Journal Issue
Share
Article

Angola: Selected Issues and Statistical Appendix

Author(s):
International Monetary Fund
Published Date:
September 2003
Share
  • ShareShare
Show Summary Details

I. A Note on Inflation1

Executive Summary

1. This note documents recent developments with inflation in Angola. First, it briefly presents previous work that has been done on Angolan inflation. Second, it characterizes the inflation phenomenon and presents some basic statistics. And third, it analyzes the impact of monetary developments on inflation.

2. Despite important progress towards stabilization in recent years, inflation remains quite high by international standards. Indeed, inflation has fallen from 329 percent in 1999, to 116 percent in 2001 and to 106 percent in 2002. Nevertheless, the monthly average rate of inflation remains above 6 percent to date.

3. Broader monetary aggregates including dollar denominated assets seem to display a closer relationship with inflation than reserve money. Statistical evidence robustly suggests that both in the short- and in the long-run, M2 and M3 are more informative regarding inflation than reserve money, even when assuming different lag structures for the latter. This high information content of broader monetary aggregates has been found in other highly dollarized economies (such as Argentina, Peru and Turkey). In this sense, targeting kwanza-denominated monetary aggregates might not be enough to reduce inflation. Increasing surveillance over broader aggregates might help in this regard.

4. Several years of high inflation rates have negatively affected economic conditions in the country: real wages have been eroded, real money balances have declined and dollarization has reached very high levels.

A. Introduction

5. After several years of political conflict and poor economic performance, Angola has embarked on a gradual process of economic stabilization and reform that has yielded some advances. In particular, inflation has fallen from 329 percent in 1999, to 268 percent in 2000, 116 percent in 2000, and 106 percent in 2002 (Figure I.1). As suggested by some statistics, both the average and the volatility of the inflation rate have fallen (Table I.1). While, in 1999 and 2000, the average and the standard deviation of monthly inflation were 12.3 percent and 6.2 percent respectively, in 2001 and 2002, they fell to 6.5 percent and 3.1 percent, respectively.

Figure I.1.Inflation (12-month percent change), Jan. 1998 – Dec. 2002

Sources: Angolan authorities; and Fund staff.

Table I.1.Inflation (in percent)
Monthly12-month change
AverageStandard DeviationAverageStandard Deviation
Jan 1997 - Dec 20028.25.9191.7101.5
Jan 1997 - Dec 19985.85.8103.933.8
Jan 1999 - Dec 200012.36.2288.084.6
Jan 2001 - Dec 20026.53.1139.346.7
Jan 2002 - Dec 20026.21.5109.39.2
Sources: Angolan authorities; and Fund staff.
Sources: Angolan authorities; and Fund staff.

6. Although a large reduction has been achieved, inflation still remains relatively high with respect to international standards (Figure I.2.). For the period, January 2002 to December 2002, the average monthly rate of inflation has been 6.20 percent, while the average 12-month rate reached 109.30 percent. At the same time, the monthly rates of inflation in the last months of 2002 were consistently higher than those in 2001. For instance: 3.95 percent versus 7.30 percent in July; 3.41 percent versus 5.20 percent in September; 4.10 percent versus 6.16 percent in November.

Figure I.2.Inflation (12-month percent change), Jan, 2001 – Dec. 2002

Sources: Angolan authorities; and Fund staff.

7. The rest of the note is organized as follows. Section II describes previous analytical work done on Angolan inflation. Section III characterizes the inflation phenomenon and presents some basic statistics. Section IV elaborates on the costs of inflation. Section V presents a number of econometric tests used to analyze the basic relationships between inflation monetary, money growth, and exchange rate depreciation. A main finding from the analysis is that the broader the monetary aggregate, the stronger its link with inflation in the short- and the medium-run. Section VI discusses some policy issues. Finally, Section VII advances some final concluding remarks. The various annexes to the note present the technical details of the econometric work; the Angolan authorities’ view on the inflation phenomenon as presented in an analytical paper furnished to staff;2 and some general remarks on high inflation episodes in other countries around the world.

B. Previous Analysis

8. Inflation has long been recognized as the main macroeconomic problem in Angola. In particular, Aguilar (1994) found an important relationship between inflation and the parallel exchange rate, and a strong structure of expectations and inflation inertia. Regarding trends with relative prices, the author argued that while the prices of food and beverages grew faster than the general level, prices of housing, utilities and commuting grew at a slower pace on account of price controls and/or regulations. Interestingly, Aguilar found that the informal goods market was more (price) competitive than the formal one, which had important monopolistic structures supported by the government.

9. The inflation phenomenon has also been extensively documented in previous analysis by the staff. For instance, Ize et al. (2001, 2002) suggest that, despite some advances, inflation in Angola remains quite high. Ize’s analysis suggests that: (i) excess kwanza issues, induced by large fiscal and quasi-fiscal deficits are at the root of the high inflation problem; and (ii) kwanza issues affect prices directly and indirectly via the exchange rate’s depreciation.

10. Ize et al. propose a number of recommendations to secure macroeconomic stabilization, namely: (i) increased controls over fiscal or quasi-fiscal imbalances; (ii) reductions in the central bank’s operational deficit; (iii) enhanced treasury and central bank coordination; (iv) and the establishment of a central bank policy of ex-ante sterilization of the government’s and/or Sonangol’s foreign exchange operations.

11. More recently, the staff’s analysis in the context of the 2002 Article IV consultation, noted that the main source of inflation has been the insufficient control of public spending, including notably large extra budgetary expenditures. Against this background, the staff emphasized the need to implement a consistent stabilization program and introduce structural reforms based on prudent monetary policy supported by fiscal consolidation, a flexible exchange rate policy, and appropriate transparency and governance measures.

