Assessing the Resilience of the Canadian Banking System

A resilient banking system can handle major adverse events

The stability of the Canadian financial system, as well as its ability to support the Canadian economy, depends on the ability of financial institutions to absorb and manage major shocks. This is especially true for large banks, which perform services essential to the Canadian economy. Using a suite of models called FRIDA (the Framework for Risk Identification and Assessment), we assess how these banks would perform in a major adverse scenario. This is a kind of stress test for the Big Six Canadian banks and the Desjardins Group (referred to collectively as domestic systemically important financial institutions, or D-SIFIs).1 Together these D-SIFIs account for 90 per cent of the total assets of Canadian deposit-taking institutions. However, the goal of this exercise is not to evaluate outcomes for individual institutions but, rather, to understand the resilience of the banking system as a whole.

We find that although the D-SIFIs would suffer significant losses in this severe scenario, they would remain resilient. Their capital buffers and ability to generate income would maintain capital levels well above the regulatory minimum.

A severe but plausible risk scenario

This exercise considers a three-year adverse scenario developed in collaboration with the International Monetary Fund (IMF), as part of the IMF’s 2019 assessment of Canada under the Financial Sector Assessment Program. The scenario considered in this analysis is not a prediction or expected outcome. Rather, it is a hypothetical episode, specifically designed to investigate the potential impacts of severe but plausible macroeconomic and financial conditions.

The scenario begins with an escalation of global trade tensions, which leads to a decline in international trade and global productivity and a period of persistently above-target inflation. To stabilize inflationary expectations, some major central banks tighten monetary policy at a faster-than-expected pace. This is accompanied by an unexpected and disorderly repricing of risks. Global financial conditions tighten significantly, setting off downturns in the housing market and credit cycle. Consequently, global economic activity weakens markedly, and the deflationary pressures induced by the recession prompt accommodative monetary policy in later years.

In Canada, the scenario is a combination of the three key risks highlighted in the Financial System Review: (i) a severe nationwide recession leading to a widespread increase in financial stress, (ii) a house price correction in overheated markets, and (iii) a sharp increase in long-term interest rates driven by higher global risk premiums. Because Canada has not experienced a severe downturn comparable with those experienced in the United States and Europe, the scenario is designed to be more severe than previous Canadian recessions in terms of the decline in real gross domestic product (GDP), the duration of the recession and the increase in the unemployment rate (Table 1). Notably, while the house price decline is much larger than in previous Canadian recessions, it is comparable to the decline experienced in the United States during the Great Recession.

Table 1: The risk scenario is more severe than any of the three previous major recessions

Sources: Statistics Canada and Bank of Canada calculations

Banking system capital would be diminished

To assess the impact on the banking system, we estimate the credit and market losses expected under the adverse scenario. We then weigh these losses against the banking system’s loss-absorbing capacity,2 which includes initial capital buffers and income generated over the course of the scenario.

We find that the estimated credit and market losses would erode aggregate capital in the banking system over the three-year period (Chart 1) but that capital levels would nonetheless remain well above the regulatory minimum. The aggregate common equity Tier 1 (CET1) capital ratio of the D-SIFIs declines quickly in the first year of the scenario and breaches the capital conservation buffer (CCB) threshold of 8 per cent early in the second year. After the CCB is breached, banks face automatic restrictions on dividend distributions, as per capital rules, which subsequently reduces the rate of capital depletion.3 After reaching a trough at the end of the second year, the aggregate CET1 ratio begins to recover in the third year as the D-SIFIs return to profitability. Notably, the aggregate CET1 ratio remains well above the regulatory minimum of 4.5 per cent for the entire period of stress.

Chart 1: Evolution of aggregate CET1 ratio in adverse scenario

Chart 2 decomposes the peak-to-trough impact on the aggregate CET1 ratio. D-SIFIs start the scenario with CET1 capital equal to 12.1 per cent of risk-weighted assets.4 Income earned before credit and market losses, referred to as pre-provision net revenue (PPNR),5 contributes an additional 4.2 per cent to loss-absorbing capacity. The combination of credit and market losses results in a depletion of capital equal to 5.8 percentage points of risk-weighted assets leading into the trough. The aggregate CET1 ratio is reduced by a further 2.8 percentage points as a result of both an increase in risk-weighted assets and dividends that were distributed before capital levels fell below the CCB threshold.

