## Annex I) ## Source details **Canonical URL:** [Annex I)](https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021017-print-pdf.pdf) ## Other formats - [Markdown version](/-/media/files/publications/wp/2021/english/wpiea2021017-print-pdf.pdf.md) - [Structured JSON version](/-/media/files/publications/wp/2021/english/wpiea2021017-print-pdf.pdf.json) --- ### Empirical methodology - Objective: Estimate the dynamic response of base money, inflation, exchange rate, and broad money to a shock in central bank financing. - Identification and estimation: - Method: Local projections (LP) method of Jordà (2005) with country and time fixed effects to estimate impulse response functions (IRFs). - Motivation: LP preferred over a fully-specified VAR (Sims (1980)) because a VAR would be prohibitively high-dimensional in the panel setting; LP allows inclusion of additional control variables and reasonable degrees-of-freedom. - Baseline specification: separate regression for each horizon h; dependent variable is the macroeconomic variable at t+h; key regressor is the ratio of total central bank loans to GDP; Controls are all control variables from baseline regression equation (1). - Inference: Standard errors clustered by country and time. ### Descriptive statistics (Table 3: Sub-Saharan Africa, 2001-17; in percent) - Base Money: Number of observations 729; Mean 11.0; Standard Deviation 7.4; Minimum 0.1; Maximum 52.8 - Exchange rate: Number of observations 764; Mean 4.7; Standard Deviation 16.8; Minimum -28.1; Maximum 295.5 - Inflation: Number of observations 764; Mean 8.2; Standard Deviation 18.0; Minimum -72.7; Maximum 357.3 - Broad Money: Number of observations 764; Mean 32.4; Standard Deviation 24.2; Minimum 3.1; Maximum 150.8 - Notes: base money and broad money defined as ratios to nominal GDP; exchange rates are annual percent change in national currency per USD; inflation is annual growth rate of CPI. ### Key empirical results (impulse response evidence) - Impact on base money: - Sign: positive and contemporaneous. - Statistical significance: no statistical significance observed. - Impact on exchange rate: - Magnitude: An increase in central bank credit to the government by one percentage point of GDP is associated with a depreciation of the exchange rate by one percentage point contemporaneously. - Timing: immediate (contemporaneous). - Statistical significance: immediate and statistically significant. - Impact on inflation: - Magnitude and timing: An increase in central bank credit by one percentage point of GDP is associated with an increase in inflation by half a percentage point a year later. - Channel: evidence suggests the impact on inflation operates mostly through the exchange rate channel. - Aggregate demand channel: evidence of credit growth (resulting in an increase in aggregate demand) appears absent. - IRF presentation note: IRFs shown for a one unit innovation in the ratio of central bank loans to GDP with 68 and 90 percent confidence intervals. ### Summary findings and policy-relevant quantitative facts - Prevalence and magnitude: - Central bank financing of government deficits has been common in SSA. - It has been increasing in the past few years. - It has been large at around 2 percent of GDP on average. - It is quantitatively more important relative to other parts of the world. - Legal limits and fiscal dominance: - Majority of SSA countries now have formal limits on central bank lending to governments; limits have become both more numerous and stricter over time. - Definition used: fiscal dominance = central bank lending to government for fiscal purposes beyond legal limits. - Incidence: central bank lending above the legal limit observed in between 9 and 29 countries in 2017. - Magnitude of breaches: the amount by which central bank lending exceeds legal limits has declined over time. - Policy determinants: - Countries borrow less from central banks when they have stricter legal limits (or IMF programs that restrict lending) and