Regional Spillovers in Sub-Saharan Africa

 Regional Spillovers in Sub-Saharan Africa Exploring Different Channels Francisco Arizala, Matthieu Bellon, Margaux MacDonald, Montfort Mlachila, and Mustafa Yenice I N T E R N A T I O N A L M O N E T A R Y F U N D ©International Monetary Fund. Not for Redistribution Copyright ©2018 International Monetary Fund Cataloging-in-Publication Data Names: Arizala, Francisco. | Bellon, Matthieu. | MacDonald, Margaux. | Mlachila, Montfort. | Yenice, Mustafa. | International Monetary Fund. Title: Regional spillovers in Sub-Saharan Africa : exploring different channels / Francisco Arizala, Matthieu Bellon, Margaux MacDonald, Montfort Miachila, and Mustafa Yenice. Other titles: Spillover notes (International Monetary Fund) ; 12. Description: Washington, DC : International Monetary Fund, 2018. | Spillover notes / International Monetary Fund ; 12 | August 2018. | Includes bibliographical references. Identifiers: ISBN 9781484367148 Subjects: LCSH: Africa, Sub-Saharan—Economic integration. | Fiscal policy—Africa, Sub-Saharan. | Financial institutions—Africa, Sub-Saharan. | Africa, Sub-Saharan—Economic conditions. Classification: LCC HJ1445 .R433 2018 Publication orders may be placed online, by fax, or through the mail: International Monetary Fund, Publication Services P.O. Box 92780, Washington, DC 20090, U.S.A. Tel. (202) 623-7430 Fax: (202) 623-7201 E-mail: [email protected] www.imfbookstore.org www.elibrary.imf.org ©International Monetary Fund. Not for Redistribution CONTENTS Abstract 1 Introduction and Summary 1 Regional Trade Links Gaining Strength 4 Intraregional Trade Links are Growing 4 Global Comparisons 5 Banking Interdependence Becoming More Subregional 16 Presence of Foreign Banks Headquartered in Sub-Saharan Africa 16 African Bank Behavior and Trends in Correspondent Banking Relationships 20 The Dominant Role of South Africa Sovereign Spread Spillovers 21 Cross-Country Co-movement and Its Drivers 21 Estimating the Impact of South Africa 22 The Changing Pattern of Remittance Flows 22 The Foreign Direct Investment Channel—South Africa Rules the Roost 29 The Fiscal Channel—The Role of Unintended Consequences 34 Importance of Customs Unions 34 Unintended Spillovers from Nigeria’s Fuel Pricing Policies to Its Neighbors 36 The Rising Socioeconomic Impact of Forced Migration 38 Concluding Remarks 40 Appendix 41 References 43 Boxes Box 1. Gravity Equation Estimation for 2010–16 Trade Flows 10 Box 2. GDP Growth Elasticities to the Growth of Trading Partners 13 Box 3. Sovereign Yield Spread Spillovers 23 Box 4. Gravity Equation Estimation for 2010–15 Remittance Flows 27 Box 5. Spillover Effects from Countries Sending Remittances 31 Box 6. South African Investment in Sub-Saharan Africa 33 Box 7. SACU Revenue-Sharing Formula 35 Figures Figure 1. Sub-Saharan Africa: Trade, Banking, and Remittance Channels 2 Figure 2. Sub-Saharan Africa: Intraregional Trade, 1980–2016 4 Figure 3. Sub-Saharan Africa: Intraregional Trade, Percent of GDP, 1980–2016 4 International Monetary Fund | August 2018 iii ©International Monetary Fund. Not for Redistribution SPILLOVER NOTES Figure 4. Intraregional Exports by Region, 2016 5 Figure 5. Sub-Saharan Africa: Intraregional Exports, 2016 6 Figure 6. Sub-Saharan Africa: Intraregional Imports, 2016 6 Figure 7. Sub-Saharan Africa: Share of Intraregional Imports, 2016 7 Figure 8. Major Intraregional Trade Links 7 Figure 9. Countries with Substantial Trading Relationships from the Perspective of the Exporter 8 Figure 10. Major Intraregional Trade Relationships 8 Figure 11. Sub-Saharan Africa: Subregional Trade, 2016 9 Figure 12. Sub-Saharan Africa:Intraregional Trade by Subregions, 2016 9 Figure 13. Sub-Saharan Africa: Intraregional Exports, by Country Groups 12 Figure 14. Sub-Saharan Africa: Exports to the Rest of the World, by Country Groups 12 Figure 15. PAB and Subregional Bank Presence in SSA,2007–16 16 Figure 16. Sub-Saharan African PABs and Subregional Banks: Home and Host Countries 17 Figure 17. Systemic Foreign Owned PAB andSubregional Banks, 2007–16 18 Figure 18. Sub-Saharan Africa: Real, Financial, and Cross-Border Links 19 Figure 19. Bank Deposits in PABs and Subregional Banks,2007–16 19 Figure 20. Sub-Saharan Africa: Sovereign Spread Correlations, 2012–16 21 Figure 21. Sub-Saharan Africa: External Flows and Remittances 25 Figure 22. Remittance Inflows in Emerging and DevelopingCountries 2010–15 25 Figure 23. Sub-Saharan Africa: Remittance Outflows and Inflows, 2010–15 26 Figure 24. Sub-Saharan Africa: Major Remittance Corridors, 2010–15 27 Figure 25. Percentage Cost of Sending US$200 across Region and over Time 29 Figure 26. Total Average Cost by Remittance Sending Provider 30 Figure 27. Selected Sub-Saharan African Countries:Intraregional Foreign Direct Investment Stock Positions 30 Figure 28. SACU Revenues and Selected Macroeconomic Indicators 36 Figure 29. Differentials between Nigerian Gasoline Prices and Those of Benin and Togo 37 Figure 30. Sub-Saharan Africa: Within Migration, Refugees 38 Figure 31. Sub-Saharan Africa and Selected Countries:Internally Displaced Persons 38 Figure 32. Selected Sub-Saharan African Countries: Civil Unrest and Terroris 39 Box Figure 3.1. South African News and Sovereign Spreads 24 Box Figure 6.1. South Africa: Outward FDIin Sub-Saharan Africa 34 Tables Table 1. Loan-to-Deposit Ratios, Largest Sub-Saharan African Countries, 2015 20 Table 2. Fuel Prices Correlation in Togo, Benin and Nigeria, 2008–17 37 Table 1.1. Determinants of Trade Flows 11 Table 2.1. GDP Growth Elasticities to the Growth of Trading Partners 14 Table 2.2. Sub-Saharan Africa and Other Developing Countries: GDP Growth Elasticities to the Growth of Trading Partners 15 Table 3.1. Impact of Global, Regional, and Domestic Factors on Sovereign Spreads, 2012–16 24 Table 4.1. Determinants of Average Remittances Flows 28 Table 5.1. Spillover Effects from Countries Sending Remittances 32 Table 6.1. South Africa: Major Multinationals 34 Appendix Table 1. Sub-Saharan Africa: List of Country Abbreviations 41 Appendix Table 2. Sub-Saharan Africa: Member Countries of Groupings 41 Appendix Table 3. Sub-Saharan Africa: Member Countries of Regional Groupings 42 iv International Monetary Fund | August 2018 ©International Monetary Fund. Not for Redistribution REGIONAL SPILLOVERS IN SUB-SAHARAN AFRICA: EXPLORING DIFFERENT CHANNELS Abstract After close to two decades of strong economic activity, overall growth in sub-Saharan Africa decelerated mark- edly in 2015–16 as the largest economies experienced negative or flat growth. Regional growth started recovering in 2017, but the question remains of how trends in the economies stuck in low gear will spill over to the countries that have maintained robust growth. This note illuminates the discussion by identifying growth spillover channels. The focus is on trade, banking, financial, remittance, investment, fiscal, and security channels, which are the most prominent and most likely to transmit growth trends across borders. In addition to bringing together findings from a broad array of existing research, the note identifies countries that are the most likely sources of regional spillovers and those that are most likely to be impacted, and provides estimates for the size of these channels. It finds that intraregional trade and remittance flows are an important channel for growth spillovers, while banking channels are less important but will remain a risk going forward. Finally, the note documents other important spillover channels through financial markets contagion, revenue-sharing arrangements in fiscal unions, commodity-pricing poli- cies, corporate investment, and forced migration. The main takeaway is that the level of interdependence among sub-Saharan countries is higher than is generally assumed. Consequently, there is a need for additional emphasis on regional surveillance and spillover analysis, along with traditional bilateral surveillance. Introduction and Summary failed to show large spillovers across the region. It also After close to two decades of strong economic activ- assesses a variety of spillover channels, whose impor- ity, overall growth in sub-Saharan Africa decelerated tance varies across countries, reflecting the heterogene- markedly in 2015–16, to its lowest level in more than ity of economic structures in sub-Saharan Africa. 20 years at 1.4 percent. However, this average masked The note covers well-known channels as well as substantial heterogeneity across the region. While the those that have received less attention in the literature. largest economies (Nigeria and South Africa) experi- It does this by using several methodologies and draw- enced negative or flat growth, a third of the countries ing on studies that have identified transmission chan- in the region continued to grow at 5 percent or more nels and mechanisms, and by breaking new ground during the period. As growth has begun to recover empirically in areas in which the existing literature is since 2017 on the back of a more favorable external silent. In all cases, the note systematically updates pre- environment, the question remains: to what extent do viously known stylized facts and empirical estimates. growth trends in the largest economies spill over to the It identifies countries that are likely to be the origin of rest of sub-Saharan Africa? In particular, will trends in economic spillovers and countries more likely to be at the economies stuck in low gear spill over to countries the receiving end, provides new empirical estimates of that have maintained robust growth? the size of various spillover channels, and documents This note focuses on identifying the channels and new channels of transmission not previously identified impacts of intraregional spillovers in sub-Saharan in the literature (Figure 1). Africa. The note goes beyond existing studies that Regional trade links are steadily gaining strength. rely on aggregate growth data and that have typically Countries that absorb most intraregional exports and hence have the highest potential to generate regional spillovers are identified, as well as countries that are The authors would like to thank Anne-Marie Gulde-Wolf for her overall guidance of the project, as well as the Spillover Task Force, more exposed to spillovers from other countries in the Céline Allard, Jesus Gonzalez-Garcia, Miguel Pereira Mendes, and region. The note also discusses the following findings: several IMF colleagues for very helpful comments and suggestions. •• Intraregional trade has steadily increased in intensity Natasha Minges provided excellent editorial assistance, Joe Procopio edited the manuscript, and Heidi Grauel provided layout. over time. It represented 6 percent of total exports International Monetary Fund | August 2018 1 ©International Monetary Fund. Not for Redistribution  SPILLOVER NOTES Figure 1. Sub-Saharan Africa: Trade, Banking, and Remittance •• Some countries are highly exposed to intraregional Channels demand. Exports to the top 10 destinations More likely to generate spillovers More likely to suffer from spillovers represent between 5 percent and 10 percent of More likely to generate spillovers and suffer from spillovers source-country GDP. 1. Trade Channel •• Subregional trade accounts for most of sub-Saharan African regional trade. Southern Africa Customs Union (SACU) subregional trade alone represents half of total sub-Saharan Africa intraregional trade. Moreover, in the cases of the Southern Africa Development Community (SADC) and the SACU, subregional trade represents more than 80 percent of their member countries’ intraregional trade. •• Econometric analysis shows that bilateral trade is more likely to be hindered by distance and socio- cultural differences in sub-Saharan Africa than in 2. Banking Channel the rest of the world, which explains why most regional trade occurs within subregions. Moreover, econometric estimates suggest that about half of the growth in regional trade over 1980–2016 stems from subregional trade integration, in particu- lar within the East African Community (EAC) and the SADC. •• The growth of regional trading partners has a significant effect on individual countries’ growth, even after controlling for variables capturing co-movement at the global and regional levels. 3. Remittances Channel Econometric estimates suggest that a 5 percentage point increase in the export-weighted growth rate of intraregional partners is associated with about a 0.5 percent increase in the average sub-Saharan African country’s growth. Beyond pan-African banks, subregional banks are emerging, and South Africa plays a significant role in determining sovereign debt spreads in sub-Saha- ran Africa frontier markets. The rising importance of subregional banks is highlighted, and countries with Sources: IMF, Direction of Trade Statistics database, World Economic Outlook database; the highest exposure to pan-African banks (PABs) and and IMF staff calculations. subregional banks are identified. The note also finds: •• Strong intraregional banking links represent a growing channel of potential spillovers. In terms (1 percent of GDP) in 1980 before taking off in of market participation, the share of PABs and the early 1990s and eventually reaching 20 percent subregional banking groups in sub-Saharan African (4 percent of GDP) in 2016. financial system is increasing, following a global •• The key players in the total demand for intra- trend of banking regionalization. regional exports (that is, the countries with the •• A few sub-Saharan African countries are identified potential to generate the largest regional spillovers) as being the primary countries of origin of banking are highly concentrated. Ten countries account for spillovers. While there is considerable overlap among 65 percent of total regional demand. the countries that are home to PABs and subregional 2 International Monetary Fund | August 2018 ©International Monetary Fund. Not for Redistribution  Regional Spillovers in Sub-Saharan Africa: Exploring Different Channels banks, the spillover recipient countries are more •• Firms from South Africa that are searching for diver- widely dispersed. sification opportunities in relatively faster growing •• Growth in countries that are home to PABs and regional African markets dominate the landscape. subregional banks is associated with private sector •• The lion’s share of investment is in services, trade, credit growth in the countries in which they oper- and the financial sector. ate, which itself has reinforcing effects on growth in those countries. Significant unintended spillovers exist from fiscal •• There are links from changes in the spread on South policies in the largest countries. The note covers how African sovereign debt to other sub-Saharan African these fiscal channels develop via large fluctuations of frontier markets. The global and emerging market tax receipts in customs unions and via negative exter- financial cycles have a major impact on all issuing nalities arising from different fuel pricing policies in sub-Saharan countries, including South Africa, but neighboring countries. there is also evidence of specific spillovers from •• The SACU revenue-sharing formula ties member South Africa. countries’ fiscal revenues to economic developments in South Africa. While providing certainty for Remittances from within sub-Saharan Africa are current revenue, the formula leads to high levels of becoming relatively more important. The key players volatility over the medium term. As a result, it com- in terms of regional remittance outflows (with the plicates fiscal management in the smallest countries potential to generate the largest regional spillovers) (Lesotho and Swaziland). are identified, as well as the countries most exposed •• Subsidized fuel in Nigeria leads to widespread to remittance spillovers. In further analysis, the smuggling and to the erosion of the tax base in note also finds: Benin and Togo. For instance, for Benin, only about •• Growth in regional remittances has outpaced the 15 percent of the fuel consumed is purchased on the growth of other external sources of financing such as formal (taxed) market. aid, foreign direct investment (FDI), and remit- tances from the rest of the world. The socioeconomic costs of forced migration are •• Remittance flows are rather concentrated in a few rising. The note analyzes the socioeconomic impact corridors, and in some countries regional remittance of forced migration owing to conflict and secu- inflows represent a substantial share of income. In rity concerns. particular, Côte d’Ivoire and Ghana are important •• The share of forced migration across countries in sources for West Africa, and South Africa is an sub-Saharan Africa declined significantly through important source for Southern and East Africa. most of the 1990s and 2000s, but the pace of •• Recent reductions in the cost to send money across decline has slowed or partially reversed. borders are associated with the development of •• Terrorism and civil conflict in the Sahel, the Lake mobile money and explain part of the observed Chad area, the eastern Democratic Republic of the increase in regional remittances. The cost of sending Congo, Somalia, and South Sudan have been the remittances in sub-Saharan Africa are the highest in main drivers of involuntary migration. the world, implying that there is room for further •• The main negative spillovers of forced migration cost reductions and increases in regional remittances. studied here are reduced economic activity, human- •• Growth in countries that send remittances is found itarian damage, and the fiscal costs of hosting to be significantly associated with growth in receiv- displaced persons and fighting terrorism. ing countries. A 5 percent increase in the growth of remittance partners is estimated to raise growth by The key takeaways from this note are that regional 0.5 percent, although this is partially outweighed by integration in its various forms is more extensive than trading partners’ growth spillovers. generally assumed and that subregional integration is moving faster than overall integration. Spillover chan- South Africa is the dominant source of regional nels for the largest economies are diverse. For South FDI. This note analyzes the corporate sector and dis- Africa, spillovers are via trade, banking, and remittance cusses the following: channels, while Nigerian spillovers are mainly through banking, fuel pricing policy, and trade (in the case of International Monetary Fund | August 2018 3 ©International Monetary Fund. Not for Redistribution  SPILLOVER NOTES Figure 2. Sub-Saharan Africa: Intraregional Trade, 1980–2016 25 Regional exports to total exports Regional imports to total imports 20 Total regional trade to total trade 15 Percent 10 5 0 1980 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 2000 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 Sources: IMF, Direction of Trade Statistics database, World Economic Outlook database; and IMF staff calculations. Note: Calculation based on weighted averages. its neighboring countries). Overall, the potential for Intraregional Trade Links Are Growing spillovers—both positive and negative—is higher than The increase in intraregional trade integration in previously found in the literature. sub-Saharan Africa over the past 35 years has amplified the potential for intraregional spillovers. Sub-Saharan African economies have become more open to and Regional Trade Links Gaining Strength integrated with global and intraregional trade, a Sub-Saharan African economies have become more trend that is marked by the increase in regional trade open to and integrated with global and intraregional as a share of total trade. Regional trade represented trade. This trend has been marked by an increase in 6 percent of total exports in 1980 before taking off in regional trade as a share of total trade, which has ampli- the early 1990s and eventually reaching 20 percent in fied the potential for regional spillovers. Fluctuations in 2016 (Figure 2). These increases in regional trade have the economic activity of intraregional trading partners been significant relative to the size of sub-Saharan Afri- affect the growth of individual countries in the region. can economies; they have been faster for small coun- tries in the region, as reflected by the faster growth in the simple average level of trade integration (Figure 3). Figure 3. Sub-Saharan Africa: Intraregional Trade, Percent of GDP, 1980–2016 8 Regional exports to GDP, weighted average 7 Regional exports to GDP, simple average 6 Percent of GDP 5 4 3 2 1 0 1980 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 2000 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 Sources: IMF, Direction of Trade Statistics database, World Economic Outlook database; and IMF staff calculations. 