Friday, August 28, 2009

Why Have Canadian Banks Been More Resilient?

Posted on PrefBlog:

A VoxEU piece by Rocco Huang of the Philadelphia Fed and Lev Ratnovski of the IMF is based on an IMF working paper, Why are Canadian Banks More Resilient? that is of great interest:
Reviewing the data, we note that the pre-crisis capital and liquidity ratios of Canadian banks were not exceptionally strong relative to their peers in other OECD countries. However, Canadian banks clearly stood out in terms of funding structure: they relied much less on wholesale funding, and much more on depository funding, much of which came from retail sources such as households. We posit that the funding structure of Canadian banks was the key determinant of their resilience during the turmoil.

Although bank capital ratio taken by itself was not a robust predictor of resilience, a more specific dummy variable capturing critically low (under 4 percent) capital was a significant predictor of sharp equity declines and probability of government assistance. Low balance sheet liquidity did well in predicting extreme stress.

The second part of this paper (Section 3) reviews regulatory and structural factors that may have reduced Canadian banks’ incentives to take risks and contributed to their relative resilience during the turmoil. We identify a number of them: stringent capital regulation with higher-than-Basel minimal requirements, limited involvement of Canadian banks in foreign and wholesale activities, valuable franchises, and a conservative mortgage product market.

We measure capitalization as a ratio of total equity over total assets. This leverage-based measure has a number of shortcomings stemming from its simplicity: it is not risk-weighted and does not consider off-balance sheet exposures. However, it is well comparable across countries. We find that this simple measure of capitalization turns out to be a good predictor of bank performance during the turmoil, particularly by identifying vulnerabilities stemming from critically low bank capital (Table 2).

This last point is not particularly earth-shattering: see the first chart (reproduced from an IMF report) in the post Bank Regulation: The Assets to Capital Multiple.

We assess the impact of these ex-ante fundamentals on bank performance during the crisis. We use three objective and subjective measures of performance.

The first is the equity price decline from January 2007 to January 2009, which is an all-in summary measure of value destruction during the turmoil, resulting from credit losses, writedown on securities, and dilution from new equity issuances including government capital injections.

The second (pair) of measures are two dummy variables identifying whether that decline was greater than the median (70 percent) or extraordinarily large (85 to 100 percent), respectively.

The third measure of performance is a dummy capturing the degree of government intervention that a bank required during the turmoil: whether it was used to avoid extreme stress or to address a less dire weakness.

I have a problem with the use of equity prices as a measure of performance. It doesn’t really measure the stability of the bank, it measures the market’s perception of the stability of the bank. On the other hand, we have to live in the real world and perceptions can become reality very quickly.

We now turn to bank liquidity. We measure balance sheet liquidity as the ratio of liquid assets over total debt liabilities. We use the BankScope measure of liquid assets, which includes cash, government bonds, short-term claims on other banks (including certificates of deposit), and where appropriate the trading portfolio. BankScope harmonizes data from different jurisdictions to arrive at a globally comparable indicator. Data for bank liquidity is shown in Table 3.

Note that a large number of U.S. banks have very scarce balance sheet liquidity. The key reason is that those banks, in their risk-management, treated mortgage-backed securities and municipal bond as liquid, and reduced holdings of other more reliably liquid assets such as government securities. Our liquidity measure does not incorporate holdings of such private and quasi-private securities. With hindsight, it is fair to say that this narrow definition is a more accurate measure of liquidity during crisis.

I have a real problem with the incorporation of claims on other banks in a narrow definition of liquid assets - the same problem I have with the preferential treatment accorded bank paper in the rules for risk-weighting assets. Encouraging banks to hold each other’s paper seems to me to be a recipe for ensuring that bank crises become systemic with great rapidity.

Yet overall, balance sheet liquidity was a weaker predictor of resilience to the turmoil than the capital ratio. Although low liquidity was a clear handicap (of twelve least liquid banks, eight had equity price declines of more than 70 percent, and four required a significant government intervention), a large number of banks from different countries (U.S., UK, Switzerland) experienced significant distress despite being relatively liquid. Another way to think about the resilience effects of balance sheet liquidity is to recognize that it can provide only temporary relief from funding pressures. During a protracted turmoil, more fundamental determinants of resilience—such as capital or funding structure—should play a bigger role.

We now turn to bank funding structure (depository vs. wholesale market funding). The financial turmoil has originally propagated through wholesale financial markets, some of which effectively froze on occasions. Our measure of funding structure, a ratio of depository funding over total assets, seeks to reflect banks’ exposure to rollover risks — the wholesale market’s refusal to roll over short-term funding, often based only on very mild negative information or rumors (Huang and Ratnovski, 2008).

And it seems to me that this last paragraph supports my argument.

Canadian banks are clearly the “positive outliers” among OECD banks in the ratio of depository funding to total assets. On this ratio, almost all large Canadian banks are in the top quartile of our sample. Anecdotal evidence also suggests that a higher fraction (than in the U.S.) of Canadian bank deposits are “core deposits,” i.e., transaction accounts and small deposits, which are “stickier” than large deposits.

One likely reason for Canadian banks’ firm grip of deposit supply is their ability to provide one-stop service in mutual funds and asset management. Unlike in the U.S. Canadian banks have been historically universal banks, and there is relatively less competition for household savings from other alternative investment vehicles.

This might be used as an argument to reduce the choices available to Canadians even further. You can bet the banks’ lobbyists will have copies of this paper tucked into their briefcases during the next revision of the Bank Act.

Regression results are shown in Table 5.

The main specification (columns 1, 4, 7, 10) shows that depository funding significantly and robustly explains bank performance during the credit turmoil, consistent with initial casual observations of the data. Balance sheet illiquidity is a good predictor of particularly rapid deteriorations in bank conditions (government intervention under extreme stress or equity decline above 85 percent). However, interestingly, the capital ratio appears as an insignificant explanatory variable.

Assets-to-capital multiple. In addition to risk-based capital, Canada uses an assets-to-capital multiple (inverse leverage ratio) calculated by dividing the institution’s total assets by total (tiers 1 and 2) capital.

This is not quite correct; the ACM includes off-balance-sheet elements in the numerator.

Finally, the Canadian mortgage market is relatively conservative, with a number of factors contributing to the prudence of mortgage lending (see Kiff, 2009). Less than 3 percent of mortgages are subprime and less than 30 percent of mortgages are securitized (compared with about 15 percent and 60 percent respectively in the United States prior to the crisis). Mortgages with a loan-to-value ratio of more than 80 percent need to be insured for the whole amount (rather than the portion above 80 percent as in the United States). Mortgages with a loan-to-value ratio of more than 95 percent cannot be underwritten by federally-regulated depository institutions. To qualify for mortgage insurance, mortgage debt service-to-income ratio should usually not exceed 32 percent and total debt service 40 percent of gross household income. Few fixed-rate mortgages have a contract term longer than five years.

I suggest the last point is the most critical one here. If the CMHC had not stepped up to buy securitized mortgages at the height of the crisis, how many of these mortgages have been rolled over? That would have been catastrophic. The liquidity advantage of Canadian banks is heightened by the fact that so much of their lending has a maximum term of five years.

This research is clearly still in its early stages, but the paper is vastly superior to the OSFI puff-piece published in May.

No comments: