Theoretical Framework and Empirical Analysis

54. We postulate a demand for real money balances, which is a function of real GDP, inflation and interest rates. The model can be summarized as follows:

m = f(y, π, i)

where,

$m=\frac{M^d}{P}=$ demand for real money (M₂ aggregate in real terms)

y = real income (GDP at constant prices)

π = inflation rate (annual change in the average CPI)

i = interest rate (discount rate)

55. Annual data for the period 1974–2001 are used.²⁷ The variables are expressed in logarithmic terms, with the exception of the interest rate and inflation. The inflation rate is measured by the variation of price levels in logarithmic terms.

56. The empirical investigation starts with the analysis of the time-series properties of all variables. The augmented Dickey–Fuller (1981) (ADF) indicates that the nonstationarity hypothesis cannot be rejected at 5 percent confidence in the monetary aggregates, real GDP, the annual inflation rate, and the interest rates. However, for the first difference of the same variables, the hypothesis of nonstationarity is rejected at 5 and 1 percent confidence, suggesting that these variables are integrated of order one I(1) (Table 4).

Table 4.

Statistics for ADF(2) Unit Root Tests

Notes: Variables are as defined in the text. For each variable expressed in level (first difference), the Augmented Dickey-Fuller (1979) ADF(2) statistics tests a null hypothesis of a unit root in that variable expressed in level (first difference) against an alternative of a stationary root, The criterion for lag selection is based on the Akaike information criterion, as described by Pantula et. Al. (1994). The critical values are taken from MacKinnon (1991). + ^*, and ^** denote rejection at 10 percent, 5 percent and 1 percent critical values, respectively.

Table 4.

Statistics for ADF(2) Unit Root Tests

Variables	Level		First Difference
	Lag	t-ADF	Lag	t-ADF
m	6	-1.28	6	-1.96 *
y	6	-1.85 +	6	-3.04 **
π	6	-1.05	6	-4.52 **
i	6	-0.68	6	-6.55 **

Notes: Variables are as defined in the text. For each variable expressed in level (first difference), the Augmented Dickey-Fuller (1979) ADF(2) statistics tests a null hypothesis of a unit root in that variable expressed in level (first difference) against an alternative of a stationary root, The criterion for lag selection is based on the Akaike information criterion, as described by Pantula et. Al. (1994). The critical values are taken from MacKinnon (1991). + ^*, and ^** denote rejection at 10 percent, 5 percent and 1 percent critical values, respectively.

View Table

Table 4.

Statistics for ADF(2) Unit Root Tests

Variables	Level		First Difference
	Lag	t-ADF	Lag	t-ADF
m	6	-1.28	6	-1.96 *
y	6	-1.85 +	6	-3.04 **
π	6	-1.05	6	-4.52 **
i	6	-0.68	6	-6.55 **

Notes: Variables are as defined in the text. For each variable expressed in level (first difference), the Augmented Dickey-Fuller (1979) ADF(2) statistics tests a null hypothesis of a unit root in that variable expressed in level (first difference) against an alternative of a stationary root, The criterion for lag selection is based on the Akaike information criterion, as described by Pantula et. Al. (1994). The critical values are taken from MacKinnon (1991). + ^*, and ^** denote rejection at 10 percent, 5 percent and 1 percent critical values, respectively.

57. The long-run relationship between m₂, y, π, and i is therefore estimated in the form of a cointegrating vector. Two dummies are used as exogenous variables to capture the impact of the structural reform efforts that took place in 1988 and 1994. The Engle-Granger (1987) and Johansen’s maximum likelihood procedures (1988) are used to determine the number of cointegrating vectors among the variables.²⁸ Both procedures indicate that there is at most one co-integrating vector (at 5 percent confidence). The long-run money market equilibrium takes the following form:

58. All the coefficients of the equation have the expected sign. A coefficient higher than unity on output in the long-run money demand function shows a declining rate of velocity in the long-run. This is consistent with the experience of countries, like Algeria, where money is the principal instrument for the accumulation of financial savings by the public. The equilibrium relationship also shows that the interest rate has a very small impact on the quantity of money the public wants to hold, confirming the behavior of public enterprises that dominate the economy and are not sensitive to interest rates, as well as the absence of alternative financial instruments. The money market equilibrium is, however, very sensitive to the rate of inflation.

59. Equation (1) tracks real M₂ well and deviations are generally of a small magnitude. Figure 3 plots the actual (money stock in real terms) and the fitted values of money. In addition, since 2000, the money supply has exceeded the level predicted by the money demand function, suggesting the existence of excess money supply in the Algerian money market, and thus raising the question of possible inflationary consequences of such disequilibrium.

Fitted and Actual Real M2, 1974-2001

Citation: IMF Staff Country Reports 2003, 069; 10.5089/9781451811421.002.A003

Source; Algerian authorities and staff estimates

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Fitted and Actual Real M2, 1974-2001

Citation: IMF Staff Country Reports 2003, 069; 10.5089/9781451811421.002.A003

Source; Algerian authorities and staff estimates

• Download Figure
• Download figure as PowerPoint slide

Fitted and Actual Real M2, 1974-2001

Citation: IMF Staff Country Reports 2003, 069; 10.5089/9781451811421.002.A003

Source; Algerian authorities and staff estimates

• Download Figure
• Download figure as PowerPoint slide

60. Tests of weak stationarity indicate that y, m, and i are weakly exogenous to this co-integrating relationship, and that the inflation rate, π adjusts to the disequilibrium in the money market.²⁹

61. The estimation of a short-term single error-correction equation is used to analyze the dynamics of inflation. By using ordinary least squares, equation (2) estimates the first difference of inflation, using the first difference of real money stock, output, interest rate, the two dummy variables to take into account the two attempts of economic liberalization, and the first lag of the error term ε as explanatory variables. After removing the statistically insignificant variables from this general equation, we obtain the following equation for the period 1982–2001 (figures in parentheses represent standard errors; figures in brackets represent t-statistics):

62. The model fits the data quite well. The residuals are very small (Figure 4). Excess money and lagged real money has a positive impact on inflation, suggesting that increased monetary aggregates have inflationary consequences. The two dummies have also positive, albeit small, impact on inflation, suggesting that the economic liberalization efforts of 1988 and 1994 led to a slight increase in inflation in the short run. Finally, the lagged and twice lagged variations of interest rates enter the equation with positive coefficients, suggesting that an increase in interest rates have very small impact on inflation. This result suggests that in Algeria an increase in the interest rate has little effect on demand and that its main impact is to increase costs of production.

Actual and Fitted Changes in inflation, 1982-2001

Citation: IMF Staff Country Reports 2003, 069; 10.5089/9781451811421.002.A003

Source: Algerian authorities and staff

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Actual and Fitted Changes in inflation, 1982-2001

Citation: IMF Staff Country Reports 2003, 069; 10.5089/9781451811421.002.A003

Source: Algerian authorities and staff

• Download Figure
• Download figure as PowerPoint slide

Actual and Fitted Changes in inflation, 1982-2001

Citation: IMF Staff Country Reports 2003, 069; 10.5089/9781451811421.002.A003

Source: Algerian authorities and staff

• Download Figure
• Download figure as PowerPoint slide

63. The model predicts an acceleration of the inflation rate in the context of excess liquidity. The fitted value of the change in inflation suggests that in 2001, in the context of excess money supply, inflation should have increased above its actual level. A series of stability and break-point tests were carried out to measure the strength of the model and to gauge whether there was any structural change in inflation dynamics during the latter part of the sample (1994-2001), in the context of an excess money supply.³⁰ The two types of Chow breakpoint and Chow forecast tests did not reject the hypothesis of model constancy over the sample period. The non-acceleration of inflation in 2001 and in particular its decline in 2002 against a background of strong monetary growth is the result of factors that have not been included in the equation, such as tariff reduction in 2001 and 2002. In the future, there seems to be little prospects for much further growth in real money without upward pressure on inflation.

64. While the stability of the money demand matters if monetary policy targets a monetary aggregate, it is not sufficient for successful targeting, since the central bank does not fully control that aggregate. It only controls the reserve money. The ability of the central bank to control M2 depends, therefore, on the predictability of the money multiplier. Therefore, further research (not covered in this paper) on the relation between reserve money and the M2 aggregate is required to complete the assessment of the transmission mechanisms of monetary policy in Algeria.

Oil Price Shocks and Monetary Equilibrium

65. The impact of hydrocarbon revenues on monetary aggregates can be a source of disturbance in the money market. Analyzing the channels through which an increase in international prices for hydrocarbon affect the money market equilibrium will help design better policies. In this section, a simplified accounting framework is used to show that an increase in international oil prices creates money market disequilibrium, and that the degree of ex-ante excess money supply depends on the distribution of hydrocarbon revenues between the government and the oil companies, and on the government expenditure policy.

66. The short-term impact of a positive price shock in the hydrocarbon sector in a fixed exchange rate regime could be expressed through the following equation.

ΔM³=ΔNFA + ΔNCG + ΔCE,

where:

M^s = money supply

NFA= net foreign assets

NCG = net credit to the government

CE = credit to the economy

67. Suppose that the money market is at equilibrium and that we have a sudden increase in the international price of hydrocarbons that only affects, in the short run, the value of exports by an amount ΔR; in particular, we assume that the external demand for hydrocarbons and domestic production (as well as the domestic demand for imports) are unchanged at this stage. Other things being equal, foreign exchange reserves will also increase by ΔR.

Therefore, as a first approximation:

ΔNFA≈ΔR.

68. Before taking into account the second round effects resulting from changes in the liquidity of the banking system, banks’ credit supply is unchanged by this accumulation of net foreign assets, and:

ΔCE = 0.

Assuming that a share α of additional hydrocarbon export proceeds accrues to the government,³¹ and that it is saved, we arrive at:

ΔNCG =–αΔR.

Therefore, money supply increases ex-ante by:

ΔM³ =(1–α)ΔR.

69. The demand for money is, however, not directly affected by the shock ex-ante. Real GDP, unlike nominal GDP, is not directly impacted by a hydrocarbon price change, at least in the short term.³² The same goes for the nominal interest rate and, to a lesser extent, for the consumer price level.³³

70. An oil price shock will, therefore, create an ex-ante disequilibrium on the money market—assuming no response from monetary authorities—which can be reestablished through one (or a combination)³⁴ of the following:

• A decrease in the nominal interest rate;

• An increase in the price level; and

• An increase in real GDP.

71. The model shows that the degree of ex-ante excess money supply depends crucially on the distribution of hydrocarbon revenues. For instance, if all revenues accrue to the government, and if the latter decides to save any unexpected revenues in its central bank account we have (assuming no variation in credit to the economy):

ΔNCG = –ΔNFA = –ΔR.

Consequently:

ΔM³ = ΔR – ΔR + 0 = 0.

72. In such a context, the money market would thus remain at equilibrium. However, since in Algeria, 65 percent of hydrocarbon revenues accrue to the government, the monetary policy has to deal with the remaining 35 percent, which represents Sonatrach’s share, as well as the liquidity injected in the market resulting from the government use of hydrocarbon revenues.

Recent Economic Developments

74. Monetary expansion was relatively contained in 2000: while the increase in net foreign assets amounted to 34 percent of the initial money stock, broad money only increased by 13 percent in 2000. This control was, however not due to monetary operations, but rather to a partial sterilization of net foreign asset accumulation through fiscal policy. In fact, in April 2000, the authorities established an oil stabilization fund to save any hydrocarbon revenues in excess of those budgeted (below). Even though the fiscal policy stance was relaxed compared to 1999 (the nonhydrocarbon deficit increased by 6 percent of nonhydrocarbon GDP) and part of the non-budgeted revenues were not saved, this expansion was still small relative to the increase in net foreign assets.³⁵ This is shown by the substantial increase in overall budget balance from almost nil in 1999 to around 10 percent of GDP in 2000.

75. Monetary policy did not have adequate instruments to absorb excess liquidity in 2001, in the context of expansionary fiscal policy. Despite continuously increasing hydrocarbon revenues, total fiscal surplus shrunk substantially in 2001.³⁶ Reflecting this development, coupled with substantial debt servicing payments made by the Treasury to public banks, money growth increased rapidly from 13 percent in 2000 to more than 22 percent in 2001. This rise occurred despite the lower increase in net foreign assets in 2001 and 2002 compared to 2000 (26 percent in 2001 against 34 percent of the initial money stock in 2000). It led to higher prices, and the average inflation rate increased from nil in 2000 to 4.2 percent in 2001. The following observations are consistent with the diagnosis of excess liquidity: (a) for most of 2001, banks resorted only to the rediscount window for central bank refinancing, and the use of this facility ended in November 2001, with a significant drop in banks’ refinancing needs; (b) banks have maintained reserve balances at the central bank well in excess of required amounts; and (c) interest rates have substantially declined on the interbank money market, and fell below the rediscount rate in the autumn of 2001. As of end-2001, the Bank of Algeria did not have the appropriate instruments to withdraw liquidity from banks. The only active instrument was the rate of reserve requirements, which was raised from 3 percent to 4.25 percent during the last quarter.

76. The monetary authorities introduced a new instrument to absorb excess bank liquidity in 2002. In the wake of continuous accumulation of foreign assets, and given the further relaxation of the fiscal stance, the monetary authorities decided to introduce in April 2002 a deposit facility (or negative liquidity auctions) at the central bank to absorb banks’ excess liquidity.³⁷ However, in the initial transactions under this new facility, the Bank of Algeria has accepted only part of the liquidity offered by commercial banks, meaning that there is still some excess liquidity in the market. If this excess liquidity persists, it might lead to further expansion of the money supply through new credits, and put upward pressure on prices.

