I. Risks and Spillovers

International Monetary Fund

doi:10.5089/9781498343077.007.A001

Jump to Content Jump to Main Navigation

I. Risks and Spillovers

Author: International Monetary Fund

Publication Date:: 30 Jul 2014

eISBN:: 9781498343077

Language:: English

Keywords:: PP; Policy advice; The Fund; low-income country; policy; Fund surveillance; banking system; financial crisis; capital market; BSA risk indicator; Spillovers; Financial statements; Stress testing; Global financial crisis of 2008-2009; Credit; Global; crisis risk rating

Download PDF (2.5 MB)

2014 Triennial Surveillance Review - Staff Background Studies

Abstract

2014 Triennial Surveillance Review - Staff Background Studies

I. Risks and Spillovers¹

Executive Summary

Risk and spillover analysis is critical for the Fund’s mission. The size of shocks during the global financial crisis and the speed at which they were transmitted across sectors and borders brought the importance of this work into sharper focus. The world remains highly interconnected and, even though the crisis has subsided, country policies will continue to propagate shocks. Further developing this analysis will support the Fund’s efforts to provide advice that helps members avoid risks and mitigate their impact should they materialize.

Country authorities welcome the Fund’s efforts so far to advance this agenda, and have called for the analysis to be deepened. In the wake of the 2011 Triennial Surveillance Review (TSR), the Fund has strengthened risk assessments in all surveillance products and given more weight to spillover analysis. This has been supported by the adoption of the Integrated Surveillance Decision (ISD) and new processes, including a more systemic approach to identifying global risks. To respond to members’ calls for more in-depth analysis, the Fund will have to address challenges that have emerged due to the pioneering nature of this work, uncertainties about transmission channels, data constraints, and the perceived lack of transparency of some models.

The Fund should respond with an eclectic approach, integrating results across surveillance products, drawing from a range of analytical tools, and applying judgment as needed. Key conclusions from relevant multilateral products and cross-country exercises should be integrated into bilateral surveillance and, conversely, multilateral surveillance can draw more effectively on the granular analysis that is possible only in bilateral surveillance. The Fund’s modeling toolkit can be leveraged further if tools can be adapted to better fit country circumstances with more transparency on their key features and limitations. However, some risks do not lend themselves to formal modeling, such as those arising from financial regulatory developments, should be assessed by informed judgment.

The Fund should leverage existing tools and selectively add new ones in order to address gaps in coverage of external risks, spillover channels and cross-sectoral linkages. More extensive sharing of the analysis of outward spillover from systemic countries can help deepen analysis of external risks in other countries. The focus on global spillovers should be accompanied by more systematic work on cross-border risks that may not be globally systemic, such as regional risks. The Fund’s understanding of linkages can be greatly enhanced by reviving and adapting the analysis of national balance sheets.

Macro-financial analysis should be mainstreamed. Financial surveillance has proved challenging for country teams owing to the variety of risks that need to be assessed and broad spectrum of skills required. To make headway in mainstreaming of financial surveillance, staff should focus on investing in tools, techniques and analytical frameworks to better assess the risks posed by rapidly evolving macro-financial and capital market developments. In particular, more systematic analysis of the relationship between credit and economic developments would help pinpoint both risks building up in the baseline and likely to emerge in shocks scenarios.

Data gaps remain a significant impediment to the Fund’s efforts to strengthen risk and spillover analysis. More than five years after the global financial crisis, the Fund has been unable to obtain from the membership some of the data required to assess key systemic risks, including cross-border banking exposures and the linkages between systemic global banks. Data gaps are also holding back efforts to strengthen structural macro models used for spillover analysis.

A. Framing and Evaluating Risk and Spillover Analysis

Introduction

1. Risk and spillover analysis is critical for the Fund’s mission. This follows from its role in assessing stability, the anchoring concept for surveillance. The size of shocks during the global financial crisis, and the speed at which they were transmitted across sectors and borders, brought the importance of this work into sharp focus.² Even though the crisis has subsided, the world remains highly interconnected and country policies will continue to spillover across borders, underscoring the relevance of this analysis for Fund surveillance. Achieving a rich understanding of countries’ exposure to domestic and external risks will help the Fund advise members on policies to strengthen their capacity to avert risks and mitigate their potential impact.

2. The 2011 TSR laid out a comprehensive and ambitious strategy to advance this agenda, which has necessitated a strengthening of the Fund’s analytical toolkit. The adoption of a new legal framework, the launch of new products, and the development of internal processes have laid the foundations for an expansion in risks and spillovers coverage in core surveillance products. The need to deliver rigorous analysis has also spurred the development of a range of quantitative tools, which represents a major undertaking given the pioneering nature of this work.

3. Stakeholders and external experts would like the Fund to go even further, which will pose some challenges. Country authorities would like deeper analysis, including more in-depth discussion of the impact of systemic countries’ policies. External experts are also strongly supportive of this agenda, and provide a number of suggestions to refine the way it is carried out, including calling for greater focus on balance sheets. At the same time, staff needs to manage the practical challenges of delivering effective analysis, including the limitations and perceived lack of transparency of some of the new models, data constraints, and difficulties in gaining traction with authorities in discussions on their outward spillovers.

4. This paper provides suggestions to advance this agenda taking into account the challenges that have emerged. It proposes an organizational framework for risk and spillover analysis at the Fund and suggests how the institution’s analytical capacity can best be leveraged. It draws on the stakeholder surveys, a review of surveillance products and a parallel study by external experts. It also complements other TSR work-streams relating to the integration of bilateral and multilateral surveillance, and the scope of multilateral surveillance.

Box 1.

Possible Options to Advance Risk and Spillover Analysis

Enhance coordination and cross-fertilize ideas

Multilateral surveillance could feed more effectively into bilateral surveillance while drawing on insights from the more granular analysis possible at the bilateral level. Mechanisms for this are in place but can be strengthened to improve the sharing of information, for example by providing more details on how global and regional risks could play out, depending on conditions in individual countries.
Further mainstream analysis of financial sector risks. This involves focusing Article IV surveillance on rapidly evolving macro-financial risks, while slower moving risks arising from weaknesses in institutions are assessed at a lower frequency drawing on specialized expertise such as that provided by FSAPs. This analysis should also consider macro-prudential policy measures to address these risks.

Strengthen analysis of spillovers

Adopt an eclectic approach that combines quantitative analysis with judgment. The more rigorous analysis from models and quantitative techniques needs to be accompanied by a deeper understanding of their limitations and whether they fail to capture some spillover channels.
Integrate the analysis of outward spillovers from systemic countries into the surveillance of non-systemic countries. As outward spillovers are a source of external risk for other countries, mechanisms for effectively sharing this analysis can deepen the assessment of external risks.
Integrate spillbacks into spillover analysis. Spillbacks occur when outward spillovers boomerang back on the source country. Identification of spillbacks can enrich policy discussions with systemic country authorities by providing a basis to discuss alternative policies (as required under the Integrated Surveillance Decision).

Deepening risk analysis building on quantitative tools

Expand balance sheet analysis in surveillance. The identification of balance sheet mismatches and the analysis of linkages among sectors using new data in a matrix of balance sheet exposures can reveal how vulnerable a country is to shocks and how they will be transmitted.
Mainstream analysis of macro-financial linkages in Article IV consultations. This involves assessing the influence of the financial sector on the macroeconomic outlook. This would clarify, for example, whether the outlook is consistent with the banking system’s capacity to lend when it is deleveraging.
Make financial tool and models more accessible and transparent. Increased usage of a wider range of quantitative tools and models is impeded by concerns about their lack of transparency, which can be countered by making them widely available and easily useable.
Assess cross-border exposures using a Global Flow of Funds (GFF) framework. Surveillance of external risks can be deepened by using BIS, CPIS and IIP data to assess the cross-border transmission of financial shocks. This analysis builds on the GFF analytic framework developed at the Fund.
Extend the external debt sustainability analysis (DSA). Applying the DSA framework to cover the impact of external shocks on flows will help pinpoint external liquidity risks.
Address data gaps. Despite some improvements, data gaps have become more serious as Fund surveillance has been strengthened after the crisis. International understandings to improve data on GSIBs to assess spillovers among major banks and BIS banking statistics need to be implemented.

5. The remainder of the paper is organized as follows. This opening section provides context, an overview of the key concepts, an evaluation of progress so far, and points to the issues that should shape the Fund’s agenda on risks and spillovers in the period ahead. Section B lays out the broad contours of a framework for delivering effective coverage of risks and spillovers, while addressing the challenges that have emerged since the 2011 TSR. Section C explores how various tools fit into this framework, and how staff can respond to stakeholders’ calls to deepen their analysis, while taking into account the limitations of models.

Joining the Dots through Risk-Based Surveillance

6. Risk and spillover analysis are two very closely related agendas (see Box 2 for key definitions). Risk analysis involves identifying domestic and external risks, and assessing their probability and likely impact. External risks for one country—which are potential inward spillovers-are often associated with domestic risks in another country. Thus, there is much to be gained by effective sharing of information on risks and spillover assessments on different countries.

