February 10, 2010
Monsieur Bernard Dupont
Office of the Superintendent of Financial Institutions
Director, Capital Division
255 Albert StreetOttawa (ON) K1A 0H2
Dear Monsieur Dupont,
This letter is in response to your request for feedback on the recent Credit Risk and Market Risk
Quantitative Impact Study (QIS’s) as documented in your e-mail to Tyrone Faulds dated October 28.
The QIS’s were tabled for discussion at the November 25 meeting of the Canadian Institute of
Actuaries’ (CIA) Committee on Risk Management and Capital Requirements and several comments
were raised there.
Our overall impressions are positive. We can see that a lot of thought went into the QIS and you clearly
improved the level of risk recognition within the formulas. Some more specific feedback follows.
Comments Common to Both QIS’s
In many places in the two QIS’s you have used a 99.5% confidence level, which is consistent with what
is being discussed with Solvency II as the point of regulator intervention. In other words, 100% of the
provisions calculated at the 99.5% confidence level under Solvency II would be the point of regulator
intervention. In contrast, under the current Minimum Continuing Capital and Surplus Requirements
(MCCSR) guidelines, the point of regulator intervention is defined as 150% of the provision determined
using the factors. It is our understanding that the purpose of the QIS’s is to help define the new MCCSR
standardized approach with the implementation of the first phase targeted for 2012. It is unclear to us if
the intention under the new MCCSR is to maintain the current approach where the intervention point is
at a level of 150% of the provision determined using the factors. If the intention is to maintain the
current approach for consistency, we believe that the factors would then need to be reduced accordingly.
On a related note, you have used a 99.5% confidence level for many credit and interest rate risk factors.
For other factors, it is not clear which confidence level was used. We suggest that a consistent level and
approach be used for all risks in order to avoid inadvertently favouring one risk over another.
Finally, on page 7 of the instructions for the Credit Risk QIS, you state that actuarial liabilities contain
provisions for expected credit defaults. While that is the case currently, it is not likely to be the case
under Phase II of IFRS. In addition, we note that current CIA Standards of Practice require that the
actuary reflect expected credit risk costs as well as hold Provisions for Adverse Deviations (PfADs) for
credit risk and market risk. Many of the credit and market risk factors have been explicitly set at a high
confidence level. Regardless of the reserving methodology, we agree that a credit for any credit or
market risk provisions (expected and unexpected) held within liabilities would be appropriate.
We are supportive of the refinement to the bond factors that recognize that risk varies both by credit
rating and term to maturity. In addition, we are in agreement with using a modified Basel Foundation
approach as opposed to the Solvency II spread-based approach as this would result in increased
consistency between banks and insurance companies.
You may want to consider increasing the number of rating categories for publicly traded bonds to
include both the “+” and “-” sub-categories so as not to encourage companies to invest in the lower
ranges of your broad categories. This would also allow for a more orderly change in required capital as
an asset migrates from rating class to rating class. This may not be practical for private bonds where
information may be limited, especially in the context of the development of internal models.
With respect to private bonds, consideration should be given to whether the factors used when a bond
has a rating should be higher than publicly traded bonds. These bonds usually have higher spreads than
publicly traded bonds due to lesser liquidity. The difference in factors would be intended to reflect the
greater liquidity of the public bond market which provides a financial institution a greater opportunity to
mitigate its risk through the sale of the asset as it begins to be downgraded. Another justification for the
difference in factors is the greater uncertainty in valuing the asset. The factors to be used when a rating
cannot be determined seem conservative but are probably reasonable when one considers that the factors
are to be applied as a standard model framework and the option exists for a company to develop a more
As a final consideration for bonds, both public and private, if a company does have an internal rating
system for these assets, we recommend that you consider allowing the company to use it as part of the
advanced model framework.
For commercial mortgages, we recommend that you consider varying the factors for mortgages by term
to maturity and a measure of risk similar to what you have done for bonds. Possible measures could be
loan to value ratio or debt service coverage. By not varying the factors, you are implicitly encouraging
insurance companies to invest in long, riskier mortgages and shorter bonds.
With respect to the 6% factor itself, we have several observations:
• The 6% factor seems internally consistent with the factors for longer public bonds but perhaps
high for shorter maturities. We compared the commercial mortgage factor with that of a BBB
bond and the 6% factor seems consistent with the factors in excess of 4% for longer, BBB bonds.
• Insurance companies tend to invest in longer-term mortgages than banks. The selection of a
factor lower than the Basel II 8% seems counter-intuitive. Perhaps the differences can be
explained by the differences between insurance and bank accounting for liabilities.
• The bond factors have been set at a 99.5% confidence level. The description of the method of
selecting the 6% factor seems to indicate a lower confidence level as it seems you have set the
factor equal to an observed event in the not too distant past. Given the stated lack of data,
presumably, the capital factor at 99.5% should be higher than observed history.
