Democratic Republic of São Tomé and Príncipe: Selected Issues and Statistical Appendix

The paper reviews the background and the existing institutional framework for oil sector development in São Tomé and Príncipe and the challenges faced in implementing transparency rules in all oil-related transactions. It provides a quantitative analysis of the impact of oil sector development on government receipts, spending, and savings, and discusses the determinants of inflation from a statistical point of view. It also shows an analysis of the weak relationship between money growth and inflation.

Abstract

The paper reviews the background and the existing institutional framework for oil sector development in São Tomé and Príncipe and the challenges faced in implementing transparency rules in all oil-related transactions. It provides a quantitative analysis of the impact of oil sector development on government receipts, spending, and savings, and discusses the determinants of inflation from a statistical point of view. It also shows an analysis of the weak relationship between money growth and inflation.

III. A Note on Inflation, 1998–200520

A. Overview and Conclusions

1. This chapter discusses recent developments with inflation and monetary aggregates, and looks into determinants of inflation in São Tomé and Príncipe from a statistical point of view. The analysis is motivated by the fact that inflation has picked up since 2004, posing new challenges to policymakers in devising a strategy to keep inflation low and determining which policy variables to monitor toward this end.

2. The chapter builds upon analysis by Kuijs (2000) on the relationship between money growth, inflation, and exchange rate depreciation for the period 1992–98. The times series data have been expanded to include the disinflation period of 2000-03 and the recent inflation rebound of 2004–05. The econometric results from this chapter are consistent with those of Kuijs (2000) regarding the important role played by money growth and exchange rate depreciation in explaining inflation in São Tomé and Príncipe. The statistical analysis and results from this paper, however, would need to be revisited over time to address the lack of long time series for critical variables and remaining questions about the stability of the model on account of changing institutional relationships.

3. The chapter’s main conclusions are the following:

  • As predicted in theory, there exists a reasonable long-run relationship among prices, broad money, the exchange rate, and developments in the real economy in São Tomé and Príncipe. Over the long run, both broad money growth and changes in the exchange rate have an impact on inflation.

  • Statistical evidence suggests that exchange rate depreciation causes inflation and vice versa. Exchange rate movements may have an important effect on prices given the large share of imported products in the household consumption basket. Also, a more depreciated exchange rate reduces the demand for the dobras, thus increasing money velocity and creating inflationary pressures.

  • Our empirical results highlight the impact of exchange rate depreciation on inflation in the short run. In particular, a shock to the exchange rate has a larger and more lasting impact on inflation than an equivalent shock to broad money. Indeed, a shock to nominal money growth briefly leads to an increase in inflation, although this response is very small and statistically insignificant.

  • Empirical evidence suggests that there is substantial inflation inertia in São Tomé and Príncipe. Indeed, when inflation deviates from its long-term path, convergence to the latter persists, but the adjustment is very slow. Also, lagged inflation explains a relatively large part of current inflation.

4. A number of preliminary policy considerations could be derived from the analysis:

  • The paper finds evidence of a weak statistical relationship between money growth and inflation. This may suggest an ongoing step-increase in money demand (spurred from expectations about the growth of the domestic oil sector) which needs to be taken into account in designing monetary policy to secure adequate liquidity for non-inflationary economic growth.

  • Even assuming the existence of increasing money demand in the past, looking forward, the evidence of a relatively small but inflationary effect of money growth under such circumstances underscores the importance of controlling monetary conditions. This in turn highlights the need to maintain a sound fiscal stance and to avoid monetary financing of the budget deficits.

  • The analysis indicates that it takes around two years for a shock to the exchange rate to be fully-absorbed by inflation. This could be interpreted to mean that the formation of inflationary expectations is backward-looking, to a degree. In this regard, the development of the foreign exchange auction since 2004 could, in principle, improve the formation of expectations in the economy.

