (U.K.); Vol 16 (Autumn): pp. 23-6.

    CalderónCésarAlbertoChong and NormanLoayza1999Determinants of Current Account Deficits in Developing CountriesPolicy Research Working Paper The World Bank (Washington: The World Bank).

    CashinPaul and JohnMcDermott1998Are Australia’s Current Account Deficits Excessive?Economic Recordforth coming.

    ChangCharlesEduardoFernandez-Arias and LuisServen1999Measuring Aid Flows, A New ApproachWorld BankPolicy Research Working Paper n. 2050 (Washington: World Bank).

    CollierPaul and JanWillem Gunning1999Explaining African Economic PerformanceJournal of Economic Literaturevol. 37n1 (March) pp. 64-111.

    DebelleGuy and HamidFaruqee1996What Determines the Current Account? A Cross-Sectional and Panel ApproachIMF Working Paper 96/58 (Washington: International Monetary Fund).

    EasterlyWilliamNormanLoayza and PeterMontiel1997Has Latin America’s Post-Reform Growth Been Disappointing?Journal of International EconomicsVol. 43 (November) pp. 287-311.

    EasterlyWilliam and RossLevine1997Africa’s Growth Tragedy: Policies and Ethnic DiversitiesQuarterly Journal of EconomicsVol. 112 (November) pp. 1203-50.

    FrenkelJacob A. and AssafRazin1996Fiscal Policies and Growth in the World Economy (Cambridge, Mass.: M.I.T. Press).

    GlickReuven and KennethRogoff1995Global versus Country-Specific Productivity Shocks and the Current AccountJournal of Monetary EconomicsVol. 35 (February): pp. 159-92.

    GhoshAtish1995International Capital Mobility Amongst the Major Industrialized Countries: Too Little or Too Much?The Economic Journal 105: pp. 107-128.

    GhoshAtish and JonathanOstry1995The Current Account in Developing Countries: A Perspective from the Consumption-Smoothing ApproachWorld Bank Economic ReviewVol 9 (February): pp. 305-33.

    GrilliVittorio and GianMaria Milesi-Ferreti1995Economic Effects and Structural Determinants of Capital ControlsStaff Papers International Monetary FundVol. 42 (September) pp. 517-551.

    Holtz-EakinDouglasWhitneyDewey and Harvey S.RosenEstimating Vector Autoregressions with Panel DataEconometricaVol. 56 (November) pp. 1371-95.

    International Monetary Fund (various years)International Finance Statistics.

    KivietJan F.1995On Bias, Inconsistencies, and Efficiency of Various Estimators in Dynamic Panel Data ModelsJournal of EconometricsVol. 68 (July) pp. 53-78.

    LeidermanLeonardo and AssafRazin1991Determinants of External Imbalances: The Role of Taxes, Government Spending and ProductivityJournal of the Japanese and International EconomiesVol. 5 (December) pp. 421-50.

    LoayzaNormanHumbertoLópezKlausSchmidt-Hebbel and LuisServén1998The World Saving Database” (Unpublished; Washington: The World Bank).

    MansoorianArman1998Habits and Durability in Consumption, and the Dynamics of the Current AccountJournal of International Economics44: pp. 69-82.

    MendozaEnrique G.1991Capital Controls and the Gains from Trade in a Business Cycle Model of a Small Open EconomyStaff Papers 38 International Monetary FundVol. 38 (September) pp. 480-505.

    MendozaEnrique G.1995The Terms of Trade, the Real Exchange Rate, and Economic FluctuationsInternational Economic ReviewVol. 36 (February) pp. 101-37.

    Milesi-FerretiGianMaria and AssafRazin1998Current Account Reversals and Currency Crises: Empirical RegularitiesNBER Working Paper n. 6620 (Cambridge, Massachussets: National Bureau of Economic Research).

    ObstfeldMaurice1982Aggregate Spending and the Terms of Trade: Is There a Laursen-Harberger-Metzler Effect?Quarterly Journal of EconomicsVol 97 (May) pp. 251-70.

    ObstfeldMaurice and KennethRogoff1995The Intertemporal Approach to the Current AccountHandbook of International EconomicsVol. 3ed. by GeneGroomsman and KennethRogoff(Amsterdam: North-Holland).

    RazinAssaf1995The Dynamic-Optimizing Approach to the Current Account: Theory and EvidenceUnderstanding Interdependence: The Macro-economics of the Open Economyed. by PeterKenen(Princeton, New JerseyPrinceton University Press) pp 169-98.

    SachsJeffrey1982The Current Account in the Macroeconomic Adjustment ProcessScandinavian Journal of EconomicsVol. 84 (n. 2) pp. 147-59.

    SheffrinSteven and WingThye Woo1990Present Value Tests of an Intertemporal Model of the Current AccountJournal of International Economics 29: pp. 237-53.

    StockmanAlan1987The Equilibrium Approach to Exchange RatesFederal Reserve Bank of Richmond Economic ReviewMarch-April pp 12-31.