7. The Fund has made major organizational changes in order to deliver consistent, pointed and targeted analysis of risks and spillovers. The aim has been to ensure that surveillance focuses on the key risks, both globally and for each member. This has involved amending surveillance policies, revamping existing products and selectively introducing new ones, while also adapting internal processes to achieve consistency:

An Integrated Surveillance Decision (ISD), adopted in 2012, clarifies the scope of risk analysis and makes spillover analysis a mainstream feature of Article IV surveillance (see Annex I). It also calls for bilateral and multilateral surveillance to be more closely integrated.
More systematic risk assessments have been introduced in the WEO, GFSR and Article IV consultations.
The Spillover Report, prepared annually since 2011, initially focused on outward spillovers from the “S5” economies (China, Euro Area, Japan, United Kingdom, US). While these five economies were chosen for their size and interconnectedness, other countries are also likely to have “systemic spillovers” (and the 2014 Spillover Report takes a broader approach—see below).
The launch of the Pilot External Sector Report (ESR), to provide a comprehensive and consistent assessment of external sector positions and policies, supported by a new methodology, the External Balance Assessment (EBA).
The Early Warning Exercise (EWE), presented to the Board and IMFC twice a year, assesses low-probability but high-impact risks to the global economy and identifies policies to mitigate them.
New internal processes to spur more systematic and consistent thinking on risks and spillovers. An inter-department group is charged with developing an institution-wide view of the key global risks, which are summarized in a global risk assessment matrix (G-RAM). The internal vulnerability exercises (VE) have also been expanded to make them more relevant for Fund surveillance.

Box 2.

Risks and Spillovers 101

The following definitions of risk and spillovers form the basis for the conceptual framework:

Risks are potential shocks that have some probability of materializing with an impact on macroeconomic or financial conditions in a country or across countries. They can be domestic or external, and related to exogenous or policy shocks.
Linkages refer to the links between sectors within an individual economy or across countries, which can act as transmission channels for shocks.
Spillovers refer to the cross-border transmission of shocks. These can be global or affect one or more countries, and occur through a variety of channels. Spillovers can arise from exogenous shocks or, importantly, a country’s policies (policy spillover—see below). The impact of spillovers depends on the size and nature of the underlying shocks, the transmission channels associated with macroeconomic and financial cross-border linkages, as well as the economy’s resilience to shocks.
The analysis of inward spillovers involves evaluating the channels through which external shocks affect a country and quantifying their impact. The term encompasses actual spillovers and potential spillovers. The former is when risks have already materialized (as “shocks”), allowing the spillovers through the different channels to be observed and estimated directly. Potential spillovers correspond to risks that have not yet materialized.
Policy spillovers. The Fund has focused on identifying policies that generate spillovers. This helps increase the awareness in “source” countries of the impact of their policies on others, which can facilitate policy cooperation.
Spillbacks (or boomerang effects) are a special case of spillovers where, for example, one country’s outward spillovers affect a number of third parties, whose economic situation deteriorates, leading to adverse feedback effects (“spillbacks”) on the source country.

8. Risks and spillovers have also been made a central part of the IMF’s Financial Surveillance Strategy (FSS).³ The pillars of the FSS include: (i) strengthening the analytical underpinnings of macro-financial risk assessments and policy advice; and (ii) upgrading the instruments and products of financial surveillance to foster an integrated policy.⁴ The FSS also emphasizes that departments should work together to produce more integrated macro-financial risk assessments.

9. The Fund and the Financial Stability Board (FSB) have also launched an initiative to address data gaps in recognition of the risk that they could impede the Fund’s work. The G-20 Data Gaps Initiative (DGI), which is led by the Fund and the FSB, aims to close gaps in four areas: identifying and monitoring financial-sector risks, cross-border linkages, vulnerabilities and sectoral interconnections in domestic economies, and the communication of official statistics.

Shaping the Agenda Ahead

10. Country Authorities (CAs) welcome the changes so far, and would like the Fund to go even further.⁵ They value the quality of the Fund’s risk assessments, with around 70 percent of CAs agreeing with the Fund’s overall assessment of the risks facing their country. Spillover analysis has also improved markedly since 2011 (when it was seen as a weakness), although it lags behind other policy areas in terms of the value added it provides. For the future, the recommendations from the CAs amount to a call for the Fund to persevere with its existing strategy. They would like deeper analysis of the transmission channels of shocks, greater quantification of risk assessments, and more in-depth discussion of the impact of systemic countries’ policies on the rest of the world.

11. External experts, in a parallel study conducted for the TSR, provide specific recommendations to help the Fund shape its agenda ahead.⁶ Above all, they call for the Fund to be “joined up” in the way it tackles risks and spillovers, and to avoid overreliance on models. They offer specific recommendations for the Fund’s strategy ahead, including the following:

Strategy. Multilateral surveillance needs to feed directly into bilateral reports, while bilateral surveillance should be used to deliver more granular evidence to support analysis at the multilateral level. The Spillover Report is a key disciplining device to support this. The Fund should also make recommendations on macro-prudential policies and cover outward spillovers (including spillback effects).
Substance. Analysis of national balance sheets should be a priority. The Fund should also do more work on spillback effects, the adequacy of members’ macro-prudential frameworks, and make greater efforts to distinguish tail risks.
Research and data. The experts call for more research on the risk channels of monetary policy, cross-border capital flows, and the impact of macro-prudential policies. On the data front, balance sheet and flow of funds data should be a priority.
Communications. Senior Fund officials should give more airtime to the Spillover Report, and the Fund should in general be more open to covering tail risks that are already in the public domain.

Usage of Quantitative Tools

12. The Fund has significantly expanded its analytical toolkit for risks and spillovers, although the application of these new models has been uneven.⁷ They have been used more extensively in multilateral surveillance. The WEO, for instance, has underpinned its risk assessments with a wide range of quantitative tools, including structural models, event studies, VARs, indicator-based models and market perspectives. The Spillover Reports initially used a mixture of structural models (the GIMF and G-35S), VARs and detailed analysis of BIS and market data, but in the 2013 report relied more on structural models; with the G-35S assessing financial spillovers, and the GIMF and FSGM focusing on spillovers from macroeconomic shocks. Bilateral surveillance has drawn much less on formalized models.

13. The lower take-up of models in bilateral surveillance reflects concerns about their applicability. Some models are seen as not adequately transparent, as they are not always fully documented and can be hard to understand. In addition, models developed for one country require significant modification to make them applicable to other countries or regions with different structural features. Moreover, many of the Fund’s global models do not cover the majority of the membership.

Data Gaps: Running to Stay in the Same Place

14. Despite some improvements, data gaps remain a major obstacle to this agenda.

Expanded datasets are being used in surveillance.⁸ The Coordinated Portfolio Investment Survey (CPIS) is now reported semi-annually, improving the monitoring of cross-border investments at the sectoral level. Over 80 countries also report quarterly international investment positions (nearly twice the level in 2009), helping to improve the early detection of external vulnerabilities. Many advanced and emerging market economies (including the G20) have developed sectoral balance sheets, or are in the process of doing so, which will help improve monitoring of domestic linkages, including non-bank financial institutions.⁹ Lastly, almost 100 members now provide financial soundness indicators (FSIs), a two-fold increase since 2009, enabling better monitoring of financial sector risks.
However, demands for new analysis have increased and model developers and users indicate that data are a severe binding constraint on their work. The lack of financial sector data is a key concern: more than five years after the collapse of Lehman Brothers, the Fund still lacks aggregated data on global systematically important banks (See Box 3), and access to aggregated cross-border banking data is restricted (Box 4). Moreover, some of the Fund’s counterparts have become less willing to share data as the crisis has subsided.

Box 3.

Addressing GSIB Data Gaps Hampering IMF Surveillance

The triggering of the 2008 global financial crisis by the failure of one large U.S. financial institution with extensive linkages to other global banks revealed critical gaps in data that the IMF needed in order to assess risks to global financial stability. In 2009, the G-20 responded by launching a broad Data Gaps Initiative, led by the IMF and the FSB, that included more granular data on 29 Global Systemically Important Banks (GSIBs). These data are necessary to detect risks that are masked by the high degree of aggregation in existing data covering the whole banking system (e.g., FSIs). Supervisors of GSIBs already collect similar data but do not share them owing to confidentiality concerns. To address these concerns, the G-20 agreed to the creation of “institution-to-aggregate” data where each GSIB’s exposures are aggregated across sectors and countries to ensure confidentiality. The national authorities supervising GSIBs started sharing some data among themselves in early 2013 and are expected to have more granular data by mid-2016. The sharing of these granular data with the IMF has been agreed in principle but the implementation is still an outstanding issue.

The perpetuation of this data gap severely hampers IMF surveillance. These data are essential to identify spillover channels among GSIBs, which has become even more critical to surveillance with the adoption of the ISD and creation of the Spillover Report. They would allow recognition of the buildup in balance sheet mismatches by providing details on currency, instrument type, and maturity. Systemic liquidity risk—a key factor contributing to the crisis—is concentrated in GSIBs and cannot be assessed adequately with existing indicators such as the crude, but available, loan-to-deposit ratio. To evaluate the robustness of a financial system to liquidity shocks, and the potential for these to have spillover effects, it is essential for the Fund to obtain data on the maturity and currency of GSIBs’ liabilities, on the reliance on marketable instruments for funding, and on the geographic structure of external financing.

Coverage of these risks and spillovers depends on national authorities’ sharing their GSIBs’ data with the IMF. These data cover the countries and sectors to which the GSIBs are exposed, the financial instruments they hold, and the currency and maturity structure of their assets and liabilities, while being aggregated sufficiently to ensure confidentiality. Given the legitimate confidentiality concerns, an additional layer of confidentiality protection would be provided by internal IMF procedures governing how the data are reported and who has access to them. Discussion on the timing and extent of sharing of these data with the IMF will start in the second half of 2014.