5. The structure of this paper is as follows. Section B presents background information on the evolution of inflation and monetary aggregates in São Tomé and Príncipe. Section C presents a simple monetary model to explain inflation. The model is estimated for the period 1998–2005 using monthly data for consumer price inflation, monetary aggregates, the dobra exchange rate against the dollar, and a proxy for monthly real GDP growth. In line with standard analysis, the monetary model is first estimated without imposing any restrictions on the coefficients for the explanatory variables. As a second step, and to enhance the statistical significance of the estimated coefficients, the model is re-estimated with the necessary restrictions over the specification of the equation, mainly in terms of the number of lagged explanatory variables. Section D presents impulse responses of inflation to shocks in the exchange rate and money growth using the restricted inflation model. The objective is to gauge how changes in key explanatory variables affect inflation over the short run. Section D also includes a discussion of Granger-causality tests estimates.

B. Inflation Trends in São Tomé and Príncipe, 1998–2005

6. After picking up in the late 1990s, inflation fell to almost single-digit levels during 2000–03, before increasing again in 2004–05 (Figure III.1). A significant hike in inflation in 1997 mirrored a sharp increase in domestic financing of a sizeable fiscal deficit. During 1998–2003, however, improved government policies helped sustain output growth and lower inflation. Over this period, real GDP growth increased from 2.5 percent to 4 percent, while inflation declined from 21 percent to 10 percent. In 2004, the economy continued to grow at a moderate pace, but inflation increased to 15 percent by year-end, as bank credit to the private sector rose sharply and the government loosened fiscal policy. In particular, the government raised expenditures to an unsustainable level in anticipation of a large oil signature bonus, which in the event was not received in 2004. Inflation continued to increase in 2005, albeit in the context of a sharp increase in international oil prices which was passed-through to domestic consumers.

Figure III.1.
Figure III.1.

Inflation (12-month percentage change), Jan. 1998–Sep. 2005

(In percent)

Citation: IMF Staff Country Reports 2006, 329; 10.5089/9781451835090.002.A003

7. The volatility of inflation (measured by the standard deviation of the inflation rate) has mirrored the path of the inflation rate. Indeed, the standard deviation of the monthly and the 12-month inflation rate declined during the disinflation period of 2000–03, before increasing again during 2004–05 (Text Table III.1). While the average and the standard deviation of monthly inflation were 1.3 percent and 1.0 percent in 1998–99, they fell to 0.8 and 0.5 during 2000–03, respectively. During 2004–05, the average monthly inflation rate rebounded to 1.3 percent, while its standard deviation doubled to 1.1 percent.

Text Table III.1.

Average Inflation, Jan. 1998–Sep. 2005

(In percent)

article image
Source: IMF staff estimates.

8. The relationship between key monetary variables and inflation remains puzzling in São Tomé and Príncipe. Text Table III.2 presents correlations among key monetary variables and inflation, while Figure III.2 shows the paths of inflation and growth of monetary aggregates (i.e., reserve money (M1) and broad money (M3)) during 1998–2005.21 There is a broadly positive relationship between inflation and money supply throughout the whole sample period. However, there exists unexpected irregularity in this relationship: while inflation is positively correlated with monetary aggregates in 1998–2002, such a relationship is not captured by the data for 2003–04. In 2005, inflation is again significantly correlated with reserve money growth.

Text Table III.2.

Simple Correlations

article image
Source: Staff estimates.

9. Also, the data do not support a clear relationship between money growth and exchange rate depreciation. For example, while monetary aggregates grew at double digit rates in 2001–03, the nominal exchange rate depreciated only marginally during this period. The relatively weak correlation between money growth and the exchange rate depreciation persists even using monthly data (Text Table III.2).

São Tomé and Príncipe—Broad Money Growth and Exchange Rate Depreciation, 2001–05

(In percent)

article image
Figure III.2.
Figure III.2.

Inflation and Money Growth, Jan. 1998–Sep. 2005

(In percent)

Citation: IMF Staff Country Reports 2006, 329; 10.5089/9781451835090.002.A003

10. The weak/unstable relationship either between money growth and inflation or between money growth and exchange rate depreciation might be explained by increases in money demand stemming from expectations about the domestic oil economy. Indeed, the recent increase in the number of commercial banks in the country (from one bank in 2003 to six banks in 2005) appears to have been instrumental in supporting an increase in the ratio of monetary aggregates to GDP. Also, rigidities in the functioning of the managed float during 2003–04 may have prevented the exchange rate from reflecting any excesses of money supply over money demand. However, the introduction of a foreign exchange auction system in late 2004, and its ongoing development, should help improve the formation of exchange rate expectations in the economy and better reflect in the future the underlying relationship between money growth and the depreciation of the exchange rate.

