Buoyant Capital Spending and Worries over Real Appreciation: Cold Facts from Algeria

Oumar Diallo; Boileau Loko; Kangni R Kpodar

doi:10.5089/9781451868494.001.A001

Jump to Content Jump to Main Navigation

Buoyant Capital Spending and Worries over Real Appreciation: Cold Facts from Algeria

Author: Oumar Diallo¹, Mr. Boileau Loko², and Mr. Kangni R Kpodar

¹ 0000000404811396https://isni.org/isni/0000000404811396International Monetary Fund
| ² 0000000404811396https://isni.org/isni/0000000404811396International Monetary Fund

Contributor Notes

Authors’ E-Mail Addresses: bloko@imf.org, kkpodar@imf.org, dialloo@un.org

Publication Date:: 01 Dec 2007

eISBN:: 9781451868494

Language:: English

Keywords:: WP; exchange rate

Download PDF (173.9 KB)

The Government of Algeria has pursed a relatively expansionary fiscal policy in recent years, thanks to rising oil prices and revenues. The paper explores the potential effects of such a stance on real exchange rate and uncovers a relatively small appreciating effect of increased government capital expenditure. This is explained by the fact that a significant share of capital spending falls into tradable imported goods. However, the envisaged increase in capital spending, if well designed and implemented, might in the long-run translate into rising operations and maintenance expenditure-mostly nontradable goods-thereby causing a higher real appreciation. This implies that Algeria should carefully consider the implications of its public investment program on recurrent expenditure.

Abstract

I. Introduction

A favorable external environment and appropriate macroeconomic policies have contributed to Algeria’s encouraging economic performance over the past five years. Economic growth has relatively been brisk, inflation has remained low, and current account position has strengthened markedly. Also, the authorities decided to use part of the fiscal space created by hydrocarbon revenues to increase public spending. The government of Algeria has launched an ambitious five-year public investment program, the Growth Consolidation Plan (GCP), which aimed at improving infrastructure, housing and the delivery of public services.

The recent surge in public spending, in particular in capital expenditure, carries some risks. In particular, increased domestic use of hydrocarbon revenues may lead to an appreciation of the real exchange rate, which has the potential to undermine the competitiveness of the nonhydrocarbon tradable sector and to hamper the efforts of the Algerian government to expand economic activity beyond the hydrocarbon sector.² Therefore, understanding the impact of government spending on the economy, and particularly on the real effective exchange rate, is potentially of great importance to economic policy-makers.

The empirical work carried out in this paper suggests that, in the case of Algeria, the increase in government capital spending has had so far a limited appreciating impact on the real exchange rate. We argue that government spending generates foreign exchange outflows, insofar as public spending has high import content, attenuating somewhat the appreciating effect of the spending. However, along with ensuring fiscal sustainability and the quality of capital spending, the authorities should carefully consider the implications of the public investment program on future recurrent expenditure, as these goods are mostly nontradables.

The remaining of the paper is organised as follows. Section II provides a brief overview of fiscal policy in Algeria. Section III highlights the theoretical arguments underlying the nexus between fiscal policy and real exchange rate, and then confronts these arguments with empirical evidence, relying on a revised version of the theoretical model developed by Cashin and al. (2004). Finally, the last section concludes and draws some policy implications from our analysis.

II. Fiscal Policy Stance in Algeria

The hydrocarbon sector dominates the Algerian economy, accounting for almost 45 percent of GDP in 2005 and 98 percent of exports of goods. The state-owned petroleum company (Sonatrach) fully repatriates and surrenders foreign currency receipts generated by hydrocarbon exports to the bank of Algeria.³ Such earnings are generally divided between the government, and Sonatrach and its subsidiaries. On average, two-third of export receipts accrue to the treasury, making government revenue highly dependent on oil prices⁴ fluctuations, thus volatile. Similarly, public spending appears erratic, following very much the same features as those displayed by government revenue and oil prices.

Indeed, high hydrocarbon prices allowed the government of Algeria to increase public spending in the 1970s and in the beginning of the 1980s. Capital expenditure accounted for much of these increases. The dynamics changed when hydrocarbon prices collapsed during the second half of the 1980s. Public investment was cut significantly, but current expenditure remained high (Figure 1). The outcomes of maintaining relatively high levels of current spending were rising fiscal deficits, depleting official reserves, extensive external and internal borrowing, rising inflation, and intensifying import restrictions. The country then embarked on a Fund-supported program in 1994 that was accompanied by a fiscal tightening during 1994–98.