4 International Monetary Fund | August 2018 ©International Monetary Fund. Not for Redistribution  Regional Spillovers in Sub-Saharan Africa: Exploring Different Channels Figure 4. Intraregional Exports by Region, 2016 1. Intraregional exports 2. Intraregional exports (Percent of total exports) (Percent of GDP) 40 Exports to total exports (weighted) 10 Percent of GDP (weighted) Exports to total exports (simple) 9 Percent of GDP (simple) 35 8 30 Percent of total exports Percent of total exports 7 25 6 20 5 15 4 3 10 2 5 1 0 0 SSA MENA Emedev Asia CIS Latin America MENA CIS SSA Emedev Asia Latin America Sources: IMF, Direction of Trade Statistics database, World Economic Outlook database; and IMF staff calculations. Note: CIS = Commonwealth of Independent States; EMEDEV = Emerging and Developing economies; MENA = Middle East and North Africa. SSA = sub-Saharan Africa. The increase in global and regional trade integra- regional spillovers—is broadly in line with levels in tion is the result of global developments but also of other regions (Figure 4.1 and 4.2). Measured as a share a strengthening of institutional and macroeconomic of total exports, sub-Saharan Africa exhibits the highest conditions in the region. Part of the increase in global share of intraregional trade integration among emerg- trade can be explained by the twofold increase in the ing and developing regions, followed by the Middle relative price of commodity exports over the period East and North Africa and emerging and developing 1995–2013. Another part is explained by volumes of Asia. Relative to the size of the economy, sub-Saharan exported commodities, which increased by two and Africa is in the middle of the pack. a half times over the same period (Allard and others 2016). In addition to these supporting conditions, the Countries Most Exposed to Regional Trade Spillovers sub-Saharan Africa region has experienced a substantial Although significant heterogeneity exists among strengthening of macroeconomic policies and political sub-Saharan African countries in terms of intraregional and economic institutions over the past 20 years, along trade integration, many of them are highly connected with an abatement of internal and external conflicts to other countries in the region (Figures 5 and 6). This and, in some cases, countries exiting fragility. These is particularly the case within subregions. For example, elements all contributed to improving the business in several countries in the SACU (such as Swaziland environment and supported the deepening of regional and Lesotho) and in other small and very open econ- trade (IMF 2015a). Furthermore, the establishment omies (such as Togo and The Gambia) intraregional of regional trade agreements in various subregions has exports represent more than 65 percent of total exports contributed to regional and bilateral reductions in (IMF 2012).1 Also, export shares can be large relative tariffs, which have further supported trade integration to the size of the economy. This is the case for Zimba- (ODI 2010). Compared with advanced economies, bwe and certain SACU members (Botswana, Lesotho, intraregional trade remains low and the business Namibia), where intraregional exports represent about environment remains challenging, but the direction has 20 percent of GDP, and some Western Africa Eco- been favorable over time. nomic and Monetary Union (WAEMU) countries (Côte d’Ivoire, Guinea, Senegal), where they constitute close to 10 percent of GDP. Global Comparisons The average level of regional trade integration in sub-Saharan Africa—and thus the potential for 1The SACU comprises Botswana, Lesotho, Namibia, South Africa, and Swaziland. International Monetary Fund | August 2018 5 ©International Monetary Fund. Not for Redistribution  SPILLOVER NOTES Figure 5. Sub-Saharan Africa: Intraregional Exports, 2016 100 25 Exports to total exports 90 Exports to GDP (right axis) 80 20 Percent of total exports 70 Percent of GDP 60 15 50 40 10 30 20 5 10 0 0 SWZ ZWE GMB TGO LSO BDI SEN MWI NAM SLE KEN ZAF CAF BEN ZMB COD MOZ GIN MUS GNB MLI COG NER CMR LBR MDG ETH AGO STP GNQ COM GAB SYC ERI TCD CIV CPV RWA UGA TZA BWA BFA NGA GHA Sources: IMF, Direction of Trade Statistics database, World Economic Outlook database; and IMF staff calculations. Note: Swaziland’s exports-to-GDP ratio is not shown; it was equal to 62 percent in 2016. See page 41 for country abbreviations tables. Countries More Likely to Generate Regional chains and re-exports. In most of the top 10 importing Trade Spillovers countries, intraregional imports primarily consist of Regional demand for intraregional exports is con- manufactures, fuels, and food. An economic decelera- centrated in a very few countries. Ten sub-Saharan tion in any of these countries thus has the potential to countries represent 65 percent of total regional weaken demand for intraregional exports and may be a demand for intraregional exports, with South Africa, source of wider negative spillovers. Botswana, and Namibia accounting for the largest Imports absorbed by the top 10 regional importers shares of total regional demand, and South Africa of sub-Saharan African intraregional trade represent alone importing 15 percent of total intraregional significant shares of the economies of the exporting exports (Figure 7). These SACU member countries countries, setting the stage for potentially large spill- trade significantly among themselves, with exports con- overs. The strength of this channel is commensurate centrated mainly in manufactures, food, and machin- with the importance of the importer’s demand relative ery, often in the context of regional or global value to the size of the exporting country’s economy. For Figure 6. Sub-Saharan Africa: Intraregional Imports, 2016 90 60 Imports to total imports 80 Imports to GDP (right axis) 50 70 Percent of total imports 60 40 Percent of GDP 50 30 40 30 20 20 10 10 0 0 LSO ZWE SWZ NAM ZMB COD MLI MOZ MWI STP BDI TCD CMR GNB GMB SYC BEN COM NER SEN MUS SLE TGO ZAF COG MDG GIN KEN GNQ CAF AGO GAB ERI ETH SSD CIV CPV BWA RWA BFA UGA TZA GHA NGA Sources: IMF, Direction of Trade Statistics database, World Economic Outlook database; and IMF staff calculations. Note: Swaziland’s imports-to-GDP ratio is not shown; it was equal to 81 percent in 2016. See page 41 for country abbreviations table. 6 International Monetary Fund | August 2018 ©International Monetary Fund. Not for Redistribution  Regional Spillovers in Sub-Saharan Africa: Exploring Different Channels Figure 7. Sub-Saharan Africa: Share of Intraregional Imports, 2016 16 Other resource-intensive countries 14 Non-resource-intensive countries Oil exporters Percent of total regional imports 12 10 8 6 4 2 0 ZAF BWA NAM ZMB ZWE SWZ CIV COD MOZ COG Sources: IMF, Direction of Trade Statistics database; and IMF staff calculations. Note: See page 41 for country abbreviations table. Figure 8. Major Intraregional Trade Links (Major importers perspective) CPV COG COD GAB CMR GHA CIV GIN BWA GMB BFA GNB BEN GNQ BDI KEN AGO LBR ZWE LSO ZMB MDG ZAF MOZ UGA MUS TZA MWI TGO NAM SYC NGA RWA SEN SLE SWZ Sources: IMF, Direction of Trade Statistics database, World Economic Outlook database; and IMF staff calculations. Note: The thickness of the arrows refers to the size of bilateral exports in percent of GDP of the exporting country. The top 10 destinations are in red; the other countries in blue. See page 41 for country abbreviations table. instance, South African imports from Swaziland, Leso- sub-Saharan African intraregional exports, import more tho, Zimbabwe, and Mozambique represent between than 1 percent of GDP of their subregional trading 4 percent and 11 percent of these economies’ GDP. partners.2 Indeed, for four countries in the region, Similarly, Zimbabwe’s total demand for goods from their exports to Nigeria and Mali represent more Zambia, Malawi, and Botswana constitutes between than 1 percent of their economy, making them also 1 percent and 4 percent of these countries’ GDP potential sources of intraregional spillovers (Figures 9 (Figure 8). and 10).3 Other countries import non-negligible shares of their neighbors’ GDP and can be a substantial source 2Informal trade (not captured by official statistics) between Nige- of spillovers at the subregional level. This is the case for ria and its neighbors may be economically important for the smaller Nigeria, Mali, Ghana, and Burkina Faso, which, even countries (Balami, Ogboru, and Talba 2011). 3Similarly, for 10 countries in the region, their exports to South though they do not import substantial shares of total Africa represent more than 1 percent of their GDP. International Monetary Fund | August 2018 7 ©International Monetary Fund. Not for Redistribution SPILLOVER NOTES Figure 9. Countries with Substantial Trading Relationships from the Perspective of the Exporter 10 9 8 7 Number of countries 6 5 4 3 2 1 0 ZAF NGA MLI ZWE BWA GHA BFA COG ZMB SWZ Sources: IMF, Direction of Trade Statistics database; and IMF staff calculations. Note: In the above chart, number of countries for which the country is a destination market representing more than 1 percent of GDP of the sender’s economy. See page 41 for country abbreviations table. Figure 10. Major Intraregional Trade Relationships (Exporter’s perspective) GHA GAB CPV COG GIN COD GMB CMR GNB CIV GNQ BWA KEN BFA LBR BEN LSO AGO MDG ZWE MLI ZMB MOZ ZAF MUS UGA MWI TZA NAM TGO NER SYC NGA SWZ RWA SEN SLE STP Sources: IMF, Direction of Trade Statistics database, World Economic Outlook database; and IMF staff calculations. Note: Blue and red dots indicate exporting and importing countries, respectively. The size of the red dots is proportional to the number of countries for which the country is a destination market representing more than 1 percent of GDP of the exporter’s economy. The thickness of the arrows refers to the size of bilateral exports in percent of GDP of the exporting country. See page 41 for country abbreviations table. The Role of Subregional Trade Concentration rest of the region (Figure 11).4 In the Central African Trade spillovers are more likely to take place within Economic and Monetary Community (CEMAC) and subregions, as most regional trade is concentrated at the WAEMU regions, the concentration is less pro- the subregional level. In the SADC, the SACU, and the EAC, subregional trade accounts for more than 4The SADC comprises Angola, Botswana, Democratic Republic of 70 percent of their total trade with sub-Saharan Africa, the Congo, Lesotho, Madagascar, Malawi, Mauritius, Mozambique, Namibia, Seychelles, South Africa, Swaziland, Tanzania, Zambia, and reflecting the fact that member countries are mostly Zimbabwe. All SACU member countries are also part of the SADC. integrated within themselves rather than with the The EAC includes Burundi, Kenya, Uganda, Rwanda, South Sudan, and Tanzania. 8 International Monetary Fund | August 2018 ©International Monetary Fund. Not for Redistribution  Regional Spillovers in Sub-Saharan Africa: Exploring Different Channels Figure 11. Sub-Saharan Africa: Subregional Trade, 2016 Figure 12. Sub-Saharan Africa: (Percent of the subregional exports to sub-Saharan Africa) Intraregional Trade by Subregions, 2016 (Percent of total sub-Saharan Africa trade) 100 Export 80 90 Import Export 70 Import 80 70 60 60 50 Percent 50 Percent 40 40 30 30 20 20 10 10 0 SADC SACU EAC CEMAC WAEMU 0 SADC SACU WAEMU EAC CEMAC Sources: IMF, Direction of Trade Statistics database; and IMF staff calculations. Note: CEMAC = Central African Economic and Monetary Community; EAC = East Sources: IMF, Direction of Trade Statistics database; and IMF staff calculations. African Community; SACU = Southern African Customs Union; SADC = Southern Note: CEMAC = Central African Economic and Monetary Community; EAC = East African Development Community; WAEMU = West African Economic and African Community; SACU = Southern African Customs Union; SADC = Southern Monetary Union. African Development Community; WAEMU = West African Economic and Monetary Union. nounced, but subregional trade nonetheless represents half of their intraregional trade. These significant shares 2016). The results also suggest that subregional trade of subregional trade point to the importance of con- agreements played a major role in strengthening bilat- sidering subregions explicitly when assessing potential eral trade in the region, in particular for countries in spillovers. the SADC and the EAC (see Box 1). In absolute terms, a few subregions account for most The Role of the Economic Structure of total sub-Saharan Africa intraregional trade. The SADC and the SACU account for more than 70 per- Regional trade integration in sub-Saharan Africa cent and 50 percent of total sub-Saharan Africa trade, and its potential to generate intraregional spillovers respectively (Figure 12). Other subregions represent a varies substantially depending on the natural resource smaller share of total sub-Saharan African intraregional endowment of a country. Non-resource-intensive coun- trade, with the WAEMU, EAC, and CEMAC account- tries are the most exposed to regional demand, with ing for less than 10 percent each. The overall pattern of intraregional exports accounting for 7 percent of GDP high subregional integration reflects not only geo- and 30 percent of total exports, on average (Figure 13). graphic proximity but also infrastructure constraints They are followed closely by non-oil-resource-intensive and the impact of regional trade agreements and lower countries (“other”). In terms of demand concentration, nontariff trade barriers within subregions. these other resource-intensive countries have the largest Empirical estimates suggest that trade in the region share of imports from the region, constituting 30 per- is larger between countries that are culturally and cent of total imports. Oil-producing countries have geographically closer and that regional trade growth distinct trading relationships compared with the other over the past four decades is mostly explained by groups and are notably more oriented toward the rest subregional integration. A trade gravity equation of the world. Exports from oil-producing countries to estimation shows that bilateral trade in the region is the rest of the world amount to 25 percent of GDP, greater among countries that are separated by a smaller while intraregional exports represent only 1.5 percent distance and that share a common currency, language, of GDP (Figures 13 and 14); thus, the latter group is ethnicity, and colonial heritage. Cross-region compar- relatively less likely to suffer from intraregional spill- isons show that distance is a great barrier to trade in overs through the trade channel. sub-Saharan Africa, possibly because of the well-known infrastructure gaps in the region (Allard and others International Monetary Fund | August 2018 9 ©International Monetary Fund. Not for Redistribution SPILLOVER NOTES Box 1. Gravity Equation Estimation for 2010–16 Trade Flows A gravity model of bilateral trade flows over the ​log​F​  ijt​​  = β ​X​  ij​​  + ​θ​  o​​ ​I​  i∈SSA​​ ​. X​  ij​​  + ​θ​  d​​ ​I​  j∈SSA​​ ​. X​  ij​​  + ​γ​  it​​  + ​ period 1980–2016 is estimated to study the determi- δ​  jt​​  + ​ε​  ijt​​ . (​ 3)​​ nants of regional trade integration. In all specifications, This specification allows us to investigate whether the sample includes annual data from the Direction of distance plays a specific role for export flows within Trade Statistics (DOTS) database with all the country sub-Saharan Africa. Coefficient estimates θ​ ​​ o​​​ and θ​​ ​  d​​​ pairs in the world that exchanged goods at least once. of these interaction variables, respectively, reflect the The first specification is as follows: differential effects of distance for sub-Saharan Africa ​log​F​  ijt​​  = ​α​  i​​  + ​ϕ​  j​​  + ​θ​  t​​  + β ​X​  ijt​​  + ​β​  xr​​ ​xr​  ijt​​  + γ ​Y​  it​​  + δ ​ origin and destination countries. Table 1.1, column Z​  jt​​  + ​ε​  ijt​​ , ​(1)​​ 3, is composed of three subcolumns; the coefficient estimates ​​θ​  o​​​are reported in the second subcolumn where ​​F​  ijt​​​is the logarithm of export values in dol- while estimates θ​ ​​ d​​​are reported in the third. The lars from country i to country j; ​​X​  ijt​​​ corresponds results indicate that distance is a greater hindrance to corridor-specific variables, including geographic when exporting to sub-Saharan destinations and distance and dummy variables indicating whether the that having belonged to the same colony is a greater countries share a common official language, share the benefit in sub-Saharan Africa. Also, compared with same ethnic group, share the same colonial origin, other regions, sub-Saharan African exports are even share a common official religion, or share a currency; ​​ larger between countries that share a language and a xr​  ijt​​​ corresponds to bilateral exchange rates; and ​​Yit​  ​​​ colonial background. Finally, the distance between two and ​​Zjt​  ​​​, respectively, refer to the logarithms of GDP sub-Saharan countries deters trade significantly more per capita and population in the origin and destina- than between other countries, as shown by a Wald test tion countries. The specification includes country fixed (p-value of 0.02). effects, ​​α​  i​​​ and ϕ ​​ ​  j​​​, to control for country time-invariant Table 1.1, column 4, explores whether trade integra- characteristics, as well as time effects, θ​ ​​ t​​​, to control for tion occurred faster as of 2016 between countries that all annual shocks common to all countries. belonged to the same economic union: the WAEMU, The specification in Table 1.1, column 2 includes CEMAC, EAC, SADC and SACU (see main text country-time fixed effects, ​​γ​  it​​​and ​​δ​  jt​​​, to control for all for region definitions). The following specification is country variable characteristics. The corridor variables ​​ estimated: X​  ij​​​in equation (1) are kept but the country-level char- acteristics (population and GDP) are dropped, as these ​​log​F​  ijt​​  = ​α​  ij​​  + ​γ​  it​​  + ​δ​  jt​​  + β ​​(​​X​  ij​​  . t​)​​  + ​ϕ​  1​​ ​I​  waemu​​  . t + ​ characteristics are absorbed by the country-time fixed ϕ​  2​​ ​I​  cemac​​  . t + ​ϕ​  3​​ ​I​  eac​​  . t+ ​ϕ5​​ ​Ieac​​. t + ​​ϕ​  4​​  I​  sacu​​  . t + ​ε​  ijt​​  ,  ​(4)​​​ effects. Hence, the results presented in column 2 refer ​ ​ to the specification: where country-time, δ​​ ​  jt​​​and ​​γ​  it​​​, and country-pair, ​​α​  ij​​​, fixed effects are included; corridor fixed characteristics ​log​F​  ijt​​  = β ​X​  ij​​  + ​γ​  it​​  + ​δ​  jt​​  + ​ε​  ijt.