Journal Issue
Share
Article

São Tomé and Príncipe: Recent Economic Developments and Selected Issues

Author(s):
International Monetary Fund
Published Date:
June 2000
Share
  • ShareShare
Show Summary Details

III. Monetary Policy, The Exchange Rate and Inflation28

A. Introduction

80. During 1994-97, the monetary environment in São Tomé and Príncipe was characterized by high inflation, a large depreciation of the exchange rate, and an increasing importance of foreign currency deposits (FCDs) in the domestic banking system. Moreover, data available show that the real exchange rate (RER) depreciated significantly during that period. Data show that the exchange rate depreciation was substantially larger than the measured rise in prices. The depreciation of the RER in 1994-97 forms a departure from the period 1988-93, when the RER remained roughly constant. Recent developments are analyzed in the context of a simple theoretical framework that is estimated using monthly data from January 1992 to May 1998. In Section B, monetary and exchange rate developments in the period 1992-98 are discussed. In Sections C and D, a framework that analyzes the effects of monetary policy on prices, the exchange rate, and (the importance of) FCDs is developed and estimated. In Section E, conclusions are drawn, and some policy implications are suggested. It is argued that the high degree of dollarization and substitution between dobra-denominated and dollar-denominated assets severely limits the effectiveness of monetary policy.

B. Monetary and Exchange Rate Developments: 1992-98

81. After a few years of relatively prudent fiscal and monetary policy, government finances deteriorated again in 1994, with domestic financing of the budget deficit rising sharply. Owing to the strong monetary expansion, inflationary pressures built up. However, because of lags in the transmission mechanism between money and prices, inflationary pressures only materialized later in 1994 and peaked in early 1995, even though macroeconomic performance improved somewhat in 1995 (a tighter fiscal policy and low net bank lending to the government). In 1996 and—more importantly, in 1997—an expansionary fiscal stance and accommodating monetary policy resulted in an even larger rise in inflation. While year-on-year Ml growth peaked at 97 percent in April 1997, year-on-year inflation reached 87 percent in February 1998. At the end of 1997, the central bank mopped up a great deal of liquidity by changing the regulations on reserve requirements.29 In addition, the central bank rediscount rate became positive in real terms in 1998. The improvement in the fiscal outturn and tightening in monetary conditions for the first half of 1998 served to further temper the rise in prices.

82. FCDs have traditionally been held at local banks. However, the data on FCDs are properly incorporated in the monetary accounts only from 1994 onward. The importance of FCDs is a good indicator of the (lack of) confidence in the domestic currency. The level of FCDs decreased in 1995, in response to the improvement in macroeconomic conditions and reduction in inflation during that year. However, in 1996, with the relaxation of fiscal and monetary policies and the pickup in inflation, FCDs, expressed as a percentage of M2, rose rapidly from 32 percent in January 1996 to 45 percent in February 1998. During the first half of 1998, the drop in inflation did not immediately result in a fall in the level of FCDs. However, with the resumption of monetary stability since 1998, a reduction in importance of FCDs can be expected.

83. The response of the exchange rate to the relaxation of macroeconomic policies from 1994 through end-1997—and, indeed, the subsequent stabilization of the dobra in 1998—was particularly noteworthy. In 1994-97, the nominal exchange rate depreciated more significantly than prices rose. The real exchange rate (expressed as the ratio of domestic consumer prices to the prices of imports of goods in dobras) depreciated by some 56 percent between end-1993 and end-1997 (Figure 5).

Figure 5.São Tomé and Príncipe: Real Exchange Rate and Terms of Trade, January 1981 - May 1998

(Index, 1990=100)

Sources: São Tomé and Príncipe authorities; and staff estimates.

84. Several possible explanations for this weakening of the RER have to be discarded. First, a negative terms of trade shock is unlikely to have been the cause. The terms of trade, as calculated by the IMF (taken from the World Economic Outlook database), in fact improved slightly in 1994-98, owing to the favorable price movement for cocoa. Moreover, exports made up a small portion of total current account earnings.30 Hence, only a modest part of foreign exchange earnings was affected by terms of trade movements. Second, a widening gap between gross domestic product and gross domestic expenditure cannot have been the main reason for the weakening of the real exchange rate.31 The ratio of gross domestic expenditure (GDP + imports - exports) to GDP rose by only 4 percent during 1994-97, clearly not a large enough increase to be a determining factor in the weakening of the RER. Third, although measurement errors cannot be ruled out, and the increase in the consumer price index is reported to have been underestimated before 1996, in early 1996 the methodology used to measure inflation had improved. Nevertheless, from early 1996 through late 1997, the real exchange rate depreciated by a further 23 percent over the inflation rate.

85. In view of favorable terms of trade, the weakening of the real exchange rate during 1994-97 was mainly due to the reduction of confidence in the dobra arising from expansionary fiscal and monetary policies and attendant high inflation, In addition to a “portfolio” effect, expansionary monetary policy has a substitution/confidence effect: in an environment of rapid money growth and rising inflation, there is a strong tendency to substitute dobras for dollars.

C. Theoretical Framework

86. It is assumed that there exists a stable long-run money demand function. Money demand is taken to be a function of the volume of transactions in the economy and inflation—with the latter proxying the opportunity costs of holding money. Money supply, a policy variable, is exogenous. In the long run, with equilibrium in the money market, money demand is equal to money supply. However, at any particular time, the stance of monetary policy is reflected in an excess or a shortage of money. Once a long-run money demand function has been estimated, the stance of monetary policy can be expressed as the difference between the actual money stock and the demand for money (as modeled):

where Md is money demand, D represents nominal domestic demand (GDP-exports+imports), and INF denotes inflation. An important choice is whether to use Ml or M2. In São Tomé and Príncipe, quasi money consists mainly of FCDs. FCDs are likely to fuel inflation to a much lesser extent than dobra-denominated money. Therefore, Ml is used. D is used instead of GDP because D more closely approximates the level of transactions in the economy. The difference between D and GDP is very large in São Tomé and Príncipe, mainly owing to large external transfers, and is not constant over time. The stance of monetary policy, or excess money supply, can be written as

where ECMm is excess money supply. As mentioned in Section A, The importance of FCDs is expected to be a function of inflation:

where FCDM2 is the ratio of FCDs to total M2.

87. The RER is defined as CPI/(PM*ER), where CPI is the consumer price index, PM is the price of imports of goods, and ER is the exchange rate (in dobras per dollar). Testing whether the RER has been stationary is testing whether the purchasing power parity (PPP) holds in the long run. Although it is argued above that the RER has depreciated substantially during 1994-97, it is still postulated that the PPP (that is, constant RER) holds in the long run. As shown in Figure 5, the RER remained remarkably stationary from the start of the 1980s until 1993. Moreover, unless the RER movement is caused by supply and/or productivity factors, a secular decline in the real exchange rate is not sustainable:

where lower-case variables are in logarithm.

In order to determine the channels through which the exchange rate and the price level are affected by monetary policy, the following dynamic equations for these two variables are estimated, using the two ECM variables defined above:

where the prefix d stands for the first difference, x(l) índicates a vector of lagged values, and seas is a (potential) seasonal effect. These equations postulate that the price level and the nominal exchange rate tend to respond to deviations from the PPP (ECMrer) and to the stance of monetary policy (ECMm). Given the manner in which the ECM variables are defined, the signs of their impact are expected to be as follows: a loose monetary policy stance (ECMm>0) is expected to drive up the price level and weaken the exchange rate, whereas a tight monetary policy stance (ECMm<0) is expected to lead to a fall in the price level and/or a depreciation of the exchange rate. Dynamic effects of the price level, the exchange rate, and money growth are also to be expected. In light of the argument that expansionary monetary policy has caused the RER to depreciate, it is expected that the monetary policy stance has a larger effect on the (nominal) exchange rate than on the price level. That is, the coefficient of ECMm in equation (6) is expected to be higher than the one in equation (5).