Box 4.

Mapping the Global Banking Network in IMF Surveillance^1/

The global banking network played an important role in the propagation of the global financial crisis across countries. These spillovers can be assessed using the IMF’s Bank Contagion Model, which uses the BIS international banking statistics and individual bank data to map them.

The mapping of the global banking network is challenging due to the complicated structure of international banking groups and multiplicity of lending and funding channels. International banks can lend to a borrower country directly through cross-border loans, or indirectly through branches or subsidiaries in either the borrower country or in third countries. The funding model varies across banking groups, with some local branches or subsidiaries able to rely largely on local deposits as the main source of funding while others are heavily dependent of parent bank funding or wholesale financing. It also differs across regions, with subsidiaries in Latin America relying largely on local deposits, for example.

Two BIS datasets allow us to map these cross-border banking linkages, each with strengths and weaknesses. The BIS Locational Banking Statistics (LBS) captures the cross-border exposures of all international active offices in a reporting country, regardless of the nationality of the parent bank. Bank positions are recorded on an unconsolidated basis and includes those vis-à-vis their own offices in other countries. This artificially increases the size of financial centers. Figure 1 illustrates this case showing that the nominal amount of US banks’ claims on Cayman Islands borrowers appears much larger than for other countries when using LBS. The use of the Consolidated Banking Statistics (CBS) reduces this bias by consolidating exposures between parent banks and their foreign affiliates. The CBS foreign claims concept, however, can overstate the size of foreign exposures as it includes assets of subsidiaries funded by local deposits as well as parent banks’ resources. Figure 1 shows Spanish banks’ claims on Brazilian, Mexican and US borrowers. Using CBS, the amount of Spanish banks’ claims on Brazilian, Mexican and US borrowers is larger than when using LBS.

The RES Bank Contagion Module (Cerutti 2013) uses an adjusted consolidated measure. It is constructed using CBS, to avoid overstating the size of financial centers while adjusting the consolidated group assets to reflect the funding structure. Figure 2 shows the adjusted consolidated metric in which the banking linkage between US banks and Cayman Islands is much smaller than for LBS, while the banking linkages between Spanish banks and Brazilian and Mexican borrowers is significantly lower than for CBS.

^1/ Prepared by Eugenio Cerutti based on Cerutti and Ohnsorge (2014).

B. Consolidating the Framework for Risk and Spillover Analysis

15. The task for the Fund is to enhance its framework for risk and spillover analysis in order to deliver on country authorities’ expectations and address the challenges that have emerged. This section lays out the broad contours of the framework while the next section examines in more detail analytical techniques that can help deliver on these goals.

16. The Fund has adapted its organizational framework, with further changes underway.

The simpler pre-crisis framework (shown as boxes with solid borders in Figure 2) has been augmented with the Spillover Report and G-RAM, and the addition of inward and outward spillover analysis. This has facilitated the integration of bilateral and multilateral surveillance, with the G-RAM identifying global risks to be included in individual risk assessments and the Spillover Report contributing to outward spillover analysis for systemic countries.
The Spillover Report was reorganized in 2014 along thematic lines to focus on a few key spillovers rather than covering a fixed set of economies (e.g., the S5). Consequently, rather than providing a wide range of model based scenarios, two broad spillover themes are analyzed more thoroughly with strengthened analytical underpinnings of model simulations. In parallel, the G-RAM now includes model simulations for the key risks and pays greater attention to tail risks. Finally, the VE is drawing increasingly on quantitative tools; including, for example, models to identify vulnerabilities in resource rich countries.

An Integrated Approach to Risk Analysis

17. The Fund’s framework for risk identification has long been relatively decentralized. Individual country teams are charged with developing their own views of the major risks facing their country, informed by consultations with the authorities and private sector. This represents one of the most important sources of pointed and granular information on risks.¹⁰

18. New products and processes have been put in place in order to support bilateral surveillance with a coordinated analysis of risk. They draw on both the results of bilateral surveillance and analysis undertaken at the multilateral level: the arrows in Figure 2 show the linkages, and how these products and processes support surveillance.

The Global Risk Assessment Matrix (G-RAM) is an internal product, which lists key global and regional risks, identified and periodically updated by a Risk Group drawn from across the Fund. G-RAM risks are expected to be covered in the same way in bilateral surveillance, although their impact on each country must be assessed individually through inward spillover analysis (see below). The G-RAM also informs risk identification at both bilateral and multilateral levels. It is complemented by the Vulnerabilities Exercise, an internal exercise to assess individual countries’ vulnerabilities, and an internal group set to analyze tail risks.
The External Balance Approach provides a multilaterally consistent assessment of countries’ external sustainability that helps identify the risk of a sharp correction from an overvalued currency or unsustainable current account deficit. It feeds into the assessment of external risks in individual countries and also at the multilateral level, where it ensures that unsustainable deficits in some countries are counterbalanced by surpluses in others.

19. These initiatives are designed to support bilateral surveillance, but their effectiveness also depends on the extent to which they draw on analysis by country teams. Area department teams generally undertake more granular analysis and provide a reality check for analysis done at the multilateral level. In practice, country teams review, and have final say on, the assessments in the Vulnerabilities Exercise and External Balance Approach. Global risks in the G-RAM, for example, include scenarios that illustrate how they might play out at the country level, making it easier to incorporate them into country risk assessments. These risks and scenarios are, in turn, vetted by area department members on the Risk Group. Coordinating this global risk identification with Article IV consultation cycles involves challenges that the risk group can help manage.

External and Financial Risks

20. Processes to ensure that risks identified in multilateral surveillance are used in countries’ assessment of external risk can be strengthened. Key messages from the WEO and GFSR’s in-depth analysis of global risks are being distilled for area departments but could include more detail on how these risks might play out in different countries. Regional Economic Outlooks (REOs) can help support this. The analysis of risks and outward spillovers from surveillance in systemic countries can be shared more effectively to deepen the analysis of external risk in non-systemic countries that are associated with these potential outward spillovers.

21. Financial sector risks—both domestic and cross border—are important for many members. Crises are often sparked or magnified by financial sector developments, such as sharp shifts in global liquidity or capital flow reversals.

22. However, financial risks are often not easy to diagnose as they combine rapidly evolving market and macro-financial risks and slower-moving risks from institutional weaknesses. The Fund is putting a greater emphasis on macro-financial analysis, which involves country teams focusing on evolving risks that are macro relevant. The slower-moving risks can generally be assessed at a lower frequency, depending on whether they are a concern or if regulatory changes are being considered, and may require highly specialized knowledge in banking supervision.

Spillover Analysis

23. The goal of spillover analysis is to identify and quantify the channels through which risks are transmitted internationally. The imperative of quantifying spillovers was a factor behind that development of large, multi-country models to simulate scenarios that illustrate how outward spillovers from a shock in a systemic country are transmitted to other countries, although other, more specialized, techniques can be used as well.

24. The appropriate techniques for assessing spillovers vary according to their type and channel:

Actual spillovers are often observable in existing data, and so can be assessed at least qualitatively without sophisticated models.
Potential inward spillovers are associated with external risks that have not yet materialized; and, thus, need to be assessed by simulating the impact of shocks using a model for the country.
Potential outward spillovers are associated with risks in systemic countries that, when they materialize, trigger outward spillovers to many other countries. Outward spillovers need to be assessed by simulating multi-country models that can trace out the global impact of shocks and policies originating in the systemic country.
Potential spillbacks are outward spillovers that boomerang back to the systemic country that is the source of the outward spillover. This can trigger a policy response in the country that has the effect of modifying the impact of the original spillover. Spillbacks are often underestimated as they tend to occur through channels missing from models. Thus, expert judgment is critical for assessing this type of spillover.

25. Analysis of spillbacks can help the Fund gain more traction with systemic countries in discussions of outward spillovers. The mandates of authorities in systemic countries generally oblige them to focus on the domestic impact of shocks and not the impact of their policies on the rest of the world. This can complicate discussions with the Fund of alternative policies to mitigate outward spillovers—as long as they do not adversely affect domestic stability, as required under the ISD. However, when outward spillovers can be shown to have material second-round “spillback” effects authorities are likely to be more open to considering alternative policies.¹¹

26. While the Fund has focused on global risks, further efforts can be made to assess regional spillovers.¹² REOs and clustered report identify risks at the regional level effectively but tend to focus on real (i.e., trade) economy spillovers as they have found financial spillovers to be smaller (See Box 2). So far, this analysis has concentrated on spillovers within a region. It could be broadened to cover spillovers between regions; for example when a financial shocks originating in a large emerging market country in one region spills over into EMs in other regions.

Box 5.

Regional Spillovers

Regional macroeconomic and financial linkages are potentially significant sources of spillovers. This issue has attracted increasing interest in the context of the general slowdown of major emerging economies. Regional linkages could occur through a wide range of channels including trade, migration, remittances and tax spillovers.

Fund surveillance has used a variety of techniques to assess these linkages. A number of Regional Economic Outlooks have analyzed the impact of the regions’ leading economies on their neighbors’ output using econometric tools such as panel regressions, VARs and GVARs (see table). The results suggest that major emerging countries can generate significant regional spillovers through trade while regional financial market linkages are generally less significant.