12. The Angolan authorities have also presented its view on the inflation phenomenon in “Estudo sobre Politica de Rendimientos e Preços” prepared by “Consultores Internacionais, SA”. Briefly, the document argues that inflation is being caused by a series of factors, mainly the monetization of the fiscal deficit, the behavior of import prices, goods market imperfections, and the insufficient local supply of goods and services. Fiscal deficits and their monetization are, very likely, the main causes of the inflation phenomenon. Nevertheless, it is not clear (at least not based on the information presented in the authorities document) how market imperfections, or import price formation, might sustain quite high and persistent rates of inflation.

13. The authorities also suggest that given that the growth of kwanza issuance has been declining over time, money has been loosing importance when explaining inflation. As mentioned above, some evidence in our analysis suggests that broader aggregates that include dollar-denominated assets perform better than those that do not them when explaining inflation. Annex I presents a more extensive description of the government’s document with comments.

C. Characterization of the Inflation Phenomenon

14. Table I.2 and Figures I.3-4 present some basic statistics and show the paths of inflation and the rates of growth of monetary aggregates (reserve money, money, M2 and M3) during the last few years. In particular, the data in Table I.2. suggest that inflation and the growth rates of monetary aggregates have been converging over time, although money growth, with the exception of reserve money (RM) growth, has been faster than inflation. For instance, while the average annual rate of inflation was 109 percent in 2002, RM, money (M), M2 and M3 grew by 101, 127, 133, and 135 percent, respectively. Also, the volatility of monetary growth rates (measured by the standard deviation of annual growth rates) has been higher than that of inflation. For example, while the inflation was 9.2 percent in 2002, that of reserve money and money were 58.5 percent and 15.3 percent, respectively. When looking at monthly rates, in 2002, on average, inflation and monetary aggregates’ growth have had similar patterns. Interestingly the broader the measure, the closer the growth rate to inflation. Exchange rate depreciation has also decreased in recent years. Specifically, the 12-month nominal exchange rate depreciation declined from 214 percent in 2000 to 90 percent in 2002. While in 2000, the monthly average rate of depreciation was equal to 10 percent, in 2002 it was 5.6 percent.

Table I.2.Inflation, Money Growth and Depreciation, Jan. 1997 - Dec. 2002 1/(In percent)
Monthly Rates of Change
Jan 97 - Dec 02Jan 97 - Dec 98Jan 99 - Dec 00Jan 01 - Dec 02Jan 02 - Dec 02
AvSDAvSDAvSDAvSDAvSD
Inflation8.25.95.85.812.36.26.53.16.21.5
Reserve Money11.521.98.820.915.922.49.822.77.215.1
Money10.114.07.011.916.418.16.68.35.86.7
M210.815.47.512.017.121.87.87.06.95.5
M39.814.85.49.416.021.57.87.26.95.5
Official Exchange Rate9.722.96.010.117.637.35.43.15.20.9
Parallel Exchange Rate8.711.17.213.013.112.75.64.85.64.7
12-month Rates of Change
Jan 97 - Dec 02Jan 97 - Dec 98Jan 99 - Dec 00Jan 01 - Dec 02Jan 02 - Dec 02
AvSDAvSDAvSDAvSDAvSD
Inflation191.7101.5103.933.8288.084.6139.346.7109.39.2
Reserve Money253.2226.359.563.4438.7243.1164.582.1101.158.5
Money249.2200.490.339.4414.0222.0163.962.1126.915.3
M2282.1230.594.039.0474.3249.6183.978.3133.412.6
M3240.3177.785.331.8371.0201.9187.079.9134.713.7
Official Exchange Rate252.5260.173.551.0477.9284.6116.548.198.613.3
Parallel Exchange Rate201.1149.0106.542.0333.4150.4116.154.589.89.9
Sources: Angolan authorities; and Fund staff.

Av stands for averages and SD denotes standard deviations.

Sources: Angolan authorities; and Fund staff.

Av stands for averages and SD denotes standard deviations.

Figure I.3.Inflation and Monetary Aggregates, 1998-2002

Sources: Angolan authorities; and Fund staff.

Figure I.4.Inflation and Monetary Aggregates, Dec. 2000 – Dec. 2002

Sources: Angolan authorities; and Fund staff.

15. Table I.3 presents simple correlations between inflation and monetary aggregates. When looking at 12-month rates of growth, inflation and all monetary aggregates are highly correlated. Further, these correlations increased slightly when the sample period is restricted to the last three years. Note also that when looking at monthly rates, only broad aggregates (M2 and M3) seem to be relatively correlated with prices. Indeed, the correlation between inflation and reserve money growth is very low.3

Table I.3.Correlations, Jan. 1997 - Aug. 2002
Correlations: 12-month rates of change
Jan. 97 - Aug. 02Jan. 97 - Dec. 99Jan. 00 - Aug. 02
Inf, RM0.880.810.91
Inf, M0.870.830.88
Inf, M20.820.850.85
Inf, M30.830.830.83
Inf, EO0.670.890.68
Inf, EP0.680.830.89
RM, EO0.660.840.78
RM, EP0.620.850.83
M, EO0.830.900.92
M, EP0.770.920.90
M2, EO0.910.940.94
M2, EP0.880.950.91
M3, EO0.840.910.92
M3, EP0.780.920.90
EO, EP0.870.930.78
Correlations: monthly rates of change
Jan. 97 - Aug. 02Jan. 97 - Dec. 99Jan. 00 - Aug. 02
Inf, RM0.110.150.04
Inf, M0.300.290.32
Inf, M20.310.280.42
Inf, M30.330.300.41
Inf, EO0.230.240.46
Inf, EP0.360.250.62
RM, EO0.090.12-0.05
RM, EP0.110.100.12
M, EO0.680.740.54
M, EP0.480.440.57
M2, EO0.800.830.67
M2, EP0.480.420.71
M3, EO0.840.900.65
M3, EP0.460.400.70
EO, EP0.250.200.78
Sources: Angolan authorities; and Fund staff.
Sources: Angolan authorities; and Fund staff.

