## Regional Growth Spillovers in Sub‑Saharan Africa (wpiea2019160)

## Source details

**Canonical URL:** [Regional Growth Spillovers in Sub‑Saharan Africa (wpiea2019160)](https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019160.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2019/wpiea2019160.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2019/wpiea2019160.pdf.json)

---

### Introduction and approach
- Research question: How has intraregional integration evolved in sub-Saharan Africa in recent decades and how has this affected growth spillovers on the continent?
- Focus: trade linkages as the primary channel for regional integration and spillovers.
- Empirical strategy:
  - Local projections model (Jordà (2005)) with identification from Furceri, Jalles, and Zdzienicka (2016).
  - Panel fixed-effect model based on degree of interconnectedness (building on Arora and Vamvakidies (2005a), Dabla-Norris, Espinoza, and Jahan (2015)).
- Conceptual framework: domestic growth determined by domestic shocks, global shocks, and foreign/region shocks transmitted via channels such as trade and fiscal policy.

### Key empirical findings on spillovers
- Main magnitudes:
  - Abstract summary: Positive idiosyncratic shocks to regional trading partners’ growth significantly increase growth in the average sub-Saharan African country by 0.2-0.5 percent.
  - Local projections: A 1 percent shock in average growth of trading partners → about 0.5 percent increase in output of the average sub-Saharan African country four years after the shock.
  - Panel fixed effects: A 1 percentage point increase in the export-weighted growth rate of intra-regional partners → about 0.35 percent increase in average sub-Saharan African country growth (long-run interpretation varies by specification).
- Robustness:
  - Results robust to accounting for extra-regional factors including global growth and demand from large trading partners such as China, and to alternative model specifications including exclusion of top-decile economies.
- Comparative magnitude:
  - Long-run growth spillovers in sub-Saharan Africa are of similar magnitude to those in other emerging and developing economies.

### Stylized facts on intraregional trade integration and concentration
- Time trends and shares:
  - Regional trade as a share of total trade: 6 percent in 1980, reaching 20 percent in 2016.
  - Ten sub-Saharan countries represent 65 percent of total regional demand for intra-regional exports.
  - South Africa imports 15 percent of total regional exports.
- Per-country and sub-region measures:
  - Small, very open economies in SACU and ECOWAS (examples: Swaziland, Lesotho, Togo, The Gambia): intra-regional exports > 65 percent of total global exports.
  - Some SADC countries (Zimbabwe, Botswana, Lesotho, Namibia): intra-regional exports ≈ 20 percent of GDP.
  - Some WAEMU countries (Côte d’Ivoire, Guinea, Senegal): intra-regional exports ≈ 10 percent of GDP.
  - Swaziland’s exports to GDP ratio = 62 percent in 2016 (not depicted in figure).
  - Swaziland’s imports to GDP ratio = 81 percent in 2016 (not depicted in figure).
- Drivers of rising regional trade:
  - Global factors: two-fold increase in the relative price of commodity exports over 1995–2013 and a rise of two and a half times in volume of exported commodities.
  - Regional factors: strengthened macroeconomic policies, improved political and economic institutions, abating conflicts, improved business environment, and reductions in tariffs via regional trade agreements.

### Regional trade shares and potential spillovers (country examples & exposure)
- Exposure magnitudes:
  - South African imports from Swaziland, Lesotho, Zimbabwe and Mozambique represent between 4 and 11 percent of these economies’ GDP.
  - Zimbabwe’s total demand for goods from Zambia, Malawi and Botswana constitutes between 1 and 4 percent of these countries’ GDP.
  - Nigeria, Mali, Ghana, and Burkina Faso’s imports amount to more than 1 percent of GDP of their sub-regional trading partners.
- Implication: reductions in import demand in an importing country can have significant GDP consequences for its trading partners.

### Data and weighting scheme
- Sample and period:
  - Annual bilateral data from the IMF’s Direction of Trade Statistics, covering 45 countries in sub-Saharan Africa over 1980-2016.
- Country-pair specific weight:
  - W_ijt = (X_ijt + X_ij,t-1 + X_ij,t-2) / (∑_j X_ijt + ∑_j X_ij,t-1 + ∑_j X_ij,t-2)
  - Export shares over 2-year periods used to smooth outlier years; analogous index constructed for trading partners outside the region as a control.

