Canada
Financial Sector Assessment Program-Technical Note-Stress Testing and Financial Stability Analysis

This paper on Financial Safety Net and Crisis Management for the Canada reviews the stress testing and financial stability analysis. The paper highlights that the financial system’s performance has been strong. The insurance sector has remained financially sound even in the low interest rate environment. Major banks, life insurers, and pension funds have expanded their footprints abroad. Canada has strong financial linkages with the United States. Macrofinancial vulnerabilities—notably, elevated household indebtedness and housing market imbalances—remain substantial, posing financial stability concerns. Major deposit-taking institutions would be able to manage severe macrofinancial shocks; however, mortgage insurers would probably need additional capital. Major deposit-taking institutions also hold enough liquidity buffers to withstand sizeable funding outflows. However, increased balance sheet complexity and reliance on wholesale and foreign exchange funding, and the extensive use of derivatives are some areas of concern that would warrant closer monitoring by the competent authorities and a more comprehensive quality assurance in the context of supervisory or macroprudential stress testing exercise.

Abstract

This paper on Financial Safety Net and Crisis Management for the Canada reviews the stress testing and financial stability analysis. The paper highlights that the financial system’s performance has been strong. The insurance sector has remained financially sound even in the low interest rate environment. Major banks, life insurers, and pension funds have expanded their footprints abroad. Canada has strong financial linkages with the United States. Macrofinancial vulnerabilities—notably, elevated household indebtedness and housing market imbalances—remain substantial, posing financial stability concerns. Major deposit-taking institutions would be able to manage severe macrofinancial shocks; however, mortgage insurers would probably need additional capital. Major deposit-taking institutions also hold enough liquidity buffers to withstand sizeable funding outflows. However, increased balance sheet complexity and reliance on wholesale and foreign exchange funding, and the extensive use of derivatives are some areas of concern that would warrant closer monitoring by the competent authorities and a more comprehensive quality assurance in the context of supervisory or macroprudential stress testing exercise.

Executive Summary

The financial system’s performance has been strong. The banking sector has enjoyed solid profitability and sizeable capital buffers. The insurance sector has remained financially sound even in the low interest rate environment. Other nonbank sectors have grown considerably, with pension funds and mutual funds dominating the institutional and retail asset management landscape. Major banks, life insurers and pension funds have expanded their footprints abroad. Canada has strong financial linkages with the United States.1

Macrofinancial vulnerabilities—notably, elevated household indebtedness and housing market imbalances—remain substantial, posing financial stability concerns. During the decades-long credit upcycle, low interest rates, and low capital charges for mortgage lending, together with policies promoting housing ownership, have fueled borrowing to finance home purchases in the face of rapidly rising house prices. Downside risks to house prices in the medium term are sizeable given existing overvaluation, and Canada-specific housing finance characteristics may amplify procyclical effects of falling house prices due to borrowers’ refinancing pressures and lenders’ sudden adoption of risk-based mortgage pricing. During severe downturns, the household sector would be affected, with a significant increase in debt belonging to financially weak households, while the corporate sector would remain more robust.

Major deposit-taking institutions would be able to manage severe macrofinancial shocks, but mortgage insurers would probably need additional capital. In a severe adverse scenario, major deposit-taking institutions would be able to rebuild their capital positions to meet the regulatory requirements. Although credit and market risk related risks appear to have a substantial impact under stress, the relatively high starting point capital ratios, the robust structural profitability and the dual layered mortgage market structure that partially mitigates some of the risks provide a generous buffer for deposit-taking institutions. By contrast, mortgage insurers would face some capital shortfalls. Nevertheless, financial stability implications are limited given the government’s backstopping of mortgage insurance contracts.

Major deposit-taking institutions also hold sufficient liquidity buffers to withstand sizeable funding outflows. However, increased balance sheet complexity and reliance on wholesale and FX funding, and the extensive use of derivatives are some areas of concern that would warrant closer monitoring by the competent authorities and a more comprehensive quality assurance in the context of supervisory or macroprudential stress testing exercise.

Large life insurers appear somewhat exposed to financial market stress and lower interest rates. The solvency of some major life insurers could become under pressures during severe financial market stress, largely due to the impact of widening credit spreads and falling equity prices. This stems from the fact that life insurers hold a sizeable amount of low-rated and unrated bonds. Furthermore, life insurers’ solvency would be hit hard in a more sustained low interest rate environment.

Increased investor risk-taking, stretched asset valuations, and dependence on foreign creditors raise the risk of market volatility and a sharp tightening in financial conditions. Pension funds and other institutional investors have increased use of private market strategies and derivatives, which are adding to leverage and liquidity risks. Fixed income-focused investment funds have grown rapidly and have been increasing exposure to credit and duration risk, increasing the risk of a redemption shock. Fixed income and real estate asset valuations are stretched, while dependence on foreign investors for corporate bond market funding has increased significantly. In a market stress scenario, rising valuation and liquidity risks could magnify losses and amplify market volatility, with spillovers to banks and insurers owing to their holding of securities and derivatives. Furthermore, elevated dependence on foreign funding increases the risk that financial conditions would tighten sharply.

Additional required capital for mortgage exposures, along with measures to increase risk-based differentiation in mortgage pricing, are desirable. While banks’ overall capital buffers are adequate, lenders’ risk weights for mortgage exposures should be higher. Mortgage insurers’ capital requirements should also be tightened so that required capital is enough to absorb tail-risk shocks. In addition to properly accounting for through-the-cycle credit risk, these measures can help improve mortgage risk-pricing and limit procyclical effects of falling house prices. In addition, risk-based pricing of insured mortgages should be improved by increasing the risk sensitivity of insurers’ capital requirements or guarantee fees paid to the government and limiting insurance coverage of loans that fund insurance premiums.

Enhanced risk monitoring is essential especially in the areas of emerging vulnerabilities. These include (i) banks’ external, foreign-currency funding; (ii) extensive use of derivatives; (iii) rising risk-taking by life insurers, pension funds, and other nonbanks; (iv) non-prime mortgage lending outside the regulatory perimeter and HELOCs; and (v) spillovers from overseas operations and cross-border exposures. Continued efforts to address data gaps—particularly related to cross-sectoral exposures, unregulated nonbank financial intermediation, and funding market activities (e.g., securities lending)—would help gather a more complete picture of risk buildups.

The top-down stress testing capacity for banks and insurers should be enhanced. A priority should be given to further development of the BOC’s bank solvency stress testing framework; the lack of granular data impedes the ability to project key financial items by significant geographies and the capacity to fully model all risk components.

Given their systemic relevance, strengthening oversight of large public pension funds would be helpful. Increasing the detail, standardization, and reporting frequency of financial disclosures, as well as introducing standardized liquidity stress testing requirements, would improve risk monitoring and assessment.

Table 1.

Canada: Recommendations to Bolster the Financial System’s Resilience

article image
Note: Institutions in parenthesis are the agencies with responsibilities. In terms of priorities, H, M and L stand for high, medium and low. In terms of timeframe, I, NT, and MT stand for immediate (within one year), near-term (within 2–3 years), and medium-term (within 3–5 years).

Introduction

A. Financial System Structure

1. Canada has one of the largest and most developed financial systems in the world (Figure 1, Table 2). As of end-2018, total assets of Canadian financial institutions reached US$10.2 trillion or 626 percent of GDP, and outstanding debt securities and stock market capitalization amounted to US$2.2 and US$1.9 trillion, or 133 and 119 percent of GDP, respectively. Deposit-taking institutions, pension funds, mutual funds, and insurers dominate the financial system, accounting for about 45, 18, 17, and 13 percent of financial institutions’ total assets, respectively. Each segment of Canada’s financial system—deposit-taking, insurance, pension, asset management, and capital markets—is among the largest in the world in nominal terms.

Figure 1.
Figure 1.

Canada: Financial System Structure

Citation: IMF Staff Country Reports 2020, 016; 10.5089/9781513527116.002.A001

Sources: Bank of Canada; Bloomberg; FSB, Global Monitoring Report on Non-Bank Financial Intermediation 2018; Haver Analytics; IMF, Financial Development Index database and World Economic Outlook database; and IMF staff calculations.1/ For more details about the financial development index, see IMF SDN/15/08 and IMF WP/16/5.
Table 2.

Canada: Financial System Structure

article image
Sources: FSB, Global Monitoring Report on Non-Bank Financial Intermediation 2018; IMF, World Economic Outlook database; Haver Analytics; SNL; and IMF staff estimates.

Based on National Balance Sheet Accounts, thus not reflecting consolidated balance sheets of financial institutions that have overseas operations. This statistical concept is different from the above text chart, which is based on the consolidated balance sheet basis (a typical FSAP approach).

Only including securities firms (e.g. brokers-dealers) that are not part of banking groups.

Including captive financial institutions and money lenders (CFIMLs), which are largely set up for financial management, asset restructuring and fund-raising purposes to channel funds within corporations. In 2017, CFIMLs’ total assets amounted to Can$3.3 trillion.

Based on the FSB’s definition. In 2017, 73 percent of nonbank financial intermediation was related to collective investment schemes with features that make them susceptible to runs, and 18 percent was related to credit provision that is dependent on short-term funding.

Based on consolidated balance sheet basis.

Only representing regulated entities in federal and Québec jurisdictions.

2. The financial system has enjoyed solid overall growth and international expansion since the 2014 FSAP. Total assets of financial institutions have increased by 31 percent (since end-2013), underpinned by robust assets growth of banking sector, mutual funds and pension funds. Overall banking sector growth is partly driven by the expansion of U.S. operations, with total claims on nonresidents increasing to 41 percent of banking sector assets (from 31 percent). Royal Bank of Canada became a global systemically important bank in 2017. Mutual funds and pension funds have also expanded their cross-border investment, driving Canada’s international portfolio investment assets to 95 percent of GDP (from 60 percent). Domestically, banks finance about two-thirds of private sector credit, while bond issuance and nonbanks are important alternative funding sources.

Financial System: Size and Internationalization, 2003–18

(In percent of GDP)

Citation: IMF Staff Country Reports 2020, 016; 10.5089/9781513527116.002.A001

Source: Bank of Canada; IMF, International Financial Statistics and World Economic Outlook database; Haver Analytics.

Composition of Financial Institutions’ Total Assets, 20171

(In trillions of Canadian dollar)

Citation: IMF Staff Country Reports 2020, 016; 10.5089/9781513527116.002.A001

Source: IMF staff estimate based on data from Canadian authorities, FSB, Haver Analytics and OSFI.1/ Based on consolidated balance sheet basis.2/ Excluding captive financial institutions and money lenders.

3. The financial system is highly concentrated. The six largest banks and Québec’s major credit cooperative group—designated as domestic systemically important financial institutions (D-SIFIs)—account for about 90 percent of deposit-taking sector assets, while the three largest life insurers account for about 70 percent of total net premiums. These banks and life insurers, together with large public pension funds, are globally active and systemically relevant for Canada’s financial system. Retail and wholesale banking, wealth management, and capital markets are the main business lines of major banks; their subsidiaries are among leading securities market intermediaries and asset managers.

4. Financial markets also provide an important venue for public and private sector financing. While bond markets continue to expand by about 39 percent since end-2013, Canadian corporates and financial institutions have increasingly issued debt internationally, driving up the share of foreign-currency debt securities from 26 percent to 34 percent. The public debt market also includes provincial debt securities and government-guaranteed mortgage-backed securities (MBS), which jointly account for two-third of public debt instruments. Other core funding markets include money markets (repo, securities lending, and bankers’ acceptances) and foreign-exchange markets (spot and swap).