C. Empirical Analysis: Money, Exchange Rate and Prices

A vector error-correction model (VEC)

11. In order to explore the relationship among money, the exchange rate, real income/GDP and prices, we considered the following simple inflation forecasting model in which the price level (CPI) is determined by the behavior of the money supply and money demand:22

CPI=Φ(M3,Md(INTR,RGDP,ER))+++(1)

M3 is the money supply, which is in principle a policy variable and an increase in M3 should lead to an increase in general prices. Md is the money demand, which mitigates inflationary pressures and is assumed to be a function of the nominal interest rate (INTR), real GDP growth, and the exchange rate (ER). While higher interest rates make money holdings more costly, real economic expansion increases the transactions demand for money, thus leading to disinflation. A depreciation of the exchange rate lowers the relative returns on dobradenominated assets, reduces money demand, and fuels inflationary expectations.

12. Our statistical analysis uses a vector error-correction (VEC) methodology, which is the standard approach for the case when all dependent and independent variables in the model are co-moving with each other. We specify the above relationship as a VEC model with 7 lags, in which the equation for inflation is:23

ΔlnCPI=β0+β1kΔlnCPIt-k+β2kΔlnM3t-k+β3kΔERt-k+β4kΔlnRGDPt-k+β5ECMlnCPIt-1+εt(2)

where Δln is the first difference in logs of the variables, and ECM ln CPIt is an error correction term associated with disequilibrium from the long-term equilibrium in the money market, as defined in Equation (1):

ECMlnCPIt=α1lnCPIt-α2lnM3t-α3lnERt-α4lnRGDPt(3)

We exclude the interest rate variable from our estimated econometric equation (3) because the central bank reference interest rate had been unchanged since 2000 (albeit in the aftermath of a period of discretionary reductions in this rate during the disinflation period that started in 1998).24

Data and empirical results (unrestricted VEC model)

13. The analysis uses monthly data from January 1998 to September 2005. Inflation is measured by the 12-month growth rate for the consumer price index (CPI). Money supply (M3) is defined as currency in circulation plus dobra-denominated deposits and foreign currency-denominated deposits. The amount of imported investment goods deflated by the CPI is used as a proxy of real GDP, in the absence of monthly GDP data. This variable (RGDP) is defined as the three-month moving average of this time series to accommodate the observed volatility of customs data. The exchange rate (ER) used is the dobra against the U.S. dollar.25

14. The augmented Dickey-Fuller unit root tests indicate that all variables (in logarithms) are non-stationary in levels, but stationary in their first differences (Text Table III.3). In addition, the trace test statistics reject the null hypothesis of no cointegration in favor of one cointegrating vector at the 1 percent significance level (Text Table III.4).26

Text Table III.3.

Unit Root Test

article image

1 percent significance level.

Text Table III.4.

Cointegration tests

article image

percent significance level.

15. Given the above, we estimated an error-correction inflation equation of the unrestricted VEC model (see first three columns in Table I, Annex I). The fit of the inflation equation is satisfactory.27 The underlying cointegrating equation, which suggests a long-term equilibrium relationship among monetary aggregates, the exchange rate, real GDP and the price level, is estimated as follows:

lnCPI=0.528lnM3+0.531lnER-0.104lnRGDP-4.342(4)

All coefficients have the correct signs. Equation (4) shows that, in equilibrium, inflation tends to increase with money supply and depreciation of the exchange rate, and to decrease with real economic expansion.

A restricted VEC model

16. While the above unrestricted VEC model has reasonable coefficients and explains the systematic relationship among inflation, money supply, and the exchange rate, the estimated coefficients are only marginally significant in an econometric sense, possibly reflecting multicollinearity among variables in the sample period. Imposing constraints on statistically insignificant coefficients for a number of independent variables can result in a more concise specification of the error-correction inflation equation, while retaining the long-run relationship stated in equation (4). A restricted VEC model was estimated which significantly reduced the number of explanatory variables in the error-correction inflation equation (see last three columns in Table III.A, Annex I).

