## _wp0998

## Source details

**Canonical URL:** [_wp0998](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp0998.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp0998.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp0998.pdf.json)

---

### I. Introduction — key findings on inflation dynamics in Angola (1997–2008)
- Inflation fell from nearly 430 percent in mid-2000 to about 12 percent since mid-2006; the authorities’ inflation objective of 10 percent was not met for three consecutive years.
- Inflation reverted upwards in the second half of 2008 due to higher international food prices and strong domestic demand fueled by rapid money growth.
- Main empirical implications:
  - Strong long run relation between the exchange rate and the CPI; further exchange rate appreciation would be instrumental to moderate price growth.
  - Money growth above the estimated equilibrium level affects inflation with a lag; close monitoring of monetary aggregates is needed to avoid strong demand pressures.
  - International nonfuel commodity prices significantly impact inflation given the large share of imported food items in Angola’s CPI basket.

### II. Domestic and external developments shaping inflation
- Domestic factors:
  - 1997–2002: money growth reached three-digit levels, largely associated with BNA credit to the government and quasi-fiscal operations.
  - September 2003 stabilization ("hard kwanza"): shifted deficit funding toward foreign sources, specified an inflation objective, and enhanced liquidity management.
  - Monetary aggregates continued to grow rapidly due to financial deepening, remonetization of money demand, accumulation of international reserves, and limited sterilization; broad money growth exceeded GDP and money velocity declined.
  - GDP doubled in size since 2004 with double-digit growth rates; infrastructure constraints and labor shortages imply demand shocks can translate into price pressures.
  - Public expenditures declined from nearly 60 percent of GDP in 2000 to about 35 percent in 2008; since 2007 the declining path reversed and the non-oil primary deficit shifted upwards.
- External factors:
  - Exchange rate movements versus the U.S. dollar closely related to CPI trajectory; mid-1999 liberalization led to depreciation and high inflation until September 2003 stabilization.
  - From 2005 onward exchange rate recorded gradual appreciation reflecting increased oil revenues and rapid GDP growth; exchange rate stability and appreciation contributed to disinflation.
  - U.S. dollar depreciation against the euro over much of the sample may have put upward pressure on prices given a large share of imports from the European Union.
  - Commodity prices rose significantly from 2004; food prices rose by nearly 40 percent as of end-2007, directly affecting CPI.

### III. Cross-country evidence: other oil exporters (2005–07)
- Summary statistics (selected exact figures):
  - Average inflation rate (2005–07) sample average: 7.8 percent; above-average countries: 11.9 percent; below-average: 4.2 percent; Angola: 16.2 percent.
  - Broad money growth (average 2005–07): sample average 30.5; above-average countries 40.7; below-average 21.2; Angola 56.2.
  - Change in exchange rate (USD/national currency, average 2005–07): sample average 1.2; above-average countries 0.8; below-average 1.6; Angola 3.4.
  - Change in non-oil deficit (as percent of non-oil GDP, average 2005–07): sample average 0.7; above-average countries 1.1; below-average 0.3; Angola -1.7.
- Correlations:
  - Positive correlation 0.64 between average inflation rate and broad money growth among oil exporters.
  - Correlation between average inflation rate and average change in the nominal exchange rate (US$/national currency): -0.17.
  - Positive correlation 0.19 between average inflation rate and average change in the non-oil fiscal deficit (as percent of non-oil GDP).
- Interpretation for Angola:
  - Despite greater exchange rate appreciation and greater reduction in the non-oil deficit relative to the sample, Angola remained among the highest inflation and highest broad money growth countries, pointing to rapid broad money growth as a potential source of high inflation and implying benefits from greater monetary tightening.

### IV. Economic model and methodology
- Model specification:
  - CPI (P) modeled as a function of nominal exchange rate (ER), international non-fuel commodity prices (P_com*), excess money supply (M̂), real interest rate (R), and structural breaks: P = f(ER, P_com*, M̂, R, S).
- Econometric approach:
  - Vector Error Correction (VEC) with four endogenous variables: LnCPI, LnER, LnNGDP, LnM2; exogenous variables include structural breaks.
  - Two cointegrating relations tested:
    - Quantity-theory relation between M2 and nominal GDP with linear trend.
    - PPP-like relation between CPI and exchange rate with linear trend.
  - VEC includes adjustment coefficients (δi) to error correction components and short-run dynamics.