more developed financial markets. - Macroeconomic relevance: - Even low amounts of fiscal dominance can have important macroeconomic effects: associated with exchange rate depreciation and higher inflation subsequently. - Policy implication: fiscal dominance is a relevant macroeconomic issue that policy makers should take seriously in normal times, not only from the perspective of hyperinflation risk. ### Robustness checks (Annex II) — selected coefficients and model diagnostics 1. Treating the fiscal deficit as endogenous (dependent variable: Central bank loans/GDP) - Fiscal deficits coefficients across Models 1–5: 0.0156; -0.0329; -0.0172; -0.00771; -0.000961 - Fiscal deficits * Legal limit coefficients (Models 2–5): 0.00662**; 0.00531**; 0.00558***; 0.00527** - Lags of central bank loans/GDP coefficients (Models 1–5): 0.758***; 0.736***; 0.732***; 0.724***; 0.724*** - Lag of real GDP growth coefficients (Models 1–5): -0.0143; -0.0130; -0.0146*; -0.0163; -0.0166* - Lag of government debt/GDP coefficients (Models 1–5): -0.00220; 0.00573**; 0.00586**; 0.00512**; 0.00421** - Observations: 667 for Model 1; 596 for Models 2–5 - Number of countries: 45 for Model 1; 41 for Models 2–5 - Significance notation: *** p<0.01, ** p<0.05, * p<0.1 2. Total claims specification (dependent variable: Central bank claims/GDP) - Fiscal deficits coefficients across Models 1–5: 0.145***; 0.0375*; 0.0398*; 0.0274; 0.0298 - Fiscal deficits * Legal limit coefficients (Models 2–5): 0.00368**; 0.00382**; 0.00440**; 0.00460** - Lags of central bank claims/GDP coefficients (Models 1–5): 0.897***; 0.827***; 0.827***; 0.824***; 0.825*** - Lag of real GDP growth coefficients (Models 1–5): -0.0270; -0.0254; -0.0235; -0.0270; -0.0252 - Lag of government debt/GDP coefficients (Models 1–5): -0.0148**; -0.0105; -0.0103; -0.0129*; -0.0125 - Observations: 667 for Model 1; 596 for Models 2–5 - Number of countries: 45 for Model 1; 41 for Models 2–5 - Robust standard errors reported in parentheses; Significance notation: *** p<0.01, ** p<0.05, * p<0.1 ### 3. Changes — Baseline specification and interpretation - Baseline specification uses stock of loans because it "corresponds to the definition of the legal limit in central bank Acts." - Alternative interpretations noted: - Some countries may interpret the law differently. - Flow more closely corresponds to annual financing needs. ### 3. Changes — Empirical results (Dependent variable: Change in Central bank loans/GDP; Models 1–5) - Observations: 629 (Model 1); 561 (Models 2–5). - Number of countries: 45 (Model 1); 41 (Models 2–5). - Robust standard errors reported in parentheses. Significance: *** p<0.01, ** p<0.05, * p<0.1. Key coefficient estimates (coefficient (standard error)): - Fiscal deficits: - Model 1: 0.626** (0.304) - Model 2: 0.0935 (0.237) - Model 3: 0.102 (0.237) - Model 4: -0.0331 (0.252) - Model 5: -0.0253 (0.254) - Fiscal deficits * Legal limit: - Model 2: 0.0415*** (0.0159) - Model 3: 0.0441*** (0.0157) - Model 4: 0.0540*** (0.0205) - Model 5: 0.0565** (0.0200) - Fiscal deficits * Domestic market development: - Model 3: -0.182 (0.350) - Model 4: -0.193 (0.332) - Fiscal deficits * IMF conditionality: - Model 4: -0.243* (0.126) - Model 5: -0.237** (0.115) - Legal limit: - Model 1: 0.676 (0.563) - Model 2: 0.613 (0.501) - Model 3: 0.559 (0.496) - Model 4: 0.477 (0.442) - Domestic market development: - Model 3: 3.946 (5.673) - Model 4: 4.912 (4.964) - IMF conditionality: - Model 4: -8.823 (6.865) - Model 5: -9.282 (6.963) - Lags of change in central bank loans/GDP: - Model 1: -0.195** (0.0942) - Model 2: 0.0203 (0.0628) - Model 3: 0.0146 (0.0677) - Model 4: 0.00520 (0.0632) - Model 5: -0.00215 (0.0689) - Lag of real GDP growth: - Model 1: -0.187 (0.164) - Model 2: -0.206 (0.208) - Model 3: -0.196 (0.215) - Model 4: -0.213 (0.221) - Model 5: -0.200 (0.227) - Lag of government deposit/GDP: - Model 1: -1.134* (0.619) - Model 2: -0.814* (0.475) - Model 3: -0.771* (0.456) - Model 4: -0.739 (0.470) - Model 5: -0.689 (0.449) - Lag of government debt/GDP: - Model 1: -0.0873* (0.0468) - Model 2: -0.0132 (0.0818) - Model 3: -0.00591 (0.0852) - Model 4: -0.0373 (0.0753) - Model 5: -0.0280 (0.0771) ### 4. Additional explanatory variables — Purpose - Tests whether alternative aspects of outside financing conditions play a role: risk and external financing. ### 4. Additional explanatory variables — Empirical results (Dependent variable: Central bank loans/GDP; Models 1–3) - Observations: 596 (Model 1); 443 (Models 2–3). - Number of countries: 41 (Model 1); 35 (Models 2–3). - Robust standard errors in parentheses. Significance: *** p<0.01, ** p<0.05, * p<0.1. Key coefficient estimates (coefficient (standard error)): - Fiscal deficits: - Model 1: 0.0370 (0.0253) - Model 2: 0.0734* (0.0394) - Model 3: 0.0752* (0.0400) - Fiscal deficits * Legal limit: - Model 1: 0.00458*** (0.00154) - Model 2: 0.00562** (0.00218) - Model 3: 0.00557** (0.00217) - Fiscal deficits * Domestic market dvpt + Eurobond access: - Model 2: -0.0358 (0.0778) - Model 3: -0.0887 (0.113) - Fiscal deficits * Sovereign risk: - Model 2: -0.00537 (0.00420) - Model 3: -0.00536 (0.00418) - Legal limit: - Model 1: -0.00132 (0.0321) - Model 2: -0.0181 (0.0470) - Model 3: -0.0172 (0.0474) - Domestic market dvpt + Eurobond access: - Model 2: 0.174 (0.640) - Model 3: 0.670 (0.891) - Sovereign risk: - Model 2: -0.0541 (0.156) - Model 3: -0.0576 (0.151) - Lags of central bank loans/GDP: - Model 1: 0.795*** (0.0382) - Model 2: 0.791*** (0.0533) - Model 3: 0.789*** (0.0533) - Lag of real GDP growth: - Model 1: -0.0221** (0.0108) - Model 2: -0.0207 (0.0145) - Model 3: -0.0204 (0.0147) - Lag of government deposit/GDP: - Model 1: -0.0805 (0.0605) - Model 2: -0.127 (0.0888) - Model 3: -0.124 (0.0885) - Lag of government debt/GDP: - Model 1: -0.00343 (0.00659) - Model 2: -0.00226 (0.00677) - Model 3: -0.00225 (0.00663) Notes: - Sovereign risk measures the risk of debt distress using the ratings from the IMF’s Debt Sustainability Analysis (=0 if rating is "Low"; =1 if rating is "Moderate"=2 if rating is “High"=3 if rating is "In debt distress"). - Eurobond access is a dummy (=1 if the country has previously issued a Eurobond, 0 otherwise). ### Annex III — Local Projection sample sizes (Table A2) - Base Money (Observations then (Countries)): - Horizon 0: 580 (40) - Horizon 1: 545 (40) - Horizon 2: 507 (40) - Horizon 3: 469 (40) - Horizon 4: 430 (40) - Horizon 5: 391 (38) - Horizon 6: 354 (37) - Exchange Rate: - Horizon 0: 596 (41) - Horizon 1: 560 (41) - Horizon 2: 521 (41) - Horizon 3: 482 (41) - Horizon 4: 442 (41) - Horizon 5: 402 (39) - Horizon 6: 364 (38) - Inflation: - Horizon 0: 596 (41) - Horizon 1: 560 (41) - Horizon 2: 521 (41) - Horizon 3: 482 (41) - Horizon 4: 442 (41) - Horizon 5: 402 (39) - Horizon 6: 364 (38) - Broad Money: - Horizon 0: 596 (41) - Horizon 1: 560 (41) - Horizon 2: 521 (41) - Horizon 3: 482 (41) - Horizon 4: 442 (41) - Horizon 5: 402 (39) - Horizon 6: 364 (38) Note: "The Table summarizes the number of observations and the number of countries in each Local Projection. The number of observations in each regression are less than in the summary statistics because a full set of data is not available for all the control variables." ### Annex III — List of 45 Sub-Saharan African countries used in the analysis - Angola - Benin - Botswana - Burkina Faso - Burundi - Cabo Verde - Cameroon - Central African Republic - Chad - Comoros - Congo, Democratic Republic of the - Congo, Rep. - Côte d'Ivoire - Equatorial Guinea - Eritrea - Ethiopia - Gabon - Gambia, The - Ghana - Guinea - Guinea-Bissau - Kenya - Lesotho - Madagascar - Malawi - Mali - Mauritius - Mozambique - Namibia - Niger - Nigeria - Rwanda - São Tomé and Principe - Senegal - Seychelles - Sierra Leone - South Africa - Sudan - Swaziland - Tanzania - Togo - Uganda - Zambia - Zimbabwe *Source: wpiea2021017-print-pdf — 3. Changes; 4. Additional explanatory variables; Annex III.