Finally, we also consider the potential for second-round effects on bank capital that result from liquidity stress and deleveraging. Second-round effects arise when banks face additional increases in their funding costs due to a weak capital position or are forced to sell assets at fire-sale prices to meet a funding shortfall. However, we find that D-SIFIs have sufficient liquidity buffers to withstand funding stress over the course of the scenario. Therefore, the second-round effects represent an additional impact of only 20 basis points on average.

In the remaining sections of this note, we discuss how we estimate each of these components.

Chart 2: Decomposition of peak-to-trough decline in aggregate CET1 ratio

Chart 3: Estimated credit losses for the 2019 FSAP scenario exceed those of recent historical stress episodes

The highest intensity of credit losses is incurred on loans to domestic consumers and businesses (Chart 4). Credit card portfolios account for a small share of total loans, but they experience substantial losses due to the typically low recovery rate on defaulted credit cards and the significant amount of unutilized credit limits, which are assumed to be partially drawn down prior to default. Canadian business loans also experience high loss rates, driven by a broad increase in defaults across sectors, particularly in the construction and real estate sector. Mortgage portfolios experience elevated losses relative to historical experience due to high unemployment and the large decline in house prices. However, the absolute amount of mortgage losses remains low because of mortgage insurance coverage and because recovery rates on defaulted uninsured mortgages are relatively high, as borrower equity absorbs much of the decline in house prices.

Chart 4: The highest loss rates are incurred by Canadian business and consumer loans

Market losses

Canadian D-SIFIs hold significant portfolios of traded instruments. To the extent that these portfolios are not hedged, they are exposed to the risk of losses resulting from adverse changes in the market prices of bonds, equities, currencies and other assets.

To estimate potential market losses under the adverse-stress scenario, we revalue portfolios of securities held at fair value based on the evolution of securities prices in the scenario. We also revalue securities held at amortized cost, but we only account for losses due to changes in credit risk premiums.

In aggregate, banks experience $43 billion in market losses in the first year of the stress scenario. An important factor in the magnitude of these losses is the sharp rise in long-term interest rates and risk premiums, which depresses the prices of D-SIFIs’ bond portfolios. In the second and third years, improvements in market prices result in annual recoveries of $7 billion and $27 billion, respectively.

Capital buffers and income generation under stress

Regulation requires banks to finance their activities with a certain percentage of equity capital that can absorb unexpected losses in extreme operating environments. Moreover, these capital requirements are risk-based such that they are determined by a bank’s exposure to credit, market and operational risks.

At the beginning of the scenario, the D-SIFIs’ total amount of CET1 capital was $248 billion, or 12.1 per cent of risk-weighted assets. However, we also consider the capital generated through income earned over the course of the stress scenario. These amounts can be significant for the Canadian D-SIFIs. In fiscal year 2018, the D-SIFIs’ PPNR was $67 billion, or 3.3 per cent of risk-weighted assets. Under an adverse scenario, however, we expect D-SIFIs’ ability to generate earnings to be impaired. We estimate the additional capital generated during the adverse scenario by projecting the components of PPNR: net interest income, non-interest income and non-interest expense.

Net interest income

Net interest income is the difference between the interest earned on loans and securities and the interest paid on deposits and other interest-bearing liabilities. We estimate that the net interest income of D-SIFIs would decline by 15 per cent from 2018 levels in the first year of the stress scenario, followed by a gradual recovery in the second and third years (Chart 5).

Chart 5: Pre-provision net revenue declines sharply in the first year of the scenario

The decline relative to 2018 is driven primarily by the compression of the spread between the average interest rate earned on assets and the average rate paid on liabilities (i.e., the net interest margin). When bank funding costs increase in the scenario, we assume that banks are able to pass on the higher cost to their borrowers. However, margins are compressed in the short run because banks’ assets reprice more slowly, on average, than their liabilities. In a rising interest rate environment, the funding cost paid on most liabilities is expected to adjust quickly—particularly demand deposits and short-term wholesale funding.7 Conversely, a significant share of bank loans have fixed interest rates, which will reprice only gradually as loans mature (i.e., about 45 per cent of the mortgage book reprices in the first year of the scenario).

Moreover, the spread between bank funding costs and government bond yields widens during the stress scenario, which puts additional pressure on net interest margins. The regulatory liquidity framework for banks requires them to hold liquid assets, such as government bonds, to meet an unexpected demand for liquidity. As the bank funding spread increases, liquid assets become more expensive to hold as the yield earned on these assets falls further below the banks’ cost of funding.