16. Table I.3 also presents some results between inflation and exchange rate depreciation. Inflation and depreciation of the official (EO) and parallel (EP) exchange rates are relatively highly correlated. Monthly rates suggest that the correlation seems to have increased in the second half of the sample period (2000-02) when compared to the first half (1997-99).

D. The Perils of Inflation

17. As suggested by a number of authors very high or even moderate inflation could be harmful to economic growth.4 While some of the costs of inflation could be triggered by high inflation rates, others could be the result of high variability and/or uncertainty over inflation trends. The literature suggests that high inflation rates could negatively affect economic growth by distorting relative prices signals and making the allocation of resources more difficult. For example, Ghosh and Phillips (1998) found a significant and fairly robust negative relationship between inflation and economic growth. Barro (1995) suggests that the long-term effects of inflation on living standards are substantial. He suggests that an increase in the long-term average inflation rate of about 10 percent per year would negatively affect the level of real GDP after 30 years by 4 to 7 percent.5

18. Although a detailed effort to assess the costs of inflation in Angola would require much more research, some indicators could help understand why a reduction in inflation would be welcomed. Figure I.5 plots inflation and real GDP growth rates for the last years. It suggests that inflation is negatively correlated with economic growth. Figure I.6 presents the evolution of real wages during the last decade suggesting that real wages have been reduced by more than half against the background of recurrent (hyper) high-inflation episodes.6

Figure I.5.Inflation and Real Gross Domestic Product Growth

(In percent)

Sources: Angolan authorities; and Fund staff.

Figure I.6.Real Wages (Index, 1991 = 100), 1991 - 2000

Source: The World Bank.

19. Yet, another indicator suggesting how harmful inflation has been to the Angolan economy is given by the evolution of real money balances. Figure I.7 shows that real money balances are negatively affected by inflation: whenever inflation has increased real balances had declined. Real money balances have experienced some recovery in recent years given some degree of stabilization.

Figure I.7.Inflation and Real Money Balances, 1998Q1 - 2002Q1

Sources: Angolan authorities; and Fund staff.

20. High levels of dollarization have also been the result of high and persistent inflation. U.S. dollars are apparently a main medium of exchange in Angola. As suggested by the literature, dollarization may be costly, not only because of its revenue-loss implications (in terms of seignoriage), but also because it may generate a loss of monetary and exchange rate independence. Figure I.8 shows that the degree of dollarization of the Angolan economy is high, with about 80 percent of demand deposits being dollar-denominated. Dollarization was exacerbated during 1999 and 2000 as inflation increased. In the last years, however, dollarization levels have stabilized at very high rates. Simple correlation coefficients between inflation levels and dollarization equals 0.60.

Figure I.8.Inflation and Dollarization, Jan. 1998 - Aug. 2002

Sources: Angolan authorities; and Fund staff.

E. Empirical Analysis: Money, Exchange Rate and Prices

Data and techniques

21. Some statistical analysis of the impact of monetary developments on inflation is summarized in this section; further technical detail is included in Annex III. The empirical analysis has been performed using monthly data from January 1997 to August 2002. Given the hyperinflation experienced by the country prior to 1997, data before January 1997 were not considered. Standard and canonical time series techniques have been used in the analysis. Unit root and structural break tests, cointegration test, error correction-models, vector auto regression and impulse response and variance decomposition analysis have been applied to the data7. The consumer price index (CPI), reserve money (RM), money (M), M2, M3, the parallel exchange rate (EP), the official exchange rate (EO), in levels, monthly changes or yearly changes were variables used in the econometric tests.

22. The main results of the statistical/econometric analysis (details are extensively discussed in Annex III), are the following:

  • There exists stable long-run relationship among prices, monetary aggregates and the exchange rate in Angola. Our results suggest that, over the long-run broader monetary aggregates (such as M2 or M3) display a stronger relationship with inflation. For example, while the parameter of reserve money in the price equation equals 0.34, the parameter for M2 equal 0.56 (see Table I.8 in Annex III).
  • The changes in broad monetary aggregates seem to (statistically) cause inflation. Indeed, the broader the aggregate, the stronger the causality evidence. In terms of causality in the opposite direction, only reserve money growth seems to be caused by inflation.
  • Statistical evidence suggests that exchange rate depreciation causes inflation a-la-Granger. Exchange rate movements may have an important effect on prices given the significant level of pass-through and the high degree of dollarization of economy.
  • In terms of the relationship between monetary aggregates and exchange rate, the former seem to (weakly) cause the latter. In addition, exchange rate movements seem to affect monetary aggregates. This might be the result of monetary interventions in response to exchange rate depreciations.
  • Over the business cycle (i.e., short-term), the growth of monetary aggregate seems to have a significant effect on inflation. As in the long run, broad monetary aggregate have a stronger relationship with inflation than narrow monies8. For example, while the elasticity of prices with respect to reserve money equals 0.04, the elasticity with respect to M3 equals 0.09 (see Table I.10 in Annex III).

23. Moreover, as suggested by authors, dollarization raises the issue of the appropriate monetary aggregate in the conduct of monetary policy in floating exchange rate regimes. For instance, Berg and Borensztein (2000) find that broad aggregates that include foreign currency deposits are more informative about inflation development than those that do not.9 In the same vein, in a highly dollarized economy like Angola, some of our results suggest that broader aggregates (M2, M3) might be more informative about inflation than narrower ones.