Thereafter, the Algerian government has pursued a somewhat expansionary fiscal policy thanks to rising hydrocarbon prices. The government implemented the economic recovery Program (ERP) between 2001 and 2004. Capital expenditure increased from about 8 percent of GDP in 2001 to 11 percent of GDP in 2004. In 2005, the authorities launched the Growth Consolidation Program⁵ (2005–09) with the objective of improving infrastructure and generating employment. As the program was in the early stages of implementation, the execution rates of capital spending were low during 2005. Therefore, capital spending remained stable at about 11 percent of GDP in 2005, although the composition of public spending has shifted towards capital spending (Figure 2).

In a country that is still recovering from a period of political instability and civil strive, as it is the case in Algeria, policy makers have to weigh the medium and long-term benefits of a prudent fiscal stance against the immediate needs of a fragile economy and social demands. However, significant and rapid increases carries some risks. First with planning, implementation and management capacity being limited, there is a risk that additional fiscal resources are channelled to inefficient and unstainable capital projects and current expenditure programs. Second, the current fiscal policy stance might generate inflationnary pressures and eventually lead to real exchange rate appreciation. That may contribute to deteriorating the competiveness of the nonhydrocarbon traded goods sector.

III. Fiscal Policy and the Real Exchange Rate in Algeria

A. Analytical Framework

The theoretical framework builds on Cashin and others’ (2004) model which links the real exchange rate to real commodity prices. We extend the model by incorporating government spending so as to capture its effect on the real exchange rate.⁶

There are three goods (nontradable, exportable—primary commodity—and final tradable), two economic agents⁷ (household/firms and government). The country is a small open economy that produces a nontradable good and an exportable good that is assumed to be a primary commodity (not consumed domestically). Both goods are produced by firms using constant-returns- to scale technology.

Labor is the only factor of production and it is freely mobile between sectors, thereby ensuring that wages are equated across sectors and that only supply-side factors are relevant. We assume that labor supply is inelastic. Domestic households consume both a nontraded and a final tradable good, imported from abroad and not produced domestically. Foreign firms use the primary commodity joint with an intermediate good, produced only abroad, as inputs in the production process of the final tradable good. Foreign households consume both nontraded and final tradable goods.

We assume that the government (either domestic or foreign) consumes both a nontradable and a final tradable good just like the representative household does, but the parameters of its utility function are different. Specifically, it is assumed that the government’s preference for the nontraded good (or the traded good) differs from the households’⁸.

The real exchange rate is defined as the foreign price of the domestic basket of consumption(EP)relative to the foreign price of the foreign basket of consumption (P^*). The real effective exchange rate (REER) is given by the following equation (see Appendix 1 for more details):

\begin{matrix} \frac{EP}{P^{*}} = {(\frac{a_{x}}{a_{1}^{*}} \frac{a_{N}^{*}}{a_{N}} \frac{P_{X}^{*}}{P_{1}^{*}})}^{γ (1 - Pu) + g Pu} & (1) \end{matrix}

where $\frac{P_{X}^{*}}{P_{1}^{*}}$ represents the commodity terms of trade measured in foreign prices, $\frac{a_{x}}{a_{1}^{*}}$ denotes the productivity differentials between the export and the import sectors, $\frac{a_{N}^{*}}{a_{N}}$ captures the productivity differentials between the local and foreign nontraded sectors, γ is the elasticity of household consumption or spending to the nontraded good, g captures the elasticity of government consumption or spending to the nontraded good, and P_u is the government’s share in total aggregate demand.

The terms $\frac{a_{x}}{a_{1}^{*}}$ and $\frac{a_{N}^{*}}{a_{N}}$ embody the Balassa-Samuelson effect, whereby higher productivity in the commodity sector will tend to increase wages and the price of the nontraded good, and ultimately result in an appreciation of the real exchange rate. The parameter g (respectively γ) reflects the government (respectively households) relative preference for the nontraded good. A value of g higher than γ would imply that government spending is relatively more biased toward the nontraded good compared with household spending. Conversely a value of g lower than γ would imply that government spending is relatively biased toward the traded good.⁹

Taking the log of equation (1) yields:

\begin{matrix} Log (REER) = (γ (1 - Pu) + g Pu) Log (\frac{a_{x}}{a_{1}^{*}} \frac{a_{N}^{*}}{a_{N}} \frac{P_{X}^{*}}{P_{1}^{*}}) & (2) \end{matrix}

The marginal effect of a change in real price of the primary commodity is given by:

\begin{matrix} \frac{\partial Log (REER)}{\partial Log (\frac{P_{X}^{*}}{P_{1}^{*}})} = γ (1 - Pu) + gPu & (3) \end{matrix}

Equation (3) suggests that a rise in the relative price of the primary commodity will result in an appreciation of the real exchange rate. The increase in commodity price boosts wages in the primary commodity sectors, and as wages are assumed to equalize across sectors owing to labor mobility, both the wages and the price of the nontraded sector will increase as well.