### V. Data and estimation details and tests
- Data:
  - Quarterly data for 1997–2007 (observations = 44).
  - Nominal GDP quarterly series constructed from oil production data and seasonal pattern of agricultural output.
- Unit-root and cointegration evidence:
  - Phillips-Perron tests indicate variables are I(1); Table A3 level and first-difference p-values reported exactly as in source (e.g., LnCPI: level 1.00; first difference 0.01).
  - Johansen Unrestricted Cointegration Rank Test — Trace test (exactly as reported):
    - None *: Eigenvalue 0.69, Trace Statistic 92.84, Critical Value (5 percent) 55.24, Probability 0
    - At most 1 *: Eigenvalue 0.56, Trace Statistic 44.67, Critical Value (5 percent) 35.01, Probability 0
    - At most 2: Eigenvalue 0.21, Trace Statistic 10.97, Critical Value (5 percent) 18.39, Probability 0.39
    - At most 3: Eigenvalue 0.02, Trace Statistic 0.94, Critical Value (5 percent) 3.84, Probability 0.33
  - Johansen Maximum Eigenvalue test (exactly as reported):
    - None *: Eigenvalue 0.69, Max-Eigen Statistic 48.17, Critical Value (5 percent) 30.815, Probability 0
    - At most 1 *: Eigenvalue 0.56, Max-Eigen Statistic 33.70, Critical Value (5 percent) 24.25, Probability 0
    - At most 2: Eigenvalue 0.21, Max-Eigen Statistic 10.02, Critical Value (5 percent) 17.147, Probability 0.39
    - At most 3: Eigenvalue 0.02, Max-Eigen Statistic 0.94, Critical Value (5 percent) 3.84, Probability 0.33
- Caution:
  - Results should be viewed with caution given structural breaks and the relatively short sample for cointegration analysis.

### VI. Long-run relations and equilibrium estimates
- Two cointegrating relations identified with coefficient restrictions:
  - Relation 1 (M2 equilibrium): LnM2 coefficient restricted to 1; long-run elasticity of M2 to GDP "is not significantly different from one."
  - Relation 2 (ER–CPI relation): LnER coefficient restricted to 1; CPI coefficient significant with expected sign.
- Estimated long-run relations (exact expressions):
  - Relation 1 (M2): (2: ε1t) = LnM2t − 1*LnNGDPt − 5.1519 + 0.004*trend
  - Relation 2 (ER): (2: ε2t) = LnERt − 0.5472 + 0.272*CPIt + 0.069*trend

### VII. Short-run dynamics (VEC model, 1997Q1–2007Q4)
- Adjustment coefficients to cointegrating vectors (V1, V2) — exact reported values (standard errors in parentheses):
  - V1:
    - D(LnM2): -0.275** (0.145)
    - D(LnNGDP): 0.582* (0.241)
    - D(LnER): 0.218 (0.213)
    - D(LnCPI): 0.091 (0.079)
  - V2:
    - D(LnM2): -0.015 (0.203)
    - D(LnNGDP): -0.060 (0.337)
    - D(LnER): -0.660* (0.297)
    - D(LnCPI): 0.214** (0.110)
- Selected lagged differenced coefficients (exact values with standard errors):
  - D(LnM2(-1)): D(LnM2) -0.709* (0.167); D(LnNGDP) -0.042 (0.277); D(LnER) -0.151 (0.244); D(LnCPI) -0.056 (0.091)
  - D(LnM2(-2)): -0.187 (0.162); -0.075 (0.269); -0.019 (0.238); -0.059 (0.088)
  - D(LnNGDP(-1)): 0.397* (0.157); 0.461*** (0.261); 0.802* (0.231); 0.194* (0.086)
  - D(LnNGDP(-2)): 0.343* (0.112); 0.179 (0.187); 0.655* (0.165); 0.091 (0.061)
  - D(LnER(-1)): 0.871* (0.151); 0.585* (0.251); 0.644* (0.221); 0.118 (0.082)
  - D(LnER(-2)): -0.307*** (0.181); -0.239 (0.300); -0.619* (0.265); -0.183** (0.098)
  - D(LnCPI(-1)): -0.401 (0.325); -0.385 (0.540); -0.370 (0.477); -0.240 (0.177)
  - D(LnCPI(-2)): -0.053 (0.292); 1.237 (0.485); 0.050 (0.428); 0.266*** (0.159)
- Controls and diagnostics (exact reported values):
  - C: D(LnM2) 0.303** (0.156); D(LnNGDP) -0.145 (0.260); D(LnER) -0.013 (0.229); D(LnCPI) 0.317 (0.085)
  - TREND: D(LnM2) 0.003 (0.004); D(LnNGDP) -0.001 (0.006); D(LnER) 0.004 (0.005); D(LnCPI) -0.007* (0.002)
  - D(LnP_COM(-1)): D(LnM2) 0.937*** (0.526); D(LnNGDP) -0.158 (0.875); D(LnER) 1.677* (0.772); D(LnCPI) 0.551** (0.287)
  - DUM_POSTCONFLICT: D(LnM2) -0.053* (0.020); D(LnNGDP) 0.000 (0.034); D(LnER) -0.072* (0.030); D(LnCPI) -0.018*** (0.011)
  - DUM_STABILIZATION: D(LnM2) -0.252* (0.088); D(LnNGDP) 0.199 (0.146); D(LnER) -0.240** (0.129); D(LnCPI) -0.050 (0.048)
  - Adjusted R-squared: D(LnM2) 0.802; D(LnNGDP) 0.522; D(LnER) 0.682; D(LnCPI) 0.839
  - Note: lags of the VEC set as 2 according to Schwartz Criterion.
- Adjustment speed and stability:
  - Implied convergence: takes three to four quarters for CPI and exchange rate to converge to equilibrium; M2 and NGDP also take three to four quarters.
  - Rolling regressions: exchange rate adjustment coefficient declined on average in recent years; CPI adjustment coefficient increased from 0.3 to 0.8.
  - Interpretation: heavier exchange rate management by the BNA implies deviations are mostly adjusted through CPI.