    SvenssonLars E. O. and AssafRazin1983The Terms of Trade and the Current Account: The Harberger-Laursen-Metzler EffectJournal of Political EconomyVol. 91 (February) pp. 97-125.

    TornellAaron and PhilipLane1998Are Windfalls a Curse? A Non-Representative Agent Model of the Current Account and Fiscal PolicyJournal of International EconomicsVol. 44 (February) pp. 83-112.

    WoodAdrian1988Global Trends in Real Exchange Rates, 1960-84World Bank Discussion Paper 35.

    World Bank1997World Development Indicators 1997 (Washington: World Bank) available on CD-ROM.

    World Bank1998Assessing Aid: What Works What Doesn’t and Why (New York: Oxford University Press for the World Bank).

    ZiliakJ.(1997) “Efficient Estimation with Panel Data When Instruments Are Predetermined: An Empirical Comparison of Moment-Condition EstimatorsJournal of Business Economics and Statistics15 (4).

APPENDIX I

Dynamic Panel Data Econometric Methodology18

Recently developed dynamic panel data techniques allow us to address potential endogeneity problems, as well as possible unobserved time and country-specific effects that may produce biased and inconsistent estimates.19 This methodology formulates a set of moment conditions that can be estimated using GMM techniques in order to generate consistent and efficient estimates. Specifying the regression equation in differences allows elimination of the country-specific effect. First-differencing yields

The use of instruments is required to deal with two issues: first, the likely endogeneity of the explanatory variables, X, which is reflected in the correlation between these variables and the error term; and, second, the correlation of the new error term, (εi,t - εi,t-1), by construction with the differenced lagged dependent variable, (yi,t-1 - yi,t-2). Instead of assuming strict exogeneity (that is, the explanatory variables be uncorrected with the error term at all leads and lags), we allow for the possibility of simultaneity and reverse causation. We adopt the more flexible assumption of weak exogeneity, according to which current explanatory variables may be affected by past and current realizations of the dependent variable but not by its future innovations. Under the assumptions that (a) the error term, ε, is not serially correlated, and (b) the explanatory variables are weakly exogenous, the following moment conditions apply:

The GMM estimator simply based on the moment conditions in (2) and (3) is known as the differences estimator. Although asymptotically consistent, this estimator has low asymptotic precision and large biases in small samples, which leads to the need to complement it with the regression equation in levels.20 For the regression in levels, the country-specific effect is not directly eliminated but must be controlled for by the use of instrumental variables. The appropriate instruments for the regression in levels are the lagged differences of the corresponding variables if the following assumption holds; although there may be correlation between the levels of the right-hand side variables and the country-specific effect, there is no correlation between the differences of these variables and the country-specific effect. This assumption results from the following stationarity property,

Therefore, the additional moment conditions for the second part of the system (the regression in levels) are given by the following equations:21

Using the moment conditions presented in equations (2), (3), (5) and (6), and following Arellano and Bond (1991) and Arellano and Bover (1995), we employ a generalized method of moments (GMM) procedure to generate consistent estimates of the parameters of interest. The weighting matrix for GMM estimation can be any symmetric, positive-definite matrix, and we obtain the most efficient GMM estimator if we use the weighting matrix corresponding to the variance-covariance of the moment conditions. Since this variance-covariance is unknown, Arellano and Bond (1991) and Arellano and Bover (1995) suggest the following two-step procedure. First, assume that the residuals, are independent and homoskedastic both across countries and over time. This assumption corresponds to a specific weighting matrix that is used to produce first-step coefficient estimates. We construct a consistent estimate of the variance-covariance matrix of the moment conditions with the residuals obtained in the first step, and we use this matrix to reestimate our parameters of interest (i.e. second-step estimates). Asymptotically, the second-step estimates are superior to the first-step ones in so far as efficiency is concerned. The moment conditions are applied such that each of them corresponds to all available periods, as opposed to each moment condition corresponding to a particular time period. In the former case, the number of moment conditions is independent of the number of time periods, whereas in the latter case, it increases more than proportionally with the number of time periods. Most of the literature dealing with GMM estimators applied to dynamic models of panel data treats the moment conditions as applying to a particular time period. This approach is advocated on the grounds that it allows for a more flexible variance-covariance structure of the moment conditions. Such flexibility is achieved without placing a serious limitation on the degrees of freedom required for estimation of the variance-covariance matrix because the panels commonly used in the literature have both a large number of cross-sectional units and a small number of time-series periods (typically not more than five). We have, however, chosen to work with the more restricted application of the moment conditions (each of them corresponding to all available time periods) because of a special characteristic of our panel, namely, its large time-series dimension (for some countries in our sample, we work with as many as 20 time-series observations). This approach allows us to work with a manageable number of moment conditions, so that the second-step estimates, which rely on estimation of the variance-covariance matrix of the moment conditions, do not suffer from overfitting biases.

APPENDIX II

Countries in the Sample

Code

Country

ARG

Argentina

BDI

Burundi

BFA

Burkina Faso

BGD

Bangladesh

BLZ

Belize

BOL

Bolivia

BRA

Brazil

CAF

Central African Rep.