### Empirical methodology (identification and controls)
- Local projections:
  - Idiosyncratic growth shocks ε_ijt from g_ijt = α_t + γ_i + ε_ijt.
  - Export-weighted average growth shock ε̄_i,t = ∑_{j≠i} W_{ij,t-1} ε_jt.
  - Horizon K set to eight years (k=0,..,8); two lags of GDP growth included.
  - Robustness: exclude top decile economies (Angola, Democratic Republic of the Congo, Kenya, Nigeria, South Africa) and examine South Africa shock alone.
  - Note: method isolates country-specific shocks from global shocks; immediate domestic transmission may bias results downward (results are a lower bound).
- Panel fixed effects:
  - Dependent variable g_ijt; main regressor TTG_SSA_it = ∑_{j≠i} W_{ij,t-1} g_jt.
  - Controls X_it include investment to GDP, change in CPI, trade openness, percent change in bilateral exchange rates vis a vis the US dollar, occurrence of conflicts and war, percent change in the Fed funds rate.
  - Include average growth in sub-Saharan Africa, average global growth, and L lags of dependent variable with L=2.
  - Country fixed effects γ_i; standard errors clustered at country level.
  - Robustness: exclude top decile economies (Angola, Ghana, Kenya, Nigeria, South Africa), panel Arellano-Bond GMM estimator, add year fixed effects, panel Arellano-Bond GMM using five-year averages (include initial level of GDP per capita).

### Local projections results (impulse responses)
- Identification note: idiosyncratic shocks ε_ijt are highly correlated with GDP growth in most countries; some divergence during 2008-2010.
- Impulse response to a 1 percent export-weighted real GDP growth shock:
  - On average: increases real GDP by about 0.2 percent on impact.
  - Maximum impact after about four years: increase of about 0.5 percent in real GDP of the average sub-Saharan Africa country.
- Excluding largest decile economies (Angola, Ghana, Kenya, Nigeria, South Africa) → results highly robust.
- South Africa-specific shock:
  - A 1 percent positive idiosyncratic shock in South Africa → larger and much more persistent impact on average output in other sub-Saharan Africa countries compared with the weighted average of all SSA shocks.
  - South African shock dies out after approximately two decades when horizon extended.

### Panel fixed effects results (selected estimates and interpretations)
- Baseline panel fixed effects (Table 1):
  - A 1 percentage point increase in export-weighted growth rate of intraregional partners → estimated coefficient 0.15 on growth of the average SSA country (column 2).
  - Long-run interpretation in baseline: increase of about 0.2 percent in the long run (long run calculated as 0.15/(1-(0.283-0.0572))).
  - A 1 percentage point increase in growth rate of trading partners outside region → associated with increase of 0.35 percent in growth rate of the average SSA country in the long run (not statistically significant).
- Panel GMM and robustness:
  - Panel GMM estimate of regional spillovers: 0.21 percent; estimated growth of 0.25 percentage points higher for the average SSA country for every 1 percentage point increase in export-weighted growth of intraregional partners (column 3).
  - Excluding largest economies → little change in coefficient (column 4).
  - Introducing year effects controls for global shocks (column 5).
  - Panel GMM with 5-year averages: coefficient slightly lower than baseline.