### VIII. External influences, structural breaks, and robustness checks
- International nonfuel commodity prices significantly influence domestic prices (high food weight and imports to Luanda).
- Structural breaks captured by dummy variables:
  - DUM_POSTCONFLICT (post-conflict era) — negative and significant effects on D(LnM2), D(LnER), D(LnCPI) as reported.
  - DUM_STABILIZATION (stabilization program start) — negative and significant effects on D(LnM2) and D(LnER) as reported.
- Other examined but insignificant variables in the system: inflation in trading partners, Kwanza/EUR exchange rate, real lending rate, fiscal balance (results not shown).

### IX. Dynamic simulation, impulse responses, and single-equation estimates
- Dynamic simulation (start 1999Q1, endogenous variables solved within model except NGDP) replicates the 1999–2000 inflation hike and subsequent disinflation.
- Impulse responses:
  - A one percent standard deviation shock to endogenous variables shifts CPI higher.
  - Positive money-supply shock triggers upward adjustment in output and increases inflation.
  - Depreciation shock has direct positive but short-lived impact on inflation; variables revert to equilibrium.
- Single-equation estimation (D(LnCPI), Models 1–5) — selected exact findings:
  - Constant C: Model 1 0.259*; Model 2 0.210*; Model 3 0.215*; Model 4 0.157*; Model 5 0.078**.
  - DUM_POSTWAR: negative and significant across models (e.g., Model 1 -0.035*).
  - DUM_STAB: negative and significant in earlier models (e.g., Model 1 -0.171*), smaller in later models.
  - EC_ER(-1): positive and significant in all models (e.g., Model 1 0.183*).
  - EC_M2(-4): positive and significant (e.g., Model 1 0.158*), indicating M2 deviations affect inflation with a four-quarter lag.
  - REALINT(-6): negative and significant where included (e.g., -0.001**).
  - D(LnP_COM(-1)): positive and significant where included (e.g., Model 3 0.520***).
  - D(LnKZ_EUR): positive and significant where included (e.g., Model 4 0.194*).
  - D(LnCPI(-2)): Model 5 coefficient 0.339*.
  - Adjusted R-squared up to 0.902 in Model 5; Durbin-Watson stats reported (e.g., Model 1 1.875).