CHL

Chile

CHN

China

CIV

Côte d’lvoire

CMR

Cameroon

COL

Colombia

CRI

Costa Rica

DOM

Dominican Rep.

ECU

Ecuador

EGY

Egypt

ETH

Ethiopia

GHA

Ghana

GTM

Guatemala

HND

Honduras

HTI

Haiti

IDN

Indonesia

IND

India

IRN

Iran, Islamic Rep. of

JAM

Jamaica

JOR

Jordan

KEN

Kenya

KOR

Korea, South

LKA

SriLanka

MAR

Morocco

MDG

Madagascar

MEX

Mexico

MLI

Mali

MLT

Malta

MOZ

Mozambique

MRT

Mauritania

MUS

Mauritius

MWI

Malawi

MYS

Malaysia

NER

Niger

NGA

Nigeria

NPL

Nepal

PAK

Pakistan

PAN

Panama

PER

Peru

PHL

Philippines

PNG

Papua New Guinea

PRY

Paraguay

RWA

Rwanda

SDN

Sudan

SEN

Senegal

SLE

Sierra Leone

SLV

El Salvador

SYR

Syrian Arab Rep.

TGO

Togo

THA

Thailand

TTO

Trinidad and Tobago

TUN

Tunisia

TUR

Turkey

UGA

Uganda

URY

Uruguay

ZAR

Zaire

ZWE

Zimbabwe

APPENDIX III

Current Account Balances in Selected African countries

(in share of GDP)
(in share of GDP)
(in share of GDP)
(in share of GDP)
(in share of GDP)
(in share of GDP)
(in share of GDP)
(in share of GDP)
(in share of GDP)
(in share of GDP)

César Calderón: University of Rochester; Alberto Chong: The World Bank and Georgetown University; Luisa Zanforlin: International Monetary Fund. We would like to acknowledge comments and suggestions from Norman Loayza and Sergio Pereira Leite.

These data are available at the World Bank website at www.worldbank.org.

Countries are assigned to regions according to standard World Bank classification.

It is also positive but statistically nonsignificant for North Africa.

Also, balance of payments controls are positive and statistically significant for North Africa.

The current account deficit, as defined in Table 1, implies that positive (negative) numbers indicate a deficit (surplus).

That is, we are not decomposing for long-term, permanent effects.

The value of the coefficient for lagged CAD implies that the half-life of transitory shocks on the current account deficit is about six years.

The low persistence of current account deficits in Africa may be linked to the occurrence of the high levels of foreign debt in the context of the region’s low growth. CAD in an African country is expected to be adjusted more quickly than in an average developing country as the stock of debt makes an open position unsustainable, even in the very short term. This hypothesis is supported by the fact that, when debt is included among the regressors, the lagged CAD (persistence) becomes non significant. See column [3a] in Table 4.

This is consistent with a situation in which “productive” imports are manufactured domestically. For instance, machinery might be tied to investment projects that are not sensitive to domestic conditions.

The impact of public savings (and government expenditures) is larger than the impact of private savings on CAD in both samples. Countries with large (external and internal debts) have a greater need for adjustment; thus fiscal adjustments may affect even more the external gap. In other words, the strength of the link between the fiscal deficit and current account deficit depends on the level of debt of the country.

These additional regressions, though not reported, are available upon request.

However, for the full sample, the direction of the effect is reversed. An equivalent increase in the black market premium will yield a reduction of 0.40 percent for the full sample.

However, the developing country sample is significant at 10 percent. Notice that our proxy for balance of payment controls exhibits small variation over time and does not measure accurately the intensity of controls, only their presence (Grilli and Milesi-Ferreti, 1995).

Terms of trade are constructed from price indexes (in U.S. dollars) for exports and imports from the World Bank (1997).

Temporary output surges in industrial economies lead to a reduction of current account deficits of African economies because of the increasing demand for their exports.

Easterly, Loayza, and Montiel (1997) follow a similar approach.

This section draws heavily from Calderón, Chong and Loayza (1999). Also, we would like to thank Norman Loayza who generously contributed to this section.

Alonso-Borrego and Arellano (1999) and Blundell and Bond (1997) show that when the lagged dependent and the explanatory variables are persistent over time, lagged levels of these variables are weak instruments for the regression equation in differences. This weakness has repercussions for both the asymptotic and small-sample performance of the differences estimator. As persistence increases, the asymptotic variance of the coefficients obtained with the differences estimator rises (i.e., deteriorating its asymptotic precision). Furthermore, Monte Carlo experiments show that the weakness of the instruments produces biased coefficients in small samples. This is exacerbated with the variables’ over-time persistence, the importance of the specific effect, and the smallness of the time-series dimension.

Given that lagged levels are used as instruments in the differences specification, only the most recent difference is used as instrument in the levels specification. Other lagged differences would result in redundant moment conditions. (Arellano and Bover, 1995).

Other Resources Citing This Publication