There is scope to apply a wider range of analytical techniques—along with judgment—to assess regional spillover channels. A more systematic assessment of spillover channels, both quantitative and qualitative, would be useful. For example, many of the panel regressions provide estimates of the overall spillovers, but do not identify individual spillovers channels, and could be complemented with a VAR that models specific spillover channels. VARs can miss some channels, and judgment should play a role to ensure comprehensive coverage.

The following table summarizes a selection of recent studies on regional spillovers.

Study	Country	Spillover Channels (qualitative assessment)	Econometric Tool	Results
REO Asia and Pacific, 2014	China	• Mainly trade •Limited regional financial integration	Panel regression	Significant regional spillovers: 1 percent decline in China’s growth would lower GDP growth in the median Asian economy by 0.3 percent, twice as much as the estimated impact for the median non-Asian economy
REO, Sub Saharan Africa, 2012	South Africa	• Trade •Fiscal (SACU customs-sharing mechanism) Monetary (Lesotho, Namibia and Swaziland peg their currencies to the rand) •Remittances •Financial (FDI and subsidiaries of South African banks in the region)	Dynamic Panel Regressions and VAR	No significant regional spillovers are found in the study
REO Western Hemisphere, 2012	Brazil	•Mainly trade •Limited regional financial integration	VAR	Significant impact on the Southern Cone economies: 1 percent decline in Brazil’s growth would reduce Paraguay’s output by 0.9 percent after a year and Argentina, Bolivia and Uruguay’s output by about 0.25 percent
India Article IV report, 2014	India	•Trade •Monetary (Nepal pegs its currency to the rupee) •Financial (FDI) •Remittances	GVAR	Significant spillovers on Nepal (1 percent decline in India GDP would lower Nepal’s GDP by 0.21 percent after a year) but modest impact on the rest of the South Asian countries

Study	Country	Spillover Channels (qualitative assessment)	Econometric Tool	Results
REO Asia and Pacific, 2014	China	• Mainly trade •Limited regional financial integration	Panel regression	Significant regional spillovers: 1 percent decline in China’s growth would lower GDP growth in the median Asian economy by 0.3 percent, twice as much as the estimated impact for the median non-Asian economy
REO, Sub Saharan Africa, 2012	South Africa	• Trade •Fiscal (SACU customs-sharing mechanism) Monetary (Lesotho, Namibia and Swaziland peg their currencies to the rand) •Remittances •Financial (FDI and subsidiaries of South African banks in the region)	Dynamic Panel Regressions and VAR	No significant regional spillovers are found in the study
REO Western Hemisphere, 2012	Brazil	•Mainly trade •Limited regional financial integration	VAR	Significant impact on the Southern Cone economies: 1 percent decline in Brazil’s growth would reduce Paraguay’s output by 0.9 percent after a year and Argentina, Bolivia and Uruguay’s output by about 0.25 percent
India Article IV report, 2014	India	•Trade •Monetary (Nepal pegs its currency to the rupee) •Financial (FDI) •Remittances	GVAR	Significant spillovers on Nepal (1 percent decline in India GDP would lower Nepal’s GDP by 0.21 percent after a year) but modest impact on the rest of the South Asian countries

C. Deepening Analysis of Risks and Spillovers

27. The Fund’s strategy amounts to an eclectic analytical approach that combines quantitative analysis with judgment. The aim is to integrate the various strands of surveillance more effectively. The analytical toolkit that the Fund has developed over the past few years is diverse. It includes cross-country models and techniques that allow staff to tackle key questions with consistency and rigor. However, as emphasized in the parallel external study,¹³ formal models may fail to capture all the key features of some risks, and certain issues are best tackled by drawing on judgment and anecdotal evidence. The Fund has also developed more specialized models for analyzing certain types of risks. All of these models and techniques have limitations, and their interpretation must reflect judgment informed by in-depth knowledge of country circumstances.

28. Members’ calls for deeper analysis can be addressed by better integrating and further developing existing tools, and—where necessary—adding new tools. To a large extent, this involves refining existing diagnostic tools so they can be used more widely by staff. Large models capable of running multi-country scenarios can provide a wide range of simulations on how global shocks will impact a large number of different countries. Use of results from these models in Article IV reports has so far been limited largely to S5, systemically important economies that have been the focus of Spillover Reports until 2013. Bilateral surveillance in many more countries can draw on the Spillover Report analysis with the broadening of the country coverage in the scenarios.

Diagnosing Vulnerabilities and Risks

29. A number of specialized tools have been developed to assess risks and vulnerabilities, and can be better tailored for wider use in surveillance. For instance, tools used in the Vulnerabilities Exercises (VEs), which are internal exercises to inform management of vulnerabilities in individual countries, could be applied more systematically. Debt Sustainability Analysis (DSA), which is already widely used to assess public sector risks in surveillance, could be applied more systematically to assess external sector risks, including risks from external flows. The Balance Sheet Approach (BSA) could be used more extensively thanks to improvements in data availability, and banking sector stress testing models, which are primarily used at present for FSAPs, could be applied in Article IV consultations. Finally, there is scope to integrate analysis of credit into macroframeworks to better understand, for example, risks embedded in baseline projections if either deleveraging or rapid credit growth is taking place.

Bringing the VE Models into the Mainstream

30. The VEs¹⁴ were created to facilitate the Fund’s internal assessment of risks and vulnerabilities consistently across countries. The exercise gives countries overall vulnerability and crisis risk ratings that are built up from sector ratings (e.g., external, fiscal, corporate/real and financial), based on model results, which are then adjusted by country teams based on their local knowledge. It provides a framework for Fund staff to compare sectoral vulnerabilities across countries.

31. To promote wider use of these models in Fund surveillance, a number of issues need to be addressed:

Vulnerability ratings models. The methodologies for deriving vulnerability ratings for the VEE and VE-LIC exercises have been discussed by the Board.¹⁵ In view of the wide variety and complexity of the models used in these exercises, greater transparency on their methodologies would promote their integration in surveillance.
Crisis risk models. The crisis risk ratings rely largely on a non-parametric approach that uses cross-country data on crisis episodes to estimate crisis risk thresholds for a number of indicators. Making the underlying models public would allow them to be much more widely used in Article IV surveillance.
Political risk ratings—are based on external cross-country analysis supplemented by judgment of country desks. Political risk can be an important indicator of a country’s capacity to respond to shocks and is routinely covered in Fund surveillance.

Extending External DSAs to Assess Risks from External Flows

32. Extending the Fund’s external DSA—which currently assesses only external solvency— to cover also external flows could provide a useful indicator of the likely impact of external shocks. It can be constructed for any Fund member to provide a summary indicator of the likely impact of aggregate shocks to external flows, based on scenarios tailored to country circumstances, and with a breakdown by channel (e.g., trade, remittances, portfolio flows, FDI). An example for an emerging market (Box 6) shows some scenarios and demonstrates how some channels are considerably more powerful than others for shocks calibrated to historical experience.

33. An external DSA could be tailored closely to country circumstances. It could illustrate the impact of shocks, either derived from global scenarios (e.g., based on the Lehman crisis), or the country’s own experience in periods of stress (e.g., a standardized shock of two standard deviations). The shocks affect countries’ (i) gross financing needs (such as export volumes or the terms of trade) and (ii) access to external financing (such as FDI or portfolio flows) to cover these needs. The output of the external DSA is expressed in the form of the instantaneous financing gap that would emerge absent a move in the exchange rate or other variables.

Box 6.

Using an External DSA Model to Assess External Risks

An emerging market country example can illustrate the benefits of extending the external DSA to think more systematically about how shocks can affect external flows. Scenarios can be developed by applying calibrated shocks to the determinants of external financing needs and financing sources, with the impact expressed as a financing gap normalized by GDP. Two alternative approaches for calibrating the shocks are shown in this case:

Country-specific shocks. The country’s own experience can be used to size shocks. This is the normal practice for DSAs, which typically show scenarios based on shocks amounting to two standard deviations of individual variables, along with combined shock scenarios based on half a standard deviation shock to a number of variables. This approach gives a sense of how shocks typically play out in the country in question.
Global shock. An alternative approach is to consider how a well-articulated global shock scenario would affect a country (Figure, upper-left panel). ^1/

The figure shows the main elements of the analysis. This example consists of both standardized shocks and country-specific shocks. The size of the financing gap under each shock is measured in percent of GDP as well as against the stock of international reserves.

^1/ The global shock assumes (i) FDI inflows drop by 15 and 25 percent, respectively, in 2014 and 2015; (ii) rollover rates of both short-term debt and medium- to long-term debt drop to 90 percent in 2014 and 85 percent in 2015; (iii) outflow of portfolio equity at the magnitude of 10 percent of total stock in 2014, and 20 percent in 2015; and (iv) higher interest rates on new debt issued.

Reviving Balance Sheet Analysis

35. The balance-sheet approach (BSA) can now be used more widely as data availability has improved substantially. The BSA, developed in the early 2000s, tracks gross asset and liability positions between sectors, and can be used to assess emerging risks.¹⁶ Balance sheet analysis starts with the BSA matrix of assets and liabilities positions of the main sectors—government, financial sector, non-financial (e.g., corporates and households) and non-residents—broken down by currency and maturity. The matrix captures claims across sectors, and can help assess the transmission of shocks between sectors. The BSA has been applied intermittently to some emerging markets and a few advanced markets, and there is scope for it to be used more widely given the improvement in data.