### Econometric coefficient snapshots (preserving reported values)
- Selected coefficients from Table 1 (standard errors in parentheses):
  - SSA trading partners' growth: 0.132* (0.0743); 0.155* (0.0788); 0.207** (0.0812); 0.187** (0.0865); 0.144* (0.0864); 0.0773** (0.0344)
  - Non-SSA trading partners' growth: 0.151 (0.165); 0.271 (0.182); 0.401* (0.214); 0.370 (0.298); 0.296 (0.230); 0.00433 (0.0440)
  - Real GDP growth (t-1): 0.329*** (0.0726); 0.283*** (0.0798); 0.262*** (0.0276); 0.252*** (0.0760); 0.248*** (0.0277); 0.400*** (0.0553)
  - Conflict: -3.815*** (1.070); -3.876*** (1.230); -3.089** (1.332); -3.652*** (1.225); -3.876 (2.774)
  - Trade openness (t-1): 0.0533* (0.0299); 0.0854*** (0.0158); 0.0912 (0.0621); 0.111*** (0.0172); 0.0923* (0.0559)
  - Share of regional exports in total (t-1): 3.459* (1.946); 1.565 (1.821); 2.256 (2.767); 4.866** (1.951); -0.149 (2.053)
  - Investment, percent of GDP (t-1): 0.0479 (0.0384); 0.0610*** (0.0194); 0.0584 (0.0561); 0.0630*** (0.0193); 0.0874** (0.0397)
  - SSA average growth: 0.287** (0.120); 0.0758 (0.217); 0.0345 (0.121); -0.00671 (0.266)
  - World average growth: 0.105 (0.149); 0.200 (0.232); 0.0392 (0.250); 0.0467 (0.284)
- Selected coefficients from Table 2 (regional comparisons):
  - Regional trading partners' growth:
    - SSA: 0.207** (0.0841)
    - Latin America: 0.126* (0.0743)
    - MENA: -0.0335 (0.0768)
    - Asia: 0.256** (0.129)
    - Europe and CIS: 0.233 (0.184)
  - Non-regional trading partners' growth:
    - SSA: 0.401 (0.305)
    - Latin America: -0.0770 (0.177)
    - MENA: -0.109 (0.221)
    - Asia: 0.435*** (0.165)
    - Europe and CIS: 0.477*** (0.140)
  - Additional notable coefficients (Table 2):
    - Average regional real growth: Latin America 0.271** (0.106); MENA 0.509** (0.225)
    - Trade openness (t-1): Latin America 0.0183*** (0.00704); MENA 0.0346* (0.0204)
    - Investment share of GDP (t-1): Latin America -0.173*** (0.0350); MENA 0.235** (0.109)

### International comparisons and interpretation
- Emerging and developing Asia and Latin America: a one percent increase in real GDP of trading partners → peak increase in output of about 0.5-0.7 percent after three years.
- Emerging and developing Europe and the CIS: a 1 percent idiosyncratic shock → increase in average output of about 2 percent at maximum.
- MENA region: spillovers negligible and significantly smaller than those of SSA.
- Comparative interpretation:
  - SSA intra-regional elasticity slightly lower than some regions but higher than Latin America and substantially higher than MENA.
  - Geographic compactness and sectoral composition (e.g., oil importance) help explain cross-region differences.

### Determinants of bilateral trade (gravity estimates)
- Main results:
  - Exports increase significantly with origin and destination GDP per capita and population.
  - Sharing common language, ethnicity, and colonial heritage → positive and significant.
  - Contiguity and common currency → advantages to exporters; contiguity larger advantage when exporting to SSA.
  - Distance significant negative determinant of bilateral trade (log): -1.60*** (cols 1–2), -1.63*** (col 3).
- Key coefficients from Table A.1 (log bilateral trade flows):
  - Contiguous countries: 1.57*** (col 1), 1.53*** (col 2), 0.29** (col 3)
  - Distance (in log): -1.60*** (col 1), -1.60*** (col 2), -1.63*** (col 3)
  - Common language: 0.46*** (col 1), 0.45*** (col 2), 0.54*** (col 3)
  - Common ethno: 0.24* (col 1), 0.30** (col 2), 0.20*** (col 3)
  - Belonged to common colony: 1.44*** (col 1), 1.37*** (col 2), 1.08*** (col 3)
  - Common currency: 1.22*** (col 1), 1.28*** (col 2), 0.37 (col 3)
  - Origin GDP p.c.: 0.49*** (col 3)
  - Destination GDP p.c.: 0.48*** (col 3)
  - Origin population: 0.70** (col 3)
  - Destination population: 2.41*** (col 3)
  - Origin/destination FX rate: 0.00 (col 3)
- Sub-Saharan interaction terms (col 3):
  - Contiguous * SSA indicator: 0.95*** (col 3)
  - Common currency * SSA indicator: 0.77** (col 3)
  - Other SSA interactions reported (distance * SSA: -0.01; common language * SSA: 0.16; common ethno * SSA: 0.16; common colony * SSA: 0.10; common religion * SSA: -0.08).

### Sub-regional integration, time trends, and counterfactuals
- Time-interacted results (column 4):
  - EAC members: integration increased trade by an additional 4 percent per year on average.
  - SADC members: integration increased trade by an additional 2 percent per year on average.
- Counterfactual computation:
  - Average annual growth in regional trade would have been around 9 percent instead of 11 percent without sub-regional integration.
  - This implies counterfactual trade levels would be half as low as observed in 2015.
- Distance trend:
  - Interaction between distance and time indicates distance has increasingly become a barrier over time in SSA — infrastructure facilitating trade between economic unions has lagged relative to development of infrastructure within unions.

