## _wp14241

## Source details

**Canonical URL:** [_wp14241](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp14241.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp14241.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp14241.pdf.json)

---

### Introduction and research question
- Research objective: Use the global vector autoregression framework developed by Pesaran, Scheuermann, and Weiner (2004) (PSW) to empirically explore channels of transmission from global financial variables into sub-Saharan Africa’s (SSA) economic activity and exports over the last 20 years.
- Key background points:
  - SSA is not significantly integrated into global financial markets, but can be affected indirectly.
  - Indirect channels highlighted: foreign direct investment disruptions, effects on domestic demand of SSA’s trading partners, and commodity-price movements.
  - Global financial developments can impact availability and cost of funding at foreign investors’ headquarters and thus projects in SSA.

### Methodology and model specification
- Estimation framework and data:
  - Uses the global vector autoregression (GVAR) model following Pesaran, Scheuermann, and Weiner (2004).
  - Sample: data for 39 countries from Q1 1991 to Q2 2011; estimates presented for six countries and seven country groups.
  - Trading-partner weights constructed from bilateral exports (IMF DOTS) averaged over 2006-08; time-invariant group weights based on 2010 PPP GDP shares.
- Variables:
  - Domestic variables per country: (1) real GDP (y_i,t), (2) short-term nominal interest rates (i_i,t), (3) real domestic private credit (len_i,t), and (4) total exports in U.S. dollars (exp_i,t).
  - Global variables common to all countries: (1) Chicago Board of Options Exchange Volatility Index (VIX) (v_t), and (2) WTI Crude Oil Spot Price (comm_t) converted to SDR.
  - Foreign variables constructed as trading-partner weighted sums: x_i,t* = sum_j w_ij x_j,t.
- Model form and estimation details:
  - Country-specific VARX(p,q) with foreign variables included; error-correction (cointegration) representation used when appropriate.
  - Estimation validity conditions (PSW): model stability (eigenvalues of companion matrix inside/on unit circle), small elements in weighting matrix, and idiosyncratic errors weakly correlated across countries.
- Data properties and tests:
  - ADF unit root tests: most country-specific foreign variables and global variables (oil prices and VIX) do not reject non-stationarity at 5 percent.
  - Estimated number of cointegrating relationships varies between 1 and 3; SSA regions generally had only one, except South Africa with two.
  - Weak exogeneity not rejected in 95 percent of 76 possible cases; three exceptions noted (interest rates for the U.K., and the VIX for the euro area and AFRCOM).

### Historical and economic context for SSA
- Trade and commodity context:
  - Global trade dropped ten percentage points as a share of global GDP to 43 percent in 2009 from 53 percent in 2008.
  - For about half of SSA countries that are net commodity exporters, commodity prices are crucial; commodity prices doubled or tripled over the ten-to-fifteen years analyzed.
  - Commodity price dynamics explain most of the rapid increase in nominal exports of commodity-rich countries.
- Financial integration and lending trends:
  - Rapid increase in cross-border lending started around 2005/06 across most regions.
  - For "SSA Developing" the ratio of cross-border loans to GDP broadly fluctuated around a stable trend.
  - South Africa: outstanding BIS-reported loans in US dollars relatively constant until about 2005, then doubled; as a ratio to GDP more volatile.
  - SSA East Advanced (AFREAST: Kenya, Mauritius, Uganda): nominal cross-border lending and ratio to GDP expanded rapidly since 2005.
- Interest rates, lending, and financial development:
  - Nominal interest rates have fallen over time with lower inflation, but around the global financial crisis nominal interest rates rose in South Africa and East Africa.
  - Bank lending rose sharply in commodity-price-boom countries, remained subdued in South Africa.
  - World Economic Forum Competitiveness indicators: South Africa ranked among the top ten in the world; East African countries in the sample are better ranked on average than commodity exporters or developing SSA economies; commodity exporters and developing SSA economies have levels of financial development below the world median.

### Generalized impulse response analysis (GIRFs) — methodological note
- GIRFs (Pesaran and Shin, 1998) are invariant to ordering, incorporate correlated shocks via historical error correlations, and do not require identification restrictions.
- GIRFs do not imply causality but reveal potential linkages and propagation patterns.
- GVAR impulse-response estimates often have wide error bands; median point estimates interpreted with caution.