* ### Annex I). ### Annex I) ### Empirical methodology - Objective: Estimate the dynamic response of key macroeconomic variables (base money, inflation, exchange rate, and broad money) to a shock in central bank financing. - Identification and estimation: - Method: Local projections (LP) method of Jordà (2005) combined with country and time fixed effects (used to estimate impulse response functions (IRFs)). - Motivation: LP preferred over a fully-specified VAR (Sims (1980)) because a VAR would be prohibitively high-dimensional in the panel setting, while LP allows inclusion of additional control variables and reasonable degrees-of-freedom. - Baseline specification: separate regression estimated for each horizon h; dependent variable is the macroeconomic variable measured at time horizon t+h; key regressor is the ratio of total central bank loans to GDP; Controls are all control variables from baseline regression equation (1). - Inference: Standard errors clustered by country and time. ### Descriptive statistics (Table 3: Sub-Saharan Africa, 2001-17; in percent) - Base Money: Number of observations 729; Mean 11.0; Standard Deviation 7.4; Minimum 0.1; Maximum 52.8 - Exchange rate: Number of observations 764; Mean 4.7; Standard Deviation 16.8; Minimum -28.1; Maximum 295.5 - Inflation: Number of observations 764; Mean 8.2; Standard Deviation 18.0; Minimum -72.7; Maximum 357.3 - Broad Money: Number of observations 764; Mean 32.4; Standard Deviation 24.2; Minimum 3.1; Maximum 150.8 - Notes: base money and broad money defined as ratios to nominal GDP; exchange rates are annual percent change in national currency per USD; inflation is annual growth rate of CPI. ### Key empirical results (impulse response evidence) - Impact on base money: - Sign: positive and contemporaneous. - Statistical significance: no statistical significance observed. - Impact on exchange rate: - Magnitude: An increase in central bank credit to the government by one percentage point of GDP is associated with a depreciation of the exchange rate by one percentage point contemporaneously. - Timing: immediate (contemporaneous). - Statistical significance: immediate and statistically significant. - Impact on inflation: - Magnitude and timing: The same increase in central bank credit (one percentage point of GDP) is associated with an increase in inflation by half a percentage point a year later. - Channel: Evidence suggests the impact on inflation operates mostly through the exchange rate channel. - Aggregate demand channel: Evidence of credit growth (resulting in an increase in aggregate demand) appears absent. - IRF presentation note: The figure shows IRFs for a one unit innovation in the ratio of central bank loans to GDP with 68 and 90 percent confidence intervals. ### Summary findings and policy-relevant quantitative facts (concluding remarks) - Prevalence and magnitude: - Central bank financing of government deficits has been common in SSA. - It has been increasing in the past few years. - It has been large at around 2 percent of GDP on average. - It is quantitatively more important relative to other parts of the world. - Legal limits and fiscal dominance: - Majority of SSA countries now have formal limits on central bank lending to governments; limits have become both more numerous and stricter over time. - Definition used: fiscal dominance = central bank lending to government for fiscal purposes beyond legal limits. - Incidence: central bank lending above the legal limit observed in between 9 and 29 countries in 2017. - Magnitude of breaches: the amount by which central bank lending exceeds legal limits has declined over time. - Policy determinants: - Countries borrow less from central banks when they have stricter legal limits (or IMF programs that restrict lending) and more developed financial markets. - Macroeconomic relevance: - Even low amounts of fiscal dominance can have important macroeconomic effects: associated with exchange rate depreciation and higher inflation subsequently. - Policy implication: fiscal dominance is a relevant