5. The government plays a central role in housing finance. The government provides mortgage insurance through CMHC and backstops private insurers’ mortgage insurance (subject to 10 percent deductibles of original mortgage principal). Furthermore, CMHC provides a timely payment guarantee for securitization of qualified insured mortgages. As of 2018Q3, insured mortgages and government-guaranteed MBS (i.e., National Housing Act (NHA) MBS) amounted to Can$723 and Can$485 billion, respectively.

6. Various parts of the financial system are directly exposed to the housing market and/or linked through housing finance. Total residential and nonresidential mortgage credit amounted to Can$1.8 trillion, or 81 percent of GDP, at end-2018. Mortgage credit is provided by banks (69 percent) and other financial institutions; households are the main borrowers (81 percent), followed by corporates. Life insurers and pension funds have increased their investment in commercial real estate, while financial institutions hold around Can$180 billion in NHA MBS. The government’s central role in housing finance fortifies the financial-sovereign nexus.

B. Macrofinancial Conditions

7. The economy regained momentum following a slowdown driven by low oil prices (Figure 2, Table 3). Canada has enjoyed macroeconomic stability since the global financial crisis (GFC). Amid a sharp decline in oil prices, real GDP growth moderated significantly in 2015, with resource-rich provinces being particularly hard hit. The economy recovered during 2016–17, led by robust private consumption, and performed well in the first three quarters of 2018. With weak performance in recent quarters, real GDP growth is projected to be at 1.5 percent in 2019 before picking up to 1.9 percent in 2020, respectively. The medium-term outlook looks less promising, with growth expected to slow to around 1.6 percent by 2024, reflecting longstanding structural problems related to low labor productivity growth, population aging, and deteriorating international competitiveness.

Figure 2.
Figure 2.

Canada: Macrofinancial Developments

Citation: IMF Staff Country Reports 2020, 016; 10.5089/9781513527116.002.A001

Sources: Haver Analytics; and IMF staff estimates.1/ Capturing term premiums, interbank spreads, real long-term interest rates, bond and equity returns and corresponding volatility measures, all in Canada, as well as global financial conditions.2/ Showing percentage balance, with a positive (negative) value indicating tightening (loosening) conditions.
Figure 2.
Figure 2.

Canada: Macrofinancial Developments (concluded)

Citation: IMF Staff Country Reports 2020, 016; 10.5089/9781513527116.002.A001

Sources: Bank of Canada, Staff Analytical Note 2018–35; Haver Analytics; and IMF staff calculations.
Table 3.

Canada: Selected Economic Indicators

(Percentage change, unless otherwise indicated)

article image
Sources: Haver Analytics; and IMF staff estimates.

Excludes equity (authorities’ definition).

8. Financial conditions remain loose due to still favorable pricing of risk. In response to rising inflationary pressures, the Bank of Canada (BOC) initiated a tightening cycle in mid-2017, with five rounds of rate hikes. More recently, the BOC has communicated that an accommodative monetary policy stance is warranted. Long-term bond yields have subsequently declined following the rise during the tightening phase. Despite some bouts of market volatility in recent months, overall pricing of risk, which captures term and credit premiums, remains near historical lows.

9. Credit growth has moderated in line with the softening housing market due to monetary tightening and prudential measures. As of 2019Q1, credit growth moderated to 4.8 percent year-on-year. Several rounds of policy measures have successfully reduced insured mortgage lending and improved credit quality, with the share of banks’ new lending to highly indebted borrowers falling sharply. Meanwhile, house prices have been broadly stable in the past couple years, and housing market-related activities—including construction, inventory and sales, and mortgage lending—have also moderated. However, home equity lines of credit (HELOCs) have grown rapidly, some of which feature interest-only payment2 Borrowers may utilize available credit lines to satisfy the loan-to-value (LTV) requirements when obtaining new mortgages, consolidate existing higher-cost debt, or meet regular payments on other loans.

10. Market data suggests that systemic stress of financial institutions is low. Based on the market-based analysis of 19 large financial institutions as of December 2018, the probability that several financial institutions experience distress simultaneously was near historical lows. The systemic stress measure, which captures the number of institutions potentially becoming distressed and the system-wide expected loss, has been broadly stable over the past few years. Nevertheless, potential contagion effects appear to have risen over the past decade, reflecting interconnectedness among financial institutions and/or growing common exposures to the housing market.

Market Perception of Systemic Stress, 2005–181

Index between 0 and 100, representing from low to high systemic stress

Citation: IMF Staff Country Reports 2020, 016; 10.5089/9781513527116.002.A001

Sources: Bloomberg; Moody’s Analytics; and IMF staff estimates.1/ The analysis is based on the “Surveillance of Systemic Risk and Interconnectedness” approach. See Segoviano and Goodhart (IMF WP/09/4) and Technical Note on Systemic Risk and Interconnectedness Analysis, 2016 United Kingdom FSAP (IMF Country Report 16/164). The sample includes 10 deposit taking institutions, 7 insurers, and 2 other nonbank entities.2/ Cascade effects capture the probability that at least another institution became distressed given than a particular institution became distressed.3/ The systemic stress measure comprises (i) number of institutions to become distressed given than at least one became distressed; and (ii) expected loss related to 1st- percentile tail risk. Both indictors are combined based on their percentile ranking.

11. Macrofinancial vulnerabilities have declined recently but are still substantial. Given relatively limited fiscal and external vulnerabilities (Figure 3), financial stability risks remain heightened mainly due to:

  • High household indebtedness. Household debt reached 96 percent of GDP at end-2018. Canadian households are among the most indebted in advanced economies. Their debt-servicing obligations, already relatively large, could increase as interest rates rise. Households as a whole have large buffers, with net wealth of 489 percent of GDP. However, pockets of financially weak households in terms of excessive indebtedness or weak debt servicing capacity exist.

  • Persistent housing market imbalances. Overvalued house prices (relative to fundamentals such as income or rent) continue to underpin the imbalances. While house prices have largely stabilized over the past year, imbalances continue to persist in major cities such as Toronto and Vancouver.

  • Growing corporate debt. Corporate debt has risen rapidly to 111 percent of GDP at end-2018, largely driven by debt issuance (including in foreign currency) and non-mortgage borrowing. Overall profitability has recovered from the economic slowdown, but firms in the oil and gas and mining sectors continue enduring weak earnings. The rapid increase in debt of firms in the real estate sector raises a concern, especially given their weak income growth.

Figure 3.
Figure 3.

Canada: External and Fiscal Vulnerabilities

Citation: IMF Staff Country Reports 2020, 016; 10.5089/9781513527116.002.A001

Sources: IMF, World Economic Outlook database and International Financial Statistics; and IMF staff calculations.

12. Growth-at-risk analysis points to substantial downside risk to growth due to significant macrofinancial vulnerabilities. Growth-at-risk analysis provides a distribution of real GDP growth forecasts conditional on financial conditions and macrofinancial vulnerabilities. The latter captures corporate and household sector vulnerabilities, housing market imbalances, and credit-to-GDP gap. As of 2018Q3, the analysis suggests a 5 percent probability that real GDP growth would be -1.7 percent or less over the next year, and -1.6 percent (annualized) over the next three years. Downside risk to growth has declined over the past year due to some reductions in housing market imbalances and credit-to-GDP gap.

Distribution of GDP Growth Forecasts, 2018Q3

Based on growth over the next 4 quarters; annualized

Citation: IMF Staff Country Reports 2020, 016; 10.5089/9781513527116.002.A001

Source: IMF staff estimates.

Financial Conditions and Macrofinancial Vulnerabilities, 1992–2018

Citation: IMF Staff Country Reports 2020, 016; 10.5089/9781513527116.002.A001

Source: IMF staff estimates based on data from Haver Analytics.

C. Scope of the Financial Stability Analysis in the FSAP

13. The FSAP took a comprehensive approach to assess financial stability risks (Figure 4). The analysis presented in this technical note included assessing key macrofinancial vulnerabilities, the resilience of deposit-taking institutions and insurers against macrofinancial shocks, and risk-taking in nonbank financial sectors and markets. Additional analysis on intra-system and cross-border interconnectedness, systemic liquidity, and housing finance is presented in other FSAP documents. The FSAP exercise comprised the following modules:

  • Analysis of macrofinancial vulnerabilities. The analysis focused on vulnerabilities in the corporate and household sectors as well as those related to the housing market. The FSAP team employed the debt-at-risk approach to quantify the amount of debt-at-risk, which is owed by financially weak corporates and households. The FSAP team also performed house price-at-risk analysis to gauge downside risk to house prices.

  • Bank solvency stress tests. The FSAP team conducted a top-down exercise based on the balance sheet approach for the seven largest deposit-taking institutions, which cover more than 90 percent of deposit-taking institutions’ total assets, using regulatory and supervisory data. In parallel, the BOC carried out a top-down exercise, using the same sample of banks, but relied on its own methodology. All tests used the same baseline (October 2018 World Economic Outlook projections) and the adverse scenario (see the next section). In addition, a set of sensitivity tests were performed by the FSAP team.

  • Bank liquidity stress tests. The FSAP team performed a top-down exercise for the seven largest deposit-taking institutions, using regulatory data. The exercise comprised Liquidity Coverage Ratio (LCR) tests and cash-flow analysis to examine the ability of banks to handle funding outflows, including in foreign currencies.

  • Life insurance solvency stress tests. The FSAP team conducted a top-down exercise for the five largest life insurers, which cover about 80 percent of life insurers’ total net premiums, using regulatory and supervisory data. The exercise assessed the instantaneous impact of macrofinancial shocks (consistent with the adverse scenario) on the solvency position of life insurers.

  • Mortgage insurance solvency stress tests. The FSAP team conducted a top-down exercise for all three mortgage insurers, using regulatory data. The exercise was carried out in a manner consistent with the bank solvency stress tests to assess system-wide capital buffers for mortgage lending. The exercise thus considered two scenarios—the baseline and the adverse scenario.

  • Analysis of risk-taking in nonbanks and financial markets. The analysis examined the landscape of nonbank financial institutions and financial market participants, analyzed risk-taking among institutional investors—particularly, pension funds—and assessed asset price valuations.

  • Investment fund liquidity stress tests. The FSAP team conducted a top-down exercise for fixed income-focused mutual funds, which account for about 20 of mutual funds’ assets under management (AUM), using publicly available data. The exercise focused on examining potential effects of large redemption shocks on the corporate bond market; the sample captured 77 percent of corporate bonds held by mutual funds. In parallel, the BOC carried out a top-down exercise, using the same sample of banks, but relied on its own methodology.

Figure 4.
Figure 4.

Canada: FSAP’s Approach to Financial Stability Analysis

Citation: IMF Staff Country Reports 2020, 016; 10.5089/9781513527116.002.A001

Source: IMF staff.

14. The remainder of this technical note is structured as follows. Section II presents key macrofinancial risks and the scenarios underpinning the financial stability analysis. Section III discusses macrofinancial vulnerabilities—particularly, in the corporate and household sectors and related to the housing market. Sections IV, V, VI, and VII cover bank solvency stress tests, bank liquidity stress tests, life insurance solvency stress tests, and mortgage insurance solvency stress tests, respectively. Sections VIII analyzes risk-taking in nonbank financial institutions and financial market activities, while section IX covers investment fund liquidity stress tests. Each section also concludes with relevant policy issues necessary to manage identified risks.