17. The restricted VEC model for inflation has broadly the correct signs of the coefficients for the explanatory variables.28 Notably, the model indicates that inflation is largely explained by the depreciation of the exchange rate and, to a lower extent, by the growth of broad money with a two-lag adjustment period. Also, there exists substantial inflation inertia. Specifically, the error correction terms enters negatively and significantly, implying that if inflation is 1 percent below its equilibrium level in one month, inflation will increase by roughly 0.02 percent in the following month. The role of inflation inertia is also highlighted by the high and significant coefficient of lagged inflation. Due to possible measurement errors, the estimated behavior of the real GDP variable remains statistically insignificant.

D. Impulse Analysis and Granger-causality Tests

18. An impulse response function, in which one-standard deviation of shock is given to an explanatory variable, is depicted in Figure III.3. The chart shows that, compared to other variables in the model, a shock to the exchange rate has the largest impact on domestic inflation in terms of magnitude, and it would take approximately two years for this transitory impact on inflation to be fully-absorbed in the inflation rate. This could be interpreted to mean that the formation of inflationary expectations is backward-looking, to a degree. A shock to nominal money growth briefly leads to an increase in inflation, although this response is very small and statistically insignificant. This outcome, along with the large increases in money growth observed in 2000–03, might be suggesting an ongoing step-increase in money demand.

19. When applying the Granger causality tests to the above restricted error-correction inflation model, we find that inflation Granger-causes depreciation of the exchange rate, and vice versa(Text Table III.5).29 This result can be interpreted to mean that inflation and depreciation expectations might be playing an important role in the economy given the past history of persistent inflation and continued exchange rate depreciation. At the same time, money supply does not seem to Granger-cause inflation.

20. Concerning the relationship between monetary aggregates and the exchange rate, the evidence indicates that there is two-way causality; while the former seems to weakly cause the latter in a statistical sense, exchange rate movements seem to affect monetary aggregates in a stronger manner.

Figure III.3.
Figure III.3.

Impulse Response Function

Citation: IMF Staff Country Reports 2006, 329; 10.5089/9781451835090.002.A003

Text Table III.5:

Granger Causality 1/ 2/

article image

** 5% significance level; *** 1% significance level.

The null hypothesis is that there is no causality.

E. References

  • Gasha, J. (2003), “A Note on Inflation,” in: Angola—Selected Issues and Statistical Appendix, IMF Staff Country Report No. 03/292, by Pastor, G. and others (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Johansen, S. (1988), Statistical analysis of cointegration vectors, Journal of Economic Dynamics and Control, 12, pp. 231254.

  • Johansen, S. (1995), Likelihood-Based Inference in Cointegrated Vector Autoregressive Models (New York: Oxford University Press).

  • Johansen, S., and Juselius, K. (1992), Testing structural hypotheses in a multivariate cointegration analysis of the PPP and the UIP for UK, Journal of Econometrics, 53, pp. 211244.

    • Search Google Scholar
    • Export Citation
  • Kuijs, A. (2000), “Monetary Policy, the Exchange Rate and Inflation,” in: São Tomé and Príncipe—Recent Economic Developments and Selected Issues, IMF Staff Country Report No. 00/69, by Thiam, I. and others (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Muñoz, S. (2005), “High Inflation and Money Demand,” in: Zimbabwe—Selected Issues and Statistical Appendix, IMF Staff Country Report No. 05/359, by Coorey, S. and others (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Nassar, K. (2005), “Determinants of Inflation in Madagascar,” in: Republic of Madagascar—Selected Issues and Statistical Appendix, IMF Staff Country Report No. 05/321, by Fayolle, A. and others (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

Chapter III Annex I

Statistical Properties of the Error-Correction Inflation Equation

1. This annex presents the econometric estimates of the error-correction inflation equation estimated for São Tomé and Príncipe (Table III.A). As noted in the text, our error-correction inflation equation could be stated as follows:

Table III.A:

Coefficients of Error-Correction Inflation Equation 1/ 2/

article image

The dependent variable is ΔlnCPI.