### X. Policy conclusions and recommendations (exact emphases)
- Angola reduced inflation substantially but has not achieved price stability; further steps needed to reach low single-digit inflation consistent with SADC convergence criteria.
- Key policy messages from econometric results (with stated caution on sample limitations):
  - Inflation path over the last decade was largely affected by exchange rate movements.
  - Medium-term: greater exchange rate flexibility that facilitates gradual appreciation would moderate price growth by reducing import prices and limiting liquidity injection by the BNA.
  - If the exchange rate remains heavily managed or pegged, real exchange rate adjustment from high oil revenues will come through domestic prices, potentially creating persistent inflation momentum after oil prices stabilize.
- Exchange rate flexibility trade-offs and expected effects:
  - Must be weighed against benefits of a stable kwanza.
  - Further nominal appreciation could enhance confidence in the kwanza, accelerate de-dollarization, and strengthen monetary policy effectiveness.
  - Nominal appreciation could anchor inflation expectations and affect inflation inertia in the long run.
  - Impact on external competitiveness likely muted because expected moderation of inflation should leave REER broadly neutral and non-oil competitiveness constraints are mainly structural.
- Role of money and liquidity management:
  - Inflation responds to deviations of M2 from its equilibrium level (excess liquidity), not money growth per se.
  - Excess liquidity adds demand pressures and contributes to inflation with a lag.
  - Importance of closely monitoring broad money growth and improving liquidity management.
  - Greater sterilization efforts by the BNA are warranted to curb rapid money growth.
  - Fiscal policy should share the burden of disinflation by ensuring public spending aligns with macroeconomic and administrative capacity.

### XI. Annex: summary statistics and unit-root tests (selected exact figures)
- Table A1 summary statistics (observations = 44):
  - CPI: Mean 71.04; Median 43.05; Maximum 194.77; Minimum 0.22; Std. Dev. 70.11; Skewness 0.42; Kurtosis 1.56; Jarque-Bera 5.10; Probability 0.08; Sum 3125.
  - ER: Mean 44.29; Median 45.32; Maximum 89.20; Minimum 0.20; Std. Dev. 35.93; Skewness -0.07; Kurtosis 1.23; Jarque-Bera 5.78; Probability 0.06; Sum 1948.
  - GDP*: Mean 368.41; Median 132.50; Maximum 1666.35; Minimum 0.38; Std. Dev. 455.99; Skewness 1.16; Kurtosis 3.29; Jarque-Bera 10.04; Probability 0.01; Sum 16209.
  - M2*: Mean 186552; Median 62735; Maximum 855069; Minimum 94.61; Std. Dev. 243450; Skewness 1.33; Kurtosis 3.60; Jarque-Bera 13.54; Probability 0.00; Sum 8208320.
  - P_COM: Mean 93.47; Median 86.77; Maximum 143.49; Minimum 70.68; Std. Dev. 20.64; Skewness 1.24; Kurtosis 3.41; Jarque-Bera 11.55; Probability 0.00; Sum 4112.
  - G/GDP: Mean 32.61; Median 33.03; Maximum 49.42; Minimum 14.26; Std. Dev. 8.34; Skewness -0.06; Kurtosis 2.15; Jarque-Bera 1.36; Probability 0.51; Sum 1434.
  - Note: * Measured in billions of kwanzas.
- Table A2 correlations (exact reported):
  - CPI–ER correlation 0.92.
  - CPI–GDP correlation 0.94.
  - CPI–M2 correlation 0.92.
  - M2–P_COM correlation 0.93.
- Table A3 Phillips-Perron tests (1997–2007), unit-root null (constant and trend included) — exact p-values:
  - LnCPI: level 1.00; first difference 0.01
  - LnER: level 0.99; first difference 0.00
  - LnNGDP: level 0.96; first difference 0.00
  - LnM2: level 0.97; first difference 0.00
  - LnP_COM: level 0.56; first difference 0.00
  - LnEUR: level 0.131; first difference 0.00
  - REALINT: level 0.102; first difference 0.00
  - Note: Estimation used Newey-West bandwidth.

*Source: _wp0998 (IMF staff estimates and reported tables in provided document).*

### References..............................................................................................................

### _wp0998 - References

### I. Introduction — key findings on inflation dynamics in Angola (1997–2008)
- Inflation fell from nearly 430 percent in mid-2000 to about 12 percent since mid-2006; the authorities’ inflation objective of 10 percent was not met for three consecutive years.
- Inflation reverted upwards in the second half of 2008 due to higher international food prices and strong domestic demand fueled by rapid money growth.
- Main empirical implications highlighted:
  - Strong long run relation between the exchange rate and the CPI; further exchange rate appreciation would be instrumental to moderate price growth.
  - Money growth above the estimated equilibrium level affects inflation with a lag; close monitoring of monetary aggregates is needed to avoid strong demand pressures.
  - International nonfuel commodity prices significantly impact inflation given the large share of imported food items in Angola’s CPI basket.