### Results — investor uncertainty channel (VIX shocks)
- A one standard error positive shock to the VIX:
  - Lowers oil prices (comm) by about SDR 3.3, about US$4.7 at the exchange rate for end-2011.
  - Leads to a median drop of 11.8 percent in nominal USD exports for AFRCOM after four quarters.
  - Leads to export shortfalls in South Africa of 7.5 percent after four quarters.
  - Leads to export declines of 2 percent for AFRDEV and 1 percent for AFREAST after four quarters.
- Median GDP impacts over a four-quarter horizon (median responses):
  - ASIA: -1.0 percent
  - EURO: -1.1 percent
  - USA: -0.8 percent (implied for South Africa; South Africa GDP -0.8 percent reported)
  - South Africa (ZAF): -0.8 percent
  - AFREAST: -0.2 percent
  - AFRCOM and AFRDEV: essentially untouched (median responses near zero)
- Interpretation and transmission:
  - VIX shocks reduce commodity prices and global economic activity, which then reduce SSA exports—especially in commodity-reliant groups.
  - Impact on domestic GDP is concentrated in more financially integrated SSA economies (South Africa and AFREAST); commodity exporters show muted GDP responses but significant export revenue declines driven by commodity price drops.
  - Confidence bands often wide; many responses lie mostly in the negative spectrum, but for some regions (AFRCOM and AFRDEV) confidence intervals spread around zero, indicating statistical uncertainty.

### Results — bank deleveraging channel (credit shocks in U.S./Europe)
- A one s.e. negative shock to U.S. credit supply:
  - Largest drop in real credit activity observed in OTHER and LATAM with 0.9 and 0.7 percent respectively.
  - South Africa experiences a 0.8 percent reduction in credit.
  - For all SSA groupings except South Africa, point estimates fall within error bands around zero — indicating no significant relationship.
- Output effects:
  - Direct impact of U.S. credit contraction on foreign output limited to a few regions—Europe and South Africa among them.
  - For SSA outside South Africa, median impulse responses for output are marginally different from zero with wide error bands — inference: negligible impact.
  - Shock to European credit: estimated impacts more uncertain with wide error bands; less systematic relation picked up between European credit and local credit conditions.
- Business cycle synchronicity:
  - Europe plays a clearer role than the U.S. in transmitting output shocks (narrower error bands and predominantly negative impulse responses across many regions).
  - Within SSA, South Africa most vulnerable to European output shocks, followed by AFREAST; AFRDEV appears immune; AFRCOM shows weak negative correlation.
- Tentative explanations for muted spillovers to less developed and commodity-exporting SSA countries:
  - Financial and investment engagement in these countries tends to be driven more by foreign direct investment and long-horizon arrangements (less "hot" capital).
  - Over the sample period, many SSA countries resolved violent conflicts and benefited from debt relief, which may have dominated global influences on economic activity.

### Synthesis of main empirical findings
- Overall conclusion:
  - Sub-Saharan African countries have not been immune to global financial spillovers, but spillovers vary across countries and variables.
- Shocks to the VIX:
  - Affect the external environment via commodity prices, external trade, and trading-partner economic activity.
  - Translate into strong export effects for commodity-reliant SSA (South Africa and AFRCOM).
  - Translate into significant GDP impacts mainly in more financially developed SSA economies (South Africa and AFREAST).
- Shocks to credit conditions in the U.S. or the Euro area:
  - Do not seem to have a significant effect on most SSA groupings, except South Africa.
- Overall implication:
  - Spillovers from global investor uncertainty (VIX) onto SSA economic activity often operate through effects on trading partners and commodity prices rather than via direct credit-lending channels to SSA, with financial integration determining the strength of GDP transmission.

*Source: _wp14241 - Section V concludes.*

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

### References

### Introduction and research question
- Research objective: Use the global vector autoregression framework developed by Pesaran, Scheuermann, and Weiner (2004) (PSW) to empirically explore channels of transmission from global financial variables into sub-Saharan Africa’s (SSA) economic activity and exports over the last 20 years.
- Key background points:
  - SSA is not significantly integrated into global financial markets, but can be affected indirectly.
  - Indirect channels highlighted: foreign direct investment disruptions, effects on domestic demand of SSA’s trading partners, and commodity-price movements.
  - Global financial developments can impact availability and cost of funding at foreign investors’ headquarters and thus projects in SSA.