Non-interest income

In the adverse scenario, we estimate that aggregate non-interest income declines by 17 per cent in the first year relative to 2018 levels, with a modest recovery in the third year.

The decline is driven primarily by a reduction in market-sensitive revenues such as wealth management fees, underwriting and advisory fees, and commissions on securities. We assume that these revenues decline in proportion to bond and equity markets, reflecting a reduction in assets under management and a slowdown in activity in capital markets (i.e., bond and equity issuance, mergers and acquisitions).

A reduction in non-market-sensitive revenues, including deposit account and loan fees, also contributes to the decline in non-interest income. These revenues are expected to fall as loan balances decline and competition for clients reduces fee income.

Non-interest expense

We assume that banks are unable to reduce expenses during the stress scenario and that the non-interest expense is fixed at 2018 levels. While we expect that some reduction in expenses could be achieved by restricting bonuses or through other cost-cutting measures, we assume these savings would be offset by increased expenses elsewhere, such as higher servicing costs associated with increases in impaired loans.

Total capacity to absorb losses

In total, we estimate that banks generate an average of $48 billion per year (28 per cent less than in 2018) over the three-year stress horizon, for a cumulative PPNR of $144 billion. Taken together with the initial level of capital, the total capacity of the D-SIFIs to absorb losses in the context of this scenario is $391 billion.

Risk-weighted assets

In a risk-based capital framework, the amount of capital required for a given loan is sensitive to the borrower’s default risk. In an adverse macroeconomic environment, the deterioration in the credit quality of banks’ assets—even those that are not impaired—can cause a further decline in the CET1 capital ratio due to an increase in risk-weighted assets.

In this scenario, we expect an increase in risk-weighted assets as the decline in loan balances is offset by higher capital requirements on the remaining loan portfolio. After the three-year stress scenario, we estimate that risk-weighted assets would increase by $355 billion.

Even in a severe scenario, the banking system is resilient

The D-SIFIs remain resilient in the face of this severe but plausible risk scenario. Despite large losses, the D-SIFIs are able to continue to generate income. Combined with their diversified asset base and the support of the Canadian mortgage insurance system, backstopped by the Government of Canada, this helps to contain the negative effects. Most importantly, the capital buffers required by Canadian regulations provide the loss-absorbing capacity that allows these institutions to weather the storm. As in any modelling exercise, however, the results depend on a set of assumptions and methodological choices that could affect the outcomes either positively or negatively.

For example, the assumptions regarding losses on loan portfolios are conservative and may overestimate actual losses. And while we find that the impact of second-round effects is limited, our analysis assumes that bank creditors are well informed and forward looking, such that a strong capital position deters them from withdrawing funding. However, an environment of extreme uncertainty in which bank creditors are highly risk averse could result in a disproportionate loss in confidence. While this type of behaviour is not captured by our models, it could lead to more severe outcomes.

Further, it is plausible that in such a scenario, banks would take actions to conserve capital and liquidity during periods of financial stress. This behaviour, while rational at the level of the individual bank, would lead to a tightening of financial conditions and an amplification of the macroeconomic downturn. Higher funding costs for banks and a deterioration of household and business financial positions would translate into stricter lending criteria and a pass-through of higher interest rates to borrowers.

Importantly, this exercise does not consider any exceptional actions that financial institutions or authorities could take to support the resilience of the banking system. Banks can take a range of measures to support their capital positions, including selling business units or raising additional equity capital in the market. Moreover, authorities have a crisis management tool kit they can deploy to reduce contagion and support the functioning of the financial system.

In any stress-testing exercise, there is significant uncertainty around estimates of the financial position of banks in unusual operating environments. Modelling challenges are especially important in the Canadian context, where the banking system has not experienced a severe downturn in recent history. To compensate for the lack of severe outcomes in the historical data, we benchmark our model outputs to severe outcomes observed in other countries, when appropriate given the economic conditions in the scenario. Hence, the conclusions of this exercise offer one perspective on the resilience of the banking system, but they should be considered along with other inputs and judgments.

Endnotes

Acknowledgements

This note would not have been possible without the advice and analytical contributions of Don Coletti, Pradhayini Ganeshanathan, Alejandro Garcia, Harsimran Grewal, Grzegorz Halaj, Grahame Johnson, Alexandra Lai, Sofia Priazhkina and Joshua Slive.