​​ (​ 2)​​ are interacted with a time trend, (​​​​ ​​X​  ij​​  . t​)​​​​; and member- Estimation results in the first two columns of ships to a common subregion are respectively inter- Table 1.1 show that distance significantly hampers acted with a time trend. trade flows across countries. The first column addition- As shown in column 4, there is statistically signif- ally shows that exports increase significantly with both icant evidence that integration among members of population and GDP per capita of both the origin the EAC and the SADC was particularly successful and destination countries, and are also higher between in fostering trade. This relationship holds even after partners that share a common language, ethnicity, and controlling for developments in individual countries colonial heritage. Bilateral exchange rates do not have when country-time effects are introduced. Quantita- a significant effect on bilateral trade flows. tively, trade among members of the EAC increased In Table 1.1, column 3, all countries in the world by an additional 4 percent per year on average while (including those not in sub-Saharan Africa) are trade among members of the SADC increased by an included. The specification is also richer, as interaction additional 2 percent per year. Using these estimates variables between measures of distance and a dummy to compute what trade would have been without variable for either sub-Saharan Africa origin countries subregional integration, we finds that average annual (​​Ii∈SSA ​  ​​ ​. X​ ij​​​) or sub-Saharan Africa destination countries growth in regional trade would have been about (​​Ij∈SSA ​  ​​  . ​Xij​  ​​​) are introduced: 9 percent instead of 11 percent, thereby translating 10 International Monetary Fund | August 2018 ©International Monetary Fund. Not for Redistribution  Regional Spillovers in Sub-Saharan Africa: Exploring Different Channels Box 1. Gravity Equation Estimation for 2010–16 Trade Flows (continued) into trade levels that would be half as low as those barrier, meaning that the infrastructure facilitating observed in 2015. In addition, the coefficient on the trade among economic unions has lagged relative to interaction between distance and time suggests that in the development of infrastructure within unions. sub-Saharan Africa, distance has increasingly become a Table 1.1. Determinants of Trade Flows Dependent variable: logarithm of bilateral trade flows (3) (1) (2) Coefficient estimates for (4) Country controls Country FE All countries SSA origin SSA destination Time trends Contiguous countries 1.57*** 1.53*** 0.48*** 20.00 (0.21) (0.21) (0.11) (0.01) Distance (in log) 21.60*** 21.60*** 21.60*** 0.18 20.40*** 20.02*** (0.09) (0.09) (0.04) (0.11) (0.11) (0.00) Common language 0.46*** 0.45*** 0.54*** 0.17 20.11 0.01 (0.15) (0.14) (0.10) (0.14) (0.15) (0.01) Common ethnicity 0.24* 0.30** 0.21** 20.27** 0.21* 0.01 (0.14) (0.13) (0.09) (0.14) (0.12) (0.01) Belonged to common 1.44*** 1.37*** 0.81*** 1.16*** 0.74*** 20.00 colony (0.13) (0.12) (0.11) (0.20) (0.19) (0.01) Common religion 0.14 0.14 0.25*** 20.01 20.04 0.00 (0.14) (0.13) (0.06) (0.16) (0.13) (0.01) Common currency 1.22*** 1.28*** 0.40 0.99** (0.31) (0.31) (0.29) (0.39) Origin GDP p.c. 0.49*** (0.06) Destination GDP p.c. 0.48*** (0.13) Origin population 0.70** (0.30) Destination population 2.41*** (0.24) Origin/destination 0.00 FX rate (0.01) WAEMU trend 0.00 (0.01) CEMAC trend 0.00 (0.01) EAC trend 0.04* (0.02) SADC trend 0.02** (0.01) SACU trend 0.09 (0.07) Observations 92,132 95,711 556,476 95,108 R-squared 0.52 0.57 0.73 0.77 Year FE YES NO NO NO Country FE YES NO NO NO Country-time FE NO YES YES YES Country-pair FE NO NO NO YES Sources: IMF, Direction of Trade Statistics database, World Economic Outlook database; and IMF staff calculations. Note: In all specifications, the standard errors are clustered at the destination country level to control for possible unobserved correlation within importing countries. CEMAC 5 Central African Economic and Monetary Community; EAC 5 East African Community; SACU 5 Southern African Customs Union; SADC 5 Southern African Development Community; WAEMU 5 West African Economic and Monetary Union. Clustered standard errors in parentheses (Destination country) ***p  0.01, ** p 0.05, *p  0.1. International Monetary Fund | August 2018 11 ©International Monetary Fund. Not for Redistribution SPILLOVER NOTES Figure 13. Sub-Saharan Africa: Intraregional Exports, by Country Groups (Simple average) 35 Percent of total exports in 2016 Percent of GDP in 2016 30 Percent of total exports in 2000–15 Percent in GDP 2000–15 25 20 Percent 15 10 5 0 Non-resource intensive Other resource-intensive Oil exporters Sources: IMF, Direction of Trade Statistics database, World Economic Outlook database; and IMF staff calculations. Note: See page 41 for country groupings table. Intraregional Trade and Growth China, whose increased market for sub-Saharan Afri- Sub-Saharan African countries’ growth depends can exports has been shown to be an important driver on fluctuations in the economic activity of regional of growth in the region (Chen and Nord 2017). trading partners and global developments. A panel Consistent with having comparable shares of intra- regression model for all sub-Saharan African countries regional trade, the trade channel spillovers seem to be for the period 1980–2016 suggests that a 1 percent- similar in sub-Saharan Africa and in other emerging age point increase in the export-weighted growth and developing economies. Panel regression estimates rate of intraregional partners is associated with about for countries in the Middle East and North Africa, a 0.11 percent increase in the average sub-Saharan Latin America, and emerging and developing Asia sug- African country’s growth (Box 2). These results were gest that sub-Saharan African countries have a slightly obtained after accounting for extraregional factors such lower intraregional elasticity of growth compared with as terms of trade movements and demand from trading peers in Latin America and emerging and developing partners outside the region, including countries such as Asia, but a higher one than those estimated for coun- tries in the Middle East and North Africa (Box 2). Figure 14. Sub-Saharan Africa: Exports to the Rest of the World, by Country Groups (Simple average) 45 2016 2000–15 40 35 30 Percent of GDP 25 20 15 10 5 0 Oil exporters Other resource-intensive Non-resource intensive Sources: IMF, Direction of Trade Statistics database, World Economic Outlook database; and IMF staff calculations. Note: See page 41 for country groupings table. 12 International Monetary Fund | August 2018 ©International Monetary Fund. Not for Redistribution  Regional Spillovers in Sub-Saharan Africa: Exploring Different Channels Box 2. GDP Growth Elasticities to the Growth of Trading Partners A panel fixed effect model is estimated to study the column 8), as is customary in the growth literature. elasticity of GDP growth in sub-Saharan African coun- The latter allows the model to minimize serial cor- tries to the growth of the trading partners inside and relation, which is likely to be present in annual data. outside the region. The model is specified as follows: This specification includes the initial level of GDP per capita to capture growth convergence. Country fixed ​​RealGDPgrowth​  it​​  = α + ​γ​  i​​  + ​β​  1​​ ​ effects, ​​γ​  i​​​, are included to control for time-invariant RealGDPgrowth​  t−1,i​​  + ​β​  2​​ ​RealGDPgrowth​  t−2,i​​​​​ country-specific heterogeneity, and standard errors are + ​β​  3​​ ​SSAtradingpartnergrowth​  t,i​​  + ​β​  4​​  Non − ​ clustered at the country level to control for possible SSAtradingpartnergrowth​  t,i​​  + ​β​  5​​ ​X​  it​​  + ​e​  it​​             ​(​​1​)​​​​ unobserved correlation across countries. where the left-hand variable is the annual rate of The estimation includes all sub-Saharan African growth of real GDP, sub-Saharan African trading countries over the period 1980–2016. The lagged partner growth refers to the weighted average rate time-varying weights reflect the annual share of of growth of the trading partners from sub-Saharan exports going to each specific partner as reported in Africa for each country at time t. Non-sub-Saharan the IMF Direction of Trade Statistics. The baseline Africa trading partner growth corresponds to the specification suggests that a 1 percentage point increase weighted average rate of growth of the trading partners in the export-weighted growth rate of intraregional from outside the region for each country at time t partners is associated with about a 0.11 percent (Arora and Vamvakidis 2005; Chen and Nord 2017). increase in the average sub-Saharan African country The weights used in averaging refer to export shares growth rate (Table 2.1, column 1). This estimate in the previous year. The vector X of controls includes is robust across multiple specifications (Table 2.1, variables capturing regional and global growth dynam- column 5). The baseline specification also finds that a ics, as measured by the average rate of growth of the 1 percentage point increase in the growth rate of the region and the world. trading partners outside the region is associated with The model includes country-specific controls in X: an increase of 0.34 percent in the growth rate of the lags of the dependent variable, the rate of investment average sub-Saharan African country. This coefficient to GDP, the inflation rate, and the level of trade seems to capture other external environment factors, openness. This allows the coefficient of interest to such as the changes in the terms of trade and the more clearly identify the effect of the regional trading degree of openness of the economy, as evidenced by partners’ growth on individual countries, once the the results in column 2. average growth co-movements in the continent and in These results are robust to the inclusion of other the world are isolated. In addition, the model controls controls that capture structural factors such as for the occurrence of conflict and war, as captured investment-to-GDP and population growth, and by the Uppsala Conflict Data Program. As part of other monetary controls such as global liquidity, the control vector X, variables capturing the external inflation, and exchange rate movements. The regres- environment are also included: fluctuations in the sions results are also robust to implementing a panel terms of trade, and the degree of trade openness of generalized method of moments (GMM) estimation the economy and the share of regional exports in total to account for the endogeneity in a dynamic-panel exports. Other controls include a measure of interna- context (Table 2.1, column 5). Results are robust to tional liquidity as captured by the change in the Fed the exclusion of large economies (Table 2.1, column funds rate, a measure of the monetary policy stance in 6). Introducing interaction variables fails to capture the United States, the country inflation rate, and the heterogeneous effects, as the new variables are insig- change in the bilateral exchange rate with respect to nificant (Table 2.1, column 7). This means that more the US dollar. open economies are not significantly more exposed As a robustness check, the 10 percent largest econo- to spillovers, suggesting that the nature of trade in mies are excluded from the estimation sample: Angola, more open economies prevents them from being Ghana, Kenya, Nigeria, and South Africa (Table 2.1, more exposed to variation in partners’ demand. Using column 6); an interaction term between countries’ five-year averages to analyze the medium-term deter- openness and their partners’ average growth is added minants of economic growth finds that the growth (Table 2.1, column 7); and a specification using of the regional trading partners continues to play an five-year averages of the data is estimated (Table 2.1, important role in individual countries’ rate of growth, International Monetary Fund | August 2018 13 ©International Monetary Fund. Not for Redistribution SPILLOVER NOTES Box 2. GDP Growth Elasticities to the Growth of Trading Partners (continued) Table 2.1. GDP Growth Elasticities to the Growth of Trading Partners Dependent variable: real GDP growth (2) (3) (4) (5) (6) (7) (8) (1) Openness Structural Monetary GMM Excluding Exposure GMM, 5–year Baseline controls controls controls estimation 10% largest heterogeneity averages Real GDP growth (t–1) 0.318*** 0.299*** 0.292*** 0.284*** 0.264*** 0.255*** 0.264*** 0.410*** (0.0821) (0.0734) (0.0771) (0.0789) (0.0678) (0.0749) (0.0681) (0.0632) Real GDP growth (t–2) 20.0180 20.0374 20.0452* 20.0551** 20.0770* 20.0736* 20.0863* (0.0202) (0.0247) (0.0244) (0.0260) (0.0405) (0.0413) (0.0454) SSA trading partners’ growth 0.113* 0.114* 0.111* 0.120* 0.133** 0.110* 0.184*** 0.0761** (0.0636) (0.0597) (0.0573) (0.0605) (0.0610) (0.0622) (0.0658) (0.0354) Non2SSA trading partners’ 0.348* 0.265 0.251 0.252 0.334 0.306 20.0842 0.0118 growth (0.184) (0.161) (0.161) (0.164) (0.238) (0.238) (0.249) (0.0437) SSA average growth 0.288** 0.0870 0.124 0.0739 0.0526 0.0124 0.0265 20.117 (0.115) (0.202) (0.192) (0.214) (0.280) (0.260) (0.278) (0.268) World average growth 20.0391 0.210 0.198 0.221 0.108 0.107 0.0794 4.943*** (0.139) (0.201) (0.224) (0.233) (0.255) (0.269) (0.235) (1.761) Conflict year, Uppsala 24.548*** 23.722*** 23.527*** 23.810*** 23.913*** 23.119** 23.717*** 24.267 database (0.900) (1.047) (1.029) (1.081) (1.287) (1.377) (1.392) (2.871) Terms of trade, percent 0.138*** 0.153*** 0.0936* 0.0881* 0.0928 0.110* 20.0361 change (0.0457) (0.0529) (0.0470) (0.0485) (0.0594) (0.0662) (0.374) Trade openness (t–1) 0.0524* 0.0466* 0.0522* 0.0844 0.0908 0.0710 0.0915 (0.0305) (0.0270) (0.0296) (0.0585) (0.0624) (0.0500) (0.0557) Share of regional exports in 3.108* 3.153* 3.426* 1.561 2.267 0.448 20.311 total exports (t–1) (1.604) (1.750) (1.961) (2.888) (2.801) (1.876) (2.103) Investment, percent of GDP 0.0460 0.0483 0.0608 0.0587 0.0638 0.0877** (t–1) (0.0378) (0.0393) (0.0579) (0.0576) (0.0573) (0.0383) Percent change in 0.376 0.394 0.489 0.478 0.448 0.440*** population (0.391) (0.443) (0.341) (0.355) (0.349) (0.159) Change in US Federal Funds 20.000131 20.000256 20.000285 20.000405 rates (%) (0.000555) (0.000497) (0.000539) (0.000485) Inflation 20.00362*** 20.00311*** 20.00198 20.00265** (0.00118) (0.00102) (0.00306) (0.00113) Inflation (t–1) 20.00103 20.000293 20.00268 20.000313 (0.000748) (0.00154) (0.00239) (0.00154) Foreign exchange rate, % 0.180*** 0.155*** 0.101 0.133** change (0.0584) (0.0484) (0.145) (0.0530) Foreign exchange rate, % 0.0619 0.0226 0.137 0.0251 change (t–1) (0.0373) (0.0783) (0.119) (0.0772) SSA trading partners’ 21.808 growth interaction with the (2.277) lag share of regional exports Non–SSA trading partners’ 2.839 growth interaction with the (1.919) lag share of extraregional exports Share of regional exports in 17.61 GDP (t–1) (11.56) Share of extraregional 27.143 exports in GDP (t–1) (7.592) GDP per capita at the 24.95e206*** beginning of the 5y periods (5.07e207) Constant 0.758 22.943 24.541* 24.781* 26.869 219.31** 219.31** 219.31** (0.689) (2.660) (2.486) (2.513) (5.382) (8.033) (8.033) (8.033) Observations 1,345 1,344 1,344 1,301 1,252 1,118 1,252 159 R2squared 0.159 0.180 0.188 0.187 Number of countries 45 45 45 45 45 40 45 43 Source: IMF staff calculations. Note: Robust standard errors in parentheses. ***p , 0.01, **p , 0.05, *p , 0.1. 14 International Monetary Fund | August 2018 ©International Monetary Fund. Not for Redistribution  Regional Spillovers in Sub-Saharan Africa: Exploring Different Channels Box 2. GDP Growth Elasticities to the Growth of Trading Partners (continued) although the estimated coefficient is smaller. Similarly, of growth compared with peers in Latin America and a statistically significant impact of the global economic emerging and developing Asia but a higher elasticity environment continues, as captured by world average than the one estimated for countries in the Middle growth (Table 2.1, column 8). East and North Africa and in emerging and developing Estimating the specification in equation (1) for Europe, which is not statistically different from zero countries in the Middle East and North Africa, Latin (Table 2.2). The results for the Middle East and North America, emerging and developing Asia, and emerging Africa countries can be explained by the importance of and developing Europe, shows that sub-Saharan developments in oil markets. African countries have a lower intraregional elasticity Table 2.2. Sub-Saharan Africa and Other Developing Countries: GDP Growth Elasticities to the Growth of Trading Partners Dependent variable: real GDP growth (1) (2) (3) (4) (5) SSA Latin America MENA Asia Europe and CIS Real GDP growth (t–1) 0.264*** 0.197*** 20.000418 0.170** 0.325*** (0.0678) (0.0407) (0.0351) (0.0843) (0.103) Real GDP growth (t–2) 20.0770* 20.00800 20.196** 20.0487 20.0406 (0.0405) (0.0367) (0.0805) (0.0329) (0.0890) SSA trading partners’ growth 0.133** 0.157** 20.00140 0.157* 0.204 (0.0610) (0.0693) (0.0629) (0.0921) (0.163) Non–SSA trading partners’ growth 0.334 20.0482 20.107 0.492*** 0.335** SSA average growth 0.0526 0.254*** 0.476** 0.322*** 0.429** (0.280) (0.0928) (0.219) (0.117) (0.184) World average growth 0.108 0.394** 0.192 20.148 20.0342 (0.255) (0.163) (0.348) (0.173) (0.201) Conflict year, Uppsala database 23.913*** 22.588 24.737 20.913* 20.674 (1.287) (1.734) (3.427) (0.477) (1.253) Terms of trade, percent change 0.0881* 0.281*** 0.0319 0.121 0.0694 (0.0485) (0.0537) (0.0420) (0.0858) (0.0963) Trade openness (t–1) 0.0844 0.0188*** 0.0347* 0.00545 0.0229 (0.0585) (0.00690) (0.0206) (0.0112) (0.0147) Share of regional exports in total 1.561 22.190 2.662 20.988 2.864 exports (t–1) (2.888) (1.360) (2.870) (1.881) (3.652) Investment share of GDP (t–1) 0.0608 20.161*** 0.239** 0.0442 20.0103 (0.0579) (0.0312) (0.109) (0.0287) (0.0861) Percent change in population 0.489 0.223 20.530 0.108 0.162 (0.341) (0.272) (0.372) (0.534) (0.566) Percent change in US Federal 20.000256 0.00117* 0.000884 0.000218 20.00111 Funds rates (0.000497) (0.000672) (0.000958) (0.000631) (0.000715) Inflation 20.00311*** 20.00326* 20.0669*** 20.00808 20.00327 (0.00102) (0.00182) (0.0218) (0.0130) (0.00203) Inflation (t–1) 20.000293 27.14e205 0.0836** 20.00229 26.71e205 (0.00154) (0.000366) (0.0390) (0.0140) (0.00135) Foreign exchange rate, percent 0.155*** 0.314 20.829 21.672 20.140 change (0.0484) (0.200) (0.508) (1.052) (0.107) Foreign exchange rate, percent 0.0226 20.00983 0.757 0.661 20.00828 change (t–1) (0.0783) (0.00869) (0.770) (0.477) (0.121) Constant 26.869 2.959*** 23.668* 21.05 22.297 (5.382) (0.937) (2.032) (2.777) (2.913) Observations 1,252 941 560 494 420 Number of countries 45 30 20 18 18 Source: IMF staff calculations. Note: CIS 5 Commonwealth of Independent States; MENA 5 Middle East and North Africa. SSA 5 sub-Saharan Africa. Robust standard errors in parentheses. ***p , 0.01, **p , 0.05, *p , 0.1. International Monetary Fund | August 2018 15 ©International Monetary Fund. Not for Redistribution SPILLOVER NOTES Banking Interdependencies Becoming Figure 15. PAB and Subregional Bank Presence in SSA, More Subregional 2007–16 Banks headquartered in sub-Saharan Africa account 1. Assets in host countries as a percent of host country assets for an increasingly large share of the regional financial sys- PAB: Total assets in hosts (Percent of total host assets) 12 Subregional banks: Total assets in hosts 1200 tem.5 Their headquarters are based in a small number of (Percent of total host assets) countries, and their subsidiaries and branches are hosted 10 Total SSA Assets 1000 across a wide range of countries—a situation that creates (Billions USD, RHS) important spillover channels. Additionally, as regional 8 800 USD billion banks’ activities have increased over time, so has financial Percent sector depth, which itself has been associated with higher 6 600 GDP growth in the region. 