Macrofinancial Risks Underpinning Stress Testing and Financial Stability Analysis

A. Overview of Key Risks

15. Canada’s financial system faces a confluence of domestic and external risk factors that could amplify existing financial sector vulnerabilities (Table 4). The key external risks are tighter global financial conditions, significant slowdowns in the euro area and China, and rising protectionism and retreat from multilateralism. On the domestic front, a sharp house price correction could occur on the back of rising unemployment and higher funding costs. Cyber-attacks could also pose significant risk to the financial system.

Table 4.

Canada: FSAP Risk Assessment Matrix (RAM)

article image
article image
The RAM shows events that could materially alter the baseline. The relative likelihood is the staff’s subjective assessment of the risks surrounding the baseline (“low” is meant to indicate a probability below 10 percent, “medium” a probability between 10 and 30 percent, and “high” a probability between 30 and 50 percent). The RAM reflects staff views on the source of risks and overall level of concern as of the time of discussions with the authorities. Non-mutually exclusive risks may interact and materialize jointly.

B. Macrofinancial Scenarios

16. The financial stability analysis was conducted broadly based on two macrofinancial scenarios (Figure 5, Table 5). One is the baseline, which is based on the October 2018 World Economic Outlook projections. The baseline envisages a continued, stable macrofinancial environment, with economic growth of about 1.9 percent (annualized) over the three-year stress testing horizon (i.e., 2019–21). In the baseline, interest rates are expected to rise in line with monetary policy tightening. The other is the adverse scenario, which was designed based on key macrofinancial risks outlined in the RAM. The Global Macrofinancial Model—a structural macroeconometric model of the world economy, was used to simulate the adverse scenario.3

Figure 5.
Figure 5.

Canada: Key Macrofinancial Variables in the Baseline and Adverse Scenarios

Citation: IMF Staff Country Reports 2020, 016; 10.5089/9781513527116.002.A001

Sources: Haver Analytics; IMF, World Economic Outlook database; and IMF staff estimates.
Table 5.

Key Macrofinancial Variables in the Baseline and Adverse Scenarios

(In percent; unless indicated otherwise)

article image
Source: IMF staff estimates.

Interbank spread is defined as the difference between interbank rate and policy rate.

17. The adverse scenario assumes a severe recession that would occur concurrently with significant financial stress and a sharp housing market correction. The initial trigger would be disruptions in international trade and global production chains, followed by disorderly financial market adjustments and heightened policy uncertainty. As a result, global financial conditions would tighten significantly, setting off global housing market and credit cycle downturns. The impact of these negative external shocks would be amplified by existing macrofinancial vulnerabilities such as elevated household debt and housing market imbalances, resulting in a sharp housing market correction and deterioration in bank asset quality. Significant financial stress, together with large exchange rate depreciation, would also materialize in Canada. All these shocks would underpin a snapback of interest rates in Canada, driven by monetary tightening to stabilize inflation expectations (following rising inflation stemming from global supply shocks and Canadian dollar depreciation at the initial phase of the adverse scenario). Accommodative monetary policy would follow in later years due to recession-induced significant deflationary effects. The scenario is severe but plausible, with cumulative real GDP growth of -2 percent (annualized) during 2019–21, equivalent to 3 standard deviations; based on growth-at-risk analysis, the likelihood of such a severe growth outcome is 3.8 percent.

Analysis of Macrofinancial Vulnerabilities

A. Overview

18. Elevated household indebtedness and housing market imbalances continue to pose financial stability concerns. During the decades-long credit upcycle, low interest rates and low capital charges for mortgage lending, together with policies promoting housing affordability, have fueled borrowing to finance home purchases in the face of rapidly rising house prices. Risk mispricing has contributed to debt accumulation among financially weak households, with problems more exacerbated in regions experiencing larger housing market imbalances. During severe downturns, Canada-specific housing finance characteristics may amplify procyclical effects of falling house prices, and the impact on growth could be protracted due to household balance sheet adjustments.4

19. The FSAP examined vulnerabilities in the corporate and household sectors as well as those related to the housing market. The FSAP employed the debt-at-risk approach to quantify the amount of debt owed by financially weak corporates and households. A set of sensitivity tests were conducted to assess how corporate and household balance sheets would be affected by macrofinancial shocks. The FSAP also performed house price-at-risk analysis to gauge downside risk to house prices.

B. Household Sector Vulnerabilities

20. Canadian households have substantially increased their debt obligations in the past decade, stretching their balance sheets (Figure 5). Household indebtedness peaked at 97 percent of GDP in 2016Q4 but has leveled off over the past two years. In the aggregate, households hold a sizeable amount of wealth, with net wealth of 489 percent of GDP as of 2018Q4. A bulk of household wealth is in the form of nonfinancial assets such as real estate.

21. Canadian households are more indebted and have larger debt servicing obligations than advanced economy peers. Only Australia and the Netherlands have higher household debt-to-GDP ratios and aggregated debt servicing-to-income ratios. Despite asset buffers, the average Canadian’s debt servicing obligations are already large and could increase further as financial conditions tighten.

22. The household sector vulnerabilities index is close to its historical peak (Figure 6).5 For instance, household debt-to-disposable income is roughly 30 percent higher compared to its historical average. At the same time, the average debt servicing ratio seems to be slightly below the historical trend, reflecting the protracted low interest rate environment. One of the main culprits explaining the elevated household sector vulnerability is the high level of household indebtedness combined with the recent deterioration of average debt service. The latter reflects relatively higher interest rates and tighter lending standards over the past one–two years.

Figure 6.
Figure 6.

Canada: Household Indebtedness

Citation: IMF Staff Country Reports 2020, 016; 10.5089/9781513527116.002.A001

Sources: Statistics Canada’s Survey of Financial Security; IMF staff calculations.Note: DTI ratio is calculated as total amount of debt divided by household disposable income, including government transfers. TDS is the ratio of mortgage payments plus credit card payments divided by disposable income, including government transfers. LTV is the ratio of mortgage loan (final value) to the value of the principal residence. The household sector vulnerabilities index incorporates household indebtedness information such as leverage, debt servicing, and debt repayment capacity.

23. The portion of household debt held by financially weak borrowers increased substantially over the past two decades.6 Typically, financially weak households are defined as those households whose debt-to-income (DTI) ratio is above 4.5 or total debt service-to-income (TDS) is above 40 percent. The debt of these financially weak households is considered at risk. The share of debt held by households with DTI above 4.5 amounted to more than 30 percent of total household debt at end-2016, three times larger than the share in 1999. When other definitions of debt-at-risk are considered, the share of debt-at-risk ranges between 17 and 42 percent at end-2016. For instance, the share of debt-at-risk was about 17 percent when only borrowers with TDS above 40 percent were considered. At the same time, the share of debt-at-risk was about 42 percent when households with DTI above 350 percent were considered.

24. The balance sheets of low-income borrowers (those below the 20th percentile of the income distribution) are more exposed to financial distress. For instance, total debt service as share of income among low-income borrowers stood at about 55 percent in 2016, compared to 25 percent for the average borrower. In addition, the amount of debt accumulated by poorer borrowers climbed from about 3 times their annual income in 1999 to about 8 times in 2016.

25. The assumptions of the household debt-at-risk analysis follow closely those of the stress testing adverse scenario. However, this analysis only considers household income loss, house price declines, and interest rate shocks. On average, house prices in the adverse scenario are expected to decline by 40 percent over the next two years (i.e., by end-2020), in line with the adverse scenario. In addition, a decline in household disposable income of 15 percent and an increase in interest rates up to 230 basis points (bps) are considered. Depending on the renewal profile of borrowers, interest rates are adjusted upwards relative to 2016 level as follows: in 2017 by 10 basis points, in 2018 by 70 basis points, in 2019 by 220 basis points, and in 2020 by 230 basis points.

26. In the adverse scenario, house prices in Canada cumulatively decline, on average, by 40 percent by 2020.7 However, this analysis assumes heterogenous house price shocks across provinces, to better reflect valuation differences and other regional factors (e.g., supply constraints). The house price-at-risk analysis is used to inform the calibration of the housing shocks in each province or region. The decline in house prices in Ontario, British Columbia, Quebec, Alberta, and the rest of Canada ranges between 35 and 45 percent, reflecting regional differences (e.g., with Ontario and British Columbia most vulnerable).

27. Some parts of the household sector look particularly vulnerable to severe macro-financial conditions. Assessing households’ financial strength based on debt servicing-to-income, the share of debt-at-risk would increase from 17 percent in 2016 to 29 percent in the 2020 in the adverse scenario. Debt-at-risk is disproportionally higher for households with low income, low net wealth, and no full-time job. For instance, the share of debt-at-risk held by the least wealthy borrowers (i.e., in the bottom 40 percent) increases from 16.2 percent in 2016 to 31 percent in 2020 in the adverse scenario (compared with 21 percent in the baseline). Households in British Columbia and Ontario are seemingly more financially vulnerable than those in Quebec and Alberta, in part reflecting higher indebtedness levels and housing market overvaluation in the former provinces.

28. Financial stability implications of household sector vulnerabilities could be substantial given the ample share of debt-at-risk not covered by real estate assets in the adverse scenario. The share of debt-at-risk not covered by real estate assets surges from less than 0.5 percent in 2016 to about 4.5 percent by 2020 in the adverse scenario.8 The magnitude of the increase in debt-at-risk not covered by real estate assets is greatest for borrowers in Ontario and British Columbia, where the share rises by more than eight times.

C. Housing Market Imbalances

29. Alongside the household debt buildup, Canada has experienced a surge in house prices across a broad spectrum of cities. Canada-wide house prices have trended upward over the past five years, increasing by more than 40 percent. After a housing boom between 2011 and 2017, the housing market started to cool off in 2018, reflecting tighter monetary policy, stricter lending standards, and other housing market measures.9 Over the past year, certain housing-related activities have slowed substantially, including construction activity, sales, and mortgage lending. The general slowdown partially reflects the use of macroprudential measures to safeguard financial stability and other measures to mitigate speculative activity and improve affordability in regional markets.

30. House price-to-income and house price-to-rent ratios in major cities (e.g., Toronto and Vancouver) show persistently overstretched valuations. While house prices have largely stabilized over the past year, imbalances continue to persist in major metropolitan areas such as the Greater Toronto Area (GTA) and Greater Vancouver Area (GVA). For the Canada-wide housing market, the house price-to-income ratio is about 1.3 times the long-term average, with GTA being the most expensive.

31. House price-at-risk analysis shows significant downside risks to house prices in Canada. The analysis was based on individual cities, using quarterly data from 1980 to 2018. Downside risks to house prices in Canada appear to have substantially increased over the past decade, partially reflecting higher house price overvaluations.10 Supply-side drivers such as residential investment play an important mitigating role to forward-looking house price risks. In addition, capital flows are found to be associated with downside risks to key residential housing markets, but the net effect depends on the type of flows and varies across cities. Financial conditions, which reflect both domestic and external factors, are an additional relevant factor to housing market risks in Canada.