### Annex material (impulse responses and dynamics)
- Annex I: Evolution of real GDP growth and idiosyncratic shocks for SSA countries (figures).
- Annex III impulse responses:
  - Figure AIII.1: Effect of a one percent weighted average shock to output in SSA countries (weighted by bilateral trade).
  - Figure AIII.2: Same as AIII.1 but excluding largest economies (Angola, Ghana, Kenya, Nigeria, South Africa).
  - Figure AIII.3: Effect of a one percent shock in South Africa to output in SSA countries.
  - Impulse response axes express impacts in Percent over Year horizons (scales such as 0, .2, .4, .6, .8, 1 Percent up to multi-year horizons).

### Policy implications and recommendations
- Incorporate intraregional spillovers into medium-term planning and surveillance.
- Support further continent-wide integration to capture growth spillovers (e.g., deepen trade networks).
- Design growth-friendly policies that:
  - Capture benefits of intraregional spillovers,
  - Limit exposure to transmission risks from regional partners,
  - Build precautionary cushions and monitor cross-border links,
  - Reduce tariff and non-tariff barriers and improve ease of doing business,
  - Favor infrastructure development to facilitate trade between countries and between sub-regions.
- Emphasize regional surveillance and spillover analysis alongside traditional bilateral surveillance.
- Structural transformation and diversification to reduce overreliance on few products and partners; African Continental Free Trade Agreement cited as supportive for trading more products with more diverse partners.

_Source: WP/19/160, "Regional Growth Spillovers in Sub‑Saharan Africa" (Francisco Arizala, Matthieu Bellon, Margaux MacDonald)._

### Section 1

### Regional Growth Spillovers in Sub-Saharan Africa

### Introduction: research question and approach
- Research question: How has intraregional integration evolved in sub-Saharan Africa in recent decades and how has this affected growth spillovers on the continent?
- Focus: trade linkages as the primary channel for regional integration and spillovers.
- Empirical strategy:
  - Local projections model (Jordà (2005)) with identification from Furceri, Jalles, and Zdzienicka (2016).
  - Panel fixed-effect model based on degree of interconnectedness (building on Arora and Vamvakidies (2005a), Dabla-Norris, Espinoza, and Jahan (2015)).
- Conceptual framework: domestic growth determined by domestic shocks, global shocks, and foreign/region shocks transmitted via channels such as trade and fiscal policy.

### Key empirical findings on spillovers
- Abstract summary: Positive idiosyncratic shocks to regional trading partners’ growth significantly increase growth in the average sub-Saharan African country by 0.2-0.5 percent.
- Local projections estimates:
  - A 1 percent shock in average growth of trading partners is associated with an increase of about 0.5 percent in output of the average sub-Saharan African country four years after the shock.
- Panel fixed effects estimates:
  - A 1 percentage point increase in the export-weighted growth rate of intra-regional partners is associated with about 0.35 percent increase in the average sub-Saharan African country growth.
- Robustness:
  - Results are robust to accounting for extra-regional factors including global growth and demand from large trading partners such as China, and to alternative model specifications.
- Comparative magnitude:
  - Long-run growth spillovers in sub-Saharan Africa are of similar magnitude to those in other emerging and developing economies.