4 400 2 200 Presence of Foreign Banks Headquartered in Sub-Saharan Africa 0 0 2007 08 09 10 11 12 13 14 15 16 The expansion of banks headquartered in sub-Saharan Africa has contributed to the deepening 2. Deposits in host countries as a percent of host country deposits of financial systems across the region and represents an PAB: Total deposits in hosts (Percent of total host deposits) 14 Subregional banks: Total deposits in hosts 740 important vector for economic and financial spillovers. (Percent of total host assets) Banks headquartered in sub-Saharan Africa are gen- 12 640 Total SSA Deposits erally well capitalized and have expanded throughout (Billions USD, RHS) 540 10 the region mainly via subsidiaries. The subsidiaries are 440 USD billion funded primarily using local sources, which limits the 8 Percent 340 risk of spillovers in the short term (IMF 2012, 2015c; 6 Mecagni, Marchettini, and Maino 2015). However, 240 as the financial sector continues to expand on the 4 140 continent, sub-Saharan Africa–based banks constitute 2 40 an increasingly important transmission channel for real economic activity in the medium term. This channel 0 –60 2007 08 09 10 11 12 13 14 15 16 is particularly relevant today as vulnerabilities among sub-Saharan Africa–based banks have increased owing Sources: Fitch Connect; IMF, International Financial Statistics; and IMF staff calculations. to their high exposure to sovereign debt and to the Note: PAB = pan-African banks; SSA = sub-Saharan Africa, RHS = right hand side. commodity export sector, more nonperforming loan levels in some cases, and the often limited capacity for governments to support or resolve troubled banks (Kinda, Mlachila, and Ouedraogo 2016; IMF 2017b). Africa–based banking groups that own subsidiaries or Banks headquartered in sub-Saharan Africa branches in three to nine sub-Saharan African coun- expanded rapidly throughout sub-Saharan Africa fol- tries (referred to as subregional banks) grew through lowing the global financial crisis. Pan-African Banks— the postcrisis period before growth slowed down defined here as sub-Saharan Africa–based banking more recently. groups that own subsidiaries or branches in 10 or more This growth coincided with the retrenchment of sub-Saharan African countries—expanded rapidly European and American banks on the continent. from 2007 through 2013 before expansion decelerated In terms of both asset and deposit shares, PABs and in recent years (Figure 15). Similarly, sub-Saharan subregional banks control a roughly equal share of the foreign markets in which they are present (Fig- 5This section focuses on banks that are both headquartered in ure 15).6 In terms of market participation, the share sub-Saharan Africa (and thus subject to sub-Saharan African banking regulations) and majority owned by sub-Saharan African groups, 6In the literature (notably IMF 2015c), subregional banks are because of the implications for the home countries in terms of defined as having a presence in five or more countries. The definition regional spillovers. has been expanded in this note to capture important subregional 16 International Monetary Fund | August 2018 ©International Monetary Fund. Not for Redistribution  Regional Spillovers in Sub-Saharan Africa: Exploring Different Channels Figure 16. Sub-Saharan African PABs and Subregional Banks: Home and Host Countries (PAB or subregional banking group foreign assets and deposits, measured as indicated) 1. PAB Home Countries 2. PAB Host Countries (Percent of total PAB assets or deposits outside parent bank country) (Percent of total domestic assets or deposits) 50 70 Assets Assets 60 40 Deposits Deposits 50 Percent 30 40 Percent 20 30 20 10 10 0 0 South Africa Togo Nigeria LSO CAF BFA GNB BEN MLI SWZ SEN UGA NAM 3. Sub-Regional Banking Groups Home Countries 4. Sub-Regional Banking Groups Host Countries (Percent of total SRB assets or deposits outside parent bank country) (Percent of total domestic assets or deposits) 35 45 Assets 40 Assets 30 Deposits Deposits 35 25 30 20 Percent 25 Percent 15 20 15 10 10 5 5 0 0 ZAF TGO KEN NGA BWA MUS GAB CMR SWZ LSO BWA CAF COG TCD SLE TZA ZMB UGA NER Sources: Fitch Connect, IMF, International Financial Statistics; and IMF staff calculations. Note: PAB = pan-African banks; SRB = sub-regional banks. See page 41 for country abbreviations table. of PABs and subregional banking groups in the total concentrated in three countries, with South Africa sub-Saharan African financial system is increasing, fol- and Togo each home to about 40 percent of all PABs, lowing a global trend of banking regionalization (IMF leveraging their roles as subregional hubs for finan- 2015b). Nevertheless, assets and deposits have recently cial services (Figure 16.1).7 In contrast, subregional declined, consistent with the decline in regionwide parent banks are only about half as concentrated. PAB assets that coincided with the economic deceleration and subregional banking group parents are based in in the region. a similar set of countries (Figure 16.3), but they are hosted in different sets of countries. While the foreign Country Exposure to Sub-Saharan African Banks subsidiaries of PABs are spread across the region Both PABs and subregional banks are highly (Figure 16.2), countries that host subregional banks are concentrated in terms of their home countries but notably more concentrated (Figure 16.4). While the have foreign entities that are widely dispersed across majority of South African subregional bank subsidiar- sub-Saharan Africa, thus representing a potentially 7Ecobank in Togo is a special case. Only the holding company, important spillover channel. PAB parent banks are Ecobank Transnational Incorporated, is headquartered in Togo, and it has the status and privileges of a nonresident supranational finan- banks that have a presence in fewer countries, particularly those in cial institution. The de facto economic headquarters of Ecobank is in East Africa. Nigeria, where its largest subsidiary is located (IMF 2015c). International Monetary Fund | August 2018 17 ©International Monetary Fund. Not for Redistribution SPILLOVER NOTES ies are located in SADC countries, Kenyan subregional Figure 17. Systemic Foreign Owned PAB and banks are mostly situated in EAC countries, Nigerian Subregional Banks, 2007–16 subregional banks are mainly in West African coun- 1. Average Bank Deposits, by Country, 2007–16 tries, and Gabonese and Cameroonian subregional (Percent of total deposits in country) 70 banks are mostly in neighboring CEMAC countries. Average deposit ratio, PAB This means that developments in a few countries home 60 Average deposit ratio, SRB to most PABs and subregional banks can have substan- 50 tial economic spillovers in a variety of countries. Spillover channels are potentially numerous and run 40 Percent in both directions between the parent bank and its 30 subsidiaries, as well as across subsidiaries and branches of the same banking group (IMF 2015c). As described 20 in IMF (2015c), subsidiaries and branches could be affected if they are connected to their parent through 10 the placement of deposits and credit or via deficien- 0 LSO CAF SWZ GNB BEN TCD NAM BFA MLI SLE SEN COG UGA TZA cies in governance, perceptions of mismanagement, or other reputational concerns at the group level. 2. Cross-Country Average Bank Deposits, By Bank Type, 2007–16 On the other hand, parent banks themselves may be (Percent of total deposits in country) exposed to risks in any of the countries hosting their 20 Average deposit ratio, PAB foreign subsidiaries and branches, depending on the 18 Average deposit ratio, SRB systemic importance of these entities to the local 16 economy, the liquidity-sharing arrangements across the 14 banking group, and the size of the foreign subsidiary 12 Percent or branch operation relative to the group. While the 10 subsidiary model minimizes contagion risk, it does not 8 eliminate it completely, as subsidiaries may still have 6 exposure to their parent or other entities in the bank 4 group (Mecagni, Marchettini, and Maino 2015). For 2 example, there may be important risks stemming from 0 2007 08 09 10 11 12 13 14 15 16 syndicated loans between subsidiaries or branches. Finally, host countries of systemic PABs and subre- Sources: Fitch Connect; IMF, International Financial Statistics; and IMF staff gional bank subsidiaries and branches may face risks calculations. Note: PAB = pan-African banks; SRB = sub-regional banks. See page 41 for arising from unilateral or uncoordinated actions taken country abbreviations table. by the banks’ home authorities or parent banks, which can have implications for financial stability in the host jurisdiction (Mecagni, Marchettini, and Maino 2015). Nigeria and South Africa) or because the foreign enti- In many countries, banks headquartered in ties were not systemic (IMF 2012, 2015c). In either sub-Saharan Africa are systemically important, increas- case, a banking crisis in the headquarters country did ing the potential for cross-border spillovers. The ratio not typically affect ratios at the macro level in host of total deposits in foreign African subsidiaries or countries. However, this does not rule out the possi- branches of PABs or subregional banks to total depos- bility that a PAB or subregional banking group parent its by country—a measure of systemic importance—is bank could be hit by a shock that is transmitted across highest in small countries (Figure 17). Also, the degree borders. If its foreign subsidiary or branch is systemi- of systemic importance is larger for PAB subsidiaries cally important, such a shock could have real effects on and branches than for subregional banks. In the past, the host economy. This situation would be difficult for spillovers from banking crises in African countries host country policymakers to foresee. were limited, either because the subsidiaries in host countries were mainly funded by local deposits and therefore did not significantly depend on funding from their parent (for example, banks headquartered in 18 International Monetary Fund | August 2018 ©International Monetary Fund. Not for Redistribution  Regional Spillovers in Sub-Saharan Africa: Exploring Different Channels Figure 18. Sub-Saharan Africa: Real, Financial, and Cross-Border Links 1. Cross-Country Average Total Private Credit, 1960–2016 2. Private Credit Growth and GDP Growth in Sub-Saharan Africa, 2008–15 (Percent of GDP) (Average) 80 8 Average, SSA 70 PAB home countries 7 SRB home countries 60 Major PAB host countries 6 Major SRB host countries 50 5 GDP Growth Percent 40 4 30 3 20 2 10 1 0 0 –4 0 4 8 12 16 20 24 28 32 36 40 1960 63 66 69 72 75 78 81 84 87 90 93 96 99 2002 05 08 11 14 Private Credit Growth Sources: Fitch Connect; IMF, International Financial Statistics; and IMF staff calculations. Cross-border Banking and Growth Figure 19. Bank Deposits in PABs and Subregional Banks, Financial markets have deepened in countries home 2007–16 to pan-African and subregional banks, and this deepen- 700 Total deposits, oil exporters ing has been associated with higher real GDP growth. Total deposits, non-oil exporters 600 Over the long term, financial markets in sub-Saharan Africa have gradually deepened, as measured by credit 500 to the private sector (Figure 18.1). This is especially Billions of USD 400 true in countries that are home to PABs and subre- gional banks compared with those that are primarily 300 hosts to these banking groups. At the same time, 200 evidence has shown that financial development and deepening have supported growth and reduced growth 100 volatility in sub-Saharan Africa (IMF 2016a), as can be 0 seen from the strong positive association between GDP 2007 08 09 10 11 12 13 14 15 16 growth and private credit growth (Figure 18.2). Sources: IMF, International Financial Statistics; and IMF staff calculations. PABs and subregional banks could spur cross-border Note: See page 41 for country groupings table. growth spillovers. Depending on the funding arrange- ments within these banking groups, lower growth in the parent banks’ home countries could lower credit excess liquidity to their parent, which may limit credit and deposit growth in their foreign African subsid- growth in the parent country. iaries and branches, as it does at home (IMF 2012). The source of economic deceleration is important This would be the case if the parent supplies a signif- for the financial sector. Countries in the region that icant portion of liquidity to the subsidiary or branch, were hard hit by the commodity price decline have although evidence suggests that bank funding is mostly experienced credit growth deceleration and a decline local in the largest countries (IMF 2012, 2015c). On in deposits (Figure 19) (Agrawal, Duttagupta, and the other hand, lower growth in host countries also Presbitero 2017). This is partly a result of the tendency limits the prospects for new cross-border expansion of African banks to be highly exposed to the commod- opportunities for the parent bank and constrains ity sector (IMF 2017c). Reinforcing factors such as a the ability of subsidiaries and branches to repatriate slowdown in economic activity and a buildup in gov- ernment arrears to contractors can further exacerbate International Monetary Fund | August 2018 19 ©International Monetary Fund. Not for Redistribution SPILLOVER NOTES Table 1. Loan-to-Deposit Ratios, Largest Sub-Saharan African Countries, 2015 (Percent) Foreign-owned Foreign-owned Countries pan-African banks subregional banks All banks* Kenya 61.3 77.1 88.9 Tanzania 57.4 79.1 73.6 Ghana 51.5 63.8 71.6 Côte d’Ivoire 58.2 63.3 80.9 Cameroon 57.3 79.6 90.6 Uganda 50.2 83.4 79.7 Zambia 48.8 59.7 69.9 Mali 41.1 57.3 95.0 Botswana 61.7 73.7 79.6 Mozambique 53.9 68.3 69.8 Burkina Faso 56.5 64.3 93.9 Sources: IMF, International Financial Statistics; and IMF staff calculations Note: *Aggregate loan-to-deposit ratio measured using IFS bank credit-to-deposit ratio. any rise in nonperforming loans, lower bank profits, and the shadow banking sector is about 40 percent or increased solvency risk (IMF 2017b). Depending of the entire financial sector (IMF 2016b). In South on the structure of cross-border banking groups, these Africa, NBFIs hold about two-thirds of all financial credit supply and solvency issues can have cross-border assets, with pension funds holding assets equal to implications. 110 percent of GDP (versus banking assets equal to 112 percent), and long-term insurers holding assets equal to 64 percent of GDP (IMF 2014). PABs gener- African Bank Behavior and Trends in Correspondent ally have large shares of ownership in NBFIs, thereby Banking Relationships increasing the risk of spillovers from developments Both PABs and subregional banks are relatively in the real economies to the banking sectors. These less active lenders in the countries where they oper- risks are compounded by a lack of regulation in the ate. In almost all countries that host PABs, the PAB nonbank financial sector and low levels of compliance subsidiaries and branches seem to be less aggressive in the banking sector. in their lending practices on average compared with The withdrawal of correspondent banking rela- other banks. Across hosting countries, the average tionships (CBRs) in sub-Saharan Africa presents loan-to-deposit ratio for PABs is about 34 percent less another potentially important channel for spillovers. than the country-level average (Table 1). Subregional After increasing between 2011 and 2014, CBRs have banks are also more restricted in their lending, with been under pressure in recent years. Since 2011, average loan-to-deposit ratios about 22 percent less sub-Saharan Africa has seen a 4 percent decline in the than country-level averages. These patterns reflect a number of active correspondent banks and a 9 per- combination of supervisory limits, preference to act as cent decline in the number of counterparty countries deposit-taking institutions with limited lending to the (FSB 2017). The decline has been driven by a range private sector, and greater exposure to sovereigns, as of factors, including weaknesses in controls at respon- has been documented in East Africa (Cihak and Pod- dent banks, inadequate supervision and regulation, piera 2005) and across the continent (IMF 2015c). country risk, and profitability. While much of the PABs and subregional banks have become increas- decline appears to be concentrated in small banks, ingly complex by integrating nonbank activities, some subregional banks have also lost their CBRs, and increasing the potential for spillovers. This is partic- in some cases respondent banks have terminated CBRs ularly the case in the southern part of the continent, with their own subsidiaries. The termination of a CBR where these activities tend to have a regional scope, affects a bank’s ability to extend credit and transfer such as insurance and securities dealing (IMF 2015c). international payments, which has direct effects on For instance, in Namibia, nonbank financial institu- growth, trade, and internal and external stability. If a tions (NBFIs) have gross assets four times those of CBR is terminated with a PAB or regional banking traditional banks (equivalent to 330 percent of GDP), 20 International Monetary Fund | August 2018 ©International Monetary Fund. Not for Redistribution  Regional Spillovers