macroeconomic issue that policy makers should take seriously in normal times, not only from the perspective of hyperinflation risk. ### Robustness checks (Annex II) — selected coefficients and model diagnostics 1. Treating the fiscal deficit as endogenous (dependent variable: Central bank loans/GDP) - Fiscal deficits coefficients across Models 1–5: 0.0156; -0.0329; -0.0172; -0.00771; -0.000961 - Fiscal deficits * Legal limit coefficients (Models 2–5): 0.00662**; 0.00531**; 0.00558***; 0.00527** - Lags of central bank loans/GDP coefficients (Models 1–5): 0.758***; 0.736***; 0.732***; 0.724***; 0.724*** - Lag of real GDP growth coefficients (Models 1–5): -0.0143; -0.0130; -0.0146*; -0.0163; -0.0166* - Lag of government debt/GDP coefficients (Models 1–5): -0.00220; 0.00573**; 0.00586**; 0.00512**; 0.00421** - Observations: 667 for Model 1; 596 for Models 2–5 - Number of countries: 45 for Model 1; 41 for Models 2–5 - Significance notation: *** p<0.01, ** p<0.05, * p<0.1 2. Total claims specification (dependent variable: Central bank claims/GDP) - Fiscal deficits coefficients across Models 1–5: 0.145***; 0.0375*; 0.0398*; 0.0274; 0.0298 - Fiscal deficits * Legal limit coefficients (Models 2–5): 0.00368**; 0.00382**; 0.00440**; 0.00460** - Lags of central bank claims/GDP coefficients (Models 1–5): 0.897***; 0.827***; 0.827***; 0.824***; 0.825*** - Lag of real GDP growth coefficients (Models 1–5): -0.0270; -0.0254; -0.0235; -0.0270; -0.0252 - Lag of government debt/GDP coefficients (Models 1–5): -0.0148**; -0.0105; -0.0103; -0.0129*; -0.0125 - Observations: 667 for Model 1; 596 for Models 2–5 - Number of countries: 45 for Model 1; 41 for Models 2–5 - Robust standard errors reported in parentheses; Significance notation: *** p<0.01, ** p<0.05, * p<0.1 *Source: Annex I, wpiea2021017-print-pdf (Annex I).* ### 3. Changes. We use stock of loans in our baseline specification as it corresponds to the ### 3. Changes ### Baseline specification and interpretation - The baseline specification uses stock of loans because it "corresponds to the definition of the legal limit in central bank Acts." - The text notes alternative interpretations: - Some countries may interpret the law differently. - Flow more closely corresponds to annual financing needs. ### Empirical results — Dependent variable: Change in Central bank loans/GDP (Models 1–5) - Observations: 629 (Model 1); 561 (Models 2–5). - Number of countries: 45 (Model 1); 41 (Models 2–5). - Robust standard errors reported in parentheses. Significance: *** p<0.01, ** p<0.05, * p<0.1. Key coefficient estimates (coefficient (standard error)): - Fiscal deficits: - Model 1: 0.626** (0.304) - Model 2: 0.0935 (0.237) - Model 3: 0.102 (0.237) - Model 4: -0.0331 (0.252) - Model 5: -0.0253 (0.254) - Fiscal deficits * Legal limit: - Model 2: 0.0415*** (0.0159) - Model 3: 0.0441*** (0.0157) - Model 4: 0.0540*** (0.0205) - Model 5: 0.0565** (0.0200) - Fiscal deficits * Domestic market development: - Model 3: -0.182 (0.350) - Model 4: -0.193 (0.332) - Fiscal deficits * IMF conditionality: - Model 4: -0.243* (0.126) - Model 5: -0.237** (0.115) - Legal limit: - Model 1: 0.676 (0.563) - Model 2: 0.613 (0.501) - Model 3: 0.559 (0.496) - Model 4: 0.477 (0.442) - Domestic market development: - Model 3: 3.946 (5.673) - Model 4: 4.912 (4.964) - IMF conditionality: - Model 4: -8.823 (6.865) - Model 5: -9.282 (6.963) - Lags of change in central bank loans/GDP: - Model 1: -0.195** (0.0942) - Model 2: 0.0203 (0.0628) - Model 3: 0.0146 (0.0677) - Model 4: 0.00520 (0.0632) - Model 5: -0.00215 (0.0689) - Lag of real GDP growth: - Model 1: -0.187 (0.164) - Model 2: -0.206 (0.208) - Model 3: -0.196 (0.215) - Model 4: -0.213 (0.221) - Model 5: -0.200 (0.227) - Lag of government deposit/GDP: - Model 1: -1.134* (0.619) - Model 2: -0.814* (0.475) - Model 3: -0.771* (0.456) - Model 4: -0.739 (0.470) - Model 5: -0.689 (0.449) - Lag of government debt/GDP: - Model 1: -0.0873* (0.0468) - Model 2: -0.0132 (0.0818) - Model 3: -0.00591 (0.0852) - Model 