The marginal impact of a change in government spending on the real exchange rate is as follows:

\begin{matrix} \frac{\partial Log (REER)}{\partial Pu} = α Log (\frac{a_{x}}{a_{1}^{*}} \frac{a_{N}^{*}}{a_{N}} \frac{P_{X}^{*}}{P_{1}^{*}}) & (4) \end{matrix}

With α = g − γ

Since much of the current public spending and a non-negligible share of capital spending fall on nontradables in Algeria, we assume that the share of non-traded goods in government spending is much higher than that the share of non-traded goods in private spending (α ≻ 0), just as in many other developing countries.

Equation (4) indicates that increasing government spending has an appreciating effect, although the magnitude of such an appreciating effect depends on relative share of on non-traded goods in government spending. The higher is this share, the more pronounced is the appreciation of the real exchange rate. In addition, an increase in the relative share of the nontraded good in public spending with an unchanged level of public spending leads unambiguously to an appreciation of the reel exchange rate as suggested by equation (5).

\begin{matrix} \frac{\partial Log (REER)}{\partial g} = Pu Log (\frac{a_{x}}{a_{1}^{*}} \frac{a_{N}^{*}}{a_{N}} \frac{P_{X}^{*}}{P_{1}^{*}}) & (5) \end{matrix}

These findings are similar to those of Sachs and Wyplosz (1984), who developed a model where the effect of fiscal policy changes on the real exchange rate lies on the composition of government spending relative to that of the private sector. If the public marginal propensity to import is lower than that of the private sector, the fiscal expansion alters the composition of spending by shifting demand away from foreign goods toward domestic goods, thus leading to a real exchange rate appreciation.

B. Empirical Model Specification

Based on the above analytical framework, we assume that, in a commodity dependant economy such as Algeria, three factors are likely to be important in explaining the real effective exchange rate:

The real price of the primary commodity (the world price of oil): Given that hydrocarbons account for a substantial fraction of Algeria’s exports earnings, we use world price of oil in real terms, as a proxy for external terms of trade. Higher oil prices lead to higher wages which in turn would result in higher prices for nontradables and thus in real appreciation.

Differential productivity: Productivity convergence, according to the Balassa-Samuelson effect, would be associated with a real exchange rate appreciation. We expect increased relative productivity in the economy, particularly in the tradable sector, to lead to a real appreciation.

Government expenditure: We expect higher oil prices to lead to higher government spending, which in turn would appreciate the real exchange rate. However, the magnitude of the appreciation lies on a number of factors including the share of tradables and nontradables in government spending.

The empirical model for the real exchange rate (REER) can be represented by:

\begin{matrix} {REER}_{t} = f ({PROD}_{t}, {Oil}_{t}, {GK}_{t}) & (6) \end{matrix}

where REER represents the real effective exchange rate, PROD, the differential productivity between the home country and abroad, Oil, the real oil price, and GK is government capital spending as a share of GDP.

Following the Johansen (1988, 1995) maximum likelihood methodology, the complete empirical model specification can be represented as a vector autoregressive (VAR), with the following form:

\begin{matrix} X_{t} = ϕ^{0} + \sum_{j = 1}^{q} ϕ_{j} X_{t - 1} + ω_{t} & (7) \end{matrix}

where $X_{t} = [\begin{array}{l} {REER}_{t} \\ {PROD}_{t} \\ {Oil}_{t} \\ {GK}_{t} \end{array}]$ while ω_t, a (4*1) vector, captures the errors terms, ϕ_j denotes a (4*4) matrix of coefficients on lags of X_t; ϕ⁰ represents a (4*1) vector of deterministic variables, and q is the number of lags. Further, we assume ω_t to be a vector of independent and identically distributed errors (iid) that follows a normal distribution, with zero mean and covariance matrix Ω.