### Main empirical findings
- Overall conclusion:
  - Sub-Saharan African countries have not been immune to global financial spillovers, but spillovers vary across countries and variables.
- Findings about channels and variables:
  - Global uncertainty, proxied by the Chicago Board of Options Exchange Volatility Index (VIX), has a significant impact on SSA countries through the export channel.
  - There is no significant relationship between credit supply in key advanced economies and SSA’s output developments.
  - Shocks to global investor sentiment (VIX):
    - Affect exports of the region, especially in countries that rely heavily on commodities.
    - Impact on economic activity is significant only in the more financially developed economies within SSA.
  - Shocks to credit conditions in the U.S. or the euro area:
    - Do not seem to have a significant effect on sub-Saharan Africa, except in South Africa.

### Interpretation and mechanisms
- Suggested transmission mechanism:
  - Spillovers from the VIX onto economic activity in SSA may come primarily from effects on trading partners (demand/commodity-price channels) rather than direct effects on domestic economic activity in the region.
- Heterogeneity in impact:
  - Countries with more developed and globally integrated financial markets (e.g., South Africa) experienced impacts comparable to other emerging markets or developed economies.
  - Countries with significantly less developed and globally integrated financial markets were affected to a much lower degree.

### Historical context: global financial crisis evidence
- Empirical context from the global financial boom and bust of the late 2000s:
  - Global trade dropped ten percentage points as a share of global GDP to 43 percent in 2009 from 53 percent in 2008.
  - Direct financial channels during the crisis included sharp reductions in cross-border lending and sudden stops or reversals in other capital flows.
  - Indirect channels included sharp though temporary reductions in oil and other commodity prices, and broader declines in confidence affecting households’ savings and investment decisions.
  - The impact on SSA countries during the crisis varied with the degree of financial sector development.

### Methodology and paper organization
- Estimation framework:
  - Uses the global vector autoregression (GVAR) model following Pesaran, Scheuermann, and Weiner (2004).
- Empirical tools:
  - Generalized impulse response functions (GIRFs) are used to describe global spillovers from global financial variables.
- Paper structure:
  - Section II: Brief presentation of the estimation procedure in a global vector autoregression model.
  - Section III: Specification of the model to identify effects of global financial spillovers onto sub-Saharan Africa.
  - Section IV: Description of global spillovers from global financial variables using generalized impulse response functions.

*Source: Excerpt from _wp14241 - References (IMF PDF) provided in the prompt.*

### Section V concludes.

### _wp14241 - Section V concludes.

### Literature, economic context, and model — key points
- Global vector autoregressions (GVARs) model cross-country spillovers by including foreign variables in each country regression and exploiting cross-sectional relationships (PSW).
- Stylized empirical findings from prior GVAR literature:
  - Shocks that lowered U.S. equity prices pulled down global equity prices and, in most regions, also economic activity, although the effect on activity was small (PSW).
  - Shocks leading to higher German output tended to increase equity prices in Germany and most of Europe, with the exception of the UK (PSW).
  - Equity and bond markets in the U.S. and the euro area follow each other closely; U.S. monetary policy shocks have insignificant effects on euro-area output and inflation (Dées, di Mauro, Pesaran, and Smith, 2007).
  - Firm-level expected default frequencies react most to shocks to GDP, exchange rate, oil prices, and equity prices (Castren, Dées and Zaheri, 2010).
  - GVARs have been used to model global trade flows, systemic economy cycles, oil shocks, and to extend spillover analysis to sub-Saharan Africa (Bussière et al. 2009; Cashin et al. 2012; Drummond and Ramirez 2009; Samake and Yang 2011; Gurara and Ncube 2013; Canales-Kriljenko et al. 2012).