36. The BSA can deepen our analysis of risk by generating sector balance sheet risks indicators and mapping linkages among sectors. The BSA risk indicators are standard—the leverage ratio, maturity mismatch indicators to capture rollover risk, and currency mismatch measures—but have the advantage of being granular, allowing more precise identification of risks in particular sectors. The most important benefit, however, is that the inter-sectoral linkages mapped by the BSA matrix make it possible to trace out how shocks could be transmitted between sectors. Figure 3 illustrates how this matrix could be used: an initial shock from, say, a rise in contingent liabilities originating in the government sector is transmitted to the banking sector through losses on its holdings of government debt, eroding its capital base. The shock is then transmitted to the corporate and household sectors through a reduction of credit from the banking sector.

37. National balance sheets can now be constructed for many countries. This can be done for the breakdowns of the four sectors shown in Figure 3 by combining data from several sources. The basic matrix can serve as a starting point for a more in-depth analysis drawing on more detailed sector breakdowns provided by national sources (e.g., for the household and corporate sectors separately), depending on the data availability. For countries that report all the financial, government, and external balance sheet data requested in Fund templates, complete balance sheet matrices are available, while partial balance sheets can be constructed for other countries depending on how much they report. Efforts to improve this reporting across countries will expand the scope for BSA. The analysis would also need to consider whether data are sufficiently granular to capture all relevant risks, such as those that arise in the shadow banking system or from derivatives. Given that the build-up in balance sheet mismatches tends to occur slowly, the depth of coverage could vary from year-to-year.

Mainstreaming Macro-Financial Analysis

38. Macro-financial analysis assesses the linkages between the financial sector and macro-economy. Much of the work in Article IVs so far has focused on shocks that lead to deviations from the baseline macroeconomic outlook rather than how developments in the financial sector influence the baseline itself, reflecting the lack of a standard framework for this. Articulating assumptions about macro-financial relationships in the baseline would help put the spotlight on emerging risks and the credibility of medium-term projections. Further research would be helpful for this to be mainstreamed. The Fund has developed a number of tools to assess the impact of financial shocks, including the balance sheet approach and bank stress tests (see below), but there may be scope to further develop its macro framework to better delineate the role of the financial sector in shaping the macroeconomic outlook.

39. This could help deepen Fund analysis of leverage cycles. It could assess, for example, whether the macroeconomic outlook is consistent with the capacity of a banking system to provide credit when it is deleveraging. The need for a richer treatment of how the determinants of credit in a financial system impact an economy was highlighted in the aftermath of the global financial crisis as bank deleveraging in advanced markets led to a much weaker recovery than many projected. In addition to such downside risks, where credit growth is insufficient to support growth, this analysis can also help identify the risk of unsustainable credit growth that could generate asset price bubbles and inflation. A recent example of how to integrate an analysis of the leverage cycle into the outlook is provided by Spain (Box 7). The relationship between output and financial variables could also be explored more closely through empirical methods, which would help assess the internal consistency of baseline projections (Box 8).

Box 7.

Incorporating Macro-Financial Linkages in the Baseline: The Case of Spain

After a decade of strong growth driven in large part by a credit-fueled housing boom, Spain was hard hit by the global financial crisis. The subsequent real estate bust, financial market turmoil, and rising borrowing costs pushed the economy into a sharp recession. Banks could not roll over their liabilities to foreign creditors, and refinanced them through the European Central Bank (ECB), to whom their liabilities approached 40 percent of the GDP by mid-2012.

Macro-financial linkages have continued to have powerful effects in the aftermath of the crisis. The Spanish economy is undergoing a large and prolonged adjustment to address its imbalances. Bank credit to the private sector has contracted sharply, reflecting not only weak demand but also supply factors such as pressure on the banks to reduce their liabilities to the ECB. The persistence of credit constraints is expected to weigh on domestic demand and output.

To highlight the consequences of private sector deleveraging on output, the Fund has integrated macro-financial linkages into its baseline projections for Spain. The 2013 Article IV consultation report ^1/ included five-year projections for private sector debt, anchored on an assessment of the financial sector as well as the borrowing patterns of households and non-financial corporations:

Households. The team envisaged subdued household consumption due to (i) ongoing deleveraging preventing a further decrease of the saving rate, (ii) the likelihood that households have already significantly exhausted their capacity to reduce their equity holdings and other financial assets.
Corporates. Although their capacity to invest or hire was restrained by lack of credit, at the same time investment has rebounded as operational margins have increased.

Incorporating macro-financial linkages in the baseline has helped staff avert potential inconsistencies in macroeconomic forecasts and policies during the financial sector program.^3/ Projections by banks of their balance sheets assumed they would rapidly reduce both their exposure to the sovereign and their reliance on ECB financing. Integrating this with staff’s macro scenario, which envisaged continued increases in public debt and no significant pick up in other sectors’ demand for government bonds, suggested that either credit would contract faster than banks projected (which would be negative for growth) or that ECB financing would have to contract less fast. This in turn, staff pointed out, would argue for continuation of supportive ECB policies and avoiding stigma being associated with the use of ECB facilities.

^1/ See Spain – Staff Report for 2014 Article IV Consultation.2/ SAREB (i.e., Spain’s bank asset management corporation).^3/ See Spain – Financial Sector Reform: Third Progress Report.

Box 8.

Assessing Macro-Financial Linkages Between Credit and Growth

Efforts to deepen analysis of credit and leverage in surveillance can draw on recent empirical research. At the IMF, for instance, the relationship between financial variables and macroeconomic activity has been comprehensively documented for more than a hundred countries from 1970 to 2013 using a crosscountry framework.^1/ This shows that the main macroeconomic indicators (GDP and its components, industrial production and employment) can be predicted by financial variables in a forecasting regression framework. Private sector credit growth, housing prices and equity prices have the strongest predictive power among financial variables. The relationship varies between advanced, emerging market, and low income countries, but remains strong.^2/

This research provides benchmark estimates of macro-financial linkages that can be useful in analyzing the consistency between credit developments and baseline macroeconomic projections. In particular, the regression results provide country teams with information on the distribution of key macro variables in response to shocks to financial variables. They can only serve as a guide, as the relationship between financial and macro variables will depend on country circumstances, and can change in countries implementing major financial reforms or undergoing financial deepening.

^1/ See Chen and Ranciere (2014, forthcoming).^2/ See Claessens et al., Lessons and Policy Implications from the Global Financial Crisis IMF WP/10/44.

Broadening Use of Bank Stress Testing

40. Bank stress tests can be used more widely in Fund surveillance to quantify macro-financial linkages and inform macroprudential policy advice. In countries that have recently had an FSAP, the FSAP stress tests can be updated, either by the authorities or the Article IV mission team, to give a sense of the evolution of risks. In other cases, a basic version of the well-established FSAP stress testing model can be applied in surveillance.¹⁷

41. A simple, top-down credit risk stress test can be carried out using existing spreadsheet tools. It assesses the impact of hypothetical macroeconomic shocks on bank asset quality and capital ratios (Figure 4). This focus on credit risk reflects the fact that it is for most countries the primary source of bank stress. The effect on the soundness of a financial system is evaluated by comparing the lower, post-shock capital ratio to its regulatory minimum. The stress test has two main elements: (i) a simple econometric model relating growth and other macro variables to NPLs; and (ii) a spreadsheet tool incorporating simplified accounting relationships that link NPLs to bank capital.¹⁸ This can typically be done with published data on balance sheets, NPLs and regulatory capital for individual banks that make up a large part of a financial system. More sophisticated stress tests are possible—and carried out on FSAPs—and Figure 4 illustrates how the basic framework can be extended.

42. While bank stress tests vary in sophistication, they have limitations. First, while they are designed to capture the impact of macroeconomic and capital market shocks on banks, they often fail to capture the macro-financial linkages that cause significant feedback effects. The models typically lack, for example, a channel that captures the effect of a fall in the capital ratio on banks’ decision to deleverage and curb private sector credit. Second, the reliability of the stress tests depends on data quality, which in turn reflects the quality of supervision and accounting systems, which can be distorted by factors such as ever-greening. Third, national regulations—which govern how potential losses change regulatory capital and liquidity ratios—vary across countries and may not be adequately captured by the accounting relationships in the model. There is no canonical stress testing model; rather, each model needs to be tailored to the country based on a set of guiding principles.

Putting it all Together: Applying Models for Spillover Analysis

43. The goal of spillover analysis is to identify and quantify the channels through which risks are transmitted internationally. A defining feature of this analysis is the use of models containing a large number of countries, reflecting the IMF’s unique responsibility for global stability. This need to quantify spillovers leads to a heavy reliance on large, multi-country models that can simulate the scenarios that illustrate how outward spillovers from a shock in a systemic country are transmitted to other countries. This contrasts with the more specialized empirical tools used for diagnosing risk outlined above.

44. The specification and properties of these models are improving steadily but have limitations that need to be taken into account. Country coverage is small relative to the membership (at most 40 countries) and their capacity to incorporate differences in economic structure is limited. Some spillover channels are difficult to capture in models and, thus, may be overlooked, such as the impact of regulatory changes (discussed below). Also, the impact of a shock may be underestimated when shocks have non-linear effects that are not captured well by most models, which are largely linear in specification. Finally, there are significant data constraints that can limit the application of these models (see Box 9). Thus, surveillance of spillovers must use models cautiously and rely on judgment to identify potential spillovers that are not adequately captured by models.

Box 9.