### II. Domestic and external developments that shaped inflation
- Domestic factors
  - From 1997 to 2002 money growth reached three-digit levels, largely associated with BNA credit to the government and quasi-fiscal operations.
  - Stabilization introduced in September 2003 (hard kwanza policy) shifted deficit funding toward foreign sources, specified an inflation objective, and enhanced liquidity management.
  - Monetary aggregates continued to grow rapidly due to financial deepening, remonetization of money demand, accumulation of international reserves, and limited sterilization; broad money growth exceeded GDP and money velocity declined.
  - GDP doubled in size since 2004 with double-digit growth rates; infrastructure constraints and labor shortages imply demand shocks can translate into price pressures.
  - Public expenditures declined from nearly 60 percent of GDP in 2000 to about 35 percent in 2008; since 2007 the declining path reversed and the non-oil primary deficit shifted upwards.
- External factors
  - Exchange rate movements versus the U.S. dollar have been closely related to CPI trajectory; liberalization in mid-1999 led to depreciation and high inflation until the hard kwanza policy in September 2003 stabilized the rate.
  - From 2005 onward the exchange rate recorded gradual appreciation reflecting increased oil revenues and rapid GDP growth; exchange rate stability and appreciation contributed to disinflation.
  - Depreciation of the U.S. dollar against the euro over much of the sample may have put upward pressure on prices given a large share of imports from the European Union.
  - Commodity prices rose significantly from 2004; food prices rose by nearly 40 percent as of end-2007, directly affecting CPI.

### III. Cross-country evidence: experiences of other oil exporters (2005–07)
- Empirical correlations across 23 oil-exporting emerging and developing countries (sample list provided in source):
  - Average inflation rate (2005–07) sample average: 7.8 percent; countries with inflation above sample average: 11.9 percent; below average: 4.2 percent; Angola: 16.2 percent.
  - Broad money growth (average 2005–07): sample average 30.5; above-average countries 40.7; below-average 21.2; Angola 56.2.
  - Change in exchange rate (USD/national currency, average 2005–07): sample average 1.2; above-average countries 0.8; below-average 1.6; Angola 3.4.
  - Change in non-oil deficit (as percent of non-oil GDP, average 2005–07): sample average 0.7; above-average countries 1.1; below-average 0.3; Angola -1.7.
- Key patterns identified:
  - Positive correlation of 0.64 between average inflation rate and broad money growth among oil exporters.
  - Calculated correlation between average inflation rate and average change in the nominal exchange rate (US$/national currency): -0.17.
  - Positive correlation of 0.19 between average inflation rate and average change in the non-oil fiscal deficit (as percent of non-oil GDP).
- Interpretation for Angola:
  - Angola allowed greater exchange rate appreciation and greater reduction in the non-oil deficit compared to the sample average; yet it remained among the highest inflation and highest broad money growth countries, suggesting rapid broad money growth as a potential source of high inflation and the potential benefit of greater monetary tightening.

### IV. Economic model and methodology
- Model specification for CPI determinants:
  - CPI (P) is modeled as a function of nominal exchange rate (ER), international non-fuel commodity prices (P_com*), excess money supply (M̂), real interest rate (R), and structural breaks (S): P = f(ER, P_com*, M̂, R, S) (expressed in source as P = f(SR, P*, MER, M̂) in text).
- Econometric approach:
  - Vector Error Correction (VEC) model with four endogenous variables: LnCPI, LnER, LnNGDP, LnM2; exogenous variables include structural breaks.
  - Two cointegrating relations tested:
    - Long-run relation between M2 and nominal GDP (money quantity theory) with linear trend: Trend + LnNGDP − LnMV = ε (presented as equation (2)).
    - Long-run relation between CPI and exchange rate (PPP-like) with linear trend: Trend + LnCPI − LnER = ε (presented as equation (3)).
  - VEC representation includes adjustment coefficients (δi) to the error correction components and short-run dynamics across variables (equation (4) in source).