### Developments affecting sub-Saharan Africa (SSA) — empirical observations
- SSA countries are treated as small open economies that take most macroeconomically relevant prices as given.
- For about half of SSA countries that are net commodity exporters, commodity prices are crucial; commodity prices doubled or tripled over the ten-to-fifteen years analyzed, producing windfall gains and higher purchasing power abroad.
- Global financial environment dynamics:
  - Chicago Board of Options Exchange Volatility Index (VIX) fell sharply from 2003 through 2007, rose rapidly with the global financial crisis, then fell to more contained levels.
  - Low global interest rate environment since the onset of the global financial crisis expanded access for less financially developed economies to international financial markets, conditional on low uncertainty.
- Cross-border lending (BIS reporting) trends:
  - Rapid increase in cross-border lending started around 2005/06 across most regions.
  - For "SSA Developing" the ratio of cross-border loans to GDP broadly fluctuated around a stable trend.
  - Commodity-exporting SSA countries historically borrowed more abroad, but their relative exposure has dropped due to fast economic growth and alternative financing instruments (e.g., foreign-currency debt issuance).
  - South Africa: outstanding BIS-reported loans in US dollars relatively constant until about 2005, then doubled; as a ratio to GDP more volatile.
  - SSA East Advanced (AFREAST: Kenya, Mauritius, Uganda): both nominal cross-border lending and ratio to GDP expanded rapidly since 2005.
- Export dynamics:
  - Commodity price dynamics explain most of the rapid increase in nominal exports of commodity-rich countries.
  - Commodity-exporting countries have been among the fastest-growing economies in the region, followed by East African economies; many post-conflict and debt-relieved countries have increased productivity from very low levels.
- Nominal interest rates and lending:
  - Nominal interest rates have fallen over time with consistently lower inflation, but country differences reflect diverse monetary and exchange rate frameworks.
  - Around the global financial crisis, nominal interest rates rose in South Africa and East Africa, while being more subdued in commodity exporters and developing economies.
  - Bank lending rose sharply in commodity-price-boom countries, remained subdued in South Africa, and lending in South Africa has remained fairly stable since the crisis.
- Financial development indicators (World Economic Forum Competitiveness):
  - South Africa ranked among the top ten in the world.
  - East African countries in the sample are better ranked, on average, than commodity exporters or developing SSA economies.
  - Commodity exporters and developing SSA economies have levels of financial development below the world median.

### The GVAR specification used in this study
- Sample and aggregation:
  - Empirical model estimated with data for 39 countries from Q1 1991 to Q2 2011.
  - For tractability, estimates presented for six countries and seven country groups.
  - Trading-partner weights constructed from bilateral exports (IMF DOTS) averaged over 2006-08.
  - Time-invariant group weights (for country groups) based on 2010 PPP GDP shares.
- Domestic variables (per country): 
  - (1) real GDP (y_i,t), (2) short-term nominal interest rates (i_i,t), (3) real domestic private credit (len_i,t), and (4) total exports in U.S. dollars (exp_i,t).
  - Real domestic private credit deflated by CPI; except interest rates, all series are log-normalized; real GDP seasonally adjusted; some annual data interpolated to quarterly.
  - For Latin America, domestic nominal interest rate excluded due to hyperinflation in the early 1990s.
- Global variables common to all countries:
  - (1) Chicago Board of Options Exchange Volatility Index (VIX) (v_t), and (2) WTI Crude Oil Spot Price (comm_t) converted to SDR.
- Foreign variables for each country: trading-partner weighted sums of partner domestic variables (real GDP, interest rate, real domestic private credit, exports).
- Model form:
  - Country-specific VARX(p,q) with domestic endogenous variables and foreign variables constructed as weighted sums: x_i,t* = sum_j w_ij x_j,t.
  - Error terms ε_i,t assumed iid(0, Σ_i) and weakly correlated across countries.
  - Error-correction (cointegration) representation used to account for potential I(1) variables.
- Estimation validity conditions (PSW requirements):
  - Model stability after normalizing by Q^{-1} (eigenvalues of companion matrix inside/on unit circle).
  - Weights used in foreign-variable construction must be relatively small (elements at most of order 1/N and squared-sum tending to zero as N grows).
  - Idiosyncratic error terms only weakly correlated across countries.

### Data, statistical properties, and tests
- Data span: Q1 1991–Q2 2011 for 39 countries; grouped into six countries and seven country groups for reporting.
- Weighting matrix robustness: alternative weights computed for 1990–2000 used for robustness; no substantial changes to main results.
- Statistical properties:
  - ADF unit root tests indicate most country-specific foreign variables integrated of order one (unit root not rejected at 5 percent); global variables oil prices and VIX also do not reject non-stationarity at 5 percent.
  - Estimated number of cointegrating relationships varies between 1 and 3 in most regions; SSA regions had only one, except South Africa with two.
  - AFREAST cointegration relationships manually reduced from 3 to 1 after additional testing and profile analysis.
  - Rank determination used Johansen trace statistics (preferred for small samples).
- Weak exogeneity tests:
  - Weak exogeneity not rejected in 95 percent of 76 possible cases.
  - Three exceptions: interest rates for the U.K., and the VIX for the euro area and AFRCOM (commodity exporters), suggesting possible feedback from these to global variables.