D. Estimation

88. The co-integration estimation procedure has become common. For a discussion of the methodology, the reader is referred to Engle and Granger (1987) and Johansen (1988). Unit root tests have been carried out to determine the order of integration of the variables. As indicated in Table 6, all variables are nonstationary, integrated of order one -I(l).32 For the set of variables (er, cpi, pm, INF, m, and d), the Johansen procedure was used to determine the number of co-integrating vectors; the trace statistic indicated that there is only one co-integrating vector (Table 7). Using the knowledge that the RER is not stationary over the estimation period (January 1992-May 1998),33 we assume that the one co-integrating vector represents the money demand function. For this to be true, the vector should include only m, d, cpi, and INF. Although the imposition of a zero coefficient for er and pm was not formally accepted, those restrictions were imposed on the grounds that the estimated co-integrating vector (the money demand equation) has stationary residuals, which is evidence of a proper co-integrating vector. As is shown in Table 7, after the imposition of those two restrictions, the co-integrating vector takes the form of a money demand function whose coefficients have correct signs and orders of magnitude. Moreover, the impact of the restricted vector on the price level and the exchange rate has the correct sign. The final money demand equation is34

Table 6.Order of Integration: Unit Root ADF Test Statistics
LevelFirst Difference
LagTest statisticLagTest statistic
er24.493-2.60**
cpi15.051-2.66**
Pm6-2.52*5-5.10**
INF1-0.480-6.67**
m25.482-2.76**
d60.305-3.84**
Notes: Variables are as defined in the text.Asterisks * and ** denote rejection of the null hypothesis of a unit root at the 5 percent and 1 percent significance levels, respectively.
Notes: Variables are as defined in the text.Asterisks * and ** denote rejection of the null hypothesis of a unit root at the 5 percent and 1 percent significance levels, respectively.
Table 7.Co-integration Analysis, January 1992 - May 1998
Hypothesisr=0r<=1r<-2r<=3r<=4r<=5
Trace statistic 1/103.30 **50.8427.7912.482.290.04
95 percent critical value94.2068.5047.2029.7015.403.80
Normalize around m
Standardized eigerivectors
ercpipmINFmd
4.093-8.2296.5157.8001.000-5.663
Restrictions: pm and d are weakly exogenous: chi-square(2) = 14.57 [0.000] **.
Standardized eigenvectors
ercpipmINFmd
4.068-8.2295.7547.7231.000-5.972
Restriction:
Coefficient of pm is 0: chi-square (1) = 8.1 [0.004]**.
Coefficients of er is 0: chi-square (1) = 14.5 [0.000]**.
Standardized cigenvectors
ercpipmINFmd
0.000 [0.00]-1.277 [0.12]0.000 [0.00]1.532 [0.32]1.000 [0.00]-1.446 [0.74]
Standardized impact coefficients
ercpipmINFmd
0.040 [0.03]0.028 [0.01]0.000 [0.00]-0.047 [0.03]-0.065 [0.02]0.000[0.00]
Notes: Variables are as defined in the text. Asterisks * and ** denote rejection of the null hypothesis at the 5 percent and 1 percent significance levels, respectively. Numbers between square brackets denote standard errors.

Adjusted for the number of degrees of freedom.

Notes: Variables are as defined in the text. Asterisks * and ** denote rejection of the null hypothesis at the 5 percent and 1 percent significance levels, respectively. Numbers between square brackets denote standard errors.

Adjusted for the number of degrees of freedom.