### V. Data and estimation details
- Data frequency and sample:
  - Quarterly data for 1997–2007.
  - Nominal GDP quarterly series constructed from oil production data and seasonal pattern of agricultural output.
- Statistical properties and tests:
  - Phillips-Perron unit root tests indicate all variables are I(1) (see Annex reference in source).
  - Cointegration tests (Johansen) broadly confirm two long-run relations; results (Unrestricted Cointegration Rank Test) shown below exactly as in source:
    - Trace test:
      - None *: Eigenvalue 0.69, Trace Statistic 92.84, Critical Value (5 percent) 55.24, Probability 0
      - At most 1 *: Eigenvalue 0.56, Trace Statistic 44.67, Critical Value (5 percent) 35.01, Probability 0
      - At most 2: Eigenvalue 0.21, Trace Statistic 10.97, Critical Value (5 percent) 18.39, Probability 0.39
      - At most 3: Eigenvalue 0.02, Trace Statistic 0.94, Critical Value (5 percent) 3.84, Probability 0.33
    - Maximum Eigenvalue test:
      - None *: Eigenvalue 0.69, Max-Eigen Statistic 48.17, Critical Value (5 percent) 30.815, Probability 0
      - At most 1 *: Eigenvalue 0.56, Max-Eigen Statistic 33.70, Critical Value (5 percent) 24.25, Probability 0
      - At most 2: Eigenvalue 0.21, Max-Eigen Statistic 10.02, Critical Value (5 percent) 17.147, Probability 0.39
      - At most 3: Eigenvalue 0.02, Max-Eigen Statistic 0.94, Critical Value (5 percent) 3.84, Probability 0.33
- Cautions:
  - Results should be viewed with caution given the sample period includes structural breaks in economic environment and policies and the relatively short sample for cointegration analysis.

*Source: _wp0998 - References (IMF PDF content provided).*

### 10. The estimated long-run equations provide additional support for the existence of

### 10. The estimated long-run equations provide additional support for the existence of

### Long-run relations and estimated equilibria
- Two cointegrating relations are identified.
- First long-run relation:
  - LnM2 coefficient restricted to 1.
  - Estimated coefficient of output indicates the long-run elasticity of M2 to GDP and "is not significantly different from one."
  - Positive coefficient of trend may reflect both financial deepening and the decline in velocity during the sample period.
- Second long-run relation:
  - LnER coefficient restricted to 1.
  - Coefficient of CPI is significant with the expected sign.
- Estimated long-run relations and deviations from equilibrium:
  - Relation 1 (M2): (2: ε1t) = LnM2t − 1*LnNGDPt − 5.1519 + 0.004*trend
  - Relation 2 (ER): (2: ε2t) = LnERt − 0.5472 + 0.272*CPIt + 0.069*trend

### Short-run dynamics (VEC model, 1997Q1–2007Q4)
- Table 3 summary (dependent differences: D(LnM2), D(LnNGDP), D(LnER), D(LnCPI)):
  - Adjustment Coefficient V1 (for cointegrating vector V1):
    - D(LnM2): -0.275** (0.145)
    - D(LnNGDP): 0.582* (0.241)
    - D(LnER): 0.218 (0.213)
    - D(LnCPI): 0.091 (0.079)
  - Adjustment Coefficient V2 (for cointegrating vector V2):
    - D(LnM2): -0.015 (0.203)
    - D(LnNGDP): -0.060 (0.337)
    - D(LnER): -0.660* (0.297)
    - D(LnCPI): 0.214** (0.110)
  - Selected lagged differenced coefficients (with standard errors in parentheses):
    - D(LnM2(-1)): D(LnM2) -0.709* (0.167); D(LnNGDP) -0.042 (0.277); D(LnER) -0.151 (0.244); D(LnCPI) -0.056 (0.091)
    - D(LnM2(-2)): -0.187 (0.162); -0.075 (0.269); -0.019 (0.238); -0.059 (0.088)
    - D(LnNGDP(-1)): 0.397* (0.157); 0.461*** (0.261); 0.802* (0.231); 0.194* (0.086)
    - D(LnNGDP(-2)): 0.343* (0.112); 0.179 (0.187); 0.655* (0.165); 0.091 (0.061)
    - D(LnER(-1)): 0.871* (0.151); 0.585* (0.251); 0.644* (0.221); 0.118 (0.082)
    - D(LnER(-2)): -0.307*** (0.181); -0.239 (0.300); -0.619* (0.265); -0.183** (0.098)
    - D(LnCPI(-1)): -0.401 (0.325); -0.385 (0.540); -0.370 (0.477); -0.240 (0.177)
    - D(LnCPI(-2)): -0.053 (0.292); 1.237 (0.485); 0.050 (0.428); 0.266*** (0.159)
  - Constants, trend, and other controls:
    - C: D(LnM2) 0.303** (0.156); D(LnNGDP) -0.145 (0.260); D(LnER) -0.013 (0.229); D(LnCPI) 0.317 (0.085)
    - TREND: D(LnM2) 0.003 (0.004); D(LnNGDP) -0.001 (0.006); D(LnER) 0.004 (0.005); D(LnCPI) -0.007* (0.002)
    - D(LnP_COM(-1)): D(LnM2) 0.937*** (0.526); D(LnNGDP) -0.158 (0.875); D(LnER) 1.677* (0.772); D(LnCPI) 0.551** (0.287)
    - DUM_POSTCONFLICT: D(LnM2) -0.053* (0.020); D(LnNGDP) 0.000 (0.034); D(LnER) -0.072* (0.030); D(LnCPI) -0.018*** (0.011)
    - DUM_STABILIZATION: D(LnM2) -0.252* (0.088); D(LnNGDP) 0.199 (0.146); D(LnER) -0.240** (0.129); D(LnCPI) -0.050 (0.048)
  - Adjusted R-squared:
    - D(LnM2): 0.802
    - D(LnNGDP): 0.522
    - D(LnER): 0.682
    - D(LnCPI): 0.839
  - Significance notation: * Significant at 1 percent. ** Significant at 5 percent. *** Significant at 10 percent.
  - Note: The lags of the VEC are set as 2 according to Schwartz Criterion.