F. Policy Discussion

24. The following guidelines for policy discussion could be inferred from the analysis:

  • Monetary & fiscal policies should be closely coordinated given the links between monetary growth induced by large fiscal and quasi-fiscal imbalances, and relatively high and persistent rates of inflation. In that sense, reinforcing control over budgetary and extra-budgetary deficits should help. The central bank needs accurate information about fiscal developments to improve its intervention in the money market and mop up excessive liquidity caused by the fiscal stimulus.
  • Increasing surveillance over broad monetary aggregates is needed given that broad money aggregates are more informative about inflation than narrow monies. To date, central bank intervention in foreign exchange markets has been the primary tool used to manage excess liquidity. An additional instrument to manage monetary aggregates, in the margin could be reserve requirements, especially those on dollar-denominated bank deposits.
  • The BNA has been trying to achieve both monetary and exchange rate targets using only one instrument: foreign exchange interventions. Credit ceilings, previously imposed, have not been binding, and therefore, are of no use. In this context, a more active use of alternative instruments such as reserve requirements and a further development of a BNA bills market may be important in managing liquidity and monitoring the exchange rate, simultaneously.
  • U.S. dollars are a main medium of exchange in Angola. The rapid depreciation of the kwanza, and the prevalence of negative real interest rates on kwanza-denominated assets further contributes to the dollarization of the economy. In this sense, given the significant share of imported goods in the consumer price basket and the important pass-through effect between exchange rate and prices, excessive exchange rate depreciation should be avoided in order to control inflationary pressures.

G. Conclusions

25. After experiencing episodes of hyperinflation and high inflation, Angola is engaged in a process of stabilization and structural reforms that has delivered some results. In particular, inflation has been reduced, but not yet to international levels. Further efforts to bring inflation down to international levels would help improve the overall macroeconomic outlook, avoiding further real wages erosion, or negative effects on growth.

26. Measures to reduce inflation need to focus on strengthening control over budgetary and extra-budgetary deficits, and over monetary aggregates. The reduction in the growth rates of narrow monetary aggregates, has delivered some reduction in inflation. Nevertheless broader monetary aggregates which include dollar denominated assets are more correlated to inflation. Thus, policies aimed at improving surveillance over those aggregates should help reduce inflation further.

ANNEX I: Authorities’ View on Inflation

1. Chapter C in the government’s paper titled “Study of Pricing and Profitability Policies” (Estudo Sobre Politica de Rendimentos e Preços) contains an analysis of inflation in Angola. The chapter describes recent inflation trends and discusses the sources and effects of inflation.

2. Regarding inflation’s sources and effects, the chapter presents a number of hypotheses. It argues that although monetary financing of central government deficits was the main source of inflation in Angola, things changed after 1999. The paper highlights an important fiscal tightening since 1999, which reduced the central government deficits and changed the composition of deficit financing. The paper argues that since 1999 the main sources of inflation have been: (i) the high level of extra budgetary spending; (ii) the quasi-fiscal operations of the central bank, and (iii) the monetization of large capital inflows which are caused by growing oil export revenues accruing to Sonangol, and by large imports of funds by multinational oil corporations used to pay taxes.

3. The paper also acknowledges that high inflation is a regressive tax on the population and that it is highly volatile, thus leading to the formation of adverse inflation expectations and inflation inertia. The paper then proposes a number of ideas on how to address the inflation problem in Angola:

  • First, the paper argues that inflation is the result of an under-supply of goods and services in the market. Imports of capital to pay oil corporations’ taxes, together with commodity import privileges accruing to these enterprises to furnish their operational and investment needs, depress the demand and supply of domestic goods and services, and generates unemployment. To correct this situation, the paper recommends that all capital inflows be intermediated by the domestic banking system which would, in turn, lend on those funds for productive domestic activities leading to an increase in aggregate supply. According to the paper, the current practice of central bank monetization and sterilization of capital inflows complicates monetary control and induces increases in aggregate demand without a corresponding increase in aggregate supply.
  • Second, the paper refers to a number of market imperfections as additional inflationary sources. Statistics on the high concentration of import permits in Angola are revealed: less than 6 percent of the total number of authorized importers handle more than 70 percent of total commodity imports in the country. Also, Angola’s direction of trade on the import side is highly concentrated with six countries mainly Portugal, Spain, South Africa, and Brazil. The high import concentration (i.e., among few authorized importers and suppliers) results in very large sales mark-ups in prices which are even further compounded by the bureaucratic red-tape (and bribing) behind the existing import system in Angola.
  • Third, the paper notices that revenue from seigniorage accruing to the BNA is way above the “1.5 percent of GDP limit recommended by the IMF.” At current seigniorage rates equivalent to about 12 percent of GDP in Angola according to the authorities, the paper acknowledges the high risks of entering an “uncontrollable” inflation path, leading to important macroeconomic imbalances.
  • Fourth, the paper expresses concern about the relatively high ratio of currency in circulation to nominal GDP in Angola. It argues that, despite recent declines in the ratio, these notes and bills in the hands of the public are a potential source of inflation. Given the current weaknesses of the Angolan banking system (in terms of low interest rates paid on deposits) no easy solutions to this problem exist. Notably, the paper claims that currency in the hands of the informal economy poses no inflationary threat to the economy, as these money balances are used in “productive” activities.
  • Finally, the paper argues that the composition of government spending, generally in low-productive activities, does not increase the economy’s aggregate supply, but increases aggregate demand, with the resulting inflationary impact.