89. The elasticity of money demand with respect to nominal demand (d*cpi) is, although larger than one, within the range found for other countries. The negative effect of inflation on money demand is large and highly significant: a 10 percentage point rise in the inflation rate is estimated to reduce money demand by 4.7 percent. The residuals of this equation are stationary at the 5 percent significance level.35 The ECM variable measuring the stance of monetary policy on the basis of the money demand equation, according to equation (2), is shown in Figure 6. The movement of the ECM variable is in line with the discussion of the stance of monetary policy during 1994-97 (Section B). The estimated dynamic equations for the price level and the exchange rate are

Figure 6.São Tomé and Príncipe: Monetary Policy Stance and Inflation, January 1992 - May 1998

Sources: São Tomé and Príncipe authorities; and staff estimates.

where S_3 is a dummy for the month of March. Overall, the equations are in line with expectations. An expansionary monetary policy stance leads to a rise in the price level and a depreciation of the exchange rate. Moreover, the effect on the exchange rate is larger, which implies that the above-mentioned hypothesis on the effect of monetary policy on the RER is confirmed. The ECM variable representing PPP has an impact on the price level, but, somewhat surprisingly, its impact on the exchange rate is not sígnificant. Hence, when the PPP relationship is perturbed because of a shock, the PPP is restored through an adjustment of the domestic price level, but not through an adjustment of the (nominal) exchange rate. As regards dynamic effects, an increase in Ml has an immediate impact on the price level, but not on the exchange rate, and a depreciation of the exchange rate leads, with a lag of four months, to a rise in the price level. As evidenced by the size of the standard errors, the fit of the price equation is substantially better than that of the exchange rate equation. The residuals of both equations are stationary at the 1 percent significance level, and, as the Durbin-Watson statistic indicates, there is no serial correlation of the residuals. However, as is shown in Figures 7 and 8, a few large outliers (possibly caused by data inconsistencies) distort the distribution of the residuals. To the extent that the equations fail to pass other diagnostic tests, this is due to these outliers. After the inclusion of two dummies to account for the largest two outliers in both equations, the equations passed all diagnostic tests (Table 8).

Figure 7.São Tomé and Príncipe: Actual and Fitted Values of Consumer Price Index, January 1992 - May 1998

Sources: São Tomé and Príncipe authorities; and staff estimates.

Figure 8.São Tomé and Príncipe: Actual and Fitted Values of Nominal Exchange Rate, January 1992 - May 1998

Sources: São Tomé and Príncipe authorities; and staff estimates.

Table 8.Diagnostic Tests for Dynamic Equations
Consumer Price Index EquationExchange Rate Equation
AR 1-4 1/F(5, 59) = 1.1536[0.3428]F(5, 63) = 2.9351 [0.0191) *
ARCH 4 2/F(5, 54) = 1.9439[0.1021]F(5, 58) = 1.0831 [0.3794]
Normality 3/Chi2(2)= 6.6672[0.0357] *Chi2(2)= 6.5919 [0.0370] *
Xi2 4/F(15, 48) = 1.1534[0.3389]F(9, 58) = 0.702216 [0.7042]
Xi* Xj5/F(36, 27) = 2.177 [0.0195] *F(12, 55) = 1.0902 [0.3865]
Reset 6/F(1, 63) = 5.5291 [0.0218] *F(1, 67) = 1.5744 [0.2139]
Note: Asterisks * and ** denote rejection of the null hypothesis at the 5 percent and 1 percent significance levels, respectively.

Test for serial correlation of residuals (HO: no autocorrelation).

Test for autoregressive conditional heteroscedasticity (HO: no heteroscedasticity).

Test for normality of distribution of residuals (HO: normality).

Test for heteroscedasticity (HO: no heteroscedasticity).

White’s cross-product test for heteroscedasticity (O: no heteroscedasticity).

Test for general misspecification of equation (HO: no misspecification).

Note: Asterisks * and ** denote rejection of the null hypothesis at the 5 percent and 1 percent significance levels, respectively.

Test for serial correlation of residuals (HO: no autocorrelation).

Test for autoregressive conditional heteroscedasticity (HO: no heteroscedasticity).

Test for normality of distribution of residuals (HO: normality).

Test for heteroscedasticity (HO: no heteroscedasticity).

White’s cross-product test for heteroscedasticity (O: no heteroscedasticity).

Test for general misspecification of equation (HO: no misspecification).