### Adjustment speed and stability
- Adjustment to long-run equilibrium is relatively fast:
  - Implied it takes three to four quarters until the CPI and the exchange rate converge to their equilibrium level.
  - M2 and NGDP also take three to four quarters to converge to their equilibrium levels, with deviations adjusted by both M2 and output.
- Rolling regressions evidence:
  - Exchange rate adjustment coefficient has declined on average in recent years.
  - CPI adjustment coefficient increased dramatically from 0.3 to 0.8 (Figure 7).
  - Interpretation: recent heavy management of the exchange rate by the BNA implies deviations are mostly adjusted through the CPI.
- The short-run dynamics of nominal output transmit adjustments to CPI via GDP deflator movements (inflation inertia) and demand pressures from money growth.

### External influences, structural breaks, and robustness
- International nonfuel commodity prices significantly influence domestic prices due to high share of food in Angola’s CPI and significant imports to Luanda.
- Two dummy variables confirm structural breaks:
  - DUM_POSTCONFLICT (post-conflict era)
  - DUM_STABILIZATION (period starting with implementation of stabilization program)
- Other variables examined but insignificant in the system: inflation in trading partners, Kwanza/EUR exchange rate, real lending rate, fiscal balance (results not shown).

### Dynamic simulation and impulse responses
- Dynamic simulation (starting 1999Q1, endogenous variables solved within the model except NGDP) replicates:
  - The inflation hike in 1999–2000 and the subsequent disinflation process.
- Impulse response analysis:
  - A one percent standard deviation shock to endogenous variables shifts the CPI higher.
  - Shock to money supply (positive deviation from equilibrium) triggers upward adjustment in output and increases inflation.
  - Shock to exchange rate (depreciation) has a direct positive impact on inflation but is short-lived as exchange rate and CPI revert to new equilibrium.

### Single-equation estimation (Table 4)
- Dependent variable: D(LnCPI)
- Main findings across Models 1–5:
  - C: Model 1 0.259*; Model 2 0.210*; Model 3 0.215*; Model 4 0.157*; Model 5 0.078**.
  - DUM_POSTWAR: negative and significant across models (e.g., Model 1 -0.035*).
  - DUM_STAB: negative and significant in earlier models (e.g., Model 1 -0.171*), smaller in later models.
  - EC_ER(-1) (exchange rate error-correction term lagged 1): positive and significant in all models (e.g., Model 1 0.183*).
  - EC_M2(-4) (M2 error-correction term lagged 4): positive and significant (e.g., Model 1 0.158*), indicating M2 deviations affect inflation with a four-quarter lag.
  - REALINT(-6): negative and significant in models where included (e.g., -0.001**), indicating higher real lending rate reduces inflation with six-quarter lag.
  - D(LnP_COM(-1)): positive and significant in models including it (e.g., Model 3 0.520***).
  - D(LnKZ_EUR): positive and significant in models including it (e.g., Model 4 0.194*).
  - D(LnCPI(-2)): included in Model 5 with coefficient 0.339*.
  - Adjusted R-squared increases across models up to 0.902 in Model 5.
  - Durbin-Watson stats reported for each model (e.g., Model 1 1.875).