In light of these observations, we recommend further research be conducted to ensure the factors are
For uninsured residential mortgages, we suggest that you vary the factors by term to maturity and loan to
value ratio. For mortgages that are insured by entities other than Canada Mortgage and Housing
Corporation (CMHC), consideration should be given to varying the factors by the credit rating of the
Default Provisions within Liabilities
We would like to point out a technical point regarding the C1 provisions in actuarial liabilities. On page
7, you state that the actuarial liabilities contain a provision for expected losses. In addition to a provision
for expected losses, our standards of practice require us to hold a Margin for Adverse Deviations
(MfAD) so we are holding a provision for some unexpected losses as well. This should be considered in
setting the capital factors and would suggest, for similar assets, the capital factor for an insurance
company should be lower than other financial institutions.
Interest Rate Risk
We are supportive of the increased level of sophistication for measuring the interest rate risk within an
insurance company as compared to current MCCSR approaches. In response to your proposal, we have
the following observations and comments.
Your approaches are consistent with the approaches being discussed under International Financial
Reporting Standards (IFRS). We also note that there is a fundamental difference between your capital
formulas and the current Canadian Asset Liability Method (CALM) liability methodology. While it is
your intention to implement this with the introduction of IFRS Phase II, we recommend that care be
taken should it be necessary to apply the results of the interest rate scenarios to a balance sheet
containing CALM liabilities as an interim step.
The CALM projection methodology looks at the long-term risks of an insurance product whereas your
capital formulas are focused on the risk over a shorter (12 month) horizon, including the costs of closing
out the risk after a year. As a result, the best estimate liability under CALM will be different from the
present value of future cash flows under your base scenario. This is because longer-term CALM takes
into consideration the historical average of reinvestment rates, whereas your approach is based on
current rates. A simple addition of the impact of the shock to the CALM best estimate liability may
overstate or understate the actual total balance sheet requirement depending on the relationship of the
interest rates in the two base scenarios.
Perhaps the best way to address this issue is to utilize a total balance sheet approach to calculate the
solvency buffer for interest rate risk. By calculating the solvency buffer as the difference between two
scenarios defined by yourself, there is always the risk that your defined base scenario is significantly
different from the opening liabilities, even in an IFRS Phase II world. By calculating the solvency buffer
as the difference between an Office of the Superintendent of Financial Institutions (OSFI) determined
scenario and the company’s liabilities, you eliminate this problem.
Some additional comments:
• The exclusion of non-fixed cash flows is a significant departure from current approaches allowed
under CALM and many companies’ approaches to asset-liability management (ALM), but
consistent with the anticipated direction of IFRS. There was no consensus among our group as to
which approach was appropriate. We provide the following comments for your consideration:
o The exclusion of non-guaranteed dividend and rental income seems conservative for a
diversified portfolio of equities and/or real estate. Inclusion of some conservative
projection of dividend or rental income is probably appropriate as it is unlikely that these
cash flows will disappear completely and therefore appropriate for matching future
liability cash flows.
o Similarly, it is unlikely that a diversified portfolio of equities and/or real estate will have
no realizable value out in the future and some partial recognition may be appropriate.
o The introduction of additional asset classes into an investment portfolio creates an
element of diversification of risk.
o The inclusion of surplus fixed income assets into the calculation does allow the company
an opportunity to compensate for the loss of equities in the calculation by investing them
suitably long to match the long liability cash flows and reduce the risk, and
o Much of the earlier arguments are based on the consideration that insurance contracts are
long-term products and an insurance company will have sufficient time to recover. The
proposed treatment of fixed and non-fixed cash flows is consistent with a shorter-term
focus including the closing out of the risk at the end of 12 months. If the decision were
made to include non-fixed income asset classes into the calculation, the scenarios
probably will have to be expanded in number and complexity in order to reflect the
interrelationships between bonds and other asset classes and accurately measure the risk.
• The use of cash flows including MfADs is consistent with CALM and the approach most
companies use to manage interest rate risk. These MfADs may not exist in the same form under
IFRS Phase II. Some experts do feel that the inclusion of MfADs in ALM analysis is sub-optimal
and you may wish to consider the appropriateness of this for your purposes.
• Your document provides for special treatment for Universal Life (UL) and participating (par)
products. All adjustable products should be treated similarly.
• Your document is unclear with respect to the treatment of tax timing differences. It is our
opinion that these cash flows should be included.
• Principal component analysis of the movement of the yield curve indicates that when the yield
curve flattens or steepens, the range of movement is much less than if a parallel shift were to
occur. As a result, some of your “twist” scenarios where you shock the curve in opposite
directions at the 99.5% confidence level have a probability of occurring much less than 1 in 200.