4. A number of issues are worth raising on the authorities’ views on inflation and recommendations:

  • First, the authorities’ analysis seems to confuse inflation with step-up increases in domestic prices (vis-à-vis international commodity prices) due to market imperfections such as the monopolistic behavior of companies in the import and domestic distribution channels.
  • Second, the authorities’ argument that inflation since 1999 has not been caused by central government deficits, but rather by growing quasi-fiscal deficits (including large extra budgetary expenses and central bank losses), ignores the fact that money is a fungible commodity. In general, the fiscal impulse stemming from fiscal or quasi-fiscal deficits would have similar impacts on inflation, especially at very high inflation rates as in Angola.
  • Third, the authorities implicitly assume that money demand in Angola could be determined independently of the level of inflation and the real interest rate when they argue that banking sector intermediation of financial flows (currently accruing to the central bank) would yield an increase in aggregate supply, rather than putting pressure on prices and the exchange rate, as is now the case. Somehow, they believe that the level of real money balances held by the population could be altered through institutional changes orchestrated by the government or the domestic banking sector.
  • Fourth, the authorities’ conclusion that very high rates of seigniorage revenue could lead to high rates of inflation seems reasonable. For one, seigniorage revenue levels in Angola are relatively high when compared to levels in other African economies (see Table I.4, below). The most common measure of seigniorage is the ratio of the change in reserve money divided by nominal GDP, rather than the change in money divided by nominal GDP as measured in the authorities’ paper. Yet, under either definition, seigniorage in Angola is rather high by international standards, although it has declined somewhat in recent years.
Table I.4.Seignoriage: Angola and Other African countries, 1996 - 2002
AngolaBotzawanaCameroonDRCNamibiaNigeriaSouth Africa
199610.450.340.843.200.620.400.55
19974.720.671.401.500.600.300.47
19983.400.670.4761.320.121.170.23
19999.990.470.3851.531.331.550.88
20004.790.201.59-4.47-0.252.830.27
20014.600.391.270.262.550.68
20023.720.253.350.60
Memorandum item
Average (1996 - 2002)5.950.430.9319.060.451.730.53
Source: IMF, World Economic Outlook database.

Seignoriage is defined as the change in reserve money divided by nominal GDP.

Source: IMF, World Economic Outlook database.

Seignoriage is defined as the change in reserve money divided by nominal GDP.

ANNEX II: Some International Inflationary Experiences

1. On a general note, high and persistent inflation has been extensively documented and studied. In a recent example, Fischer et al. (2002) characterize a number of high inflation episodes and present a number of findings. Notably: (i) higher inflation tends to be more unstable; (ii) the relationship between fiscal balance and seignoriage is strong (both in the short- and the long-run); (iii) inflation inertia decreases as inflation rises; and (iv) in general, high inflation is associated with poor macroeconomic performance.

2. Even though the inflation mechanism is complex, extensive theoretical and empirical research has been devoted to it. As suggested by Fischer et al (2002) consensus seems to prevail on several points: (i) money and inflation are highly correlated in the long-run; (ii) fiscal imbalances usually explain (trigger) high inflation; and (iii) inflation may be the unintended consequence of inappropriate monetary policies targeting unemployment, exchange rates, and/or interest rates.

3. Even though a number of high inflation episodes have taken place in Africa, the inflation phenomenon has been studied much less than in other regions. Yet, one recent study by Reinhart and Rogoff (2002) examines the experiences of some African countries with inflation. They report that from 1970 to 2001 five countries (Democratic Republic of Congo, Angola, Ethiopia, Uganda and Zambia) have had an average annual rate of inflation higher than 40 percent. They suggest that, aside from political conflicts, inflation and fiscal problems seem to be related.

4. A number of market economies have experienced in recent history high and persistent inflation, and in some rare cases, hyperinflation.10 Based on Calvo and Vegh (1992) and Fischer et al. (2002), some quantification of inflation episodes is presented below (Table I.5). We also present the type of stabilization programs that were implemented (Table I.6).

Table I.5.Selected High Inflation and Hyperinflation Episodes
High Inflation Episodes 1/
CountryDate of EpisodeDuring High Inflation /212 Month After /2
StartEndDur. (months)CumulativeAverageHighestAverageHighest
AngolaJan-91Jun-9778287,726,17221.084.11.83.0
ArgentinaJul-74Oct-912083,809,187,961,39612.4196.61.43.0
BoliviaAug-81Aug-86615,220,26119.5182.80.72.4
BrazilApr-80May-9518220,759,903,275,65115.480.71.74.4
ChileOct-71May-7768127,95811.187.53.04.2
Congo Dem. Rep.Feb-88Jul-89182026.320.43.15.9
GhanaMay-82Feb-84222435.823.40.34.9
Guinea-BissauSep-86Feb-88181465.125.04.612.6
IsraelDec-78Mar-8688109,1878.327.51.73.3
MexicoDec-85Aug-88337246.615.51.32.5
PeruDec-86Mar-926425,392,22321.5397.03.54.8
Sierra LeoneFeb-89Dec-91356896.119.92.55.9
SomaliaMar-83Jun-84161405.619.62.79.0
UgandaFeb-84Dec-88599,0718.037.93.86.9
Hyperinflation Episodes 1/
CountryDate of EpisodeDuring High Inflation /212 Month After /2
StartEndDur. (months)CumulativeAverageHighestAverageHighest
AngolaDec-94Jun-961962,44640.384.19.538.1
ArgentinaMay-S9Mar-901115,16758.0196.612.027.0
BoliviaApr-84Sep-851897,28246.6182.85.733.0
BrazilDec-89Mar-90469367.880.814.821.5
Congo Dem. Rep.Nov-93Sep-941169,50281.3250.012.926.2
NicaraguaJun-86Mar-915811,895,866,14337.8261.21.820.3
ArmeniaOct-93Dec-941534,15847.6437.82.47.8
SerbiaFeb-93Jan-9412156,312,790228.2175092.81.012.4
UkraineApr-91Nov-94441,864,71525.0285.310.928.4

Monthly inflation statistics.