### Policy conclusions and recommendations
- Angola has reduced inflation substantially in recent years but has not achieved price stability; further steps needed to reach low single-digit inflation consistent with SADC convergence criteria.
- Econometric results (with caution due to short sample) highlight:
  - Inflation path over the last decade was largely affected by exchange rate movements.
  - From a medium-term perspective, greater exchange rate flexibility that facilitates gradual appreciation would moderate price growth by reducing import prices and limiting liquidity injection by the BNA.
  - If the exchange rate remains heavily managed or pegged, real exchange rate adjustment from high oil revenues will come through domestic prices, potentially creating persistent inflation momentum after oil prices stabilize.
- Exchange rate flexibility considerations:
  - Must be weighed against benefits of a stable kwanza.
  - Further nominal appreciation could enhance confidence in the kwanza, accelerate de-dollarization, and strengthen monetary policy effectiveness.
  - Nominal appreciation could anchor inflation expectations and affect inflation inertia in the long run.
  - Impact on external competitiveness likely muted because expected moderation of inflation should leave REER broadly neutral and non-oil competitiveness constraints are mainly structural.
- Role of money and liquidity management:
  - Inflation responds to deviations of M2 from its equilibrium level (excess liquidity), not money growth per se.
  - Excess liquidity adds demand pressures and contributes to inflation with a lag.
  - Importance of closely monitoring broad money growth and improving liquidity management.
  - Greater sterilization efforts by the BNA are warranted to curb rapid money growth.
  - Fiscal policy should share the burden of disinflation by ensuring public spending aligns with macroeconomic and administrative capacity.

### Annex: summary statistics and unit-root tests
- Table A1 summary statistics (observations = 44):
  - CPI: Mean 71.04; Median 43.05; Maximum 194.77; Minimum 0.22; Std. Dev. 70.11; Skewness 0.42; Kurtosis 1.56; Jarque-Bera 5.10; Probability 0.08; Sum 3125.
  - ER: Mean 44.29; Median 45.32; Maximum 89.20; Minimum 0.20; Std. Dev. 35.93; Skewness -0.07; Kurtosis 1.23; Jarque-Bera 5.78; Probability 0.06; Sum 1948.
  - GDP*: Mean 368.41; Median 132.50; Maximum 1666.35; Minimum 0.38; Std. Dev. 455.99; Skewness 1.16; Kurtosis 3.29; Jarque-Bera 10.04; Probability 0.01; Sum 16209.
  - M2*: Mean 186552; Median 62735; Maximum 855069; Minimum 94.61; Std. Dev. 243450; Skewness 1.33; Kurtosis 3.60; Jarque-Bera 13.54; Probability 0.00; Sum 8208320.
  - P_COM: Mean 93.47; Median 86.77; Maximum 143.49; Minimum 70.68; Std. Dev. 20.64; Skewness 1.24; Kurtosis 3.41; Jarque-Bera 11.55; Probability 0.00; Sum 4112.
  - G/GDP: Mean 32.61; Median 33.03; Maximum 49.42; Minimum 14.26; Std. Dev. 8.34; Skewness -0.06; Kurtosis 2.15; Jarque-Bera 1.36; Probability 0.51; Sum 1434.
  - Note: * Measured in billions of kwanzas.
- Table A2 correlation highlights:
  - CPI–ER correlation 0.92.
  - CPI–GDP correlation 0.94.
  - CPI–M2 correlation 0.92.
  - M2–P_COM correlation 0.93.
- Table A3 Phillips-Perron tests (1997–2007), unit-root null (constant and trend included):
  - LnCPI: level 1.00; first difference 0.01
  - LnER: level 0.99; first difference 0.00
  - LnNGDP: level 0.96; first difference 0.00
  - LnM2: level 0.97; first difference 0.00
  - LnP_COM: level 0.56; first difference 0.00
  - LnEUR: level 0.131; first difference 0.00
  - REALINT: level 0.102; first difference 0.00
  - Source: IMF staff estimates. The estimation used Newey-West bandwidth.

*Source: IMF staff estimates in the provided document.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp0998.pdf_