Data Gaps Impeding Structural Macroeconomic Modeling

The Fund’s efforts to further develop structural macro models are running into data constraints. In response to user demand, model builders have expanded country coverage, especially of emerging markets, and included new sectors to capture important economic relationships. Such models critically depend on interrelated national, government, and balance of payments accounting identities. It will be difficult to further develop these models without confronting data gaps, which exist across economies and sectors. Key gaps include the following:

Expenditure breakdown. The calibration or estimation of these models depends on a minimal expenditure breakdown of the quarterly national accounts, which remains unavailable even for some G20 economies. In the absence of data, model builders use estimates, which introduce measurement error.
Macro-financial linkages. Decompositions of private consumption (nondurables versus durables) and private investment (residential versus business) are vital for understanding linkages, because these subcomponents respond very differently to changes in financial conditions.
Capital stock data. These are important for behavioral relationships in large macro models and it is critical to obtain data on residential and business capital stocks for a larger group of countries. Likewise, a lack of data on the stock of public infrastructure impedes some fiscal applications.
Import content and trade linkages. The import content of expenditure components varies markedly, which can interact with domestic demand changes to generate shifts in trade flows. The lack of a full expenditure breakdown and incomplete import content data in many advanced economies (and most EMs) means that modeling of many interconnected economies must be done at a higher level of sectoral aggregation, and risks in subsectors will not be captured by the models.
Trade linkages. In view of the growing importance of trade in services, it would be desirable to complement existing bilateral export and import data for goods with comparable data for services. Without this detail, model builders assume that services trade patterns match those of goods—which can significantly distort estimated spillovers.
Financial linkages. Attempts to model these have focused on cross-border balance sheet exposures and contagion effects. Accurately representing these transmission channels and accounting for their impacts requires measures of private and public sector borrowing costs at different maturities. While sovereign borrowing costs are readily available, corporate yields are typically only available for individual securities, and data are needed on internationally comparable benchmarks or economy wide indexes. Accurately accounting for the associated financial income flows requires data on the maturity structures of government debt and net foreign asset positions reported in a consistent and internationally comparable way.

45. The IMF deploys three broad classes of multi-country models for spillover analysis: structural macro models (known as Dynamic Stochastic General Equilibrium—DSGE—models); Global Vector Auto-Regression (GVAR) models; and balance sheet contagion models. Each has advantages that are outlined below. Structural models have a well developed theoretical structure that makes them well suited to policy analysis and facilitates the interpretation of spillovers results. The GVAR can be easily implemented for different sets of countries and variables, making it a highly flexible tool but the less developed theoretical structure makes the results less amenable to interpretation. The balance sheet contagion models are essentially matrices of data on cross-border exposures that can be used to trace spillovers between countries but are based largely on accounting identities and do not capture behavioral relationships well.

Structural Multi-Country Models

46. The Fund has developed a number of structural, multi-country macro models for spillover analysis that each embodies different trade-offs in model design. Specifically, the large size of these models leads to a trade-off between the degree of country differentiation and the variety of spillover channels (Annex II). These differences in models need to be taken into account in how they are used. Specific types of scenarios are more suited to one model than another and can be expected to generate different simulation results, reflecting the different design choices. Some parameters in these models must be imposed rather than estimated, which needs to be taken into account when assessing the results.

The Flexible System of Global Models (FSGM) involves design choices that allow a greater degree of differentiation across countries and, thus, better captures the differential impact of a shock on each individual country in the model. The model can accommodate up to 24 countries. In order to achieve this diversity while keeping the dimensionality of model to a size that allowed to be simulated, the specification of the cross-country linkages were simplified. Specifically, the complex set of bilateral trade linkages between countries was replaced by a single reduced-form equation, and financial linkages with the rest of the world are based on uncovered interest parity.
The G-35S Model made a different design choice: to have more detailed bilateral linkages among countries and to limit differentiation across the 35 countries. Thus, it includes a wider variety of financial contagion channels but imposed common structural parameters across countries, obtained from panel estimation, to keep the dimensionality of the model to a size where it can be simulated.
The Global Integrated Monetary and Fiscal Model (GIMF) has bilateral trade linkages and a rich dynamic structure that includes a household sector with overlapping generations and firms with a financial accelerator; but, to achieve this complexity, it can accommodate a maximum of six countries/regions.

Box 10.

Applying the CCA-GVAR: An Italian Case Study

The GVAR can also be combined with the Contingent Claims Analysis (CCA) model to deepen the analysis of financial spillovers. CCA produces an indicator of financial distress than can be used as dependent variables in the GVAR. Other variables such as bank equity prices, NPL ratio, and CDS spreads could be used, but CCA variables efficiently combines forward-looking information from financial markets and firms’ balance sheet data. The GVAR-CCA model has been used to quantify domestic sovereign-bank linkages and outward spillover effects for Italy, drawn from the recent Italian FSAP.^1/ This GVAR includes 17 countries and CCA-based financial distress indicators for banks, non-financial corporate, and the sovereign, as well as the growth rates for bank credit and real GDP as variables.

Cross-sector linkages within Italy (Figure, panels 1–3): Italian banks are exposed to other sectors (panel 2). However, the effects from banks to the rest of the economy seem relatively small: bank distress generally causes a smaller impact on the sovereign credit spread and GDP growth (“bank” bars in panels 1 and 3) compared to the shocks to the other sectors.
Outward spillovers from Italy (Figure, panels 4–6): There are spillover effects from an increase in Italian sovereign credit spreads to sovereign spreads of other stressed economies (panel 4), although its effects on the real sector are small (panel 5). Cross-border spillovers to GDP are mostly channeled though real eco nomic linkages.

Source: Gray, Gross, Sydow and Paredes “Modeling the Joint Dynamics of Banking, Sovereign, Macro and Financial Risk using CCA in a Multi-country Global VAR,” IMF Working Paper (forthcoming).^1/Italy: Technical Note on Interconnectedness and Spillover Analysis, IMF, November 2013.2/ Cumulative 24-month response to a one standard deviation shock in a sector. Based on GVAR estimate for 16 countries and five variables (sovereign, bank, corporate, real GDP, and bank credit) using data for 2002–12. The estimation reflects the average relationship over the past decade (which includes a long period of tranquility, especially regarding the sovereign credit risk indicator) vis-à-vis a standardized shock. Other stressed economies include Ireland, Portugal, and Spain.

Global Vector Auto Regression models

47. The GVAR model provides a practical alternative to structural macro models. They can be estimated quickly and for groups of countries that may not be covered by the structural macro models. They also provide the flexibility to vary the choice of variables to incorporate spillover channels that may not be adequately captured in structural models. An example, discussed in Box 10, is the addition to a GVAR model of a variable capturing market perceptions of the risk of a systemic crisis derived from a Contingent Claims Analysis (CCA) model.

48. GVAR models have the advantage that they estimate jointly the linkages within an economy and spillovers between countries. GVARs typically report results in the form of impulse response functions, which are standardized shocks (e.g., of one standard deviation) that facilitate comparison of the impact across countries. The estimated model can be used to run policy scenarios but the results are harder to interpret given that they are a reduced-form model with a less developed theoretical structure (than structural models). The estimated statistical relationships in GVAR can be prone to parameter instability, which can lessen their reliability.

Balance Sheet Contagion Models

49. Bank contagion analysis applies stress testing concepts to assess spillovers between national banking systems. It relies on balance sheet identities that link banking systems across countries, which allow the impact of a shock in one banking system to be traced to others. For example, losses in one banking system could cause it to deleverage by cutting international interbank lending and by selling asset in global markets which, in turn, triggers losses and deleveraging in other systems. This model is described in Box 4, which also highlights how our capacity to do this analysis is constrained by the lack of sufficiently detailed disaggregation in the BIS banking statistics.

50. Progress is continuing, with the development of a Global Flow of Funds (GFFs) by the IMF in cooperation with a number of jurisdictions.¹⁹ This involves expanding the external sector of the BSA matrix with data on the bilateral exposures between the sectors among countries, as illustrated for a pair of countries in Figure 5. This work is still in the early stages and a full GFF matrix for multiple countries cannot yet be produced. However, existing BIS international banking statistics and IMF IIP data can be used to partially construct a GFF matrix for cross-border bilateral exposures through banking and portfolio channels, which can allows mapping of financial spillovers. This detailed disaggregation of a country’s external sector into a network of bilateral exposures makes the GFF into a useful surveillance tool that permits a much more granular mapping of potential spillover channels, and would greatly enhance the Pilot ESR’s coverage of external balance sheets and their linkages.

Spillovers from Financial Regulation and Macroprudential policies

51. Financial regulatory reforms and macroprudential policies could potentially have large spillovers—and spillbacks—but existing models are not well suited to capture these effects. These reforms strengthen the resilience of the financial sector, but they could have significant but unintended adverse cross-border spillovers.²⁰ Scenarios driven by regulatory reform therefore involve a more substantial element of judgment. Specifically, the increase in capital and liquidity buffers under Basel III could lead to deleveraging with adverse impacts on credit. Moreover, it could make G-SIFIs less willing to hold emerging market securities and credit due to their higher capital charges, impacting global capital flows. The requirement that G-SIFIs have additional capital charges would have similar effects as Basel III.

Annex I. Implementation of the ISD: 2013 Article IVs

This annex discusses early experiences from the first year of implementing the Integrated Surveillance Decision (ISD), based on an assessment of 11 Article IV consultation reports. ²¹ All of these reports examined outward spillovers, although they varied in terms of the depth of analysis and coverage of transmission channels. Most of the reports spelled out alternative policies, although they did not necessarily document spillovers from the country’s existing policies. It is unclear whether staff gained much traction with authorities in discussions of outward spillovers.