Instruments to Sterilize Excess Money Supply

77. This section analyzes only the measures that could be taken in the short term to address money market disequilibrium in Algeria, namely sterilization. Sterilization is, however, a costly process, and should not be adopted on a permanent basis. The paper does not address long term structural deficiencies hindering financial deepening, nor does it address other fundamental questions of macroeconomic policy, such as the appropriate level of government expenditure over the long run or the optimum rate of saving

78. The monetary impact of hydrocarbon export revenues cannot be analyzed separately from fiscal policy decisions. In addition to the negative liquidity facility, other monetary instruments and fiscal measures could be used by the Algerian authorities if they were to face this situation in the future.³⁸

Sterilization through monetary operations

79. In order to be able to appropriately address similar situations of excess liquidity in the future, the Bank of Algeria might wish to sterilize additional hydrocarbon-related export earnings through open market operations such as issuing treasury or central bank bills over the medium term.³⁹ By selling securities (such as treasury bills or central bank bills) on the money market, the monetary authorities might be able to mop up excess liquidity so as to prevent the nominal interest rate from falling.⁴⁰ In addition, open market operations would help develop the financial markets in Algeria.

80. Since the sterilization capacity in Algeria is currently limited by the underdevelopment of money and financial markets, the monetary authorities should, in the short term, continue resorting to negative liquidity auctions. However, they need to improve its implementation by fully absorbing the excess liquidity, instead of accepting only part of the liquidity offered by banks, which is currently the practice.

Sterilization through fiscal policy

81. Sterilization through fiscal policy is another option to rein in excess liquidity in Algeria. Since most of the hydrocarbon revenue accrues to the government, sterilization through fiscal policy is both easy and warranted to avoid the macroeconomic instability resulting from a procyclical fiscal policy that would transmit the swings of the word oil market.

82. Following the 1994–98 programs, when automatic adjusters on overall fiscal targets helped sterilize excess revenues, a stabilization fund was established in 2000. The Algerian authorities, aware of the costs of high uncertainty, decided in 2000 to adopt a longer term perspective which, in principle, should have led them to sterilize the impact of oil price shocks through fiscal policy.⁴¹ The oil stabilization fund was established, and the supplementary budget law for the 2000, and 2001 and 2002 budgets was based on conservative oil price assumptions.⁴² The purpose was to set expenditures in a multi-year perspective and to avoid, in particular, having to adjust them regularly—and sometimes drastically—because of oil price fluctuations.

83. In the period ahead, it is important that Algeria avoid a procyclical fiscal stance, and that the Treasury coordinate with the Bank of Algeria on liquidity management. During periods of abundant hydrocarbon revenues, the government should be prudent in relaxing the fiscal stance, and when monetary sterilization cannot mop-up the total excess liquidity, fiscal sterilization should complement the Bank of Algeria’s efforts. Therefore, Algeria’s high dependency on oil requires that monetary policy be coordinated with other policies (fiscal in particular) aimed at managing the excess liquidity of the economy.

Sterilization through foreign exchange interventions

84. Sales of foreign exchange are a possible instrument to sterilization. However, in the context of an economy where there is no complete capital account convertibility, such interventions imply significant exchange rate volatility.

Sterilization of Sonatrach’s share of the hydrocarbon revenue

85. Even if Sonatrach actually “sterilizes” its deposits by holding them in commercial banks rather than adjusting its expenditures and financing plans in reaction to changes in its own liquidity, its large share of hydrocarbon revenues still affects the total liquidity of the banking system. As discussed previously, the share of the hydrocarbon revenue accruing to Sonatrach is not negligible. Therefore, large fluctuations in hydrocarbon prices can have significant consequences on its financial situation and, in fine, on the domestic banking sector’s liquidity, since Sonatrach holds its deposits in dinar in the banking system.

86. The case could therefore be made for trying to sterilize in the short term at least part of Sonatrach’s windfall gains consecutive to an increase in oil prices. This could be done either through the issuance of Treasury bills to back these deposits, or by allowing Sonatrach to hold foreign exchange deposits in the domestic banking system. This is why this measure might be resisted by the authorities, especially when the traded goods sector is already being “cannibalized” by the hydrocarbon sector.⁴³ Therefore, an exchange rate adjustment should probably not be the first line of defense, but rather one which is envisaged after sterilization has already been tried.