4: -0.0373 (0.0753) - Model 5: -0.0280 (0.0771) --- ### 4. Additional explanatory variables ### Purpose - Tests whether alternative aspects of outside financing conditions play a role: risk and external financing. ### Empirical results — Dependent variable: Central bank loans/GDP (Models 1–3) - Observations: 596 (Model 1); 443 (Models 2–3). - Number of countries: 41 (Model 1); 35 (Models 2–3). - Robust standard errors in parentheses. Significance: *** p<0.01, ** p<0.05, * p<0.1. Key coefficient estimates (coefficient (standard error)): - Fiscal deficits: - Model 1: 0.0370 (0.0253) - Model 2: 0.0734* (0.0394) - Model 3: 0.0752* (0.0400) - Fiscal deficits * Legal limit: - Model 1: 0.00458*** (0.00154) - Model 2: 0.00562** (0.00218) - Model 3: 0.00557** (0.00217) - Fiscal deficits * Domestic market dvpt + Eurobond access: - Model 2: -0.0358 (0.0778) - Model 3: -0.0887 (0.113) - Fiscal deficits * Sovereign risk: - Model 2: -0.00537 (0.00420) - Model 3: -0.00536 (0.00418) - Legal limit: - Model 1: -0.00132 (0.0321) - Model 2: -0.0181 (0.0470) - Model 3: -0.0172 (0.0474) - Domestic market dvpt + Eurobond access: - Model 2: 0.174 (0.640) - Model 3: 0.670 (0.891) - Sovereign risk: - Model 2: -0.0541 (0.156) - Model 3: -0.0576 (0.151) - Lags of central bank loans/GDP: - Model 1: 0.795*** (0.0382) - Model 2: 0.791*** (0.0533) - Model 3: 0.789*** (0.0533) - Lag of real GDP growth: - Model 1: -0.0221** (0.0108) - Model 2: -0.0207 (0.0145) - Model 3: -0.0204 (0.0147) - Lag of government deposit/GDP: - Model 1: -0.0805 (0.0605) - Model 2: -0.127 (0.0888) - Model 3: -0.124 (0.0885) - Lag of government debt/GDP: - Model 1: -0.00343 (0.00659) - Model 2: -0.00226 (0.00677) - Model 3: -0.00225 (0.00663) Notes: - Sovereign risk measures the risk of debt distress using the ratings from the IMF’s Debt Sustainability Analysis (=0 if rating is "Low"; =1 if rating is "Moderate"=2 if rating is “High"=3 if rating is "In debt distress"). - Eurobond access is a dummy (=1 if the country has previously issued a Eurobond, 0 otherwise). --- ### Annex III — Local Projection sample sizes (Table A2) - Table A2 reports number of observations and number of countries in each Local Projection by dependent variable and horizon. Selected entries (Observations then (Countries)): - Base Money: - Horizon 0: 580 (40) - Horizon 1: 545 (40) - Horizon 2: 507 (40) - Horizon 3: 469 (40) - Horizon 4: 430 (40) - Horizon 5: 391 (38) - Horizon 6: 354 (37) - Exchange Rate: - Horizon 0: 596 (41) - Horizon 1: 560 (41) - Horizon 2: 521 (41) - Horizon 3: 482 (41) - Horizon 4: 442 (41) - Horizon 5: 402 (39) - Horizon 6: 364 (38) - Inflation: - Horizon 0: 596 (41) - Horizon 1: 560 (41) - Horizon 2: 521 (41) - Horizon 3: 482 (41) - Horizon 4: 442 (41) - Horizon 5: 402 (39) - Horizon 6: 364 (38) - Broad Money: - Horizon 0: 596 (41) - Horizon 1: 560 (41) - Horizon 2: 521 (41) - Horizon 3: 482 (41) - Horizon 4: 442 (41) - Horizon 5: 402 (39) - Horizon 6: 364 (38) Note: "The Table summarizes the number of observations and the number of countries in each Local Projection. The number of observations in each regression are less than in the summary statistics because a full set of data is not available for all the control variables." --- ### Annex III — List of 45 Sub-Saharan African countries used in the analysis - Angola - Benin - Botswana - Burkina Faso - Burundi - Cabo Verde - Cameroon - Central African Republic - Chad - Comoros - Congo, Democratic Republic of the - Congo, Rep. - Côte d'Ivoire - Equatorial Guinea - Eritrea - Ethiopia - Gabon - Gambia, The - Ghana - Guinea - Guinea-Bissau - Kenya - Lesotho - Madagascar - Malawi - Mali - Mauritius - Mozambique - Namibia - Niger - Nigeria - Rwanda - São Tomé and Principe - Senegal - Seychelles - Sierra Leone - South Africa - Sudan - Swaziland - Tanzania - Togo - Uganda - Zambia - Zimbabwe *Source: wpiea2021017-print-pdf — 3. Changes; 4. Additional explanatory variables; Annex III.* --- _Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021017-print-pdf.pdf_