32. Downside risks to house prices across Canadian cities have changed substantially over time, with housing markets in GTA and GVA currently the most at risk. Over the past four decades, downside risks across main metropolitan areas in Canada show several cycles, suggesting that housing markets are susceptible to booms and busts (Figure 7). Additionally, there is considerable dispersion in downside risks to house prices across cities, as reflected by the volatility of the lower bound (tenth percentile of the cross-sectional distribution). At longer horizons such as three years ahead, it is evident that Canadian downside risks have substantially increased in the past two decades, surpassing levels seen around the GFC. The average three years-ahead house price-at-risk across Canada stood at about -12 percent (annualized) at end-2018 Q3. In other words, there is a 5 percent likelihood that Canada-wide real house prices will decline by 12 percentage points or more three years from now. However, the magnitude varies substantially across cities, with tail risks to Toronto and Vancouver housing markets being close to -24 percent and -25 percent, respectively, over the same period.

Figure 7.
Figure 7.

Canada: City-Level House Prices at Risk

Citation: IMF Staff Country Reports 2020, 016; 10.5089/9781513527116.002.A001

Source: CREA; Statistics Canada; Haver; IMF staff calculations.Notes: In panels 1 and 2, p10 and p90 refer to tenth and ninetieth percentile of the cross-sectional city house prices-at-risk (HaR) distribution at each period. In panels 4, one-year ahead estimates of 5 percent HaR refer to end 2018 and 2008, respectively. In panel 5, FDI and OtherCF (other capital flows) refer to incoming foreign direct investments and capital flows other than FDI and portfolio investment, respectively.

33. Overvaluation (relative to fundamentals, such as disposable income) continues to represent a key determinant for downside risks to house prices, albeit with variations between regional markets. City-level house price valuations, proxied by the house price-to-income ratio, appear to be associated with higher one-year ahead house price-at-risk. Compared to long-term averages, housing market valuations in Canada have become more stretched over the past five years, especially in GVA and GTA.11 At the same time, tail risks to Canadian house prices have generally increased, reflecting tighter financial conditions and speculative capital inflows (see below).

34. The dynamics of housing market risks in major cities are partially correlated with capital inflows and financial conditions which seem to both amplify and mitigate downside risks to house prices across Canadian cities.12 More specifically, the sensitivity to these factors seems to differ across cities:

  • Foreign direct investment (FDI), which is typically long-term investment, is generally associated with lower future downside risks to several Canadian housing markets. For example, a one percentage point increase in Canada’s FDI inflows is significantly associated with a reduction of about 0.5–1 percentage point in house price-at-risk of cities such as Calgary, Quebec City, Ottawa, and Toronto. At the same time, other capital inflows (i.e., not FDI or portfolio flows), which are generally attributed to foreign bank transactions, are found to amplify downside risks to house prices in cities such as Toronto, Vancouver, and Calgary. The largest effects are seen in Vancouver, where speculative capital inflows might have partially contributed to the recent housing boom.

  • Tighter financial conditions, which encompass monetary policy, macroprudential measures and other factors, are negatively associated with downside risks to house prices. These effects are typically channeled through lending standards and mortgage costs. Although sensitivities to financial conditions are statistically significant for most cities, their economic magnitude is marginal. The most sensitive cities to financial conditions are Toronto, Ottawa, and Calgary.

35. Contributions to housing downside risks vary substantially across Canadian cities (Figure 8). For instance, house price-at-risk has substantially deteriorated in GTA over the past year largely due to city-specific factors such as higher valuations and constrained supply. In contrast, Ottawa’s house price at risk appears to have improved in the past five years, with loose financial conditions and balanced valuations playing important roles. At the same time, downside risks in GVA have steeply risen over the past five years, partially reflecting higher price valuations but also the negative role of other capital inflows. However, the latter contributions seem to be ameliorating over the past two years, likely reflecting the housing-related measures taken by the local government. For Canada-wide house prices, valuation net of supply-side factors such as residential investments seem to amplify downside risks, while loose financial conditions are a mitigating factor.

Figure 8.
Figure 8.

Canada: Decomposition of House Price-at-Risk

Citation: IMF Staff Country Reports 2020, 016; 10.5089/9781513527116.002.A001

Source: Canadian Real Estate Association; Statistics Canada; Haver; and IMF staff calculations.Notes: In the top-left panel, valuation is net of residential investment, which has a positive (mitigating) contribution to Canada-wide house price risks. In panels 2–6, city-level variables include house price-to-income ratio and residential investment in each city/province. FCI refers to financial conditions, which is an aggregated factor based on interest rates, volatility, and other price-based financial factors (see also IMF 2017a, Chapter 3). Capital flows include foreign investment inflows and other capital inflows. Other includes oil prices and household debt. The intercept and auto-regressive term are not depicted. All contributions refer to the quantile regression at the 10th percentile HaR (house price at risk).

D. Corporate Sector Vulnerabilities

36. Corporate sector vulnerabilities have risen in recent years. Nonfinancial corporate debt-to-GDP in Canada has grown by 22 percentage points since end-2014, to 111 percent, faster than all other G20 economies in this period except Turkey.13 Roughly two-thirds of this debt growth has been in the form of loans, with the remainder corporate debt issuance. Debt growth was fastest in 2015 and 2016, in part reflecting the impact of currency depreciation on foreign-currency borrowing; subsequently, nonfinancial corporate debt-to-GDP has largely stabilized.14 Canadian corporates also have the highest estimated aggregate corporate debt-servicing ratio (DSR) among advanced economies, according to BIS data, although this calculation likely overstates Canadian firms’ vulnerability.15

37. Debt-servicing capacity appears nonetheless to be broadly resilient for Canadian nonfinancial firms in aggregate. According to Statistics Canada data, which uses a consolidated residence-based balance sheet approach, cumulative debt growth for all nonfinancial firms has only been slightly faster than annual earnings before interest, taxes, depreciation and amortization (EBITDA) since end-2014 (16 percent versus 13 percent).16 As of 2018Q3, debt was 4.7 times EBITDA, up from 4.5 times in 2014, and 3.8 times on a net-of-cash basis. The interest coverage ratio was largely unchanged over this period, at 6.85 times.17

38. Oil-and-gas and mining firms have however seen vulnerabilities increase due to commodity price volatility. Petroleum and mining firms were particularly hard hit by the collapse in commodity prices in 2014–2015. Statistics Canada data shows these sectors saw earnings declines of 27 and 6 percent, respectively, from end-2014 to the four quarters ending in 2018Q3. Oil-and-gas firms saw debt-to-EBITDA rise to 7.4 times as of 2018Q3, from 5.8 times in 2014, and interest coverage fall to 3.4 times. Despite the recovery in oil prices, these firms may be exposed to Canada-specific oil price declines due to challenges in economically transporting oil out of Western Canada and backlogs in U.S.-based oil transportation due to problems in storage infrastructure.

39. Increasing leverage in the construction, utilities and real estate sectors merit further attention. Firms in these sectors saw relatively large increases in debt relative to earnings. Utilities firms typically operate with high leverage, reflecting highly regulated business models with predictable and steady income flows; however, firm-level data suggests that debt-servicing challenges could be rising in parts of the sector. Construction firms, which are largely unaccounted for in firm-level data, accounted for about nine percent of the increase in nonfinancial debt from end-2014 to 2018Q3. Construction firms’ nonfinancial debt also grew four times as fast as earnings, although interest coverage ratios remained high (above 6 times). Real estate, rental, and leasing firms, including real estate investment trusts and other de facto investment fund structures, are typically excluded from nonfinancial firm debt-servicing analyses but are classified as nonfinancial firms in Statistics Canada data. Notably, these firms account for 24 percent of the overall growth in nonfinancial corporate debt in this period (see Section on Risk-taking in Nonbanks and Financial Markets). Excluding these firms, nonfinancial firms’ aggregate interest coverage ratios actually improved slightly from end-2014 to 2018Q3, and the aggregate debt-to-EBITDA was unchanged.

40. Firm-level data shows the share of financially weak firms is fairly small, consistent with the aggregate data. In a sample of 314 nonfinancial, non-real-estate corporations accessed through S&P’s Capital IQ database, debt of firms with interest coverage ratios less than one— considered “debt-at-risk”—accounted for 4.9 percent of the Can$980 billion in outstanding debt across the sample.18 These financially weak firms were largely in the energy and materials (including mining) sectors. The total share of debt-at-risk was roughly unchanged from around five percent in 2016, reflecting the impact of previous years’ oil shock on corporate earnings. In 2018, several utilities firms saw debt-servicing risks increase, although some of these firms have public sector backstops. Notably, the sample does not include construction, property development and real estate firms.

41. A shock consistent with the FSAP’s adverse macrofinancial scenario would increase debt-at-risk only modestly. Simple sensitivity analyses can illustrate how the share of financially weak firms might rise in an economic downturn.19 The first element of a hypothetical shock is a sample-wide decline in EBITDA by 35 percent, which would be roughly consistent with the year-on-year GDP decline in the adverse scenario of -6.5 percent, based on the sensitivity between GDP and EBITDA observed during the GFC.20 The second element of the shock is an increase in funding costs of about 3.5 percent for all debt coming due within one year, meant to reflect the 1.7 percentage point increase in the prime lending rate and an additional increase in the interbank spread of 1.5 percentage point (a proxy for potential funding strains). The earnings shock on its own would generate an increase in debt-at-risk of about 3.6 percentage points. The interest rate shock generates an increase of 2.6 percentage points. Together, they generate an increase of 3.6 percentage points in debt-at-risk, the same size as from the earnings shock alone.

42. A standalone oil price shock would have a limited effect on energy sector debt-servicing capacity, however the analysis may underestimate potential vulnerabilities. During the oil price shock of 2014–15, when oil prices fell 75 percent, aggregate trailing 4-quarters EBITDA of Canadian energy-sector firms in the Capital IQ sample reported a decline of about 45 percent peak-to-trough (over six quarters). If energy firms faced a similar 45 percent decline in EBITDA, the share of sector debt-at-risk would only rise from about 0.6 percent to 2.0 percent, reflecting the fact that 89 percent of sector debt is owed by firms with interest coverage ratios above 3. This may however underestimate the potential rise in weak firms; an even higher percentage of energy sector debt was owed by firms with interest coverage ratios above 3 in 2014 (93 percent), but sector debt-at-risk still rose from 1.6 percent in 2014 to 7.3 percent in 2016.

43. Data limitations mean that corporate debt vulnerabilities could be underestimated. Statistics Canada data shows private nonfinancial firms have very large inter-affiliate debt liabilities, equivalent to about 50 percent of all other borrowings. While much of this may be related to genuine cross-border inter-affiliate lending relationships, it may also capture debt financing arrangements that meet the definition of inter-affiliate but are nonetheless arms-length in nature (e.g., private equity and debt funds), or are linked to arms-length obligations issued outside the statistical perimeter. This would mean the overall level of arms-length debt captured in the national balance sheet accounts is understated. There are also challenges in understanding the amount of foreign currency borrowing, and to what degree these exposures are hedged via revenue streams or in derivatives markets. Finally, the firm-level dataset is based on the globally consolidated operations of firms that report data to securities regulators, whereas Statistics Canada data may not capture the foreign activities and debt of certain non-resident Canada-incorporated firms. Thus the degree of overlap between the two data sources is unknown.

Figure 9.
Figure 9.
Figure 9.