### Results — generalized impulse response analysis (GIRFs)
- Methodological note:
  - GIRFs (Pesaran and Shin, 1998) invariant to ordering and do not require identification restrictions; incorporate correlated shocks via historical error correlations; GIRFs do not imply causality but reveal potential linkages and propagation patterns.
  - GVAR impulse-response estimates often have wide error bands; median point estimates interpreted with caution.

Investor uncertainty channel (VIX shocks)
- A one standard error positive shock to the VIX:
  - Lowers oil prices (comm) by about SDR 3.3, about US$4.7 at the exchange rate for end-2011.
  - Leads to a median drop of 11.8 percent in nominal USD exports for AFRCOM after four quarters.
  - Leads to export shortfalls in South Africa of 7.5 percent after four quarters.
  - Leads to export declines of 2 percent for AFRDEV and 1 percent for AFREAST after four quarters.
  - Median GDP impacts over a four-quarter horizon:
    - ASIA: -1.0 percent
    - EURO: -1.1 percent
    - USA: -0.8 percent (implied for South Africa; South Africa GDP -0.8 percent reported)
    - South Africa (ZAF): -0.8 percent
    - AFREAST: -0.2 percent
    - AFRCOM and AFRDEV: essentially untouched (median responses near zero)
- Interpretation and transmission:
  - VIX shocks reduce commodity prices and global economic activity, which then reduce SSA exports—especially in commodity-reliant groups.
  - Impact on domestic GDP is concentrated in more financially integrated SSA economies (South Africa and AFREAST); commodity exporters show muted GDP responses but significant export revenue declines driven by commodity price drops.
  - Confidence bands often wide; many responses lie mostly in the negative spectrum, but for some regions (AFRCOM and AFRDEV) confidence intervals spread around zero, indicating statistical uncertainty.

Bank deleveraging channel (negative shock to credit in U.S./Europe)
- A one s.e. negative shock to U.S. credit supply:
  - Largest drop in real credit activity observed in OTHER and LATAM with 0.9 and 0.7 percent respectively.
  - South Africa experiences a 0.8 percent reduction in credit.
  - For all SSA groupings except South Africa, point estimates fall within error bands around zero — indicating no significant relationship.
- Output effects:
  - Direct impact of U.S. credit contraction on foreign output limited to a few regions—Europe and South Africa among them.
  - For SSA outside South Africa, median impulse responses for output are marginally different from zero with wide error bands — inference: negligible impact.
  - Shock to European credit: estimated impacts more uncertain with wide error bands; less systematic relation picked up between European credit and local credit conditions possibly because deleveraging may occur via reduced cross-border lending while maintaining local portfolios.
- Business cycle synchronicity:
  - Europe plays a clearer role than the U.S. in transmitting output shocks (narrower error bands and predominantly negative impulse responses across many regions).
  - Within SSA, South Africa most vulnerable to European output shocks, followed by AFREAST; AFRDEV appears immune; AFRCOM shows weak negative correlation.
- Tentative explanations for generally muted spillovers to less developed and commodity-exporting SSA countries:
  - Financial and investment engagement in these countries tends to be driven more by foreign direct investment and long-horizon arrangements (less "hot" capital).
  - Over the sample period, many SSA countries resolved violent conflicts and benefited from debt relief, which may have dominated global influences on economic activity.

### Conclusion — synthesized findings
- Sub-Saharan Africa is not immune to global financial spillovers, but spillovers vary across countries and economic variables.
- Shocks to the VIX:
  - Affect the external environment via commodity prices, external trade, and trading-partner economic activity.
  - Translate into strong export effects for commodity-reliant SSA (South Africa and AFRCOM).
  - Translate into significant GDP impacts mainly in more financially developed SSA economies (South Africa and AFREAST).
- Shocks to credit conditions in the U.S. or the Euro area:
  - Do not seem to have a significant effect on most SSA groupings, except South Africa.
- Overall implication:
  - Spillovers from global investor uncertainty (VIX) onto SSA economic activity often operate through effects on trading partners and commodity prices rather than via direct credit-lending channels to SSA, with financial integration determining the strength of GDP transmission.

*Source: _wp14241 - Section V concludes.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp14241.pdf_