in Sub-Saharan Africa: Exploring Different Channels Figure 20. Sub-Saharan Africa: Sovereign Spread Correlations, 2012–16 AGO 1 CIV 0.85 1 CMR 0.93 0.96 1 ETH 0.92 0.89 0.92 1 GAB 0.93 0.92 0.98 0.92 1 GHA 0.90 0.95 0.98 0.88 0.96 1 KEN 0.88 0.91 0.95 0.91 0.96 0.93 1 MOZ –0.02 –0.17 –0.21 0.03 –0.22 –0.28 –0.24 1 NAM 0.88 0.96 0.98 0.88 0.95 0.95 0.92 –0.22 1 NGA 0.89 0.88 0.93 0.85 0.95 0.93 0.90 –0.23 0.93 1 SEN 0.87 0.95 0.98 0.90 0.97 0.96 0.96 –0.27 0.96 0.92 1 TZA 0.70 0.80 0.82 0.65 0.79 0.80 0.73 –0.25 0.85 0.80 0.78 1 ZAF 0.81 0.92 0.94 0.79 0.92 0.93 0.87 –0.29 0.95 0.93 0.93 0.90 1 ZMB 0.91 0.96 0.99 0.91 0.98 0.99 0.96 –0.23 0.97 0.95 0.98 0.81 0.95 1 AGO CIV CMR ETH GAB GHA KEN MOZ NAM NGA SEN TZA ZAF ZMB Sources: JP Morgan; and IMF staff calculations. Note: See page 41 for country abbreviations table. group parent, the impact can be felt across the entire markets and the country most highly integrated with group and multiple countries. global financial markets, may be driving movement in sovereign spreads of the region’s frontier markets (IMF 2012, 2017b). The Dominant Role of South Africa Sovereign The substantial co-movement in sovereign spreads Spread Spillovers across sub-Saharan African frontier markets may be Though highly vulnerable to global trends, frontier driven in part by conditions in South Africa. A simple markets in sub-Saharan Africa are driven in part by correlation analysis of sovereign spreads indicates trends in South Africa. This is consistent with other a high degree of co-movement across sub-Saharan emerging markets and low-income regions dominated by a African frontier markets (Figure 20).8 A principal single large and financially integrated economy. component analysis on these spreads further shows that 85 percent of this co-movement is explained by their first common factor. This first common factor is Cross-Country Co-movement and Its Drivers then itself strongly correlated with economic indica- Evidence suggests that financial market tors of a single country, South Africa, with a correla- co-movement in emerging and developing econo- tion coefficient of 0.93 against the implied volatility mies is driven by both global and regional factors. It index of the Johannesburg stock exchange (SAVI) and is well established that open economies are exposed of 0.94 against the South African sovereign spread. to a global financial cycle, and certainly sub-Saharan However, correlations alone cannot determine whether African countries are no exception (Rey 2015). the South Africa factor is driving the co-movement or However, national policies and fiscal fundamentals a confounding global factor is driving both the South can also have an impact on regional treasury bill and African spread and co-movement with sub-Saharan bond markets. In sub-Saharan Africa, there is evidence African frontier markets. of cross-country fiscal spillovers to sovereign spreads within regional economic areas (Hitaj and Onder 2013). Evidence also shows that if there is a single Estimating the Impact of South Africa dominant economy in an emerging market region, The estimated impact of cross-border spillovers of it tends to influence asset prices in smaller coun- movements in the South African spread to frontier tries within the region (see, for example, Mwase and markets of sub-Saharan Africa is considerable. The others 2016). For sub-Saharan Africa, it is natural to ask whether South Africa, the largest sovereign debt 8Mozambique is an outlier owing to public debt misreporting and emitter with the deepest and most liquid financial subsequent default, which increased its spread significantly in recent years, making it substantially out of line with regional spreads. International Monetary Fund | August 2018 21 ©International Monetary Fund. Not for Redistribution SPILLOVER NOTES South African spread explains about 6 percent more of The composition of remittances in sub-Saharan the variation in frontier market spreads than domestic Africa is shifting toward intraregional flows. While and global factors (as measured by the CBOE Volatil- total remittance inflows to sub-Saharan African coun- ity Index (VIX)) alone, while a 100 basis point change tries have remained constant at slightly over 2 per- in the South African sovereign spread is estimated cent of GDP over the past 10 years, the composition to be associated with a 20 basis point increase in the has shifted: remittances among sub-Saharan African average frontier market spread (Box 3 and Table 3.1, countries have grown faster than those from the rest columns 1 and 2). The fact that developments specific of the world in the past five years.9 Meanwhile, total to South Africa act as drivers of regional spreads sug- remittances are becoming relatively more important as gests that investors may have incomplete information other sources of external funding, such as aid and FDI, on other regional developments (for instance, owing decline (Figure 21.1). Regional remittances accounted to data availability constraints) and may proxy their for about 35 percent of the region’s total remittance portfolio allocation for the region on the basis of per- inflows in 2015 (Figure 21.2).10 ceptions of the South African economy. This has been Measured as a share of GDP, total remittance shown to be the case in other regions in which a single inflows in sub-Saharan Africa are larger than those in country is the dominant economy (Furceri, Jalles, and other emerging and developing regions (Figure 22). Zdzienicka 2016). The relative importance of intraregional inflows in Sub-Saharan African frontier market spreads are sub-Saharan Africa is the third highest, after the also influenced by emerging market trends. It is likely Commonwealth of Independent States (CIS), and the that the impact of the South African spread on other Middle East and North Africa. sub-Saharan frontier market spreads is driven by a For some countries in sub-Saharan Africa, regional confounding factor; namely, trends in emerging mar- remittance inflows account for a large share of national kets globally. However, in controlling for an index of income, setting the stage for a potentially high expo- emerging market bond spreads, the estimated impact sure to regional spillovers. In 27 of the 45 sub-Saharan of changes in South Africa’s spread remains positive African countries, regional remittance inflows exceed and significant (although, depending on the index, the interregional remittances. At the high end, Lesotho, impact may decrease in absolute value and be quan- Liberia, and Togo receive more than 5 percent of titatively less important than the emerging market GDP in remittances from other sub-Saharan Afri- trends), suggesting that movement in South African can countries. Given that remittances have been spreads explains a significant share of movement in shown to reduce macroeconomic fluctuations and sub-Saharan African frontier spreads. poverty and foster financial development, regional remittance flows can help redistribute resources from fast-growing countries to slower-growing ones. This The Changing Pattern of Remittance Flows factor was particularly helpful, for instance, in the case Regional remittances among sub-Saharan African of resource-intensive countries hit by the commodity countries are relatively large. They account for a third of price shock: Liberia, Mali, and Nigeria, with remit- total remittance inflows, and their share is growing in tance inflows of 8, 4, and 2 percent of GDP, respec- parallel with declining costs. Because of a high concen- tively (IMF 2016c; Gonzalez-Garcia and others 2016; tration of outflows from a few countries and the large Gupta, Patillo, and Wagh 2009). exposure of some recipient countries, these remittances Most remittance outflows from sub-Saharan African constitute an important spillover channel. In particu- countries are sent to other countries in the region. lar, Côte d’Ivoire and Ghana are important sources for Specifically, 31 out of 45 sub-Saharan countries send West Africa, while South Africa is the main source for more remittances to the region than to the rest of the Southern and East Africa. Econometric estimates suggest world, and three-quarters of total remittances from that growth spillovers between remittance partners are important but may be outweighed by growth spillovers 9The estimates used in this section are from official sources, which from trading partners, although trading and remittance are known to underestimate remittances. partners often coincide. 10Remittance inflows from other sub-Saharan Africa countries increased from 0.6 percent of regional GDP in 2010 to 0.8 per- cent in 2015. 22 International Monetary Fund | August 2018 ©International Monetary Fund. Not for Redistribution  Regional Spillovers in Sub-Saharan Africa: Exploring Different Channels Box 3. Sovereign Yield Spread Spillovers An examination of the movement in sovereign included in the original EMBIG index sample. X spreads of sub-Saharan frontier markets around major is country-specific controls, including inflation, the news events in South Africa shows noticeable responses exchange rate relative to the US dollar, and an index to such events. This is first studied via an ad hoc event of financial stress. analysis during the month of December 2015, a time The model is estimated in first differences follow- of high volatility in South African markets. Three ing standard tests that indicate the presence of a unit major events occurred: root in all the time series variables. Country fixed December 4 Ratings downgrade by Standard effects, ​​γ​  i​​​, are included to control for time-invariant & Poor’s country-specific heterogeneity. Standard errors are clus- December 9 Finance minister fired tered at the country level. The sample uses monthly December 1311 New finance minister appointed data from January 2012 to August 2017, and includes These three major events are plotted in Figure 3.1, all sub-Saharan African countries that have issued along with the sovereign spread of South African foreign-denominated debts on international finan- and sub-Saharan African frontier markets. All rates cial markets since 2012 (frontier markets). These are are normalized to zero on the date of the first event. Angola, Cameroon, Côte d’Ivoire, Ethiopia, Gabon, The chart shows 10 days before and 15 days after the Ghana, Kenya, Mozambique, Namibia, Nigeria, Sene- first event (and thus includes the December 9 and gal, South Africa, Tanzania, and Zambia. 14 events). It shows that there is a marked jump in The analysis augments IMF (2016) by taking advan- South Africa’s sovereign spread within 24 hours of tage of the panel dimension of the data. The estima- each announcement, which continues for up to 72 tion results for equation (1) are reported in Table 3.1, hours. Movement in other countries’ sovereign spreads and show that movements in the sovereign spreads of tend to be heterogeneous in the days leading up to sub-Saharan African frontier markets are associated the announcements, but they move in concert with with changes in the South African spread. This result is the South African spread immediately following the robust after controlling for additional global emerging announcements. This suggests that South Africa has a market factors. Across columns 2–5, the South African degree of influence over its regional peers via sovereign spread coefficient is positive and highly significant, financial markets. though its value halves and its level of significance A panel fixed effects model is estimated to study falls when emerging market factors are controlled the drivers of sovereign yield spreads. The model is for via the synthetic EMBIG index, suggesting that specified as follows: both global and South African factors are import- ant in explaining spreads across sub-Saharan Africa. ​sprea ​d​  it​​  = α + ​γ​  i​​  + ​β​  1​​  sprea ​d​  t−1,i​​  + ​β​  2​​ ​ZAF​  t​​  + ​ Additionally, based on the adjusted R2, the South β​  3​​  Globa ​l​  t​​  + ​β​  4​​ ​X​  it​​  + ​e​  it​​ (​ 1)​​ African factors combined explain 5 percent more of where spread refers to the difference between a coun- the variation in spreads (versus without them), while try’s foreign denominated bond yields with respect global emerging market factors explain only up to an to US Treasury bond yields and ZAF corresponds to additional 2 percent of the variation. The results are South African factors, including either the SAVI or consistent with the correlation analysis and indicate the sovereign spread. Global is global factors, includ- the importance of both regional and global emerging ing the VIX, oil prices, and—in columns 4 and 5 market-specific factors in driving sub-Saharan African of Table 3.1—either the Morgan Stanley Capital frontier market yields. International (MSCI) Emerging Markets index or a synthetic version of the Emerging Market Bond Index Global (EMBIG). The synthetic EMBIG spread is constructed with the same sample and weights as the actual EMBIG spread, excluding South Africa and those sub-Saharan African frontier markets that are 11The announcement was made on Sunday, December 13, 2017. Since markets are closed on Sunday, the observed movement in spreads occurred on Monday December 14, 2017, which is when the indicator is identified in Figure 3.1. International Monetary Fund | August 2018 23 ©International Monetary Fund. Not for Redistribution SPILLOVER NOTES Box 3. Sovereign Yield Spread Spillovers (continued) Box Figure 3.1. South African News and Sovereign Spreads Angola Ethiopia Gabon Ghana Côte d’Ivoire Kenya Mozambique Namibia Nigeria Senegal Tanzania South Africa Zambia 200 150 100 Level change 50 0 –50 –100 –150 –10 –9 –8 –7 –6 –5 –4 –3 –2 –1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Days Source: Bloomberg LP. Table 3.1. Impact of Global, Regional, and Domestic Factors on Sovereign Spreads, 2012–16 Dependent Variable: Country spread (t) (1) (2) (3) (4) (5) Country spread (t–1) 20.20* 20.24** 20.21* 20.24** 20.23** (0.10) (0.09) (0.10) (0.09) (0.09) FSI 0.02 20.03 0.00 20.03 20.03* (0.01) (0.02) (0.01) (0.02) (0.02) Oil Price 20.25*** 20.16* 20.19** 20.15* 20.03 (0.07) (0.08) (0.07) (0.08) (0.07) VIX 0.12*** 0.12*** 0.12*** 0.10*** 0.10** (0.03) (0.03) (0.03) (0.03) (0.03) Inflation 20.07 20.07 20.06 20.07 20.04 (0.07) (0.07) (0.07) (0.07) (0.07) Exchange Rate 0.51** 0.45*** 0.48** 0.47*** 0.17 (0.18) (0.14) (0.20) (0.14) (0.16) South Africa Spread 0.20*** 0.20*** 0.10* (0.04) (0.04) (0.05) SAVI 0.06*** (0.01) MSCI 0.03 (0.03) Synthetic EMBIG 0.16*** (0.03) Constant 20.00 20.00 0.00 20.00 0.01* (0.00) (0.00) (0.00) (0.00) (0.00) FE Yes Yes Yes Yes Yes N 641 641 641 641 626 R2 (adj) 0.25 0.31 0.26 0.30 0.32 Source: Bloomberg LP; Chicago Board Options Exchange; IMF, International Financial Statistics; and staff estimates, and calculations. Note: All variables are standardized and estimated in first differences. Clustered standard errors in parentheses. *p  0.1, **p  0.05, ***p  0.01. 24 International Monetary Fund | August 2018 ©International Monetary Fund. Not for Redistribution  Regional Spillovers in Sub-Saharan Africa: Exploring Different Channels Figure 21. Sub-Saharan Africa: External Flows and Remittances 1. Selected External Flows 2. Remittance Inflows 5 40 12 Remittances inflows Share in total SSA inflows (LHS) Aid (net official development 35 Amount of SSA regional assistance and official aid) remittances (RHS) 4 Foreign direct investment 30 Billions of US dollars 10 Percent of GDP 25 Percent share 3 20 2 15 8 10 1 5 0 0 6 2005 06 07 08 09 10 11 12 13 14 15 2010 11 12 13 14 15 Sources: IMF, World Economic Outlook database; and World Bank, World Development Indicators, Migration and Remittances database. Note: Remittances in the World Bank databases are measured as the sum of three items in the IMF’s Balance of Payments Statistics Year Book: (1) personal transfers, (2) compensation of employees, and (3) migrants’ transfers (that is, capital transfers between resident and nonresident households). LHS = left hand side; RHS = right hand side. sub-Saharan African countries are sent to other coun- Figure 22. Remittance Inflows in Emerging and Developing tries in the region.12 Countries 2010–15 Remittance outflows originate in a few countries. 1.5 The four largest senders in 2015 accounted for 50 per- From the region From rest of the world cent of total regional remittances. Remittances from Chad, Cameroon, Côte d’Ivoire, and Ghana to Nigeria alone account for 50 percent of received remittances 1.0 Percent of GDP in the region. Côte d’Ivoire and Ghana are also a large source of remittance flows to other West African countries (Figures 23 and 24). Their ongoing growth 0.5 spell has translated into steady remittance flows to the region and has contributed to economic growth in the subregion. Remittance outflows from South Africa, the other large sender, are spread across the region, making 0.0 SSA Asia Europe America MENA CIS the country of regionwide importance. In general, remittances in sub-Saharan Africa are Sources: World Bank, Migration and Remittances database, World Development Indicators; and IMF staff calculations. exchanged among the largest, wealthiest, and closest Note: CIS = Commonwealth of Independent States; MENA = Middle East and economies, but on a net basis, they flow toward poorer North Africa. SSA = sub-Saharan Africa. and more connected countries, making them more exposed to spillovers. A standard gravity equation on the 2010–15 average remittance flows shows that sending costs, especially relative to migrants’ incomes. remittances are larger for geographically and cultur- Regression results also show that higher sending costs ally close countries (Box 4 and Table 4.1, columns 1 are associated with lower remittance flows (Table 4.1, and 2). Compared with other regions in the world, column 4), even after controlling for distance and by geographical distance seems to be a greater barrier origin and destination fixed effects. in sub-Saharan Africa because of higher travel and 12Exceptions include East African countries whose remittance outflows are generally to India and China and certain francophone countries whose remittance outflows go to France. International Monetary Fund | August 2018 25 ©International Monetary Fund. Not for Redistribution SPILLOVER NOTES Figure 23. Sub-Saharan Africa: Remittance Outflows and Inflows, 2010–15 1. Share of Total Intraregional Remittance Outflows 18 Percent of total regional outflows 16 14 12 10 8 6 4 2 0 CMR CIV TCD GHA ZAF BEN GAB NGA NER KEN BFA TGO GMB SSD TZA 2. Average Remittances Outflows 30 Intraregional Interregional 25 20 Percent of GDP 15 10 5 0 GMB BEN TGO TCD CMR NER LBR CIV GAB SSD BFA GIN CAF GNB RWA 3. Average Remittances Inflows 25 Intraregional Interregional 20 Percent of GDP 15 10 5 0 LSO LBR TGO MLI SEN UGA STP BEN GMB GNB CPV NGA COM KEN MUS Source: World Bank, Migration and Remittances database. Note: The Gambia stands out as a country with large inflows and outflows of remittances. Large inflows from non-sub-Saharan African countries can be explained by the large diaspora (about 5 percent of the population lives abroad, and two out of three Gambians who graduate from foreign universities stay abroad). Large outflows are essentially directed to Senegal, which surrounds the country. Anecdotal evidence suggests that these outflows originate from the many Gambians who have family members in Senegal and those who seek better health care and transportation means in Senegal, given the superior quality of Senegalese infrastructure. See page 41 for country abbreviations table. 