Canada: Corporate Financial Soundness

Citation: IMF Staff Country Reports 2020, 016; 10.5089/9781513527116.002.A001

Sources: Capital IQ; Haver Analytics; and IMF staff estimates.Note: The sixth panel is based on Statistics Canada’s Quarterly Financial Statements data. Income is calculated as profits plus interest expense and depreciation expenses.1/ Financially weak firms are defined as firms whose earnings before interest, tax, depreciation and amortization (EBITA) is less than interest expense (including capitalized interest). Debt of these financially weak firms is considered at risk.2/ The sensitivity analysis assumes income shock (i.e., 35 percent decline in EBITDA) and funding cost shock (i.e., 3.5 percentage points increase).

Bank Solvency Stress Tests

A. Overview

44. The FSAP, in cooperation with the BOC, the Office of the Superintendent of Financial Institutions (OSFI) and Autorité des marchés financiers (AMF), assessed the banking sector’s resilience to macrofinancial risks. The analysis included solvency stress testing for all six domestic systemically important banks (D-SIBs) and Quebec’s domestic systemically important financial institution (D-SIFI), which account for more than 90 percent of deposit-taking institutions’ total assets.21 The BOC carried out a similar top-down exercise in parallel but used its own methodologies and infrastructure, which also account for contagion effects, including interaction between solvency and liquidity conditions. Both exercises were conducted based on 2018Q3 data and assumed a three-year stress testing horizon.

45. The FSAP top-down solvency stress test accounted for a comprehensive set of risks. The FSAP team used IMF’s internally developed solvency stress testing framework to capture credit risk (covering both credit impairments and scenario impact on risk-weighted assets, funding, and interest rate risk), market risk (covering repricing and spread risks for interest rate sensitive assets, as well as equity and commodity risks), and non-interest income risks.

B. Recent Performance of the Banking Sector

46. The financial system has enjoyed solid overall growth and remarkable international expansion since the 2014 FSAP. The combined size of Canadian banks’ balance sheets has increased by 52 percent since end-2013, underpinned by robust loan book growth. Banking sector growth is partly driven by the expansion of U.S. operations, with total claims on nonresidents increasing to 39 percent of banking sector assets (from 31 percent). With a targeted expansion in the U.S. being the obvious choice, each major Canadian bank has followed a different implementation plan to achieve its enlarged footprint. While it may take some time to assess the effectiveness of each strategy, the aggregate level the growth of the sector has clearly been steady in terms of balance sheet growth rates and robust in terms of aggregate profitability (Figure 10).

Figure 10.
Figure 10.

Canada: Banking Sector Recent Performance

Citation: IMF Staff Country Reports 2020, 016; 10.5089/9781513527116.002.A001

Sources: OSFI data on federally regulated entities; and IMF staff estimates.

47. Internal income generation remains robust, partly due to the exceptional non-interest stream of revenues and the contained increase of operational expenses. The favorable economic conditions have further supported growth strategies. Credit-linked impairments have been extremely low, and dividend distribution has remained at reasonable levels. As a result, retained earnings could fully finance the implementation of bank expansion plans while at the same time support the build-up of significant capital buffers above the minimum ratios.

48. Going forward, the ability of the sector to maintain a decent net interest margin might need to be reconfirmed. Growth of non-domestic asset classes might be more expensive in terms of funding and capital charges faced by banks. On the domestic front, rising competition and market maturation effects22 might create some strong challenges for the sector. Potential downside risks on the credit portfolio, related to either credit losses or higher risk weights, could have a sizeable impact on overall sector profitability. Nevertheless, the solid current capital ratios of Canadian banks could provide a mitigating feature, if such adverse developments were to materialize in an economic downturn.

C. IMF Top-Down Stress Tests: Methodology

49. A granular solvency stress test for banks was conducted using the two macrofinancial scenarios (baseline and adverse) presented in paragraphs 16 and 17. The exercise was of an accounting nature, based on the latest available regulatory reports (calendar 2018Q3). It mainly focused on system-wide projections and targeted the assessment of risks and vulnerabilities identified in the FSAP RAM that are relevant to the banking sector. An outline of the overall methodological features, the modelling focus and approach, and the main assumptions are presented below.

Balance Sheet, Income Projections, and Hurdle Rates

50. Balance sheet growth followed a quasi-static allocation assumption. The allocation of assets and the composition of funding sources remained constant as of the cut-off date. Credit exposures in bank balance sheets were projected to grow in line with the baseline and adverse macroeconomic scenario projections for retail and business credit growth. The interest sensitive liability segments were proportionally adjusted to fully reflect the changes in the asset side. Any exchange rate adjustments from the scenario were also incorporated in the balance sheet projections. Maturing capital instruments were not assumed to be rolled over under the adverse scenario, reflecting stressed conditions for banks to raise capital or renew capital instruments directly from the market.

51. The FSAP stress testing approach also accounted for the recently introduced IFRS 9 accounting standard. Accounting provisions under both scenarios were projected using stressed stage transition probability matrices for each asset segment. Starting point stage transition probabilities were used based on bank-level reported data, and a fully-fledged methodology was implemented to project lifetime Expected Credit Losses for Stage 2 assets. As a result, the solvency exercise also projected the relative size of IFRS 9 staging volumes by credit segments and accounted for both accounting impairments and regulatory expected losses. Excess credit accounting provisions over regulatory expected losses were added back to tier-2 capital up to the cap of 0.6 percent of credit RWAs. Furthermore, regulatory expected losses in excess of accounting provisions were deducted from common equity tier-1 (CET1) capital to fully recognize the treatment of internal ratings-based (IRB) shortfall or excess.

52. Projection of RWAs accounted for balance sheet growth, defaulted credit exposures formation, and changes in the exchange rate. RWAs changed due to balance sheet growth (positive or negative), adjustment in the credit quality of the performing exposures, additional formation of nonperforming exposures based on the scenario, and exchange rate fluctuations. Additional impairments attributed to credit risk were also considered for the RWAs calculation both under the standardized (STA) and advanced IRB (AIRB) approach. For market risk, the value-at-risk (VaR) RWA charge was replaced with the stressed VaR RWA charge in the adverse scenario. RWAs for operational risk and any residual RWA charge were kept constant at the level of the cut-off date for both baseline and adverse.

53. Income (profit or loss) was projected based on the impact of all risk factors considered in the stress test. Specifically, total net income reflected projections for net interest income, non-interest income and expenses, trading income, profit and loss or other comprehensive income changes due to the revaluation of the “fair value through other comprehensive income” (FVOCI) and “fair value through profit or loss” (FVTPL) portfolios, credit loss provisions, and tax charges and dividend distributions.

54. Minimum capital requirements were based on actual supervisory requirements during the stress testing horizon (2019–2021). The stress testing results were benchmarked against the regulatory minimum and target values for CET1, T1, and total capital that also includes a capital conservation buffer (CCB) of 2.5 percent (fully phased in 2019) and a D-SIB surcharge of 1 percent for all 7 banks within the scope of the exercise. An additional domestic stability buffer (DSB) of 1.75 percent23 is required for the six federally regulated D-SIBs. Table 6 summarizes the regulatory minimum and target capital requirements. Under the baseline scenario, banks are expected to maintain capital ratios that are above the total target requirements (including minimum, CCB, D-SIB surcharge, and DSB, if applicable). In the adverse scenario, the requirement on preserving the CCB and the DSB are relaxed.

Table 6.

Canada: Minimum and Target Capital Requirements for Solvency Stress Tests (effective 2019)

(In percent)

article image
Note: See Chapter 1, Section 1.5 of “OSFI Capital Adequacy Requirements (OSFI 2018).

55. Banks were assumed to comply with the regulatory dividend distribution policy and applicable restrictions under both scenarios. Dividends were assumed to be constant, anchored to the distribution of the last fiscal year; distributions were only restricted if the ending capital ratios during the scenario years was below the regulatory thresholds that are used to restrict distribution, as per Table 7.

Table 7.

Canada: Dividend Distribution Restrictions

article image
Note: The 1 percent D-SIB surcharge extends the capital conservation buffer for the purpose of capital conservation. Dividend restrictions depend on the quartile of the shortfall, i.e., full restriction applies when the ending CET1 capital ratio is below 6.125 percent, a restriction of 20 percent of earnings is imposed when the CET1 capital ratio falls into the range of 6.125–6.75, and no restriction applies when the CET1 capital ratio is above 8 percent.

Credit Risk Analysis and Estimation

56. Credit risk is the most important risk factor for the Canadian system given that credit exposures correspond to almost 51.6 percent of total assets (Figure 11). By type, retail real estate-linked products account for about 40 percent of the total exposures at default (EAD) and almost 50 percent of the total drawn lines. Corporate loans correspond to 32 percent of credit exposures and consumer credit (including credit cards) to 15.8 percent. By geography, 68.3 percent of total exposures are domestic and another 21 percent in the United States. However, individual banks appear to be quite diverse in terms of their geographical footprint. The relative importance of credit risk is also confirmed by the substantial proportion of total RWAs that are attributed to credit risk versus residual risk types (which is almost 83 percent, including charges for securitizations and counterparty credit risk). Finally, market aggregate data indicate that AIRB is the predominant regulatory approach; approximately 88 percent of exposures are reported under IRB and only 12 percent under STA.

Figure 11.
Figure 11.

Canada: Asset Side Analysis

Citation: IMF Staff Country Reports 2020, 016; 10.5089/9781513527116.002.A001

Sources: IMF staff estimates.

57. A variety of modelling approaches were used to model credit parameters under the two scenarios. Exposures were initially allocated to asset types based on an exposure segmentation scheme that included retail real-estate related classes (insured and uninsured mortgages and HELOCs), consumer retail classes (consumer credit and credit cards), and corporate exposure classes. Corporate exposures were further broken down by industry into 11 industrial asset classes. Additional segmentation layers included a breakdown by geography—Canada, United States, and Rest of the World (ROW)—and by regulatory approach (AIRB and STA). Credit parameters starting points were collected for all asset types and segments, and a suite of models were estimated for the scenario dependent projection-of-credit parameters—probability of default (PD), point-in-time (PiT), and loss given default (LGD) PiT.24 Table 8 illustrates the entire segmentation scheme, with 16 asset exposure classes for 3 geographies (Canada, United States, and ROW) and 2 regulatory approaches (AIRB and STA). This results in a total of 96 asset-class segments.

Table 8.

Canada: Credit Risk Exposure Segments

article image

58. For the projections of PDs, the stress test approach made use of satellite models to project scenario-dependent forward paths. Bayesian Model Averaging (BMA) was used as the core modelling approach in the estimation of econometric models for the projection of default probabilities under both scenarios. This framework was used for all 11 corporate segments, and the consumer and credit card segments for both Canada and the United States. For corporate PD PiT satellite models, historical default rates by industry were used. The historical default rates by industry were obtained by third party data providers.25 For credit cards, the aggregate for the market historical delinquency and loss statistics time series was provided by the Canadian Bankers Association statistical data warehouse.26 These time series were used to drive the calibration of a regression model with reasonable response. Based on the estimated econometric models, a PD path for each scenario and asset segment/class was generated. Details of the BMA PD PiT satellite models and the estimation approach can be found in Appendix III.

59. The satellite models also informed the IFRS 9 calculation engine that was used to account for the accounting layer. PiT stage transition rates were collected from the regulatory returns for all banks and all segments for the first three quarters of 2018. Based on the PiT PD paths for each scenario, a stressed transition rate matrix was produced for each year of the stress testing horizon. The estimation of stressed transition matrices facilitated the projection of exposure transitions across stages through the horizon. An additional calculation layer was used to estimate lifetime losses for Stage II assets. The accounting layer was developed to enable a full projection of accounting impairments for each asset type/segment. Under the baseline scenario, macroeconomic variables were assumed to remain constant after year 3 of the scenario; under the adverse scenario, macroeconomic variables were assumed to gradually converge back to the baseline path within a period of six years. Perfect scenario foresight was also assumed to simplify provisioning projections.