A. Scope of Spillover Analysis

All 11 Article IV staff reports contained at least some coverage of outward spillovers, although the scope of the analysis varied. Five reports covered both actual and potential spillovers from current or alternative policies, while five considered only potential policy spillovers. A number of reports discussed potential benefits from alternative policies (“positive spillovers”) without always examining the negative spillovers from current policies as envisaged by the ISD. The benefits from alternative policies may in some cases be the mirror image of the spillovers from current policies, although this was not always clarified.

Coverage across reports could be more consistent. Some spillovers were covered as inward spillovers in the reports of countries that might be affected, but received only limited coverage in the originating country. For instance, the 2013 China Article IV did not examine spillovers from a sharp slowdown in growth triggered by a build-up of excess capacity or capital flow reversals, which was identified as a risk by a number of other countries, although this issue had been discussed in the 2012 China Article IV. Similarly, some outward spillovers were discussed in the Spillover Report, but did not appear in the corresponding Article IV report. Moreover, outward spillovers of policies or shocks highlighted in the Euro Area report, such as spillovers from the stagnation of medium-term growth, did not always receive attention in the report of individual member countries where they seemed relevant.

Box. Impact of the Integrated Surveillance Decision

The ISD clarified the scope of risk analysis and made spillover analysis a mainstream feature of Article IV surveillance. It requires Article IVs to cover members’ policies that can significantly influence their own domestic or balance of payments stability.^1/ Coverage of outward policy spillovers is also required in certain situations, including where a country’s policies have spillovers that “significantly influence” global economic and financial stability, which in the remainder of this paper are referred to as “systemic spillovers”. In such cases, the Article IV should cover actual or potential outward spillovers from all of the members’ policies and through all relevant channels (BOP or non-BOP, e.g., contagion, or market pricing effects).^2/ If a country’s policies are promoting domestic stability—yet could cause adverse systemic spillovers—staff should discuss with the authorities alternative policies that minimize spillovers while continuing to promote domestic stability.

The ISD also calls for bilateral and multilateral surveillance to be more closely integrated. These two strands of surveillance are mandated by different clauses of the Fund’s Articles of Agreement; and, until recently, have been conducted in parallel. The ISD establishes a conceptual link between them, and makes the case for much closer operational integration to help strengthen the coherence of the Fund’s risks and spillover work.

^1/ Previous guidance was for Article IVs to focus only on policies affecting the member’s own economic and financial stability. Although many teams already looked at a wider range of risks and spillover channels, coverage was uneven, which may have contributed to the failings of surveillance in the run up to the crisis.^2/ Outward spillovers are deemed to “significantly influence” the effective operation of the international monetary system (hereinafter “significant impact on global stability”), if by themselves, or in combination with spillovers from other members’ policies, or through their regional impact, they enter the macro-financial policy considerations of members representing a significant portion of the global economy. Outward spillover analysis is also required for any spillovers where a country is not promoting its domestic or balance of payments stability.

B. Depth of Analysis

The reports varied in the degree to which they quantified spillovers. Half of the reports quantified outward spillovers, all drawing from quantitative models used in the Spillover Report. Two reports elaborated the models and discussed the quantitative impact, while one report mentioned the quantitative impact without discussing the models used. Some reports, including the US Article IV consultation, did not seem to draw extensively on quantitative tools.

Transmission channels were discussed in most reports, but coverage was sometimes narrow. Currently available models, such as those used in the Spillover Report, do not capture all transmission channels, and generally do not capture the impact of financial regulations or capital flow/portfolio rebalancing channels. This forced many reports to impose simplifying assumptions for some sectors, or meant that they were unable to include very rich dynamics. Nevertheless, some reports flexibly combined results from different results and applied judgment on the effects of other channels to produce a richer analysis of the impact of spillovers.

C. Integration and Traction

The outward spillover discussion was well integrated into most reports. They varied, however, in the weight they attached to spillovers, and only a minority of the reports alluded to outward spillovers in the staff appraisal. The authorities’ reactions to the spillover work (including alternative policies) were not recorded in half of the reports, which raises uncertainties about how much traction staff gained with the authorities from discussing outward spillovers.

Most Article IV consultation reports discussed the spillovers of alternative policies. Six reports provided alternative policies to mitigate outward spillovers, in most cases reflecting “positive” spillovers from policies that are designed to be welfare enhancing domestically. In some cases, staff recommended policy packages, and it was hard to discern the impact of the various components of the package.

Annex II. Lessons from a Seminar on Modeling Risks and Spillovers at the IMF

The IMF relies on a variety of economic models to assess risks and spillovers. To clarify how different models should be used in surveillance, a seminar was held in April 2014 to obtain the perspectives from seven academics and practitioners, as well as model users. They assessed the IMF’s three main structural macroeconomic models for spillover analysis and a variety of specialized models, focusing on specific applications illustrating how the models were used. This annex summarizes the conclusions from the seminar.

Challenges of Using Models in IMF Surveillance

Models provide a rigorous analysis of macroeconomic processes and help quantify the impact of shocks. They necessarily simplify the complex systems they seek to describe, which can limit their usefulness. The challenge for surveillance is to make effective use of models while taking account of their limitations. The decision regarding the extent to which one can rely on models or must rely on judgment can benefit from a deeper understanding of their strengths and weaknesses provided by the seminar. The seminar also helped identify directions in which to develop models to enhance their usefulness.

Three Multi-Country Structural Macroeconomic Models

Three large structural DSGE models have been developed. The coverage of multiple countries by these models—reflecting the global role of the IMF—makes them uniquely suited to assess international spillovers. However, their larger size relative to single-country models gives rise to tradeoffs in model design, notably between the degree of country differentiation and the variety of spillover channels. Each of the models embodies a different tradeoff that results in important differences in their properties and in the scenarios for which they are used.

The Flexible System of Global Models (FSGM) is designed to allow substantial differentiation across a large number of countries and, thus, better captures the differential impact of shocks across up to 24 countries. To allow this flexibility while keeping the model solvable, crosscountry linkages are simplified, with countries’ bilateral trade linkages replaced by a single reduced-form equation, and simple financial linkages based on uncovered interest parity. The FSGM has been used to generate a global downside scenario in which recovery in the U.S., Europe, Japan and emerging markets is weaker that the WEO baseline.
The G-35S Model made the strategic choice to have detailed and varied international linkages, including by building in financial contagion channels, but limited differentiation across the 35 countries.²² To make the model solvable, panel estimation of common structural parameters across countries is used to reduce the dimensionality of the model. The G-35S has been used to model spillovers from early exit from unconventional monetary policy (UMP) due to faster U.S. growth, which led to a sharp rise in U.S. interest rates with negative financial spillovers that partly offset the positive ones from stronger U.S. growth.
The Global Integrated Monetary and Fiscal Model (GIMF) has bilateral trade linkages and a rich dynamic structure that includes a household sector with overlapping generations and firms with a financial accelerator,²³ but this complexity limits the number of countries/regions it can handle to six. The model also has relatively simple financial linkages based on uncovered interest parity. It has been used to study a euro area fiscal consolidation in which structural reforms help offset the growth impact of the consolidation.

Directions for Future Work

The design choices embodied in these models results in differences in their coverage of international linkages and, therefore, in the types of scenarios they are most suited to assess in Fund surveillance. The outside academics and practitioners outlined the direction in which they thought IMF modeling work could be strengthened:

Efforts should focus on better capturing financial spillovers. This includes the non linear effects that amplify spillovers during episodes of financial stress. The use of event studies to calibrate spillovers is one approach but this inevitably involves a large degree of judgment.
Financial spillovers could be better captured by building balance-sheet effects into models. This would allow them to incorporate the build-up of currency and liquidity mismatches that increase vulnerability to exchange rate and capital flow volatility.
Spillover analysis using models could distinguish more clearly between spillovers from the initial shocks, the policy responses and second round spillovers (i.e., spillback effects).
More realistic modeling of capital flows and financial account openness is needed. Significant departures from uncovered interest parity occur in practice, and portfolio balance effects need to be incorporated.

The Role of Specialized Models in IMF Surveillance

The limitations of large structural macro models mean that the Fund also needs specialized models to assess specific risks or spillovers in greater depth. These typically cover only some of the risks and spillover channels and lack the rich theoretical structure of the large macro models. Specialized models rely to a greater extent on accounting identities to define spillover channels, or on econometric methods to estimate them. This limits the extent to which they can be used for policy scenarios needed to assess policy spillovers, relative to structural models. The IMF uses many specialized models and the seminar had to be selective and choose several representative models on which outside academics and practitioners could comment

Country crisis risk models that use a nonparametric statistical technique that can combine a large number of indicators of vulnerability to come up with a crisis risk rating for each country.
A bank contagion model that relies on balance sheet identities linking banking systems across countries. It enables users to trace the impact of shocks, but can capture only a limited set of spillovers and does not incorporate offsetting private sector and policy responses.
Global vector auto-regressions (GVARs) provide a flexible econometric approach to estimating spillovers for country groupings but are not well suited for policy scenarios that assess alternative policies.