26 International Monetary Fund | August 2018 ©International Monetary Fund. Not for Redistribution  Regional Spillovers in Sub-Saharan Africa: Exploring Different Channels Figure 24. Sub-Saharan Africa: Major Remittance Corridors, 2010–15 SEN KEN TCD MLI BWA LBR BEN NGA CIV SWZ ZAF BFA GHA LSO GIN TGO MOZ CMR NER Sources: World Bank, Migration and Remittances database, World Development Indicators. Note: Averages for top five remittance senders, 2010–15 (in red) (Chad, Cameroon, Ghana, Cote d’Ivoire, and South Africa). Thickness of the arrow represents the size of remittance flow in US dollars. Flows of less than US$10 million were censored. Nigeria receives substantial remittances through the compensation of Nigerians working in neighboring countries. Only officially recorded remittances sent through formal channels are recorded, which explains why some countries known for receiving large remittance inflows—such as Zimbabwe and Chad—are not included in the graph. See page 41 for country abbreviations table. Box 4. Gravity Equation Estimation for 2010–15 Remittance Flows A gravity model of bilateral remittance flows over are dropped, as these characteristics are absorbed by the period 2010–15 is estimated to study the determi- the country fixed effects. The new specification is as nants of remittances. In all specifications, the sample follows: includes annual data over 2010–15 from the World Bank Migration and Remittances database with all the ​​log​F¯ ​​  i,j​​  = ​α​  i​​  + ​γ​  j​​  + β ​Xij​  ​​  + ​ε​  ij​​ . (​ 2)​​ country pairs in the world that exchanged remittances Estimation results in the first two columns of at least once. All variables are simple averages over the Table 4.1 demonstrate that all distance measures six-year period. The standard errors are clustered at the significantly hamper remittance flows across countries. destination country level to control for possible unob- The first column additionally shows that remittance served correlation within receiving countries. flows increase significantly with both population and The first specification is as follows: the GDP per capita of both the origin and destina- tion country. It does, however, increase more with the ​​log​F¯ ​​  ij​​  = ​α​  0​​  + β ​Xij​  ​​  + γ ​​Y¯ ​​  i​​  + δ​ ¯ ​Zj​  ​​​  + ​ε​  ij​​   ​(1)​​ GDP per capita of the origin country, implying that where ​​​F¯ ​​  ij​​​refers to the logarithm of the average remit- net flows toward a poorer country are increasing with tance flow from country i to country j; ​​Xij​  ​​​ corresponds the difference in GDP per capita. to corridor-specific independent variables, including The specification in column 3 adds interaction geographic distance, the exchange rate between the variables between measures of distance and dummy currency of the origin and destination countries, and variables. These variables are introduced for either dummy variables indicating whether the countries sub-Saharan Africa origin countries (​​I​ i∈SSA​​ ​Xij​  ​​​) or share a common official language, share the same sub-Saharan Africa destination countries (​​I​ j∈SSA​​ ​Xij​  ​​​): ethnic group, had the same colonial origin, or share a ​​log​F¯ ​​  ij​​  = ​α​  i​​  + ​γ​  j​​  + β ​X​  ij​​  + ​θ​  o​​ ​I​  i∈SSA​​ ​X​  ij​​  + ​θ​  d​​ ​I​  j∈SSA​​ ​X​  ij​​  + ​ common official religion; and Y ​​​¯ ​​  i​​​ and ¯ ​Z ​​  j​  ​​​​ which, respec- ε​  ij​​ . (​ 3)​​ tively, refer to the logarithms of the average GDP per capita and the average population of the origin and This specification is appropriate to investigate destination countries. whether distance plays a specific role for remittance The specification in column 2 includes country fixed flows within sub-Saharan Africa. Coefficient estimates ​​ effects, ​​α​  i​​​ and γ​​ ​  j​​​, to control for origin country i and θ​  o​​​ and θ​​ ​  d​​​of these interaction variables, respec- destination country j time-invariant characteristics. tively, reflect the differential effects of distance for The corridor variables ​​X​ ij​​​in equation (1) are kept but sub-Saharan Africa origin and destination countries. the country-level characteristics (population and GDP) International Monetary Fund | August 2018 27 ©International Monetary Fund. Not for Redistribution SPILLOVER NOTES Box 4. Gravity Equation Estimation for 2010–15 Remittance Flows (continued) Column 3 is composed of three subcolumns; the number of days it took for the money to be trans- coefficient estimates θ​ ​​ o​​​are reported in the second sub- ferred. Only a small subset of remittance corridors column, while estimates ​​θ​  d​​​are reported in the third. is covered by the Remittance Price database, which The results indicate that distance is a greater hindrance explains the smaller sample size, the higher R2, and the for both origin and destination sub-Saharan countries, weaker significance of the results. Despite these issues, except for those that belonged to the same colony. the results indicate that costs significantly reduce In column 4, the specification reverts to equation bilateral flow. (2) but includes two supply-side variables to the set of independent variables: the median cost of sending US$200 in percent of the US$200 and the median Table 4.1. Determinants of Average Remittances Flows Dependent variable: the logarithm of the average bilateral remittance flow (3) (1) (2) Coefficient estimates for (4) Country controls Country FE All countries SSA origin SSA destination Costs Contiguous countries 2.61*** 2.40*** 2.09*** 0.56 0.94 (0.14) (0.16) (0.19) (0.36) (0.80) Distance (1000 km) 20.25*** 20.24*** 20.22*** 20.19*** 20.11*** 20.01 (0.01) (0.01) (0.01) (0.03) (0.02) (0.05) Common language 1.52*** 1.07*** 0.93*** 20.11 0.54** 20.05 (0.14) (0.17) (0.21) (0.33) (0.24) (0.41) Common ethnicity 0.64*** 0.48*** 0.65*** 0.09 20.45* 0.53 (0.14) (0.16) (0.19) (0.34) (0.23) (0.51) Belonged to common 1.58*** 1.50*** 1.59*** 20.94*** 0.57* 1.22*** colony (0.14) (0.22) (0.22) (0.31) (0.31) (0.39) Common religion 0.11* 0.61*** 0.70*** 20.14 20.11 20.20 (0.07) (0.10) (0.11) (0.34) (0.26) (0.48) Origin GDP p.c. 1.04*** (0.02) Destination GDP p.c. 0.63*** (0.02) Origin population 0.67*** (0.01) Destination population 0.82*** (0.01) Origin/destination FX rate 20.02** (0.01) Median costs (% of 20.10* amount sent) (0.06) Median completion time 20.15 (days) (0.12) Observations 10,704 10,814 10,814 220 R-squared 0.54 0.81 0.81 0.93 Country FE NO YES YES YES Sources: French Centre d’Etudes Prospectives et d’Informations Internationales (CEPII) database; IMF, World Economic Outlook database; World Bank, World Development Indicators; and remittances and migration database. Note: Clustered standard errors in parentheses (Destination country). ***p  0.01, **p  0.05, *p  0.1. 28 International Monetary Fund | August 2018 ©International Monetary Fund. Not for Redistribution  Regional Spillovers in Sub-Saharan Africa: Exploring Different Channels Figure 25. Percentage Cost of Sending US$200 across Region and over Time (Percentage of the amount sent) Global East Asia & Pacific Europe & Central Asia Latin America & Caribbean Middle East & North Africa South Asia 13 Sub-Saharan Africa 12 11 10 Percent 9 8 7 6 5 2011:Q1 11:Q3 12:Q1 12:Q3 13:Q1 13:Q2 13:Q3 13:Q4 14:Q1 14:Q2 14:Q3 14:Q4 15:Q1 15:Q2 15:Q3 15:Q4 16:Q1 16:Q2 16:Q3 16:Q4 17:Q1 17:Q2 Source: World Bank, Remittance Prices Worldwide, Issue 22, June 2017. Growth in countries sending remittances spills Developments in financial technologies (fintech) over to receiving countries. Regression results show have the potential to increase the magnitude of remit- that a 1 percent increase in GDP growth in origin tance flows within the region, strengthening them as countries is associated with a 0.1 percent increase in a channel for regional spillovers. Sub-Saharan Africa growth in a receiving country (Box 5). Interestingly, is the most expensive destination to send money to the result only holds for origin countries that belong (Remittance Price Database Report 2017) (Figures to the same region, perhaps reflecting that interre- 25 and 26). Remittance costs in 2017 were about gional remittances are dominated by those sent from 25 percent higher there than in the rest of world. advanced economies and that their variations are too However, these costs have been decreasing for the past small to have an impact on the receiving countries 10 years, partly because of the rise of mobile money through the remittance channel. For many countries, technology. Mobile money transfers are two times less remittance and trade partners tend to be the same, and expensive than those at money transfer operators and data limitations do not allow us to accurately distin- post offices and almost three times less expensive than guish spillovers occurring through one channel rather transfers through commercial banks. As mobile money than the other.13 However, estimates suggest that both technology continues to expand, and its coverage and spillovers have a similar magnitude, each accounting usage continue to increase across sub-Saharan Africa, it for half of the total effect identified in the baseline is expected to contribute to an increase in remittance estimation. Overall, regression estimates suggest that flows. On the basis of estimates presented in Box 4, growth performance of large origin countries such as and assuming there is no substitution across corridors, Côte d’Ivoire, Ghana, and South Africa can generate a decline in remittance costs to the world average growth spillovers to the rest of the region. (from 9.4 percent to 7.4 percent) could result in increases in bilateral flows of up to 20 percent. 13The correlation between the share of imports and the share of remittance flows across partner countries varies substantially. The median correlation is about 50 percent, but for some countries with correlations close to 1, like Swaziland and Lesotho, trade and remit- tance partners are almost the same (South Africa, in this case). International Monetary Fund | August 2018 29 ©International Monetary Fund. Not for Redistribution SPILLOVER NOTES Figure 26. Total Average Cost by Remittance Sending Provider (Percentage of the amount sent) 12 10 8 Percent 6 4 2 0 Bank Money transfer operator Post office Mobile operator Source: World Bank, Remittance Prices Worldwide, Issue 22, June 2017. The Foreign Direct Investment Channel—South Figure 27. Selected Sub-Saharan African Countries: Africa Rules the Roost Intraregional Foreign Direct Investment Stock Positions Foreign direct investment constitutes an important 70 Percent of total Inward 40 channel of regional integration in sub-Saharan Africa. Percent of total inward direct investment direct investment (LHS) 35 60 Percent of GDP (RHS) For some countries, inward FDI from sub-Saharan 30 Africa constitutes the largest share of total inward FDI. 50 This is the case in Togo, Rwanda, Guinea-Bissau, and Percent of GDP 25 40 Botswana, where the share of regional inward FDI 20 positions is more than 40 percent of their total stock 30 15 of FDI (Figure 27). 20 Firms from South Africa, Kenya, and Nigeria have 10 the largest presence in other sub-Saharan markets. 10 5 South African firms are the most visible, with more 0 0 than 2,400 subsidiaries in other African countries TGO GNB SYC MOZ BEN ZMB MLI MUS NER ZAF CPV RWA BWA UGA BFA NGA (Box 6), but other hubs are growing fast, such as Kenya (with an important presence in East Africa) Source: IMF, Coordinated Direct Investment Survey database. and Nigeria (in West Africa). Other subregions, such Note: See page 41 for country abbreviations table. as central Africa, have limited cross-border corporate ownership (IMF 2017c). Multinational firms on the continent can be import- Kenya, Tanzania, and Rwanda) compensated for the ant sources and transmitters of positive international weaker performance of the headquarters (IMF 2017b). spillovers. Firms’ potential to affect both their host and The risks of negative cross-border spillovers are headquarters economies increases with their size and equally important. Multinational firms can increase interconnectedness. Regional integration at the corpo- national exposure to negative shocks across borders. rate level allows firms to leverage knowledge transfer The risks are many and varied and include the follow- and country-level comparative advantages and achieve ing four examples: (1) If a firm is systemically import- diversification, all while tapping into economies of ant for an economy, performance of its headquarters scale. For instance, during the recent economic slow- or foreign subsidiaries could have macroeconomic down in Nigeria, the strong performance of foreign implications for the host country; (2) A firm’s ability subsidiaries in high-growth countries (for example, to borrow can be affected by its exposure to sovereign risk, as surges in sovereign spreads often lead to 30 International Monetary Fund | August 2018 ©International Monetary Fund. Not for Redistribution  Regional Spillovers in Sub-Saharan Africa: Exploring Different Channels Box 5. Spillover Effects from Countries Sending Remittances A panel fixed effects model is estimated to study the variables are then included to test whether sub-Saharan elasticity of GDP growth rates to the average growth countries behave differently than other countries in the among remittance partners inside and outside a region. world. The model is specified as follows: Regressions results are reported in Table 5.1. Start- ing from column 2, the specifications include controls ​​RealGDPgrowth​ it​​  = α + ​γ​  i​​  + ​β​  1​​ ​RealGDPgrowth​ t−1,i​​  + ​ for both the average growth of the sub-Saharan Africa β​  2​​  regionalremitpartners’growth​​+ region and for world growth, to ensure that coeffi- ​β​  3​​  extraregionalremitpartners’growth + cients are not capturing average co-movements in  ​β​  4​​ ​Xit​  ​​  + ​β​  5​​ ​Ii∈SSA ​  ​​  * regionalremitpartners’growth​​​+ continent or world developments. In columns 3 and  ​β​  6​​ ​Ii∈SSA ​  ​​  * extraregionalremitpartners’growth + ​e​  it​​    ​(​​6​)​​​​ 4, the lag of the dependent variable is introduced to where the dependent variable is the annual growth address serial correlation issues, and an Arellano-Bond rate of real GDP in country i; regional remittance estimation procedure is adopted to address the possible partners’ growth is the average growth rate of the remit- endogeneity of that variable. Results are robust to tance partners that are in the same region as country the exclusion of the 10 percent largest economies (in i; extraregional remit partners’ growth corresponds to column 4) and to the introduction of interaction vari- the average growth rate of the remittance partners that ables capturing heterogeneous effects (in column 5). are outside the region; and ​​I​ i∈SSA​​​is a dummy variable In the latter specification, the share of remittances in for sub-Saharan African countries that is interacted GDP is interacted with the remittance partners’ aver- with the last two variables. The growth averages are age growth to test whether more-exposed economies weighted using the share of lag remittance inflows. The are more affected. These new variables are not signifi- sample includes 2010–15 annual data from the World cant. In column 6, the specification also includes the Bank Migration and Remittances database for all growth averages of the trade partners that are inside countries in the world, as well as all the data included and outside the region of the country considered. The in Box 1. purpose of these controls is to distinguish between All specifications include the same standard spillover effects that are channeled through trade and country-specific controls that are included in those channeled through remittance inflows. Table 1.1, column 5 when studying spillovers from The estimation results indicate that growth spillovers the trade channel. Country fixed effects ​​γ​  i​​​, are also through the remittance channel are significant for all included to control for time-invariant country-specific countries in the world and that these spillovers are heterogeneity. Coefficient estimates have the expected not different in sub-Saharan Africa (columns 1–3). signs, but results are omitted from Table 5.1 for pre- The results are robust to controlling for the average sentational purposes. The share of regional remittances regional growth and serial correlation (columns 2–3). in total remittance inflows and the share of remittance Column 4 results suggest that some of the spillovers inflows in GDP are added as controls. Standard errors associated with the remittance channels may capture are robust, clustered at the country level to control for trade spillovers. The joint significance of the average possible unobserved correlation within countries. growth of trade and remittance partners is confirmed In contrast with the regression using trade data, by a Wald test with a p-value below 0.02. Despite remittance flow series are much shorter, spanning the weaker significance of each estimate separately, 2010 to 2015, which reduces efficiency in the estima- estimated values suggest that both channels have equal tion. Given the limited degrees of freedom, a lagged importance and that half of the observed spillovers can dependent variable is not included as a control. In be attributed to each channel. The lack of significance addition, as opposed to the baseline specification in of the remittance channel in that last specification Box 1, all countries of the world are included in the may be due to the small sample size and the imperfect regression to increase statistical power. Interaction measurement of remittance flows. International Monetary Fund | August 2018 31 ©International Monetary Fund. Not for Redistribution SPILLOVER NOTES Box 5. Spillover Effects from Countries Sending Remittances (continued) Table 5.1. Spillover Effects from Countries Sending Remittances Dependent variable: Real GDP growth (1) (2) (3) (4) (5) (6) Baseline World/regional GMM Excluding Exposure Trade channel shock controls estimation 10% largest heterogeneity controls Real GDP growth (t-1) 0.166 0.170 0.168 0.174 (0.123) (0.134) (0.123) (0.122) Regional remittance partners’ growth 0.0917** 0.0587 0.0999** 0.0971* 0.109** 0.0588 (0.0365) (0.0431) (0.0503) (0.0516) (0.0450) (0.0692) Extraregional remittance partners’ 20.129 0.101 20.0114 20.0907 20.0378 20.0976 growth (0.179) (0.240) (0.234) (0.261) (0.341) (0.234) Regional remittance partners’ growth, 20.0849 20.0700 20.00218 0.0222 20.0804 0.0239 SSA differential (0.129) (0.127) (0.144) (0.146) (0.194) (0.142) Extraregional remittance partners’ 20.366 20.651 20.514 20.430 20.528 20.366 growth, SSA differential (0.376) (0.393) (0.414) (0.417) (0.542) (0.459) Conflict