60. A structural model was used to project loss rates for mortgage and real estate-related retail portfolios. Major loss events in residential housing loan portfolios are practically non-observable in the history of the Canadian market; furthermore, the currently observable default rates are particularly small. For these reasons, a satellite model calibrated on conventional econometric approaches was not feasible for fully capturing the dynamics of a major economic downturn like the one captured in the adverse scenario. A structural model, on the other hand, can fully incorporate features that mimic how defaults in residential mortgage markets may be masked by a housing market boom. Therefore, it can perfectly explain the bimodal dynamics of smaller loss rates during rising housing markets (like those observed in the Canadian market) and also provide some valuable insights on how defaults and non-performing loan (NPL) formations may drastically evolve under severely adverse macrofinancial conditions. In that context, the model can be consistent with economic realities under both scenarios.

61. The model uses borrower affordability metrics (i.e., DSR) and how they get affected under a specific scenario to infer default probabilities. House price shocks are used to estimate losses in the default event. Scenario shocks are linked to household microdata sources in order to also simulate behavioral elements for the borrowers under distress. This allows a joint distribution of the DSR, the loan-to-value (LTV) ratio of the mortgage, and the liquid wealth of the borrower to fully represent the credit quality of the portfolio. The scenario shocks (on interest rates, unemployment, and house prices) determine how the borrowers’ population is affected, how defaults are triggered, and how potential losses can be linked to such defaults. A detailed presentation of the approach underpinning the implementation and use of the structural model for mortgage portfolios can be found in Appendix IV.

62. The structural model proved to be particularly flexible and was used to assess several aspects of the mortgage market. More specifically, in the FSAP overall approach, the mortgage structural model served the following tasks. First, it was used as the primary model for the purpose of projecting PD PiTs and LGD PiTs. Second, it facilitated the task of disentangling the two-layer architecture of coexistence of banks and mortgage insurers and provided estimates for the credit parameters of both insured and uninsured mortgage portfolios. Third, it provided bank specific starting points and projections after adjusting for bank specific portfolio quality characteristics. Fourth, it provided valuable guidance in assessing the adequacy of capital requirements for both market layers, both under current and post-stress conditions. Fifth, it enabled the team to perform some sensitivity or impact attribution analysis under different set of assumptions for the interest rate, unemployment, and house price shocks.

Market Risk Analysis

63. The analysis for market risk captured the valuation risks of the securities portfolio due to changes in risk-free interest rates and credit spreads for interest sensitive instruments, as well as equity and commodity risks. An ad-hoc data collection request was made to banks which they reported their debt, equity, commodity and fund portfolios by accounting category and maturity bucket (where applicable). The macrofinancial scenario was extrapolated to produce financial variable shocks (mainly interest rate spread shocks by security type). Interest rate and spread shocks were applied to debt portfolios using the modified duration approach by debt portfolio type (CAN sovereign, other sovereign issuers, provincial and municipal governments, MBSs. corporate bonds and paper, and ABSs and ABCPs).

64. The analysis covered the impact of interest rate risks and spread risks on sovereign and corporate debt securities in all accounting portfolios: amortized cost (AC), FVOCI, and FVPL. Losses from AC and FVPL portfolios was assumed to have an impact on regulatory capital through net profits, while those for FVOCI portfolios affected capital through other comprehensive income. For interest rate sensitive positions in the AC portfolio, only the impact of credit spread changes was accounted for. For conservatism, existing hedges were assumed to be ineffective during the scenario horizon.27

65. Debt securities are primarily held in the FVOCI book and are largely dominated by domestic and international sovereign debt. Sovereign debt accounted for 57 percent of the FVOCI portfolio and 40 percent of the total securities portfolio, with provincial government debt adding an additional 12 percent in the total general government securities share. Corporate debt stood at 14 percent, with an additional 12 percent of asset-backed securities (ABS) and mortgage-backed securities (MBS). In terms of portfolio allocation, 12.4 percent of securities are in the AC portfolio, 33.6 percent in the FVOCI portfolio, and the residual in the FVPL book. Figure 12 illustrates the composition of the three accounting portfolios and a breakdown by asset/security type.

Figure 12.
Figure 12.

Canada: Securities Holdings Analysis

Citation: IMF Staff Country Reports 2020, 016; 10.5089/9781513527116.002.A001

Source: IMF staff estimates.Note: FVOCI = Fair-value Through Other Comprehensive Income; FVTPL = Fair-value Through Profit or Loss.

66. Market valuation losses from interest rate risks in the debt portfolios were derived using a modified duration approach. First, the analysis captured the re-pricing losses in the FVOCI and the FVTPL books due to shocks to sovereign yield curves. Second, it also accounted for the valuation impact in all accounting books due to shocks to the spread of debt securities. Spread projections on provincial government debt, corporate bonds, and ABS/MBS debt were computed based on average yield per maturity tenor of the relevant type of security from Bloomberg and was anchored to reflect the macrofinancial conditions described by the two scenarios.

67. Commodity price risks and mutual fund exposure risks were also accounted for using the shocks provided in the scenarios. The banking sector’s exposure to these types of risks is relatively limited when compared to interest rate and spread risks (Figure 2).

Net Interest Income

68. Econometric models were estimated to project the scenario impact on a set of interest sensitive liability and asset segments. Given that historical price series of effective interest rates were available for six asset and six liability interest bearing segments (Table 9), the FSAP approach was to implement econometric models to project effective interest rates under the two scenarios. First, a system-wide historical time series of effective interest rates was estimated for all twelve segments using quarterly historical data on effective balances and interest rates. Second, an econometric model for each segment was calibrated using BMA techniques. The BMA approach operates with a pool of equations (several hundreds or thousands) per dependent variable, to which weights are assigned that reflect their relative predictive performance, which then results in a “posterior model” equation. The pool of equations contains a large number of equations for every single interest rate risk indicator (per portfolio segment), by considering all possible combinations of predictors from a pool of potential predictor variables, including variables such as real GDP, investment, consumption, price inflation, short- and long-term interest rates, and credit spreads. Detailed satellite estimations on effective interest rates can be found in Appendix V.

Table 9.

Canada: Net Interest Income Segments

article image

69. The distance of bank specific starting points by segments informs the projection for the bank-specific interest rate forward paths. Bank-specific interest rate percentage deltas from the system-wide composite interest rate per segment at the starting point were used to adjust composite interest rate projections (from the BMA satellite model output) and produce bank-specific interest rate projection paths for both scenarios and across all segments.

70. Maturity gap analysis was also used to reflect pass-through constraints for mortgages and consumer credit. The short-term maturity profile of most liability segments implies a fast repricing of most segments, with no significant pass-through constraints. For these segments a direct net interest income/expense projection was based on the effective interest rate projected by the satellite model. For mortgages and consumer credit, after a thorough analysis of the repricing profile in both the domestic and non-domestic dimension, it was decided to also employ an additional constraint to account for the imperfect pass-through conditions (due to the longer-term nature of the repricing gap). Therefore, new mortgage origination rates, as projected in the two macrofinancial scenarios, were used in combination with the repricing gap to derive bank-specific effective interest rate paths. An additional cap of 330 basis points for the spread of the mortgage rate over the personal demand deposit rate was used to ensure conservatism and reflect structural competition factors that would not allow the two segments to reprice independently. For consumer credit, the effective interest rate projected by the econometric model was used as a proxy of the effective rate to be used for the part of the portfolio that is repriced during this period (with no additional constraints). After enforcing these two pass-through constraints, tighter interest margins are observed in the adverse scenario during the first two years of the horizon.

Non-Interest Income Analysis

71. Non-interest income is a significant component of Canadian banks’ overall profitability and internal capital generation capacity. Non-interest total income reached a level of almost 75 percent compared to the sector’s net interest income in 2018 and has been a particularly stable source of profitability over the past seven years. Growth of non-interest income has followed in magnitude the growth of the sector in terms of total asset size and geographical concentration, and the composition of non-interest income by business lines has been broadly stable (Figure 13).

Figure 13.
Figure 13.

Canada: Non-interest Income Analysis

Citation: IMF Staff Country Reports 2020, 016; 10.5089/9781513527116.002.A001

Source: IMF staff estimates.

72. In the adverse scenario, non-interest income was stressed using a relatively simplified approach based on the historical variance of the non-interest income components by income activity. Historical data were collected on total non-interest income and then broken down by specific activity for the last five years. A conservative estimate of projected bank-specific income was produced by adjusting yearly profit by activity: under the adverse scenario, profits from each business activity were projected to be equal to the latest income minus one standard deviation of the historical variability of the income. This approach ensures that projections for banks with higher levels of dispersion in terms of activity income (as a percentage of total assets) would be more negatively impacted. As an additional constraint, a minimum drop in yearly income was introduced for each activity to account for its sensitivity to adverse macrofinancial conditions. To reduce balance sheet sensitivity, and to fully account for the non-domestic expansion of Canadian banks, bank-specific standard deviations were calculated after dividing historical profits by bank total assets each year.

73. The approach allows for different sub-components to be stressed in accordance with the degree of their market sensitivity or to the level of historical variability. Income breakdown across activities serves as the starting point that generates different projections for each bank. The foreign-currency composition of each revenue stream in both scenarios is also accounted for by considering the exchange rate appreciation/depreciation in the scenario and the currency breakdown at the cut-off date. Operating expenses are assumed to remain constant through the scenario, allowing only for the impact of exchange rate appreciation/depreciation. Additional information on the business activity lines that were considered, and the minimum scaling factors used, can be found in Appendix VI.

D. IMF Top-Down Stress Tests: Results

Main Results

74. Banks appear resilient to severe macrofinancial shocks, with all banks meeting the hurdle rates by the end of the stress testing horizon (Figure 14).

  • In the baseline, the system-wide aggregate CET1 capital ratio would be on an upward trajectory due to banks’ solid revenue-generating capacity.

  • In the adverse scenario, the aggregate CET1 capital ratio would decline by 4.8 percentage points to 7.4 percent in 2020 before recovering to 9.6 percent in 2021. During the stress testing horizon, most banks would run down conservation capital buffers, subjecting them to dividend restrictions (partially or fully).

  • The worst outcome in 2020 largely reflects the dynamics of the adverse scenario that features a strong recovery with a more favorable interest rate environment in 2021. During 2019–20, in the adverse scenario, cumulative pre-loss income would amount to 2.9 percent of RWAs, compared with losses due to credit and market risks of 3.9 and 1.8 percent of RWAs, respectively. Deteriorating credit quality of loan portfolios, as reflected in higher RWAs, would also have an impact of 1.2 percentage points in the final CET1 ratio.

  • The key factors behind the larger capital depletion in the adverse scenario (relative to the baseline) are substantially larger credit impairments, lower net interest income and non-interest income, and increased RWAs. The difference between the baseline and adverse aggregate CET1 capital ratio at the end of the stress testing horizon is 5.6 percentage points.

Figure 14.
Figure 14.

Canada: Bank Solvency Stress Test Results

Citation: IMF Staff Country Reports 2020, 016; 10.5089/9781513527116.002.A001

Source: IMF staff estimates.