### Stylized facts on intraregional trade integration
- Trend in intra-regional trade:
  - Regional trade as a share of total trade: 6 percent in 1980, reaching 20 percent in 2016.
- Relative and per-country measures:
  - The increase in regional trade was faster for small countries, reflected in faster growth in the simple average level of trade integration.
  - When measured as a share of total exports in 2016, sub-Saharan Africa exhibits the highest share of intra-regional trade integration among emerging and developing economies (relative ranking versus Middle-East and North Africa, and emerging and developing Asia).
  - Relative to economy size, sub-Saharan Africa is in the middle of the pack among regions.
- Country- and sub-region-specific integration:
  - In small and very open economies in SACU and ECOWAS (examples: Swaziland, Lesotho, Togo, The Gambia), intra-regional exports represent more than 65 percent of these countries’ total global exports.
  - In some SADC countries (Zimbabwe, Botswana, Lesotho, Namibia), intra-regional exports represent about 20 percent of GDP.
  - In some WAEMU countries (Côte d’Ivoire, Guinea, Senegal), intra-regional exports are close to 10 percent of GDP.
  - Swaziland’s exports to GDP ratio is equal to 62 percent in 2016 (not depicted in figure).
  - Swaziland’s imports to GDP ratio is equal to 81 percent in 2016 (not depicted in figure).
- Concentration of regional demand:
  - Ten sub-Saharan countries represent 65 percent of total regional demand for intra-regional exports.
  - South Africa imports 15 percent of total regional exports.
- Drivers of rising regional trade:
  - Global factors: two-fold increase in the relative price of commodity exports over 1995–2013 and a rise of two and a half times in volume of exported commodities (Allard and others, 2016).
  - Regional factors: strengthened macroeconomic policies, improved political and economic institutions, abating conflicts, improved business environment, and reductions in tariffs via regional trade agreements.

### Policy implications and recommendations
- Policymakers should factor intraregional spillovers into medium-term planning and surveillance.
- Need to support further continent-wide integration to capture growth spillovers.
- Design growth-friendly policies that:
  - Capture benefits of intraregional spillovers.
  - Limit exposure to transmission risks from regional partners.
- Emphasize regional surveillance and spillover analysis alongside traditional bilateral surveillance.

*Source: WP/19/160, "Regional Growth Spillovers in Sub-Saharan Africa" (Francisco Arizala, Matthieu Bellon, Margaux MacDonald).*

### Section 2

### Section 2

### Regional trade shares and potential spillovers
- South African imports from Swaziland, Lesotho, Zimbabwe and Mozambique represent between 4 and 11 percent of these economies’ GDP.
- Zimbabwe’s total demand for goods from Zambia, Malawi and Botswana constitutes between 1 and 4 percent of these countries’ GDP.
- Nigeria, Mali, Ghana, and Burkina Faso’s imports amount to more than 1 percent of GDP of their sub-regional trading partners.
- Implication: any reduction in import demand caused by an economic downturn in an importing country could have significant consequences for GDP growth in its trading partners.

### Data and weighting scheme
- Data: annual bilateral data from the IMF’s Direction of Trade Statistics, covering 45 countries in sub-Saharan Africa over the period 1980-2016.
- Country-pair specific weight (equation (1)):
  - W_ijt = (X_ijt + X_ij,t-1 + X_ij,t-2) / (∑_j X_ijt + ∑_j X_ij,t-1 + ∑_j X_ij,t-2)
  - Weight interprets country i’s exports to country j over years t, t-1, t-2 as a share of total regional exports of country i over the same years.
  - Export shares over 2-year periods used to smooth outlier years; analogous index constructed for trading partners outside the region as a control.

### Empirical methodology
- Local projections (Jordà (2005) panel adaptation; equations (2) and (3)):
  - Idiosyncratic growth shocks defined as residual ε_ijt from g_ijt = α_t + γ_i + ε_ijt (equation (2)).
  - Use export-weighted average growth shock ε̄_i,t = ∑_{j≠i} W_{ij,t-1} ε_jt in equation (3).
  - Horizon K set to eight years (k=0,..,8).
  - Two lags of GDP growth included; robustness checks: exclude top decile economies (Angola, Democratic Republic of the Congo, Kenya, Nigeria, South Africa) and examine South Africa shock alone.
  - Note: method isolates country-specific shocks from global shocks, with the cost that immediate domestic transmission may bias results downward (results are a lower bound).
- Panel fixed effects (equation (4)):
  - Dependent variable g_ijt; main regressor TTG_SSA_it = ∑_{j≠i} W_{ij,t-1} g_jt (equation (5)).
  - Controls X_it include investment to GDP, change in CPI, trade openness, percent change in bilateral exchange rates vis a vis the US dollar, occurrence of conflicts and war, percent change in the Fed funds rate.
  - Include average growth in sub-Saharan Africa, average global growth, and L lags of dependent variable with L=2.
  - Country fixed effects γ_i; standard errors clustered at country level.
  - Robustness checks: exclude top decile economies (Angola, Ghana, Kenya, Nigeria, South Africa), panel Arellano-Bond GMM estimator, add year fixed effects, panel Arellano-Bond GMM using five-year averages (include initial level of GDP per capita).