year, Uppsala database 24.980** 24.878** 25.081*** 25.327*** 25.276*** 25.032*** (1.941) (1.899) (1.283) (1.104) (1.323) (1.250) Share of regional remittances in total 20.0356 20.0347 20.0523** 20.0587** 20.0536** 20.0515** inflows (t-1) (0.0257) (0.0254) (0.0256) (0.0276) (0.0270) (0.0251) Share of remittances inflows in 20.0626 20.0250 0.127 0.136 0.0728 0.128 GDP (t-1) (0.0900) (0.0958) (0.118) (0.123) (0.178) (0.119) Region average growth 0.424*** 0.485*** 0.501*** 0.500*** 0.474*** (0.138) (0.164) (0.165) (0.172) (0.162) World average growth 20.601 20.490 20.972 20.510 21.056 (0.971) (0.949) (1.061) (0.953) (1.006) Share of regional imports in total 20.633 20.245 0.561 0.206 0.547 0.265 inflows (t-1) (3.060) (3.140) (3.203) (3.110) (3.127) (2.994) Trade openness (t-1) 0.0291 0.0183 0.0156 0.0101 0.0136 0.00987 (0.0238) (0.0261) (0.0334) (0.0342) (0.0333) (0.0331) Percent change in the terms of trade 0.0918 0.0948 0.196 0.237 0.208 0.205* (0.0949) (0.0974) (0.121) (0.152) (0.128) (0.121) Investment share of GDP (t-1) 0.0601 0.0698 0.0672 0.0737 0.0631 0.0742 (0.0948) (0.0966) (0.0955) (0.0984) (0.0939) (0.0951) Percent change in population 0.524* 0.615** 0.751** 0.726** 0.708** 0.687** (0.286) (0.297) (0.323) (0.350) (0.324) (0.306) Percent change in US Federal Funds 0.0100** 0.00718 0.00588 0.00741 0.00588 0.00624 rates (0.00482) (0.00471) (0.00405) (0.00470) (0.00410) (0.00403) Inflation 20.152 20.180 20.225* 20.223* 20.224* 20.223* (0.127) (0.125) (0.121) (0.135) (0.124) (0.121) Inflation (t-1) 0.0507 0.0290 20.0131 20.0299 20.0143 20.00990 (0.0757) (0.0732) (0.0681) (0.0763) (0.0679) (0.0680) Percent change in the foreign exchange 1.141 2.383 2.371 4.949 2.144 2.233 rate (3.504) (3.620) (3.278) (4.181) (3.321) (3.270) Percent change in the foreign exchange 20.886 0.817 3.334 4.065 3.409 3.128 rate (t-1) (2.628) (2.442) (2.450) (2.765) (2.472) (2.454) Regional remittance partners’ growth 21.222 interaction with the lag share of (0.943) remittance inflows Extraregional remittance partners’ 0.523 growth interaction with the lag share of (4.190) remittance inflows Regional remittance partners’ growth 6.046 interaction with the lag share of (6.585) remittance inflows, SSA differential Extraregional remittance partners’ 0.308 growth interaction with the lag share of (9.081) remittance inflows, SSA differential Regional trading partners’ growth 0.0497* (0.0302) Extraregional trading partners’ growth 0.480* (0.255) 32 International Monetary Fund | August 2018 ©International Monetary Fund. Not for Redistribution  Regional Spillovers in Sub-Saharan Africa: Exploring Different Channels Box 5. Spillover Effects from Countries Sending Remittances (continued) Table 5.1. Spillover Effects from Countries Sending Remittances (continued) Dependent variable: Real GDP growth (1) (2) (3) (4) (5) (6) Baseline World/regional GMM Excluding Exposure Trade channel shock controls estimation 10% largest heterogeneity controls Regional trading partners’ growth, SSA 20.0152 differential (0.329) Extraregional trading partners’ growth, 20.470 SSA differential (0.427) Constant 1.582 1.888 0.659 2.520 1.374 1.861 (2.872) (3.682) (3.687) (3.810) (3.833) (3.790) Observations 565 565 448 393 448 448 R-squared 0.161 0.181 Number of countries 117 117 117 103 117 117 Sources: French Centre d’Etudes Prospectives et d’Informations Internationales (CEPII) database; IMF, World Economic Outlook database; World Bank, World Development Indicators; and remittances and migration database. Note: Robust standard errors in parentheses. ***p  0.01, **p0.05, *p  0.1. Box 6. South African Investment in Sub-Saharan Africa South African companies have become an increas- South African investment is highly regarded and an ingly important source of investment in sub-Saharan important driver of growth. Survey evidence suggests Africa since the mid-1990s, having expanded through that about 80 percent of sub-Saharan Africans who joint ventures, greenfield investments, and mergers and interact with South African firms find them to have a acquisitions (Table 6.1) (IMF 2012; Games 2017). better reputation than local firms in the same industry About 75 percent of investment from South Africa to (DNA Economics 2013). Further empirical evidence the continent is in the services, trade, and financial shows that outward foreign investment from South sectors (IMF 2016c). Africa has significant impacts on local growth rates in Outward foreign direct investment (FDI) from the southern continent, in particular the rates of con- South Africa is increasing. The total stock of FDI from vergence to South African GDP per capita. Countries South Africa to sub-Saharan African countries was with a high stock of South African FDI converge more equivalent to 6.8 percent of South African GDP in rapidly to South African per capita income levels, 2015, up from 4.9 percent of GDP in 2010 (Fig- while countries with low bilateral FDI stocks vis-à-vis ure 6.1). In receiving countries, South Africa’s invest- South Africa show no evidence of convergence (Dunne ments represented as much as 3.2 percent of GDP (in and Masiyandiam 2015). Thus the deceleration of the Mauritius), with an average of 0.4 percent across all South African economy in recent years could spill over sub-Saharan African countries in which it invested in to other countries that have large stocks and flows of 2015. South African FDI and could manifest as both lower FDI and lower GDP growth in these countries. International Monetary Fund | August 2018 33 ©International Monetary Fund. Not for Redistribution SPILLOVER NOTES Box 6. South African Investment in Sub-Saharan Africa (continued) Box Figure 6.1. South Africa: Outward FDI in Sub-Saharan Africa (Percent of GDP) 8.00 7.00 6.00 5.00 4.00 2010 11 12 13 14 15 Source: IMF, Coordinated Direct Investment Survey database. Table 6.1. South Africa: Major Multinationals Sub-Saharan Africa Reach: Global Reach (Number of countries with (Total number of countries Name Sector operations, excluding South Africa) with operations) Bidcorp Food services 16 41 First Rand Financial services 8 16* MTN Mobile phone operator 16 21 Naspers Internet and media 37 130 Sanlam Financial services 12 18 Sasol Chemical 7 33 Shoprite Groceries, furniture, restaurants, pharmaceutical, 15 15 logistics, property, hospitality, ticketing, liquor, money markets Standard Bank Financial services 16 30 Group Note: *FirstRand has eight full service banking locations in Africa. It has one investment banking licence (Nigeria) and two representative offices (Angola and Kenya) on the continent, plus five branches and representative offices globally. increases in corporate spreads; (3) Financial channel could reduce the attractiveness of such countries as an spillovers can manifest as counterparty spillovers, where FDI destination. the default of a firm causes financial distress for its foreign creditors, shareholders, or parent (Jorion and Zhang 2009); and (4) The imposition of restrictions, The Fiscal Channel—The Role of Unintended such as those on foreign exchange, can hurt local and Consequences foreign-owned firms by disrupting their production Spillovers can arise from large fluctuations in fiscal processes and revenue remittances. The subsequent revenues in the context of customs unions and from negative performance of the local subsidiary could commodity pricing policies in neighboring countries. These pose risks for the parent company (IMF 2017c), which 34 International Monetary Fund | August 2018 ©International Monetary Fund. Not for Redistribution  Regional Spillovers in Sub-Saharan Africa: Exploring Different Channels Box 7. SACU Revenue-Sharing Formula All customs and excise revenues collected in the imports to the country from all other SACU members, Southern African Customs Union (SACU) coun- less re-exports, divided by the value of imports less tries are pooled and managed by the South African re-exports for all SACU countries; ​​y​ i​​​is the share of Revenue Fund, then distributed to member countries GDP of the country in the SACU GDP; and ​​h​ i​​​ the according to a revenue-sharing formula. The 2002 level of GDP per capita in the country divided by the agreement describes how revenues are distributed: average across SACU members. •• One hundred percent of customs revenue is distrib- Given the structure of the formula, fluctuations uted on the basis of intra-SACU imports. in customs revenues have a larger impact on smaller •• Eighty-five percent of excise revenue is distributed countries but are partially compensated by transfers on the basis of members’ GDP. associated with the development component (Bas- •• Fifteen percent of excise revenue is distributed devant 2012). equally through a development component, with an Another source of volatility in the revenues stems adjustment inversely proportional to the member’s from the forecasting and adjustment mechanism GDP per capita. (Honda and others 2017). Transfers in any given year As described in Basdevant (2012), the revenue (R) correspond to the forecasted value a year earlier, and sharing formula is any discrepancies between forecast and actuals are compensated the following year. Cuevas and others ( ) ​h​  ​​  − 1 ​​​Ri​  ​​  = ​a​ i​​  C + 0.85 ​y​ i​​  E + 0.15 ​ _1 ​  E​ ​​1– ​ _ i 10  ​​ ​​​​ (2012) argue that this approach increases the predict- 5 ability of the revenues in the short term but over the where C refers to custom duties and E to excises. medium term increases the variance of SACU transfers For each country i, ​​a​ i​​​is the value at the border of beyond the variance of the underlying revenue pool. developments have an impact on economic activity and with the procyclical structure of the sharing formula, fiscal sustainability. creates strong economic links between SACU member countries and can complicate the conduct of fiscal pol- icy and macroeconomic management (Basdevant 2012; Importance of Customs Unions Honda and others 2017). For countries that share a customs union, fluc- Given the relative importance of the South Afri- tuations in trade revenues constitute an important can economy in the union, the state of its economic spillover channel. In the case of the SACU, member cycle is the main driver of the total amount of union countries pool their trade-related revenues and excises receipts to be shared. Fluctuations in South Africa’s and distribute them following a proportionality rule consumption and imports have a large impact on (Box 7). Fluctuations in SACU revenues have a larger total SACU trade-related and excise revenues. Indeed, impact on the fiscal revenues of smaller member coun- South African imports alone generate more than tries because they represent a relatively larger portion 90 percent of the SACU revenue pool (Basdevant of their total public revenues. For example, in Swa- 2012; IMF 2012). During periods of economic ziland and Lesotho, SACU revenues constitute more expansion in South Africa, when both total demand than 40 percent and 50 percent of total public reve- and import demand increase, there is an increase nues, respectively (Figure 28, panel 1) (IMF 2016c). in total union-wide SACU revenues. On the other The volatility of customs duties and excises can hand, SACU revenues experience large declines when have substantial economic and fiscal impacts. SACU economic conditions deteriorate in South Africa, as receipts as a share of government revenues or GDP was the case during the global financial crisis and more can be highly volatile in the case of the smallest SACU recently as growth slowed in South Africa (Figure 28, members. For instance, in Lesotho and Swaziland, panel 3). Declines in SACU revenues have led to sharp the standard deviation of receipts as a share of GDP deteriorations in the fiscal position of SACU member was between 5 percent and 7 percent for the period countries, with Lesotho and Swaziland most affected 2000–16 (Figure 28, panel 2). This volatility, together (Figure 28, panel 4). The deterioration was particularly International Monetary Fund | August 2018 35 ©International Monetary Fund. Not for Redistribution  SPILLOVER NOTES Figure 28. SACU Revenues and Selected Macroeconomic Indicators 1. SACU Revenues, 2000–16 2. Volatility of SACU Revenues, 2000–16 60 8 Percent of total government revenue Percent of GDP, standard deviation 7 50 6 40 5 30 4 3 20 2 10 1 0 0 Swaziland Lesotho Namibia Botswana South Africa Lesotho Swaziland Botswana Namibia South Africa 3. SACU Revenues and South Africa Growth, 2000–16 4. Overall Fiscal Balance in SACU, 2006–16 SACU revenues, maximum and minimum South Africa Botswana 60 Median 6 16 Lesotho Namibia 8 South Africa growth (RHS) 5 12 Swaziland South Africa 50 growth (RHS) Change, percent of GDP 4 8 4 40 Percent of GDP 3 4 30 2 0 0 1 –4 20 0 –8 –4 10 –1 –12 0 –2 –16 –8 2000 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 2006 07 08 09 10 11 12 13 14 15 16 5. Fiscal Balance, 2016 6. Current Account Balance, 2016 0 30 Current account excluding SACU transfers –5 20 Current account balance –10 10 Percent of GDP Percent of GDP –15 0 –20 –10 –25 Fiscal balance –20 –30 Fiscal balance excluding –35 SACU transfers –30 –40 –40 Lesotho Swaziland Namibia Botswana South Africa Lesotho Namibia Swaziland South Africa Botswana Sources: IMF, country team databases, World Economic Outlook. pronounced in 2016, when the overall fiscal balance in revenues constitute an important source of foreign Lesotho and Swaziland worsened by 9.5 percent and exchange inflows and contribute to the stock of 6 percent of GDP, respectively. international reserves (Honda and others 2017). For Given the contribution of SACU revenues to fiscal example, in Namibia, these transfers are an important and external deficits, spillovers originating from rev- source of foreign income, with the current account enue fluctuations can have considerable implications (excluding SACU transfers) at 31 percent of GDP for fiscal and external sustainability. For instance, (Figure 28, panel 6). In the context of the Common in Lesotho, SACU revenues are critical to fiscal Monetary Area—in which Lesotho, Namibia, and sustainability—in 2016 the fiscal deficit excluding Swaziland have pegged their currency to the South regional revenues was 34 percent of GDP (Figure 28, African rand—fiscal sustainability has direct implica- panel 5). In addition to the fiscal implications, SACU tions for external stability and the sustainability of the 36 International Monetary Fund | August 2018 ©International Monetary Fund. Not for Redistribution  Regional Spillovers in Sub-Saharan Africa: Exploring Different Channels Figure 29. Differentials between Nigerian Gasoline Prices and Those of Benin and Togo 1. Benin versus Nigeria Gasoline Prices 2. Togo versus Nigeria Gasoline Prices 700 700 600 600 500 500 400 400 300 300 200 200 Benin Formal Togo Formal 100 Nigeria 100 Nigeria 0 0 Jan. 2008 Jul. 08 Jan. 09 Jul. 09 Jan. 10 Jul. 10 Jan. 11 Jul. 11 Jan. 12 Jul. 12 Jan. 13 Jul. 13 Jan. 14 Jul. 14 Jan. 15 Jul. 15 Jan. 16 Jul. 16 Jan. 17 Jul. 17 Jan. 2008 Jul. 08 Jan. 09 Jul. 09 Jan. 10 Jul. 10 Jan. 11 Jul. 11 Jan. 12 Jul. 12 Jan. 13 Jul. 13 Jan. 14 Jul. 14 Jan. 15 Jul. 15 Jan. 16 Jul. 16 Jan. 17 Jul. 17 Sources: Country authorities; and IMF staff calculations. Table 2. Fuel Prices Correlation in Togo, Benin and Nigeria, 2008–17 International oil Nigeria’s domestic Price differential prices prices with Nigeria Benin Official prices 0.72 0.34 Informal market prices 0.3 0.63 Togo Formal prices 0.49 0.31 Informal market prices 0.2 0.72 Domestic consumption 20.03 0.43 20.85 Nigeria Official prices 0.01 Sources: Country authorities; and IMF staff calculations. monetary arrangement (Basdevant 2012; Honda and average 2 percent of GDP per year), policies that do others 2017). not reduce them are problematic. Nigeria’s fuel subsidies are a quintessential example of negative fuel pricing spillovers, which had serious Unintended Spillovers from Nigeria’s Fuel Pricing fiscal impacts on Benin and Togo.14 Benin and Togo Policies to Its Neighbors set about reforming their fuel pricing policies from Fuel subsidies in sub-Saharan African countries have 2008 through 2012, while Nigeria continued to pro- large fiscal costs and may have considerable negative vide subsidies. Inevitably, a significant price differential spillovers to neighboring countries. Subsidies that arose between official fuel prices in these countries lower fuel consumption costs in one country below and those in Nigeria, leading to increased operating those in border countries tend to have the unintended margins for smugglers and more fuel smuggling (Fig- consequence of cross-border fuel smuggling (IMF ure 29). The level of fuel sold on the formal (taxed) 2012). The negative externalities of smuggling are market declined precipitously in Benin to only 15 per- particularly acute if the receiving country has an auto- cent of total consumption and was much less than it matic fuel price adjustment mechanism that prevents should have been in Togo. This led to a smaller fuel tax authorities from lowering fuel prices in line with smug- gled prices. Given that fuel subsidies are expensive (on 14Inflation correlations between Nigeria and its neighboring coun- tries are also strong, especially in food prices (IMF 2012). International Monetary Fund | August 2018 37 ©International Monetary Fund. Not for Redistribution  SPILLOVER NOTES Figure 30. Sub-Saharan Africa: Within Migration, Refugees Figure 31. Sub-Saharan Africa and Selected Countries: 14 Internally Displaced Persons Refugees Rest of migrants 13 1.3 1. SSA IDPs, 2010–16 12 1.1 11 8 10 7 Millions of persons 9 8 2.2 6 7 Millions of persons 6 11.8 5 3.1 11.2 5 4 4 3 6.7 3 2 4.1 1 2 0 1990 2000 10 13 1 Sources: United Nations High Commissioner for Refugees database; and World 0 2010 11 12 13 14 15 16 Bank, Migration and Remittances database. 2. IDPs by country, 2016 2.5 16 Number of IDPs (LHS) base for legally consumed fuel in these two countries IDPs percent of 14 total population (RHS) (Mlachila, Ruggiero, and Corvino 2015). 