Monthly inflation statistics.

Table I.6.Stabilization Programs
Some Experiences on Exchange Rate-Based Stabilization Programs 1/
CountryBeginning and Ending DatesExchange Rate ArrangementsInflation Rate
InitialInflation LowestRate Date Achieved
IsraelJun 85 - Sep 86Fixed, crawling peg1128.950.1Jun-86
BrazilFeb 86 - Nov 86Fixed286.076.2Nov-86
MexicoDec 87 - Dec 94Fixed, crawling peg, band159.06.7Sep-94
UruguayDec 90 - presentBand133.724.4Dec-96
ArgentinaApr 91 - 01Currency Board267.0-0.3May-96
Some Experiences on Money-Based Stabilization Programs 1/
CountryBeginning and Ending DatesExchange Rate ArrangementsInflation Rate
InitialLowestDate Achieved
ChileApr 75 - Dec 77Monetary anchor394.363.4Dec-77
ArgentinaDec 89 - Feb 91Liquidity cut/floating rate4923.3287.3Feb-91
BrazilMar 90-Jan 91Tight monetary policy5747.3119.5Jan-91
Rep. Dom.Aug 90 - presentER unification and floating60.02.5Nov-93
PeruAug 90 - presentMonetary anchor/Dirty Floating12377.810.2Sep-95

5. A main message from the literature is that, aside from the nominal anchor used in the stabilization program, key elements common to successful programs were: (i) the control of fiscal imbalances; and (ii) policy credibility to reduce inflationary expectations.

ANNEX III: Unit Roots, Structural Breaks, and Cointegration

1. Individual unit-root and structural break tests were performed for each variable. Also, following Perron (1997), tests for contrasting unit roots and possible structural breaks of unknown dates were performed. Table I.7 presents the results of some of the test. Note that in all but two cases, no statistical evidence for rejecting the hypothesis of unit root was found. All variables, in levels (i.e., logarithms), were found to be integrated of order one, and no (strong) evidence of structural breaks were found for the period considered.11

Table I.7.Unit Root/Structural Break Tests
Unit Root/Structural Break Test
Break in:CPIRMMM2M3EPEO
Intercept-5.70** (1999.07)-4.63-4.51-4.39-4.41-3.51-9.15*** (1999.03)
Slope-3.02-3.46-3.4-3.25-3.04-2.98-2.98
Both-4.16-5.16-4.22-4.36-4.06-4.13-8.59*** (1999.03)
Notes:1. Variables are in logarithms.2. Test suggested in Perron (1997) when the date of possible break is unknown.3. *, **, *** indicates 10%, 5% or 1% of significance respectively.4. When evidence of structural break is found, the date is presented in parenthesis.
Notes:1. Variables are in logarithms.2. Test suggested in Perron (1997) when the date of possible break is unknown.3. *, **, *** indicates 10%, 5% or 1% of significance respectively.4. When evidence of structural break is found, the date is presented in parenthesis.

2. Cointegration tests among these 1(1) variables were then performed following the standard Johansen (1991) procedure. Some of the results are presented in Table I.8. For the tests, the following ordering is used: CPI, the monetary aggregate, and parallel exchange rate (all in logarithms). In addition, a linear trend and one lag was considered.12

Table I.8.Cointegration: Money, Exchange Rate and Prices
Cointegration Tests
StatisticRMMM2M3
Trace9.6171***12.5173***10.8049**11.3196**
Max-Eigen.9.4120***12.0547***10.3450**10.6249*
Cointegrating Vectors1111
Cointegrating Coefficients
LCPI1111
L Monetary Aggregate0.3470***0.5443***0.5651***0.4849***
(0.0488)(0.0789)(0.0898)(0.0690)
LEP0.6084***0.3859***0.3192***0.4610***
(0.0542)(0.0877)(0.1066)(0.0755)
Notes:1. Variables are in logarithms.2. The following ordering is considered: CPI Monetary Aggregate, EP.3. Linear trend and one lag have been used in the tests.4. *, **, *** indicates 10%, 5% or 1% of significance respectively.5. The cointegrating coefficients reported are those of prices as a function of monetary aggregates and exchange rates.6. Standatd errors are reported in parenthesis.
Notes:1. Variables are in logarithms.2. The following ordering is considered: CPI Monetary Aggregate, EP.3. Linear trend and one lag have been used in the tests.4. *, **, *** indicates 10%, 5% or 1% of significance respectively.5. The cointegrating coefficients reported are those of prices as a function of monetary aggregates and exchange rates.6. Standatd errors are reported in parenthesis.

3. Regardless of the monetary aggregate considered evidence suggest that there exists a long/medium run stable (cointegrating) relationship among prices, monetary aggregates, and the parallel exchange rate. Note that one cointegrating relationship has been found in each case. Results suggest that, in the long/medium run broader monetary aggregates (such as M2 or M3) display a stronger relationship with inflation. For example, while the parameter of reserve money in the price’s equation equals 0.34, the parameters in M2 and M3, equal 0.56 and 0.48, respectively.