### Local projections results
- Identification: idiosyncratic shocks ε_ijt are highly correlated with GDP growth in most countries; some divergence during 2008-2010 (global financial crisis).
- Impulse response to a 1 percent export-weighted real GDP growth shock:
  - On average, increases real GDP by about 0.2 percent on impact.
  - Maximum impact occurs after about four years and is equal to an increase of about 0.5 percent in real GDP of the average sub-Saharan Africa country.
- Excluding largest decile economies (Angola, Ghana, Kenya, Nigeria, South Africa) yields results highly robust to baseline.
- South Africa-specific shock:
  - A one percent positive idiosyncratic shock in South Africa generates a larger and much more persistent impact on average output in other sub-Saharan Africa countries compared with the weighted average of all sub-Saharan Africa shocks.
  - Extending the horizon shows that the South African shock dies out after approximately two decades.

### Panel fixed effects results
- Baseline panel fixed effects (Table 1):
  - A 1 percentage point increase in the export-weighted growth rate of intraregional partners is associated with an estimated coefficient of 0.15 on the growth of the average sub-Saharan African country (column 2).
  - Interpretation: implies an increase of about 0.2 percent in the long run (long run calculated as 0.15/(1-(0.283-0.0572))).
  - A 1 percentage point increase in the growth rate of trading partners outside of the region is associated with an increase of 0.35 percent in the growth rate of the average sub-Saharan Africa country in the long run (not statistically significant).
- Panel GMM and robustness:
  - Panel GMM estimate of regional spillovers: 0.21 percent, or estimated growth of 0.25 percentage points higher for the average sub-Saharan Africa country for every 1 percentage point increase in export-weighted growth of its intraregional partners (column 3).
  - Excluding largest economies yields little change in coefficient (column 4).
  - Introducing year effects controls for global shocks (column 5).
  - Panel GMM with 5-year averages: coefficient estimate slightly lower than baseline, suggesting regional trading partners remain important but slightly less so when annual variation is smoothed.

### International comparisons
- Methods re-estimated for Middle-East and North Africa, Latin America, Emerging and Developing Asia, and emerging and developing Europe using both local projections and panel GMM (for panel fixed effects).
- Impulse response functions for the four non-sub-Saharan Africa regions (local projections) are shown in Figure 12.
- Finding: intraregional spillovers in other emerging and developing regions are of approximately the same magnitude as those in sub-Saharan Africa.

*Source: wpiea2019160 - Section 2 (IMF staff calculations and cited equations and figures from the source).*

### Section 3

### wpiea2019160 - Section 3

### Growth spillovers across emerging and developing regions
- Emerging and developing Asia and Latin America: "the estimated impact of a one percent increase in real GDP is at its largest after three years and equal to an increase in output of about 0.5-0.7 percent."
- Emerging and developing Europe and the CIS: "a 1 percent idiosyncratic shock to export-weighted GDP associated with an increase in average output of about 2 percent at its maximum."
- MENA region: "spillovers in the MENA region are negligible and significantly smaller than those of sub-Saharan Africa."

### Key findings for sub-Saharan Africa (SSA)
- A 1 percent shock in the weighted average trading partners' growth is associated with an increase of about 0.2-0.5 percent in output of other sub-Saharan African countries.
- South Africa is likely the driving force behind much of these spillovers (largest economy in the region with the largest share of regional trade).
- Trade integration trend: "trade integration in sub-Saharan Africa is now at comparable levels with developing and emerging market economies in other regions." SSA is more integrated in terms of growth spillovers than MENA, and roughly in line with Latin America, emerging and developing Europe and CIS, and Asia.