E. Conclusions and Policy Implications

90. As long as the dobra remains the main form of legal tender for domestic transactions, there will be demand for dobra-denominated money. However, as the above analysis has shown, owing to the high degree of dollarization in São Tomé and Príncipe, the demand for dobra-denominated money is highly sensitive to the stance of monetary policy. When monetary expansion is too rapid, as it was in 1996-97, the sharp reduction in the willingness to hold dobras leads to an equally sharp depreciation of the exchange rate and a strong increase in the importance of FCDs. With FCDs reaching more than 40 percent of M2 (at the end of 1997 and beginning of 1998) and dollar notes circulating widely as well, the control of the monetary authorities over monetary conditions is limited.

91. In general, the main economic advantage of having a national currency is that the monetary authorities can respond to shocks impinging on the economy by adjusting the exchange rate (or allowing it to adjust), the (short-term) interest rate, or monetary aggregates, thereby limiting the cost of shocks in terms of output and employment. As shown above, in São Tomé and Príncipe the ability to affect real dobra-denominated monetary aggregates is limited by the sensitivity of prices and foreign currency holdíngs to the monetary stance. Nevertheless, the exchange rate is particularly sensitive to monetary policy, so that the authorities can affect the RER, at least in the short to medium run. This ability to affect the RER provides, in principle, for a useful policy tool. In the absence of a manufacturing sector in São Tomé and Príncipe, the importance of real exchange rate developments should be reviewed mainly for the emerging tourism sector. The question is whether the ability of monetary authorities to affect the real exchange rate constitutes a major policy advantage in the current economic structure.

92. The costs of maintaining a national currency backed by an independent monetary policy in a small country such as São Tomé and Príncipe are high. The operational expenses of the central bank are high relative to GDP, and banks devote substantial resources to exchange rate related activities, at the expense of the allocation of savings and credit. In addition, the economy suffers from the costs of a national currency, while the central bank’s ability to use monetary policy effectively is lìmited by the high degree of dollarization and substitution between currencies. In the circumstances, two opposing policy options should be explored further. If the authorities deem the ability to respond to shocks as important, the strategy should be to strive to increase control over monetary conditions. However, if the ability to respond to shocks is judged as relatively unimportant, either the adoption of a currency board, the full dollarization of the economy, or a peg to the euro-for example, through the arrangements of neighboring CEMAC countries-could be considered.

References

    EngleRobert F. and C.W.J.Granger1987“Co-integration and Error Correction: Representation, Estimation, and Testing”,EconometricaVol. 55 (March) pp 25176.

    • Search Google Scholar
    • Export Citation

    JohansenSoren1988“Statistical Analysis of Cointegration Vectors”Journal of Economic Dynamics and ControlVol. 12 (June-September) pp. 23154.

    • Search Google Scholar
    • Export Citation
28Prepared by Aloysius Kuijs.
29From November 1997 onward, reserves of both domestic and foreign deposits were required to be constituted in local currency, and, from March 1998 on, government’s bank deposits became subject to a 100 percent reserve requirement. These changes forced the commercial banks to buy dobras to be deposited at the central bank.
30Some 28 percent in 1997. Exports of goods constituted only 12 percent of total current account earnings. The dominant part of current account earnings stems from official transfers.
31In the case of an increasing divergence between national production and national expenditure (part of which is on imports), a depreciation of the RER restores equilibrium on the foreign exchange market by making imports more expensive and exports more competitive. Diverging productivity trends (between a country and its trading partners) could also induce RER changes. However, the movement registered since 1994 is too abrupt and substantial to be explained by such factors.
32The variable pm is 1(0) at the 5 percent significance level, but 1(1) at the 1 percent level.
33This implies 77 observations, of which 2 were lost owing to the inclusion of lags.
34The numbers in parentheses represent i-statistics.
35The unit root augmented Dickey-Fuller test statistic is -2.44, compared with a 5 percent, critical value of-1.95.

Other Resources Citing This Publication