Directions for Future Work

The IMF’s specialized models offer greater flexibility than structural models to delve more deeply into specific risks such as, for example, tail risks scenarios. Future work needs to take account of these more diverse objectives while helping to strengthen these models. Models could be developed in the following directions:

Allowing for “time variation” between crisis and non-crisis episodes would help to assess the impact of shocks more accurately. Failure to allow for this leads to underestimation of the role of country vulnerabilities and a failure to appreciate the larger spillovers that can occur when these are weak.
Models need to incorporate liquidity risk more effectively. This involves capturing better the sensitivity of external funding (in the form of portfolio or international banking flows) to shifts in investor perceptions of default risk in the banking, sovereign and corporate sectors.

The leverage cycle in financial sectors driven by global liquidity conditions needs to be better incorporated into models. This involves more accurate measurement of leverage (i.e., to capture leverage from corporate bonds issued offshore) and an assessment of the risk that capital supporting this leverage may prove inadequate in the face of adverse shocks.

Prepared by a staff team comprising Ding Ding (APD), Qianying Chen (EUR), Hiroko Oura and Francis Vitek (both MCM), Eugenio Cerutti, Ayhan Kose and Esteban Vesperoni (all RES), Evrim Bese Goksu, Noriaki Kinoshita and Manik Shrestha (all STA), Sean Craig, Gavin Gray (lead), Jean Frederic Noah Ndela Ntsama, Rania Papageorgiou, Michael Perks and Veronique Salins (all SPR). David Marston (RMU) provided general guidance.

At the same time, calls for work on spillovers date back to at least the 1990s, when the Asian financial crisis demonstrated the significance of cross-border linkages. See IEO (1999).

Financial Surveillance Strategy, IMF 2012.

The third pillar of the FSS was to engage more effectively with stakeholders in order to improve the traction of financial sector surveillance.

TSR background papers on “Stakeholders’ Perspectives on IMF Surveillance” and a “Review of IMF Surveillance Products” provide more detail on country authorities’ views and on gaps identified in a review of surveillance products.

See the TSR External Study on “Risks and Spillovers” by David Li and Paul Tucker.

Section C discusses the individual models in greater detail.

Examples of usage in Fund research include Financial Soundness Indicators and the Characteristics of Financial Cycles, Lane R. P. and Gian Maria Milesi-Ferretti Cross-Border Investment in Small International Financial Centers, IMF WP No. 10/38, and Brushko, Iuliia and Yuko Hashimoto (2014) The Role of Country Concentration in the International Portfolio Investment Positions for the European Union Members, IMF WP No. 14/74.

Available data are posted on the Principal Global Indicators website.

As David Li and Paul Tucker note: “Bilateral surveillance will often be closer to the vulnerabilities and pathologies that give rise to international spillover risks and problems.”

Examples of spillbacks are provided in the TSR External Study on “Risks and Spillovers” by David Li and Paul Tucker.

A parallel staff study on the scope of “Surveillance in Low Income Countries (LICs)” covers the relevance of risks and spillover work for this group of countries.

See the TSR External Study on “Risks and Spillovers” by David Li and Paul Tucker.

The VE comprises three separate exercises covering advanced economies (VEA), emerging markets (VEE) and low-income countries (VE-LIC).

See Managing Volatility: A Vulnerability Exercise for Low-Income Countries (March 9, 2011), and The IMF-FSB Early Warning Exercise: Design and Methodological Toolkit.

Key papers developing the BSA are Allen, et al, IMF WP/02/210 and Mathisen and Pellechio, IMF WP/06/100.

Stress tests assess the soundness of the system as a whole. Staff should be careful about how they are communicated to avoid making implicit judgments about the health of individual banks.

These relationships are typically approximations that may not fully reflect regulatory treatment of banks in a country. If estimation is difficult, one can consider some arbitrary shocks (for instance, that the NPL ratio increases by 5 percentage points). A top-down test, which can be carried out by a country-authority or IMF staff using supervisory data, will usually be more accurate. A bottom-up test is a test carried out by banks themselves. See recent IMF policy paper on stress testing for typology and basic structures of stress tests, as well as a survey of stress testing practice of country authorities.

See Mapping the Shadow Banking System Through a Global Flow of Funds Analysis, IMF WP/14/10.

For instance, Santos and Elliott, 2012, Estimating the Costs of Financial Regulation, SDN/12/11; and Viñals, Pazerbasioglu, Surti, Narain, Erbenova, and Chow, 2012, Creating a Safer Financial System: Will the Volcker, Vickers, and Liikanen Structural Measures Help? SDN/12/4.

The reports covered China, Euro Area, France, Germany, Greece, Italy, Japan, Spain, Switzerland, United Kingdom, and United States.

See Policy Analysis in the World Economy: A Panel Unobserved Components Approach.

See The Global Integrated Monetary and Fiscal Model (GIMF)—Theoretical Structure, and Getting to Know GIMF: The Simulation Properties of the Global Integrated Monetary and Fiscal Model.

Other Publishers

Save
Cite
Email this content

Share Link

Copy this link, or click below to email it to a friend
Email this content
or copy the link directly:

https://www.elibrary.imf.org/view/journals/007/2014/059/article-A001-en.xml
The link was not copied. Your current browser may not support copying via this button.

Link copied successfully

2014 Triennial Surveillance Review - Staff Background Studies

Author: International Monetary Fund

Volume/Issue:: Volume 2014: Issue 059

Publisher:: International Monetary Fund

ISBN:: 9781498343077

ISSN:: 2663-3493

Pages:: 119

DOI:: https://doi.org/10.5089/9781498343077.007

Keywords
I. Risks and Spillovers1
Annex I. Implementation of the ISD: 2013 Article IVs
Annex II. Lessons from a Seminar on Modeling Risks and Spillovers at the IMF

View raw image

Quantitative Tools Used in 2013 Article IV Reports^1/

(percent of reports reviewed)

View raw image

Mapping Bank Inter-Linkages Using LBS And CBS

View raw image

Mapping Bank Inter-Linkages Using Adjusted CBS

View raw image

The IMF’s Surveillance Framework

View raw image

View raw image

Balance Sheet Approach (BSA) Matrix

View raw image

Spain: Credit to Private Sector

(percent growth)

View raw image

Spain: Savings & Investment

(percent of GDP)

View raw image

Structure of a Macro Scenario Test

View raw image

View raw image

Global Flow of Funds: Bilateral Cross-Border Exposures between Country Sectors

View raw image

Coverage of Actual and Potential Spillovers (Number of reports)

View raw image

Quantification of Outward Spillovers (Number of reports)

View raw image

Integration of Outward Spillover Analysis (Number of reports)

View raw image

Alternative Policy Options to Reduce Outward Spillovers (Number of reports)

Study	Country	Spillover Channels (qualitative assessment)	Econometric Tool	Results
REO Asia and Pacific, 2014	China	• Mainly trade •Limited regional financial integration	Panel regression	Significant regional spillovers: 1 percent decline in China’s growth would lower GDP growth in the median Asian economy by 0.3 percent, twice as much as the estimated impact for the median non-Asian economy
REO, Sub Saharan Africa, 2012	South Africa	• Trade •Fiscal (SACU customs-sharing mechanism) Monetary (Lesotho, Namibia and Swaziland peg their currencies to the rand) •Remittances •Financial (FDI and subsidiaries of South African banks in the region)	Dynamic Panel Regressions and VAR	No significant regional spillovers are found in the study
REO Western Hemisphere, 2012	Brazil	•Mainly trade •Limited regional financial integration	VAR	Significant impact on the Southern Cone economies: 1 percent decline in Brazil’s growth would reduce Paraguay’s output by 0.9 percent after a year and Argentina, Bolivia and Uruguay’s output by about 0.25 percent
India Article IV report, 2014	India	•Trade •Monetary (Nepal pegs its currency to the rupee) •Financial (FDI) •Remittances	GVAR	Significant spillovers on Nepal (1 percent decline in India GDP would lower Nepal’s GDP by 0.21 percent after a year) but modest impact on the rest of the South Asian countries

Topics

Countries

Series

About

Resources

Abstract

I. Risks and Spillovers1

Executive Summary

A. Framing and Evaluating Risk and Spillover Analysis

Introduction

Joining the Dots through Risk-Based Surveillance

Shaping the Agenda Ahead

Usage of Quantitative Tools

Data Gaps: Running to Stay in the Same Place

B. Consolidating the Framework for Risk and Spillover Analysis

An Integrated Approach to Risk Analysis

External and Financial Risks

Spillover Analysis

C. Deepening Analysis of Risks and Spillovers

Diagnosing Vulnerabilities and Risks

Bringing the VE Models into the Mainstream

Extending External DSAs to Assess Risks from External Flows

Reviving Balance Sheet Analysis

Mainstreaming Macro-Financial Analysis

Broadening Use of Bank Stress Testing

Putting it all Together: Applying Models for Spillover Analysis

Structural Multi-Country Models

Global Vector Auto Regression models

Balance Sheet Contagion Models

Spillovers from Financial Regulation and Macroprudential policies

Annex I. Implementation of the ISD: 2013 Article IVs

A. Scope of Spillover Analysis

B. Depth of Analysis

C. Integration and Traction

Annex II. Lessons from a Seminar on Modeling Risks and Spillovers at the IMF

Challenges of Using Models in IMF Surveillance

Three Multi-Country Structural Macroeconomic Models

Directions for Future Work

The Role of Specialized Models in IMF Surveillance

Directions for Future Work

Same Series

Other IMF Content

Other Publishers

Asian Development Bank

Inter-American Development Bank

The World Bank

Share Link

Table of Contents

Headings

Figures

Metrics

I. Risks and Spillovers¹