### Econometric results — selected coefficient estimates (preserving reported values)
- From Table 1 (GDP Growth Elasticities of Growth of Trading Partners), selected estimates (standard errors in parentheses):
  - SSA trading partners' growth: 0.132* (0.0743); 0.155* (0.0788); 0.207** (0.0812); 0.187** (0.0865); 0.144* (0.0864); 0.0773** (0.0344)
  - Non-SSA trading partners' growth: 0.151 (0.165); 0.271 (0.182); 0.401* (0.214); 0.370 (0.298); 0.296 (0.230); 0.00433 (0.0440)
  - Real GDP growth (t-1): 0.329*** (0.0726); 0.283*** (0.0798); 0.262*** (0.0276); 0.252*** (0.0760); 0.248*** (0.0277); 0.400*** (0.0553)
  - Conflict: -3.815*** (1.070); -3.876*** (1.230); -3.089** (1.332); -3.652*** (1.225); -3.876 (2.774)
  - Trade openness (t-1): 0.0533* (0.0299); 0.0854*** (0.0158); 0.0912 (0.0621); 0.111*** (0.0172); 0.0923* (0.0559)
  - Share of regional exports in total (t-1): 3.459* (1.946); 1.565 (1.821); 2.256 (2.767); 4.866** (1.951); -0.149 (2.053)
  - Investment, percent of GDP (t-1): 0.0479 (0.0384); 0.0610*** (0.0194); 0.0584 (0.0561); 0.0630*** (0.0193); 0.0874** (0.0397)
  - SSA average growth: 0.287** (0.120); 0.0758 (0.217); 0.0345 (0.121); -0.00671 (0.266)
  - World average growth: 0.105 (0.149); 0.200 (0.232); 0.0392 (0.250); 0.0467 (0.284)

- From Table 2 (Sub-Saharan Africa and Other Developing Countries), regional trading partners' growth estimates:
  - SSA: 0.207** (0.0841)
  - Latin America: 0.126* (0.0743)
  - MENA: -0.0335 (0.0768)
  - Asia: 0.256** (0.129)
  - Europe and CIS: 0.233 (0.184)
- Non-regional trading partners' growth (from Table 2):
  - SSA: 0.401 (0.305)
  - Latin America: -0.0770 (0.177)
  - MENA: -0.109 (0.221)
  - Asia: 0.435*** (0.165)
  - Europe and CIS: 0.477*** (0.140)
- Additional notable coefficients (Table 2):
  - Average regional real growth: Latin America 0.271** (0.106); MENA 0.509** (0.225)
  - Trade openness (t-1): Latin America 0.0183*** (0.00704); MENA 0.0346* (0.0204)
  - Investment share of GDP (t-1): Latin America -0.173*** (0.0350); MENA 0.235** (0.109)

### Interpretation and comparative magnitudes
- Relative elasticities: emerging and developing Asia estimated increase of about 0.28 percentage points on average following a 1 percent increase in export-weighted growth of trading partners; Eastern Europe and CIS estimated increase of 0.33 percentage points for the same shock. SSA countries have slightly lower intra-regional elasticity of growth, but higher than Latin America (long run growth spillovers estimated to be 0.16) and substantially higher than MENA (long run impact about -0.03, not statistically significant).
- Geographic and sectoral explanations: lower spillovers in Latin America and MENA may be due to geography (less compact regions) and the high importance for trade of developments in oil markets.

### Policy recommendations and implications
- Structural transformation strategies to promote diversification and guard against spillovers from overreliance on too few products and partners.
- Deeper trade networks, as promoted by the African Continental Free Trade Agreement, can help countries trade more products with more diverse partners.
- Governments should:
  - build precautionary cushions,
  - monitor and regulate cross-border links to set the stage for growth and stability,
  - reduce tariff and non-tariff barriers,
  - improve the ease of doing business,
  - favor infrastructure development to facilitate trade between countries and between sub-regions.

### Conclusion highlights
- Evidence fills a gap: "growth of countries in sub-Saharan Africa has an effect on each country’s domestic growth rate, and the more so the greater is bilateral trade between countries."
- Spillovers are "statistically significant and economically large, and robust to various different modelling specifications."
- The rate of growth of identified key source and destination countries will have implications for their major trading partners.

*Source: wpiea2019160 - Section 3 (IMF staff calculations, tables and text in the provided PDF content)*