Causality analysis

4. In order to further assess the relationship between money, the exchange rate and prices some simple econometric exercises are performed. First, the bivariate causality test (a-la-Granger13) for the variables (in differences) is presented in Table I.9.14

Table I.9.Causality: Money, Exchange Rate and Prices
Causality Test - Granger: 12-Month Rates of Changes
Hypothesis (No Causality)Jan. 97 - Aug. 02
F-StatisticProbability
RM, CPI11.35510.0000
CPI, RM2.57500.0846
M, CPI17.91450.0000
CPI, M0.53330.5900
M2, CPI16.43250.0000
CPI, M20.07520.9277
M3, CPI17,11620.0000
CPI, M31.00290.3742
EP, CPI7.59380.0013
CPI, EP1.06430.3528
EP, RM6.90590.0023
RM, EP2.14440.1180
EP, M6.96270.0022
M, EP3.83190.0284
EP, M28.33410.0007
M2, EP1.85290.1676
EP, M36.66500.0028
M3, EP2.38440.1028

5. Monetary growth seems to Granger-cause inflation. Moreover, the statistical relationship seems to be stronger, the broader the measure. In terms of causality in the opposite direction, only reserve money growth seem to be caused by inflation. Note also that exchange rate depreciation seems to Granger-cause inflation.

6. Concerning the relationship between monetary aggregates and exchange rate, the former seem to (weakly) cause the latter. Exchange rate movements seem to affect monetary aggregates. This might be the result of monetary interventions in response to exchange rate depreciations.

Error-correction results

7. Table I.10 presents some of the results for the error-correction models derived from the cointegrating relationships discussed before.15 Note that consistent with previous results, inflation seems to be more responsive to broad monetary measures than to narrow ones. Indeed, while the elasticity of prices with respect to reserve money equals 0.04 and is barely significant, the elasticity with respect to M3 equals 0.09 and is statistically significant. As mentioned before, in the case of Angola, M2 and M3 include foreign currency deposits. Because of the high share of foreign currency deposits, broad aggregates are influenced by changes in the exchange rate. The levels of foreign currency holdings have been steadily increasing. For instance, the share of foreign currency deposits in M2 has increased from 65 percent in December 2001, to 68 percent in December 2002, to 73 percent in March 2003.

Table I.10.Error-Correction Results
Dependent Variable
CPI
Predetermined Variables
CPI (-1)0.3085***0.2941**0.2835**0.2578**
(0.1193)(0.1289)(0.1278)(0.1214)
Reserve Money0.0399
(0.0267)
Reserve Money (-1)-0.0094
(0.0334)
Money0.0719*
(0.0405)
Money (-1)-0.0585
(0.0621)
M20.0801**
(0.0400)
M2 (-l)-0.0465
(0.0559)
M30.0862**
(0.0440)
M3 (-1)0.0088
(0.0424)
Parallel Exchange Rate0.1569*0.1556*0.13030.1312
(0.0931)(0.0985)(0.0901)(0.0886)
Parallel Exchange Rate (-1)0.01430.02130.0191-0.0076
(0.0643)(0.0588)(0.0587)(0.0639)
Error Correction (-1)-0.1246***-0.1480***-0.1378***-0.1324***
(0.0479)(0.0471)(0.0427)(0.0454)
R-Squared0.52780.55320.55370.5521
No. Obs.66666666
Notes:1. Variables are differences of logarithms.2. *, **, *** indicates 10%, 5% or 1% of significance respectively.3. Standard errors are shown in parenthesis.
Notes:1. Variables are differences of logarithms.2. *, **, *** indicates 10%, 5% or 1% of significance respectively.3. Standard errors are shown in parenthesis.

8. At the same time, while money becomes more important the broader its measure, the relative impact of exchange rate depreciations on inflation declines. In addition, inflation persistence seems to be of some importance, as suggested by the coefficient and significance of lagged inflation. Lags of money growth and exchange rate depreciation are relatively unimportant.

9. Finally, as suggested by the error correction term coefficients any deviation from equilibrium would have a half-life of about 4-5 months.16

Impulse response and variance decomposition

10. In order to complement the statistical and causality results, a simple VAR model was used to analyze the relationship between money, exchange rate, and prices. Impulse responses and variance decompositions were also assessed. In this case, 12-month rates of change for the period 1998-2002 were used. In order to assess the robustness of the results with alternative monetary aggregates were considered. Key results are the following:

  • Inflation significantly reacts to monetary impulses. Again, the broader the monetary aggregate, the larger and the more lasting its impact on inflation. As shown in Figures I.9 and I.10, M3 impulses produce larger effects on inflation than reserve money impulses. Note that inflation, as suggested by the analysis presented above, reacts to changes in the exchange rate, and that, it also (weakly) reacts to its own innovations.
  • Monetary innovations affect the exchange rate. As is the case with inflation, the broader the aggregate, the larger and more lasting the effect on the exchange rate (Figures I.9 and I.10). The exchange rate also reacts to its own impulses, but not to inflation innovations. Monetary aggregates react to exchange rate impulses, and to its own innovations, but do not react to inflation ones.
  • The variance decomposition analysis tells us a similar story. As shown in Figures I.11 and I.12 both reserve money and M3 importantly explain the variance of prices. In the case of M3, its importance in the inflation variance is relatively higher than that of reserve money on inflation. The exchange rate is also important in explaining the inflation variance. Exchange rate variance is mainly explained by itself and money, in that order. Money variance is explained primarily by itself, and secondly by exchange rate movements.

Figure I.9.Impulse Response: Reserve Money, Exchange Rate and Prices

Figure I.10.Impulse Response: M3, Exchange Rate and Prices

Figure I.11.Variance Decomposition: Reserve Money, Exchange Rate and Prices

Figure I.12.Variance Decomposition: M3, Exchange Rate and Prices

Pass-through

11. Movements in the exchange rate have some impact on the inflation rate. In order to further assess the importance of these effects, simple pass-through exercises are presented. Regressions of inflation on parallel exchange rate depreciation are shown. Monthly rates of change for the period January 1997 to August 2002 have been used. Results for the two sub periods, 1997 to 1999 and 2000 to 2002 are also presented in Table I.11.