We transform relation (6) into a vector error correction model (VECM) by adding and taking away lags of X_t and obtain the following:

\begin{matrix} Δ X_{t} = ϕ^{0} + ϕ X_{t - 1} + \sum_{j = 1}^{q - 1} Δ Π_{j} X_{t - j} + υ_{t} & (8) \end{matrix}

where $ϕ \equiv (\sum_{j = 1}^{q} ϕ_{i}) - I$ and Π_j = –(ϕ_j+1+…+ϕ_q).

The number of cointegration relations is equal to the rank ofϕ. Thus, a rank ofϕ, equal to v,¹⁰ means that there are v cointegrating vectors; ϕ can then be represented as a combination of two full-column rank (4*v) matrices: φ and χ (ϕ =φχ’). Substituting such a relation in (8) leads to:

\begin{matrix} Δ X_{t} = ϕ^{0} + φ χ^{'} X_{t - 1} + \sum_{j = 1}^{q - 1} Δ Π_{j} X_{t - j} + υ_{t} & (9) \end{matrix}

It is worth noting that the matrix χ’ contains the cointegrating vectors while ϕ columns capture the feedback effects.

C. Data, Estimation, and Results

The estimation uses annual data series covering the period 1970–2006, and the variables of interest are those highlighted in the above model specification. The data are drawn mostly from the World Economic Outlook (WEO) database (See Appendix 2). In order to estimate the empirical model, we first test for the presence of unit roots in the series using the Augmented Dickey Fuller (ADF) and the Phillips-Perron (PP) tests. Both tests do not reject the null hypothesis of a unit root. All series are integrated of order one (Table 1).

Table 1.

Unit-Root Tests

Notes: LREER, LOIL, LPROD, LGK stand for the log of the real effective exchange rate, the log of the real price of oil, the log of the productivity differential, and the log of the ratio capital expenditure to GDP, respectively. The lag length is determined based on Schwarz information criterion (SIC). *Significant at 1 percent; ** Significant at 0.1 percent; *** Significant at 0.01 percent.

Table 1.

Unit-Root Tests

	ADF	ADF (with trend)	PP	PP (with trend)
*Level*
LREER	-0.175	-2.325	-0.369	-1.765
LOIL	-2.272	-2.248	-2.274	-2.248
LPROD	-0.262	-2.158	-0.179	-2.080
LGK	-2.333	-2.712	-2.263	-2.154
*First Difference*
LREER	-4.439*	-4.428*	-4.424*	-4.464*
LOIL	-6.106***	-6.030***	-6.106***	-6.030***
LPROD	-8.428***	-8.608***	-7.750***	-7.854***
LGK	-7.220***	-7.061***	-6.828***	-6.666***

Table 1.

Unit-Root Tests

	ADF	ADF (with trend)	PP	PP (with trend)
*Level*
LREER	-0.175	-2.325	-0.369	-1.765
LOIL	-2.272	-2.248	-2.274	-2.248
LPROD	-0.262	-2.158	-0.179	-2.080
LGK	-2.333	-2.712	-2.263	-2.154
*First Difference*
LREER	-4.439*	-4.428*	-4.424*	-4.464*
LOIL	-6.106***	-6.030***	-6.106***	-6.030***
LPROD	-8.428***	-8.608***	-7.750***	-7.854***
LGK	-7.220***	-7.061***	-6.828***	-6.666***

In the second step, we check for the existence of cointegration among the real effective exchange rate, the differential productivity, the real oil price and government spending using the Johansen cointegration test. However, many authors (for instance Gonzalo, 1994) indicate that the Johansen approach may be sensitive to the choice of the lag length. In this context, we use the AIC procedure to choose the number of lags (4).

Another important shortcoming with the Johansen approach is the bias associated with the use of small samples (Johansen, 2002). Many studies reveal that in small samples, the Johansen approach fails to reject the null hypothesis of noncointegration even if there is no cointegration vector. One way of addressing the bias is to correct the test statistic so that the finite sample distribution is closer to the asymptotic distribution (Reinsel and Ahn, 1988, Johansen, 2002).¹¹

To correct for a possible small sample bias, we use Reinsel and Ahn’s approach which suggests that the critical value be adjusted upward by a multiplicative scaling factor, T/(T-nj), with T the sample size, n the number of variables, and j the number of lags. After correcting the critical values, both Trace and Max-Eigen statistics reject the null hypothesis of zero cointegrating equation. Both tests indicate the presence of one cointegrating vector (Table 2).

Table 2a.

Trace Statistics for Cointegrating Rank