75. Stress test results highlight the importance of properly capturing accounting considerations, particularly related to IFRS 9. In the adverse scenario, and when compared to the baseline, credit impairments could increase significantly under IFRS 9, which introduced lifetime expected credit losses for Stage 2 assets (i.e., non-defaulted exposures showing deteriorating credit quality). In particular, aggregate Stage 2 assets and Stage 3 assets (i.e., defaulted exposures) would account for 13.6 and 8.3 percent of total exposures, respectively. Overall, accounting impairment charges in the adverse scenario would exceed regulatory provisions for most banks, which further underpins the importance of also considering the accounting layer in the stress testing exercise. As a result, excess provisions are added back to tier-2 capital, up to the cap of 0.6 percent of credit RWAs, to derive a more accurate projection of realistic CET1 projections.

76. Credit losses are the main driver of the stress test solvency result. Impairments in the adverse scenario are significant across all asset classes. The cumulative default rate at the end of the stress testing horizon stands at 6.4 percent for uninsured mortgages and HELOCs, 13.3 percent for consumer retail exposures, and 8 percent for corporate exposures. The respective figures for the baseline scenario are 0.7, 8.3, and 3.8 percent. Additional cumulative impairment flow under the adverse scenario reaches approximately Can$90 billion, with 37 percent of them corresponding to corporates, 34 percent to consumer credit asset classes, and 29 percent to uninsured mortgages and HELOCs (Figure 14).28

77. Net interest income experiences a significant drop in the adverse scenario due to the imperfect pass-through. The snapback nature of the interest rate path in the adverse scenario is a particularly challenging scenario for the system in terms of pass-through efficiency. The short-term nature of the funding mix tends to reprice faster than the asset side when interest rates are on the rise, leading to a higher increase of the interest expense component versus the interest income component (Figure 14). This translates into compressed net margin for all years. Average yearly net interest income is approximately 24 percent lower when compared to the year preceding the cut-off date (2018) and the average net interest income in the baseline. It is important to point out that the aggregate net interest income reaches the lowest value during Year 2 when net interest expense reaches the maximum value. At this point net interest income is 34 percent below the cut-off year level, and broadly recovers after that, following the favorable adjustment of interest rates.

78. In the adverse scenario the impact for non-interest income is considerable. Non-interest income projections are approximately 20 percent lower for the system aggregate in the adverse scenario relative to the baseline. Banks still manage to generate Can$168 billion of non-interest profits during the three years, or an average of Can$56 billion per year (Figure 14), versus a yearly starting point of Can$70 billion. This decline could be attributed to the conservative estimates across the various business lines and, to a lesser extent, the slowly decreasing size of interest-bearing assets and the exchange rate adjustments in accordance with the scenario. In contrast, foreign-currency related effects on non-operational expenses do not to seem to play an important role in the overall result.

79. While pre-loss income is substantially stressed it remains positive throughout the entire stress testing horizon of the adverse scenario. This is because the starting points for bank net interest income (high net interest margin in absolute terms) and non-interest income are very high compared to operational expenses. The significant level of diversification observed in the break-down of non-interest income and the concentration on business lines that are not market sensitive also have a positive contribution to the observed resiliency of the pre-loss income. Such resiliency supports the capital position of banks through positive earnings under stressed conditions.

80. The total market risk impact largely depends on the scenario path and more specifically on the interest rate path of the scenario. FVOCI and FVTPL portfolios contain significant positions of interest rate sensitive instruments. As a result, the end-horizon losses under the baseline are larger compared to the losses incurred under the adverse scenario due to the higher interest rates shock experienced. In the adverse scenario, profit-and-loss and other comprehensive income losses are not substantial (less than Can$10 billion, or approximately 0.5 percent of RWAs), but the timing of losses (and profits) is important. Given the material shock to interest rates during Year 1 in the adverse scenario, revaluation market risk losses reach Can$43 billion. Most of it is partly recovered during Years 2 and 3, as yields gradually converge to lower levels following accommodative policy changes. In the baseline, interest rate losses are more protracted: at the end of the stress testing horizon market risk losses account for Can$32.4 billion, or 1.6 percent of RWAs. Exposures on other asset types (equity risk, post recognition of current hedging relationships; commodity risks; and fund participation risks) do not seem to have a material impact on profit and loss or capital. Derivatives exposures, though significant in terms of notional amounts, were not explicitly stressed due to unavailability of sensitivity data. Data limitations made it impossible to include counterparty credit risk and credit valuation adjustment (CVA) risk in the scope of the market risk analysis.

81. Unfavorable macrofinancial conditions and the resulting deterioration of credit quality for the credit exposures have a material impact on RWAs in the adverse scenario. Overall, credit risk RWAs increase to Can$1.66 trillion in Year 2 of the adverse scenario versus a starting level of Can$1.29 trillion, reflecting the significant increase in the through-the-cycle PDs for all asset classes (Figure 15). Improving macrofinancial conditions during Year 3 have some positive impact on corporate PDs and, therefore, total RWAs are lower in the later year. If compared with the starting point (cut-off date), credit RWAs increase by 29 percent by Year 2 and 17 percent by Year 3 and have a negative contribution to the CET1 depletion of 1.2 percent and 0.8 percent of total RWAs, respectively. The impact of market risk RWA changes in the adverse is immaterial given the small relative size of market risk RWAs.

Figure 15.
Figure 15.

Canada: Stress Test Results—Credit Risk Impairments and RWA Projections

Citation: IMF Staff Country Reports 2020, 016; 10.5089/9781513527116.002.A001

Source: IMF staff estimates.

82. When compared with the respective values of the baseline case, incurred losses under the adverse scenario are (partially) offset by the lower provisions for tax and the assumed restrictions on dividend distribution. Taxes are generally lower because banks may be loss making or experience significantly lower profitability in the adverse scenario than in the baseline. Dividend distribution is also restricted as a result of the regulatory restrictions regarding CCB maintenance. These are illustrated in Table 7.29 As a result, tax-linked outflows are lower by 0.9 percent of starting RWAs and dividend distributions are lower by 0.7 percent of RWAs, partially compensating for the more severe CET1 depletion in the adverse scenario.

Sensitivity Analysis

83. There are a number of Canada-specific features that cannot be fully captured in the abovementioned set of stress test results that are based on standard assumptions, for which further sensitivity analysis was performed by the FSAP team. In agreement with the BOC, the FSAP team conducted some additional tests to quantify the impact of the following elements: (i) sizeable undrawn exposures related to HELOCs and consumer credit (not captured under a quasi-static balance sheet assumption); (ii) mortgage exposures where lifetime expected credit losses (for Stage II assets under IFRS 9) are calculated based on remaining contractual maturity, which is far shorter than the residual amortization period; and (iii) more dynamic risk-based pricing of mortgage spreads, which would be more likely to adversely affect financially vulnerable households.

Stressing Undrawn Exposures

84. The quasi-static balance sheet assumption does not fully assess the impact of undrawn off-balance sheet exposures in the banking book. A more rigorous assessment would assume that part of the undrawn committed lines of credit are utilized during the stress testing horizon. This assumption can only be explored under a fully dynamic balance sheet assumption where behavioral reactions are captured for borrowers and creditors. Given that significantly sized undrawn exposures were observed in Canada, the FSAP approach was to evaluate the impact of such exposures in the context of sensitivity analysis and compare it with the results obtained under the constrained balance sheet presented in the previous section.

85. In the above context, a certain percentage of the undrawn credit line is assumed to be utilized during each year of the horizon, and the appropriate loss rates are applied for the assumed additional drawn exposures. The analysis assumed that 19 percent of the starting undrawn notional credit line for uninsured mortgages, 13 percent for HELOCs, 3 percent for consumer credit, 6 percent for credit cards, and 9 percent for corporate is drawn each year on top of the drawn exposure path under the core solvency stress testing approach.30 Second-round effects, such as the impact on RWAs and the impacts linked to the quality of the portfolios under this “blending,” are ignored. Based on the above assumptions, total impairments in the adverse scenario were found to be Can$18.5 billion higher than the main results with constrained balance sheets. This corresponds to an additional CET1 impact of 91 basis points in the adverse scenario by the end of Year 3.

86. Sensitivity analysis results suggest that the potential impact of undrawn exposures is substantial, especially in the adverse scenario. The growing importance of segments with substantial undrawn parts has constrained creditors’ ability to proactively deleverage (mainly HELOCs but also consumer lines of credit). This potential risk needs to be monitored. Large undrawn credit lines may also make it more difficult for lenders to identify emerging credit problems. This is because borrowers can utilize available lines to mask high debt loads by consolidating high-interest loans into a secured credit line that charges a lower interest rate. As a result, lenders may not observe the initial phases of a borrower’s financial distress if borrowers use their HELOC products to make regular payments on their other loans. The interaction of the rising popularity of HELOC-type products with rising household balance sheet vulnerabilities has an amplifying effect, and close monitoring (or preventive action) may be required.

Stressing the Residual Amortization Tenor for IFRS 9-Expected Credit Loss

87. Under IFRS 9 the contractual tenor of mortgage exposures is used to calculate lifetime expected credit losses. The term contractual tenor refers to the average portfolio maturity until renewal, which is significantly shorter than the average portfolio amortization period given the five-year standard renewal cycle that prevails in the Canadian mortgage market. As part of the sensitivity analysis, the residual impact on credit losses was measured in the hypothetical case were the contractual maturity matches the residual amortization period. Under this assumption,31 Stage II mortgages would incur higher losses since the average lifetime would be significantly longer than the one considered under the contractual assumption.

88. The average lifetime for mortgages was set at 15 years as opposed to the 2.5 years used in the primary stress test approach. The yearly portfolio maturation rate was assumed to be 4 percent (corresponding to a 25-year amortization period). Due to the adjusted maturation rate, flows between IFRS 9 stages are also affected. Under these alternative assumptions the relative size of Stage II loan assets in the adverse scenario increases to 16.4 percent from 13.6 percent in the primary approach, and the average coverage ratio across all asset segments almost doubles to 3.4 percent (Figure 16). However, the impact on total credit impairment flows is not substantial since they only increase by Can$5.2 billion to Can$94.8 for Year 3 of the adverse scenario. This increase in impairments flows corresponds to an additional decline of the CET1 capital ratio by 26 basis points at the end of the stress testing horizon.

Figure 16.
Figure 16.

Canada: Sensitivity Analysis Results

Citation: IMF Staff Country Reports 2020, 016; 10.5089/9781513527116.002.A001

Source: IMF staff estimates.Notes: A=adverse scenario; B=baseline.

89. The results of this analysis should be interpreted with some caution. Given that there is no data to support a proper differentiation between the quality of the exposures that flow into Stage II (a single PD PiT is used for all performing exposures in Stage I and Stage II), the lifetime expected credit losses might be underestimated when compared with bottom-up projections.32 (The latter are obtained using a higher level of granularity in terms of credit quality of exposures that flow into Stage II.) In relative terms, the increase in provisioning does not appear to be that critical.