2.0 12 Nigeria’s price level had a large positive and sta- Percent of population tistically significant impact on foreign formal fuel 1.5 10 Millions of persons consumption. In Togo, the correlation between formal 8 market consumption and the price differential vis-à-vis 1.0 6 Nigeria is –0.85 (Table 2). In 2011 this implied an 4 implicit subsidy of about 3 percent GDP to Togo, 0.5 three-quarters of which was captured by smugglers 2 and one-quarter by Togolese consumers, as their 0.0 0 welfare increased with the purchase of fuel at lower COD NGA SSD CAF CMR BDI TCD NER MLI COG MOZ informal prices. Sources: United Nations High Commissioner for Refugees database; and World In countries that provide subsidies, such as Nige- Bank, Migration and Remittances database. ria, the first best approach is to follow international Note: See page 42 for country abbreviations table. LHS = left hand side; RHS = right hand side. fuel prices more closely by instituting an automatic fuel pricing mechanism (see IMF 2013). In addition to reducing fiscal costs, this would reduce negative spillovers to the country’s neighbors. A second-best The Rising Socioeconomic Impact of approach is more cooperation among countries to Forced Migration control borders to reduce smuggling and further har- The number of internally displaced persons has risen monization of tax policies to avoid generating negative significantly as a result of conflicts and violence, primarily regional spillovers. the rise of religious extremism affecting the Sahel region However, if there is large-scale smuggling, the first and northeastern Nigeria. These developments hurt eco- best solution may not work effectively. In the pres- nomic activity and weigh on public expenditures. ence of a porous border and large price differentials, While available statistics show evidence of a automatic price adjustment—if it increases the price significant decline in the share of forced migration differential—can lead to tax base erosion. In this case, across countries in sub-Saharan Africa through most the best strategy may be to lower the tax rate, which is of the 1990s and 2000s, there are some indications tantamount to lowering the domestic price (Mlachila, that the pace of the decline has slowed or partially Ruggiero, and Corvino 2015). However, this approach reversed. The latest available migration data show that may have substantial fiscal costs compared with a situa- the ratio of refugees to the total migrant population tion of no smuggling. declined from over 40 percent in 1990 to 10 percent 38 International Monetary Fund | August 2018 ©International Monetary Fund. Not for Redistribution  Regional Spillovers in Sub-Saharan Africa: Exploring Different Channels Figure 32. Selected Sub-Saharan African Countries: Civil Unrest and Terrorism 1. Civil Unrest, 2011–2017 2. Terrorism, 2011–2017 300 G5 Sahel 500 G5 Sahel Cameroon Cameroon Nigeria Nigeria 400 Number of incidents Number of incidents 200 300 200 100 100 0 0 2011 12 13 14 15 16 17* 2011 12 13 14 15 16 17* Sources: IMF, Corporate Services and Facilities Security Services database; and IMF staff calculations. Note: * = August 2017. in 2013 (Figure 30). This was a result of the decline in arms and the resultant elevated frequency of terrorist large-scale conflicts in Southern and West Africa and attacks in the Sahel, especially in Burkina Faso and the end of the Rwandan genocide (Gonzalez-Garcia Mali. Boko Haram attacks have been the leading and others 2016). Since then, while the number cause of displacement of people in Niger, northeastern of intraregional sub-Saharan African refugees has Nigeria, Chad, and northern Cameroon. In the other declined, the number of internally displaced persons countries, domestic and neighboring political turmoil has risen significantly (Figure 31, panel 1). The Sahel has been the driving factor of forced migration. For countries, Nigeria, Democratic Republic of the Congo, instance, Uganda currently hosts more than 1.2 mil- South Sudan, and Central African Republic, are lion refugees and asylum seekers, the highest number among the countries most affected by internal displace- in sub-Saharan Africa and the third largest in the ment triggered by conflicts and violence (Figure 31, world; they are mainly from South Sudan and Demo- panel 2).15 Furthermore, while the number of refugees cratic Republic of the Congo (UNDP 2017). has been falling, the absolute number of migrants has Forced migration reduces economic activity and has risen considerably and is currently at record levels. The considerable humanitarian and fiscal costs stemming increase in the number of migrants within sub-Saharan from both fighting terrorism and hosting displaced Africa is likely driven by individuals seeking greater persons (IMF 2016b). This is the case in the Lake economic opportunity and reflects reduced barriers to Chad area, where tourism has completely shut down. the movement of people.16 While some economic activities may not be currently There are many reasons for the increased number of affected (for example, mining in remote regions), displaced persons. Across the Sahel countries, Nigeria, domestic and foreign investment is likely to be held and Cameroon, terrorism-related events and civil con- back by the insecurity and the higher cost of doing flicts have more than doubled since 2011 (Figure 32). business. The United Nations Office for the Coordina- The collapse of the government in Libya, combined tion of Humanitarian Affairs estimates that the security with the rise of religious extremism, has undoubtedly situation will result in nearly 30 million people suffer- been the leading cause for civil conflict in the Sahel. ing food insecurity, with almost 12 million of these at The direct impact has been increased availability of crisis or emergency levels. The office reports that the Sahel region contended with approximately 4.9 million 15See the Internal Displacement Monitoring Centre website at refugees and internally displaced persons in 2017; it http://​www​.internal​-displacement​.org/​sub​-saharan​-africa/​summary/​. estimates the region’s humanitarian and financial needs 16For a detailed discussion of the economic determinants at US$2.7 billion for that year. The fiscal costs of of migration in sub-Saharan Africa, see Gonzalez-Garcia and hosting displaced populations vary; they are estimated others (2016). International Monetary Fund | August 2018 39 ©International Monetary Fund. Not for Redistribution SPILLOVER NOTES to range between 1 percent and 5 percent of GDP, can have positive effects for medium-term growth in depending on the number of displaced persons.17 sub-Saharan Africa. Indeed, Sub-Saharan Africa hosts some of the biggest The analysis presented here suggests that the level of refugee camps in the world (Gonzalez-Garcia and integration in sub-Saharan Africa is higher than com- others 2016). monly assumed, and it has the potential to deepen on many levels, supported by appropriate policies: •• Trade integration is now at comparable levels with Concluding Remarks that in other developing regions. Regional inte- Over the past few decades, sub-Saharan Africa has gration policies—such as reduction of tariff and been undergoing a process of regional integration that nontariff barriers—and improvements in transport has materialized through many dimensions, including infrastructure can facilitate intraregional trade. trade, banking, financial markets, financial innovation, Structural transformation strategies that promote and customs and monetary unions—these changes diversification could minimize spillover risks have likely increased the potential for regional spill- associated with overreliance on too few products overs. What does this imply for the economic recovery and partners. of the region now that the largest economies are expe- •• While greater integration and sophistication in the riencing a tepid pick-up in growth after an extended regional financial systems are welcome, they pose period of slow growth? As this note shows, the answer new risks, as they increase the interdependence of depends on the dynamism of growth in countries with financial markets and the potential for contagion. the highest potential to generate regional spillovers. Policies should address these risks by enforcing In the context of the trade links, countries that have existing banking regulatory frameworks, strengthen- been identified as the main destination for regional ing cooperation among supervisors, and developing imports play a significant role in the growth of their cross-border bank resolution frameworks, which are trading partners. Some of these countries, like South currently lacking. Africa, are facing persistently sluggish economic activ- •• Ongoing financial and technological developments ity, but others, such as Côte d’Ivoire, have maintained have translated into lower costs for sending remit- fast growth for a decade. In terms of banking chan- tances throughout the region and have contributed nels, PABs and subregional banks have contributed to rising regional remittance flows. It is essential to the deepening of banking systems in the region, to provide a regulatory environment for these new a development that has been associated with higher technologies that is both enabling and risk mitigat- medium-term growth. In the context of remittance ing (IMF 2017a). channels, the increase of intraregional remittance flows •• As economic integration deepens, public policies in origin countries has an impact on the economic need to be mindful of fiscal spillovers to mitigate activity of countries that receive remittances from the associated fiscal risks. This calls for greater harmoni- region. The largest countries in the region have import- zation of fiscal policies. ant fiscal policy spillovers, which can either boost or •• A recent resurgence in population displacement derail growth in neighboring countries, depending on highlights the need for policies to address the the state of their economies. main causes of forced migration, such as increased As the sub-Saharan African economy has grown, economic and physical insecurity, and put in place regional demand for intraregional exports has become a system that accommodates and integrates forced an important market for goods produced in the region, migrants in host countries in a sustainable way. regional banks have created new business opportuni- Increased international aid would greatly facilitate ties and activity in the region, and remittance flows the process. have contributed to economic activity in recipient countries. Further enhancing regional integration This note stresses the growing need for policymakers to factor in spillovers from within the region when planning for the medium term and to design policies 17For instance, the United Nations Development Programme that address increasing transmission risks. estimates the cost of hosting refugees and asylum seekers in Uganda at about US$320 million, or about 1.3 percent of GDP. 40 International Monetary Fund | August 2018 ©International Monetary Fund. Not for Redistribution  Regional Spillovers in Sub-Saharan Africa: Exploring Different Channels Appendix Appendix Table 1. Sub-Saharan Africa: List of Country Abbreviations AGO Angola ERI Eritrea MLI Mali SWZ Swaziland BDI Burundi ETH Ethiopia MOZ Mozambique SYC Seychelles BEN Benin GAB Gabon MUS Mauritius TCD Chad BFA Burkina Faso GHA Ghana MWI Malawi TGO Togo BWA Botswana GIN Guinea NAM Namibia TZA Tanzania CAF Central African Republic GMB Gambia, The NER Niger UGA Uganda CIV Côte d’Ivoire GNB Guinea-Bissau NGA Nigeria ZAF South Africa CMR Cameroon GNQ Equitorial Guinea RWA Rwanda ZMB Zambia COD Congo, Dem. Rep. of KEN Kenya SEN Senegal ZWE Zimbabwe COG Congo, Rep. of LBR Liberia SLE Sierra Leone COM Comoros LSO Lesotho SSD South Sudan CPV Cabo Verde MGD Madagascar STP São Tomé & Príncipe Appendix Table 2. Sub-Saharan Africa: Member Countries of Groupings Other resource-intensive Non-resource-intensive Oil exporters countries countries Angola Botswana Benin Cameroon Burkina Faso Burundi Chad Central African Republic Cabo Verde Congo, Rep. of Congo, Dem. Rep. of Comoros Equatorial Guinea Ghana Côte d’Ivoire Gabon Guinea Eritrea Nigeria Liberia Ethiopia South Sudan Mali Gambia, The Namibia Guinea-Bissau Niger Kenya Sierra Leone Lesotho South Africa Madagascar Tanzania Malawi Zambia Mauritius Zimbabwe Mozambique Rwanda São Tomé and Príncipe Senegal Seychelles Swaziland Togo Uganda International Monetary Fund | August 2018 41 ©International Monetary Fund. Not for Redistribution SPILLOVER NOTES Appendix Table 3. Sub-Saharan Africa: Member Countries of Regional Groupings Southern The West African Economic and Common Market African Economic Economic and Monetary Community for Eastern and East Africa Development Southern Community of Monetary Union of Central African Southern Africa Community Community Africa Customs West African (WAEMU) States (CEMAC) (COMESA) (EAC-5) (SADC) Union (SACU) States (ECOWAS) Benin Cameroon Burundi Burundi Angola Botswana Benin Burkina Faso Central African Comoros Kenya Botswana Lesotho Burkina Faso Republic Côte d’Ivoire Chad Congo, Dem. Rep.of Rwanda Congo, Dem. Namibia Cabo Verde Rep. of Guinea-Bissau Congo, Rep. of Eritrea Tanzania Lesotho South Africa Côte d’Ivoire Mali Equatorial Guinea Ethiopia Uganda Madagascar Swaziland Gambia, The Niger Gabon Kenya Malawi Ghana Senegal Madagascar Mauritius Guinea Togo Malawi Mozambique Guinea-Bissau Mauritius Namibia Liberia Rwanda Seychelles Mali Seychelles South Africa Niger Swaziland Swaziland Nigeria Uganda Tanzania Senegal Zambia Zambia Sierra Leone Zimbabwe Zimbabwe Togo 42 International Monetary Fund | August 2018 ©International Monetary Fund. Not for Redistribution  Regional Spillovers in Sub-Saharan Africa: Exploring Different Channels References Gupta, S., C. Patillo, and S. Wagh. 2009. “Effect of Remittances on Poverty and Financial Development in Sub-Saharan Allard C., J. Canales-Kriljenko, J. Gonzalez-Garcia, E. Kitsios, Africa.” World Development 37 (1): 104–15. J. Trevino, and W. Chen. 2016. “Trade Integration and Hitaj, E., and Y. Onder. 2013. “Fiscal Discipline in the Global Value Chains in Sub-Saharan Africa: In Pursuit of WAEMU: Rules, Institutions and Markets.” IMF the Missing Link.” African Department Departmental Paper, Working Paper 13/216, International Monetary Fund, International Monetary Fund, Washington, DC. Washington, DC. Agrawal, I., Duttagupta, R., and Presbitero, A., 2017. “Inter- Honda, J., F. Im, N. Koliadina, M. Morgan, M. Nose, C. national Commodity Prices and Domestic Bank Lending in Padilla, and J. Torres. 2017. “Fiscal Rules: Coping with Reve- Developing Countries”. IMF Working Paper WP/17/279. nue Volatility in Lesotho and Swaziland.” African Department Arora, V., and A. Vamvakidis. 2005. “How Much Do Trading Paper International Monetary Fund, Washington, DC. Partners Matter for Economic Growth?” IMF Staff Papers International Monetary Fund (IMF). 2012. “Nigeria and 52 (1): 24–40. South Africa: Spillovers to the Rest of Sub-Saharan Africa, Balami, D., I. Ogboru, and D. Talba. 2011. “The Cereal Econ- Chapter 2 in SSA Regional Economic Outlook. Washington, omy in Nigeria and the Sub-Regional Dimension.” Book, DC, October. Social Study Science Group, Vol. 1, No. 29. ———. 2013. “Energy Subsidy Reform—Lessons and Implica- Basdevant, O. 2012. “Fiscal Policies and Rules in the Face of tions.” Washington, DC. Revenue Volatility within Southern Africa Customs Union ———. 2014. “South Africa—Financial Stability Assessment.” Countries (SACU).” IMF Working Paper 12/93, Interna- IMF Country Report 14/340, Washington, DC. tional Monetary Fund, Washington, DC. ———. 2015a.” Global Value Chains: Where Are You? The Chen, W., and R. Nord. 2017. “A Rebalancing Act for China Missing Link in Sub-Saharan Africa’s Trade Integration.” and Africa: The Effect of China’s Rebalancing on Sub-Saharan Chapter 3 in Regional Economic Outlook: Sub-Saharan Africa. Africa’s Trade and Growth.” African Department Paper. Inter- Washington, DC, April. national Monetary Fund, Washington, DC. ———. 2015b. “International Banking after the Crisis: Increas- Cihak, M., and R. Podpiera. 2005. “Bank Behavior in ingly Local and Safer?” Chapter 2 in Global Financial Stability Developing Countries: Evidence from East Africa.” IMF Report. Washington, DC, April. Working Paper 05/129, International Monetary Fund, ———. 2015c. “Pan-African Banks: Opportunities and Chal- Washington, DC. lenges for Cross-Border Oversight.” Washington, DC. Cuevas, A., L. Engstrom, V. Kramarenko, and G. Verdier. ———. 2016a.” Financial Development and Sustain- 2012. “SACU Revenue Volatility: Roots and Options for able Growth,” Chapter 3 in Regional Economic Outlook: Mitigation.” In The Economics of Regional Integration: Current Sub-Saharan Africa, Washington, DC, April. and Future Challenges for Southern Africa. Washington, DC: ———. 2016b. “Macro-Financial Risks from Linkages between International Monetary Fund. Banks and Non-Bank Financial Institutions.” IMF Country DNA Economics. 2013. “South African Investment in Africa.” Report 16/374. Washington, DC. Pretoria, South Africa. ———. 2016c.” Time for a Policy Reset.” Chapter 1 in Dunne, P. J., and N. Masiyandima. 2017. “Bilateral Foreign Regional Economic Outlook: Sub-Saharan Africa Washing- Direct Investment from South Africa and Income Conver- ton, DC, April. gence in the SADC Region.” African Development Review 20 ———. 2017a. “Fintech and Financial Services: Initial (3): 403–15. Considerations.” IMF Staff Discussion Note,/17/05 Financial Stability Board (FSB). 2017. “FSB Correspondent Washington, DC. Banking Data Report.” Basel, Switzerland. ———. 2017b.” The Quest for Recovery.” Chapter 1 in Furceri, D., J. T. Jalles, and A. Zdzienicka. 2016. “China Regional Economic Outlook: Sub-Saharan Africa. Washington, Spillovers: New Evidence from Time-Varying Estimates.” DC, October. IMF Spillover Note 7, International Monetary Fund, ———. 2017c. “Restoring Sub-Saharan Africa’s Growth Engine. Washington, DC. “Chapter 1 in Regional Economic Outlook: Sub-Saharan Africa, Games, D. 2017. “South Africa’s Economic Engagement in Chapter 1. Washington, DC, April. Sub-Saharan Africa: Drivers, Constraints and Future Pros- Jorion, P., and G. Zhang. 2009. “Credit Contagion from Coun- pects.” Unpublished research paper, Chatham House, Royal terparty Risk” Journal of Finance 64 (5): 2053–87. Institute of International Affairs., London Kinda, T., M. Mlachila, and R. Ouedraogo. 2016. “Commodity Gonzalez-Garcia, J., E. Hitaj, M. Mlachila, A. Viseth, and M. Price Shocks and Financial Sector Fragility.” IMF Working Yenice. 2016. “Sub-Saharan Africa Migration: Patterns and Paper 16/12, International Monetary Fund, Washington, DC. Spillovers.” IMF Spillover Note 9, International Monetary Mecagni, M., D. Marchettini, and R. Maino. 2015. “Evolving Fund, Washington, DC. Banking Trends in Sub-Saharan Africa: Key Features and Challenges.” International Monetary Fund, Washington, DC. International Monetary Fund | August 2018 43 ©International Monetary Fund. Not for Redistribution SPILLOVER NOTES Mlachila, M., E. Ruggiero, and D. Corvino. 2015. “Unintended Rey, H., 2015. “Dilemma Not Trilemma: The Global Financial Consequences: Spillovers from Nigeria’s Fuel Pricing Policies Cycle and Monetary Policy Independence.” NBER Working to Its Neighbors.” IMF Working Paper 15/17, International Paper No. 21162, National Bureau of Economic Research, Monetary Fund, Washington, DC. Cambridge, MA. Mwase, N., P. N’Diaye, H. Oura, F. Ricka, K. Svirydzenka, and The World Bank, “Remittance Prices Worldwide” Issue 22, June Y. Zhang. 2016. “Spillovers from China: Financial Chan- 2017, available at http://​remittanceprices​.worldbank​.org. nels.” IMF Spillover Note, International Monetary Fund, United Nations Development Programme. 2017. “Ugan- Washington, DC. da’s Contribution to Refugee Protection and Manage- Overseas Development Institute (ODI). 2010. “Impediments ment.”UNDP, Uganda. to Intra-Regional Trade in Sub-Saharan Africa.” Lon- don, September. Presbitero, A. F., D. Ghura, O. S. Adedeji, and L. Njie. 2016. “Sovereign Bonds in Developing Countries: Drivers of Issu- ance and Spreads." Review of Development Finance 6.1: 1–15. 44 International Monetary Fund | August 2018 ©International Monetary Fund. Not for Redistribution