*10% significance level, ** 5% significance level; *** 1% significance level.

ΔlnCPI=β0+β1kΔlnCPItk+β2kΔlnM3tk+β3kΔERtk+β4kΔRGDPtk+β5ECMlnCPItk+εt(1)

where Δln is the first difference in logs of the variables, and ECM ln CPIt is an error correction term associated with disequilibrium from the long-term equilibrium in the money market.

2. A number of tests was conducted to evaluate the statistical properties of our error-correction inflation model and the estimated long-run equilibrium relationship between inflation and key explanatory variables (equation 4 in text). The diagnostic tests are, in general, encouraging:

  • The corresponding adjustment matrix indicates that the estimated eigenvector is important for all four equations of the unrestricted error-correction model (Table III.B).

  • The standard weak exogenous tests based on the adjustment speed parameters in Table II suggest that all variables included in the model are endogenously-determined as a unit system.30 Inflation, money supply, exchange depreciation and real GDP cannot be treated as “weakly exogenous.”

  • Although there remain certain concerns about stability of the estimates in the error-correction inflation equation, the null hypothesis of no autocorrelation cannot be rejected at the 5 percent significance level,31 and the standard normality tests cannot be rejected at least at the 1 percent level.32

  • The coefficient associated with the error correction term is negative, as expected, and its absolute value indicates that the speed of the adjustment process toward equilibrium would be relatively slow in the short-run dynamics.

Table III.B.

Unit Root Test

article image

:1 percent significance level.

Table 1.

São Tomé and Príncipe: Gross Domestic Product and Expenditure at Current Prices, 2001–04

(In billions of dobras, unless otherwise specified)

article image
Sources: São Tomé and Príncipe authorities; and staff estimates.
Table 2.

São Tomé and Príncipe: Gross Domestic Product and Expenditure at Current Prices, 2001–04

(In percent of GDP)

article image
Sources: São Tomé and Príncipe authorities; and staff estimates.
Table 3.

São Tomé and Príncipe: Gross Domestic Product and Expenditure at Constant Prices, 2001–04

article image
Sources: São Tomé and Príncipe authorities; and staff estimates.
Table 4.

São Tomé and Príncipe: Land Distribution, 2001–04 1/

article image
Sources: São Tomé and Príncipe authorities; and staff calculations.

The land distribution project was initiated in 1993 with the objective of distributing land from

government agricultural estates to small and medium-sized farms. At end-1992, the government estates covered 65,367 hectares, of which 33,821 hectares were cultivated.

Computed using only small and medium-sized farms. Forest and other noncultivable areas are not owned by individual households.

Table 5.

São Tomé and Príncipe: Production of Principal Agricultural Crops, 2001–04

(In metric tons)

article image
Source: São Tomé and Príncipe authorities.
Table 6.

São Tomé and Príncipe: Energy Production and Consumption, 2001–05

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

Production exceeds consumption, owing to losses in distribution.

Including the government.

Table 7.

São Tomé and Príncipe: Imports of Petroleum Products, 2001–05

article image
Sources: São Tomé and Príncipe authorities; and staff estimates.
Table 8.

São Tomé and Príncipe: Cost Structure of Petroleum Products, 2001–05

article image
article image
Source: São Tomé and Príncipe authorities.

ENCO: the National Petroleum Distribution Company.

Table 9.

São Tomé and Príncipe: Components of Official Consumer Price Index, 2001–05

(Index, 1996=100; end of period)

article image
Source: São Tomé and Príncipe authorities.
Table 10.

São Tomé and Príncipe: Monthly Movements in Official Consumer Price Index, 2001–05

(Index, 1996 = 100, unless otherwise indicated)

article image
Source: São Tomé and Príncipe authorities.
Table 11.

São Tomé and Príncipe: Monthly Movements in Components of Official Consumer Price Index, 2001–05

article image
article image
article image
article image
article image
Source: São Tomé and Príncipe authorities.