Risk-based Pricing for Mortgages

90. Risk-based pricing analysis examines the impact that additional capital charge costs (attributed to the costs arising from elevated RWAs) have on borrower affordability. Hence, borrowers of lower credit quality (as captured by a lower affordability ratio) face higher spreads over the benchmark lending rate. Concurrently, the analysis assumes that competition forces may drive spreads lower for the high-quality (prime) segments in a downturn. The FSAP performed a simulation where the same satellite model (structural model for mortgage defaults) was modified to account for an additional variable that reflects the assumed risk-based sensitive spread over the lending rate. Borrowers in the riskier DSR band (DSR ratios higher than 0.4) were offered renewals with an additional spread of 100 basis points, and borrowers in the next band (DSR>0.3) were offered a contract with an adjusted higher spread of 50 basis points. On the prime side, the highest quality borrowers (DSR<0.15) were offered a spread that was tighter by 50 basis points to the applied benchmark interest rate change. Effective interest rates for the other DSR bands were not changed and the spread add-ons were applied to all borrowers in the distribution, irrespective of their LTV bucket.

91. There is a significant impact on model projected default and loss rates under the risk-based pricing assumption. PD PiTs for both insured and uninsured segments increase by 40 to 50 percent, and the increase gets translated into loss rates that are also higher by a similar magnitude (Figure 16). Several model simulations with different risk-based pricing spreads confirm an increase in loss rates by some magnitude.

92. This result underpins the importance of having mechanisms in place that would prevent “wrong-way” risks for the system. For example, building capital buffers during good times could be used to mitigate amplifying correlation risks during an economic downturn (i.e., when house price developments and affordability concerns are contributing to credit quality deterioration of the portfolio). Shaping the institutional landscape toward a direction that favors countercyclical risk-based pricing for all mortgages would probably be a first step in this direction.

93. The Canadian mortgage market is vulnerable to the long-dated amortization profile of the entire portfolio. There is currently a significant concentration of exposures with average remaining amortization tenors between 20 and 30 years (Figure 17). This remarkably skewed profile appears to be more persistent than the impact of longer amortization periods at origination for some vintages and may also infer that the amortization period has been extended during regular contractual renewals for some parts of the existing stock of exposures. The remaining amortization tenor profile should be closely monitored by authorities as it may be used to mask increased household balance sheet vulnerabilities. In addition, it may pose significant impediments during a severe housing market shock because the system’s ability to offer the most efficient measure of temporary forbearance—the extension of the amortization period—may have already been used for most borrowers. Furthermore, it endogenously contributes to the amplification of systemic risks associated with renewal risk, especially under a scenario where banks are forced to deleverage.

Figure 17.
Figure 17.

Canada: Mortgage Portfolio Amortization Profile

Citation: IMF Staff Country Reports 2020, 016; 10.5089/9781513527116.002.A001

Source: IMF staff estimates based on data provided by OSFI, AMF, BOC.

Concentration Risk Analysis

94. Concentration risk was also part of the FSAP analysis. The analysis used large credit exposures data to non-financial corporations for all banks and assessed the impact of simultaneous defaults for the largest one, five, and ten credit exposures. Assuming an aggregate LGD of 50 percent, the default of the single largest exposure would have an impact of 38 basis points in the average CET1 capital ratio. A concurrent default of the five largest exposures would reduce the CET1 capital ratio by 1.2 percentage points, and a simultaneous default of the ten largest exposures would lead to a decline in the CET1 capital ratio by 1.8 percentage points. In each of these theoretical counterparty default cases, no individual bank would fall below any regulatory CET1 capital minimum. A more thorough review of name-specific exposures also revealed that the most significant exposures are toward well-established domestic and global corporate entities.

95. A complementary concentration analysis was conducted on interbank unsecured counterparty exposures, based on the available data from regulatory returns. At the interbank level, scattered concentrations are noted. However, exposure analysis finds that these are at the operational level (e.g., exposures with custodian banks, clearing houses, and central counterparties). In that context, the review did not find elevated or unexpected concentration risks for the Canadian banking sector.

E. Top-Down Stress Tests by Authorities

96. The BOC stress test results are largely aligned with the FSAP stress test results in terms of overall impact on capital and relative performance of individual banks. Although the approaches might be different in several aspects, projections are comparable in key segments—PDs and loss rates for credit exposures, slowdown in net interest income, and dynamics of non-interest income and operating expenses. Most of the differences can be attributed to the specific assumptions or modelling approach of either exercise. A summary of the BOC stress test results was presented in the May 2019 Staff Analytical Note 2019–1633 after the conclusion of both exercises.

F. Assessment of the Bank of Canada’s Solvency Stress Testing Framework

97. The BOC solvency stress test framework has significantly evolved in recent years. The BOC stress testing and analytical framework has its origin in the robust analysis of second-round effects at the system-wide level. It has a specific focus on contagion and spillover effects from the propagation of solvency and liquidity shocks within the stressed macroeconomic environment (MacroFinancial Risk Assessment Framework module). Recently, the BOC has developed the core layer of a unified top-down macro stress testing framework (Macro Stress Testing module) that accounts for a full top-down macroprudential stress test and can be naturally combined with the second-round analysis module.

98. Although the pace of developing a fully-fledged macroprudential stress testing capacity has been rapid, some further enhancements are yet to be implemented. More specifically, there are two major areas that need to be developed:

  • Most of the satellite models should be adjusted or further developed in a direction that can accommodate bank-specific characteristics. The framework should (i) combine the estimated satellite models and the use of bank-specific starting points to better capture idiosyncratic attributes of bank portfolios; (ii) collect a wider set of bank specific historical data points; (iii) and extend the use of models or estimation techniques that can fully capture bank specificities would be a major step in this direction.

  • The approach should become more granular so that data attributes that are really important for the domestic banking system are fully captured. Examples of potential dimensions to be explored are geographies of exposures or significant currencies. Credit and interest rate parameters are important to be tracked by geography or by significant currency. A single “blended” mix across all geographies or currencies may simply not be good enough as the baseline option going forward.

  • A direct consequence of such enhancements would be the adaptation of satellite models to produce bank-specific projections that are anchored to bank-specific parameters and not system-wide averages. This process may be further supported by ensuring that the required data are also reported by banks on a regular basis.

99. A wider coverage of risks would also be required. This might include segments that are not currently covered (for example market risk or operational risk) but it would ensure that the framework becomes more comprehensive and granular on how the overall stress test impact is projected. This direction could include more scenario-consistent projections on balance sheet evolution (quasi-static or dynamic balance sheet), a more granular coverage of RWA tracking for performing and nonperforming portfolios, a full coverage of nonperforming exposures, and the development of an accounting layer to distinguish economic losses from accounting reality (fully fledged IFRS 9 accounting layer).

G. Policy Recommendations

100. Additional required capital for mortgage exposures are desirable. While banks’ overall capital buffers are adequate, lenders’ risk weights for mortgage exposures should be higher. In particular, authorities should develop and communicate to the industry a minimum set of attributes that would be sufficient to characterize a historical period as a full economic cycle. In the absence of historical episodes of economic downturns, a representative proxy might be required; authorities should enforce the use of such downturn proxies, given that historical data may provide little or no use. Avoiding extreme procyclicality by incorporating buffers in the system (including those for banks and mortgage insurers) during good times may prove to be critical to surpass any cliff effects during an adverse macrofinancial shock. Enforcing higher levels of RWA densities for mortgage portfolios and to ensure that risk amplifications during a downturn are (at least partially) mitigated is recommended.

101. The use of “challenger” models that are conceptually orthogonal to typical econometric techniques based on historical data may be useful in this direction. Challenger models provide a reasonable alternative to classical econometric techniques when the later fail to produce rational scenario projections. They could potentially provide some end-state visibility on how portfolios would look like after a simulated shock and support the concept of artificially enlarging the available data space using simulation. Banks may be guided to use similar approaches in their internal models to ensure that significant parts of stressed history is included in their calibrations.

102. Stress testing infrastructure should be further developed but, most importantly, this should also be complemented by a fully-fledged analysis and redesign of the relevant data needs. The existing regulatory reporting framework, though relatively modern and sophisticated, may need to be revamped to better capture (at least) the following emerging needs:

  • As geographies outside Canada gradually become more important for some institutions, reporting requirements should be adjusted to better capture country-specific attributes (including foreign-currency linked dependencies).34

  • System-wide risks should also be accounted for, and additional bank-specific reporting that captures vulnerabilities (post aggregation) at the system level may also be needed.35

103. An elevated level of monitoring and additional reporting would be required on issues that are linked with the size and the complexity of the Canadian banking system. Potential areas of focus include large derivatives exposures (mainly potential foreign-currency exposure mismatches, but also interest rates exposures), significant counterparty risks, the impact of existing hedging strategies and their potential inefficiencies, and risks originating from funding or currency mix concentration. Due to the systemic nature of some of these vulnerabilities, and their potential impact on market confidence, it is recommended that the BOC, OSFI, and AMF discuss and agree on a common monitoring and reporting framework that would provide a “bird’s eye view” on the entire system. Building additional capacity to selectively include some of these themes in the regular top-down or bottom-up stress testing exercises would be the natural step to follow.

Bank Liquidity Stress Tests

A. Overview

104. This FSAP assessed liquidity risks that could arise from volatility in bank funding sources. The team used two approaches to evaluate liquidity risks—the LCR-based test and the cash-flow analysis.36 For both types of stress tests a set of scenarios reflecting a systemic liquidity stress event were derived, liquidity conditions for all banks were simulated, and the relevant liquidity metrics calculated after the application of the scenario. Scenario shocks were developed based on historical liquidity crisis episodes and were calibrated in such a way that they would target specific vulnerabilities of the domestic market.

B. Current Liquidity Conditions and Bank Liquidity Profiles

105. Canadian banks appear to be well diversified in terms of funding mix. In addition to the bulk of retail and wholesale deposits, banks also appear to depend on significantly sized repo and derivative-linked segments (Figure 18). The significant size of derivatives-related liabilities raises some concerns on the foreign-currency funding mix and its sustainability under a potential market disruption. However, if the asset side is also taken into account, there are no major mismatches to be reported. The notional size of the FX and cross-currency swaps suggests that a thorough monitoring of cross-currency funding is warranted.

Figure 18.
Figure 18.

Canada: Stress Test Results—Liquidity Analysis

Citation: IMF Staff Country Reports 2020, 016; 10.5089/9781513527116.002.A001

Sources: IMF staff estimates.

106. Consolidated regulatory reporting for liquidity monitoring purposes served as the data layer for FSAP liquidity stress tests. The existing set of liquidity monitoring regulatory returns provides a very granular and rich dataset on several liquidity monitoring metrics that facilitates the monitoring and analysis of liquidity related vulnerabilities. This data set includes information on contractual maturities for all funding segments, a granular picture of the asset side in terms of counterbalancing capacity, and a comprehensive analysis of the use of collateral by security type for repo/reverse repo/securities financing positions.

C. LCR-based Tests

107. The LCR metric measures the ability of banks to meet liquidity needs in a 30-day liquidity stress scenario by using a stock of unencumbered high-quality liquid assets (HQLA). With a regulatory standard of 100 percent, three scenarios were calibrated to measure banks’ ability to withstand a 30-day stressed run-off rate versus the run-off rate prescribed by the regulator for the standard LCR calculation.

108. Three scenarios were considered for the LCR-based tests. The first scenario draws on some stressed run-offs on retail deposits, assuming some increased retail withdrawal and measuring the relative impact on the LCR. The second scenario uses increased run-offs on wholesale sources of funding and especially on wholesale/corporate demand deposits. The third scenario, which combines the stressed run-off parameters of the first two tests, was also evaluated; however, it is not considered to have a similar level of plausibility, given that stress factors are applied in parallel for most funding segments across all banks. A full set of the stressed values that were used and their deviation from the regulatory LCR calculation parameters can be found in the first part of Appendix VII.

Figure 19.