Table I.11.Exchange Rate Pass-Through
Jan. 97 - Aug. 02Jan. 97 - Dec. 99Jan. 00 - Aug. 02
0.140.120.33

12. For the whole period, the level of pass-through equals 0.14. When dividing the sample in two sub periods, the results suggest that the pass-through have increased in the latter period. While the estimate equals 0.12 between 1997 and 1999, it equals 0.33 for the last three years. As suggested by Ize et al. (2002), evidence suggest that dollarization is increasing in the Angolan economy. This and the increasing openness of the economy may be part of the explanation of the importance of exchange rate movements on inflation.

References

    Aguilar, Renato, 1994, “Informe de Consultoria. Inflacion en Angola”, Mimeo.

    Barro, Robert, 1995, “Inflation and Economic Growth”, NBER Working Paper No. 5326.

    Berg, Andrew, andEduardoBorensztein, 2000, “The Choice of Exchange Rate Regime and Monetary Target in Highly Dollarized Economies”, Journal of Applied Economics, Vol. 3, No. 2.

    • Search Google Scholar
    • Export Citation

    Briault, Clive, 1995, “The Costs of Inflation”, Bank of England Quarterly Bulletin, No. 35.

    Calvo, Guillermo, and CarlosVegh, 1999, “Inflation Stabilization and BOP Crises in Developing Countries”, NBER Working Paper No. 6925.

    • Search Google Scholar
    • Export Citation

    Dickey, D.A. andW.A.Fuller, 1979, “Distribution of the Estimators for Autoregressive Time Series with a Unit Root”, Journal of the American Statistical Association, 74.

    • Search Google Scholar
    • Export Citation

    Driffill, John, GrayhamMizon, andAlistarUlph, 1990, “Costs of Inflation” in Friedman and Kahn, eds. Handbook of Monetary Economics, Vol. 2.

    • Search Google Scholar
    • Export Citation

    Fischer, Stanley, RatnaSahay, andCarlosVegh, 2002, “Modern Hyper and High Inflations”, NBER Working Paper No. 8930.

    Ghosh, Atish, andStevenPhillips, 1998, “Warning: Inflation May Be Harmful to Your Growth”, IMF Staff Paper, Vol. 45, No. 4.

    Ize, Alain, MagnusAlvesson, HernanMejia, JorgePerez, andGonzaloSanhueza, 2001, “Angola, Monetary and Foreign Exchange Policy Framework in Support of an Effective Disinflation Strategy”, International Monetary Fund, Unpublished Document.

    • Search Google Scholar
    • Export Citation

    Ize, Alain, PedroAlbuquerque, Carlos PerezVerdia, and MartinNaranjo, 2002, “Angola, Further Progress Towards Stabilization”, International Monetary Fund, Document.

    • Search Google Scholar
    • Export Citation

    Johansen, Søren, 1991, “Estimation and Hypothesis Testing of Cointegration Vectors in Gaussian Vector Autoregressive Models”, Econometrica, 59.

    • Search Google Scholar
    • Export Citation

    Perron, Phillips, 1997, “Further Evidence on Breaking Trend Functions in Macroeconomic Variables”,Journal of Econometrics 80.

    Reinhart, Carmen, andKenethRogoff, 2002, “FDI to Africa: The Role of Price Stability and Currency Instability”,International Monetary Fund, Mimeo.

    • Search Google Scholar
    • Export Citation
1

Document prepared by Jose Giancarlo Gasha. The paper was presented and discussed in a seminar at the Ministry of Finance in January 2003.

2

“Informe de Consultoria sobre Inflacion en Angola.”

3

Additional correlations were computed assuming different lags for the monetary aggregates, and the results were very similar to those in Table I.3. For instance, while the contemporary correlation between inflation and reserve money growth equals 0.0384 the correlation with one and two lags equal 0.0407 and 0.1168, respectively (the correlation turns negative and nonsignificant afterwards). For the other monetary aggregate similar results arise.

4

For instance, Driffill et al (1990) or Briault (1995) provide surveys of the theoretical literature on the costs of inflation.

5

Barro (1995) argues that overall, the general idea is that “business and households are thought to perform poorly when inflation is high and unpredictable.”

6

As documented for instance by Fischer et al. (2002).

7

It should be noted that these techniques are intended to be used, primarily, to perform long-run analysis. A six year sample period it is assumed to be of sufficient length in the Angolan case.

8

In the case of Angola, M2 and M3 include foreign currency deposits. Because of the high share of foreign currency deposits, broad aggregates are of course influenced by changes in the exchange rate. As suggested in section C, the levels of foreign currency holdings have been steadily increasing. For instance, the share of foreign currency deposits in M2 has increased from 65 percent in December 2001, to 68 percent in December 2002, to 73 percent in March 2003.

9

The authors analyze the cases of Argentina, Bolivia, Peru, the Philippines, and Turkey.

10

Fischer et al. (2002) define high inflationary episodes those with inflation rates above 100 percent per annun. Hyperinflation is defined a-la-Cagan: “as beginning in the month the rise in prices exceeds 50 percent and as ending in the month before the monthly rise in prices drops below that amount, and stays below for at least a year”.

12

Other specifications for the deterministic components and different number of lags were also considered, and the results were quite similar to those presented. Akaike and Schwarz criteria were used for the lag selection.

13

As usual, causality test primarily suggest statistical precedence.

14

The test were performed including one lag as suggested by the Akaike and Schwarz information criteria. Results including two lags yield similar results.

15

Error-correction (business cycle/short-run) results might be the most relevant ones for our analysis.

16

Half-life is computed according to Ln(0.5)/Ln(l+φ), where φ is the adjustment coefficient.

Other Resources Citing This Publication