### Section 4

### wpiea2019160 - Section 4

### Determinants of bilateral trade (gravity estimates)
- Main specification shows exports increase significantly with:
  - Origin and destination GDP per capita and population.
  - Sharing a common language, ethnicity, and colonial heritage.
- Bilateral exchange rates do not have a significant effect on bilateral trade flows.
- Contiguity and common currency confer advantages to exporters; contiguity is a greater advantage when exporting to sub-Saharan destinations.
- Distance is a significant negative determinant of bilateral trade (in log) with coefficient reported as -1.60*** (columns 1 and 2) and -1.63*** (column 3).
- Key coefficient estimates from Table A.1 (dependent variable: logarithm of bilateral trade flows):
  - Contiguous countries: 1.57*** (col 1), 1.53*** (col 2), 0.29** (col 3)
  - Distance (in log): -1.60*** (col 1), -1.60*** (col 2), -1.63*** (col 3)
  - Common language: 0.46*** (col 1), 0.45*** (col 2), 0.54*** (col 3)
  - Common ethno: 0.24* (col 1), 0.30** (col 2), 0.20*** (col 3)
  - Belonged to common colony: 1.44*** (col 1), 1.37*** (col 2), 1.08*** (col 3)
  - Common religion: 0.14 (col 1), 0.14 (col 2), 0.24*** (col 3)
  - Common currency: 1.22*** (col 1), 1.28*** (col 2), 0.37 (col 3)
  - Origin GDP p.c.: 0.49*** (col 3)
  - Destination GDP p.c.: 0.48*** (col 3)
  - Origin population: 0.70** (col 3)
  - Destination population: 2.41*** (col 3)
  - Origin/destination FX rate: 0.00 (col 3)
- Interaction terms for sub-Saharan Africa:
  - Contiguous countries * SSA indicator: 0.95*** (col 3)
  - Distance * SSA indicator: -0.01 (col 3)
  - Common language * SSA indicator: 0.16 (col 3)
  - Common ethno * SSA indicator: 0.16 (col 3)
  - Common colony * SSA indicator: 0.10 (col 3)
  - Common religion * SSA indicator: -0.08 (col 3)
  - Common currency * SSA indicator: 0.77** (col 3)
- Model fit and sample:
  - Observations: 92,132 (col 1), 95,711 (col 2), 556,476 (col 3), 95,108 (col 4)
  - R-squared: 0.52 (col 1), 0.57 (col 2), 0.73 (col 3), 0.77 (col 4)
  - Year FE: YES (col 1), NO (cols 2–4)
  - Country FE: YES (col 1), NO (cols 2–4)
  - Country-time FE: NO (cols 1–3), YES (col 4)
  - Country-pair FE: NO (cols 1–3), YES (col 4)
  - Clustered standard errors in parentheses (Destination country). Significance notation: *** p<0.01, ** p<0.05, * p<0.1.

### Sub-regional integration and time trends
- Specification introduces interaction of sub-region membership with time trends to assess whether trade integration occurred faster within WAEMU, CEMAC, EAC, SADC, and SACU (memberships as of 2016).
- Results (column 4) show statistically significant evidence that integration among:
  - EAC members increased trade by an additional 4 percent per year on average.
  - SADC members increased trade by an additional 2 percent per year on average.
- Counterfactual computation using these estimates:
  - Average annual growth in regional trade would have been around 9 percent instead of 11 percent without sub-regional integration.
  - This translates into trade levels that would be half as low as the observed in 2015.
- The interaction between distance and time indicates that in sub-Saharan Africa distance has increasingly become a barrier over time — infrastructure facilitating trade between economic unions has lagged relative to the development of infrastructure within unions.

### Annex I–III: Evolution of GDP growth, idiosyncratic shocks, and impulse responses
- Annex I figures:
  - Show evolution of real GDP growth (dashed lines) and idiosyncratic shocks to GDP growth (red lines) for sub-Saharan Africa countries; figures for remaining countries are available in Appendix 2.
- Annex III impulse response analysis:
  - Figure AIII.1: Effect of a one percent weighted average shock to output in sub-Saharan Africa countries (weighted by bilateral trade). Note: Shock occurs at t=0. The weighted average shock is calculated for each country pair i and j as the share of country i's imports from j in country i's total imports from sub-Saharan Africa, as defined in equation (1).
  - Figure AIII.2: Effect of a one percent weighted average shock to output in sub-Saharan Africa countries (weighted by bilateral trade, excluding largest economies). Note: Shock occurs at t=0. Angola, Ghana, Kenya, Nigeria, and South Africa excluded from sample.
  - Figure AIII.3: Effect of a one percent shock in South Africa to output in sub-Saharan Africa countries. Note: Shock occurs at t=0. The shock is derived for South Africa in equation (2).
- Impulse response axes shown in figures express impacts in Percent over Year horizons (figures indicate scales such as 0, .2, .4, .6, .8, 1 Percent and multi-year horizons up to 5 years).

*Source: IMF DOTS and WEO databases and authors’ calculations.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019160.pdf_