We note that you have used prescribed interest rates in the measurement of interest rate risk beyond year
30. We believe that the use of prescribed interest rates is a valid approach to introduce stability in the
measurement of the tail risk given that there is very little market for bonds beyond 30 years. We suggest
the following adjustments to the formula:
• Instead of a fixed 6% discount rate, use a moving average of historical long bond yields such that
the rates are similar to current market levels but have a level of stability. The use of a single 6%
discount rate tends to underestimate the risk when rates are low and overstate it when rates are
• Consider using the fixed discount rate at an earlier duration than 30 years. The Government of
Canada issues 30-year bonds periodically and the current 30-year benchmark has an actual term
to maturity of 27 years.
• We observe that some of your scenarios have the shock on the long bond rate and the 31+ year
discount rate going in opposite directions. We believe that these scenarios can be dropped since
we believe that under extreme scenarios, such as you are suggesting, there would be a high level
of correlation between these rates.
We believe that you need to expand the guidance on modeling future policyholder cash flows on
adjustable products, including par whole life and universal life. The guidance provided was unclear to
the intent of how to model future cash flows. These products are highly complicated and contain many
embedded risks. Some suggestions follow.
We believe that, in order to get an accurate measure of the risks, the actuary would have to model future
reinvestment rates, best estimate policyholder investment decisions and, in some cases, full portfolios of
assets. Given the variety of funding approaches and reinvestment strategies a policyholder may take,
significantly different future cash flow patterns may emerge. For example, your suggested approach of
using daily interest when there are no guarantees may result in Universal Life products lapsing
significantly earlier than is truly expected. This could materially understate the interest rate risk on lapse
supported products. The policy funds will likely be invested in a mix of funds that will keep it in force
for a longer period of time and the investment/risk management strategy of the insurance company will
In addition, the typical UL product in Canada offers the policyholder a multitude of fixed income and
equity investment options. The fixed income investment options typically have a two-tiered structure
that guarantees the higher of (a) a percentage (typically 90%) of the Government of Canada bond rate of
the same term and (b) a flat rate. The interaction of the two elements of the guarantee varies
significantly by term and, as such, a significantly different risk position can be obtained depending on
the investment option chosen. It is quite possible that the guarantees for one fund bucket could be
significantly in the money while not being in spread compression in another bucket. In practice, very
little money is invested in the daily interest fund. For these products, we are recommending that future
policy cash flows be modeled using a model office that reflects a reasonable mix of future policyholder
reinvestment decisions and credited rates based on the implied forward rates in your scenarios.
For products that credit interest on a portfolio average approach, we believe the only way to model
future policy cash flows would be to project future asset reinvestment as that is the only way to
determine future crediting rates.
Another challenge is with respect to products that involve equity returns being passed back to the
policyholder, whether it be directly, as is the case with indexed investment options available on most UL
and Variable Universal Life (VUL) products, or indirectly as part of a participating whole life contract.
We believe that this is another situation where it is desirable to project future policyholder
reinvestment/disinvestment decisions. In addition, the related equity assets must be included in the
projections. The reason for these recommendations is the potentially higher yields of equities will keep
the products in force longer and give a truer measure of the interest rate risk that comes from lapse
supported products. If the equity assets are not modeled, we likely overstate the risk as we have
projected policyholder cash flows using equity returns, yet discounted them back at risk free rates.
Finally, we recommend that clear guidance be given as to policyholder behaviour assumptions when the
funds exhaust, even if it is to suggest that the same methodology is to be used as was used in the
With respect to the equity scenarios, we have the following observations and comments:
• It would be helpful if we had further information on the selection of the scenarios chosen as it is
not clear as to how the equity shocks were set. Given the availability of data and mathematical
models, the development of a 99.5% confidence level scenario for equity shocks is achievable.
The specified shocks appear to be below a 99.5% confidence level.
• It was unclear to us as to how to differentiate equity index stocks from managed equity
portfolios. We presume it is meant to include broad-based indices such as the TSX-60 and the
S&P 500, but there exist a number of equity indices covering everything from niches in the
market (technology, for example) and smaller, developing economies that represent much higher
risk than a managed, North American equity portfolio. Perhaps the formula can state the factors
for broad-based indices such as the TSX-60 and all other portfolios can have their capital factors
developed based on the betas or historical volatilities of the respective portfolios.
• Similarly, consideration needs to be given to other, similar asset classes such as commodities, as
well as leveraged investments such as exchange traded funds with betas > 1.
• For pass through products, not all of the risk is necessarily being passed on the policyholders. It
is common practice to take a spread off the equity returns being credited which changes the risk
profile. If the equity portfolio suffers a significant shock, for some products (for example, UL)
the policyholder cash flows may change, resulting in a change in interest rate risk. This is
possibly best illustrated by Canadian Generally Accepted Accounting Principles (GAAP). Under
Canadian GAAP, the present value of future spreads on policyholder investments are considered
in the valuation and are available to cover non-equity cash outflows. If a shock occurs, then the
income statement may reflect a loss because the drop in the asset value may be greater than the
drop in the reserve. We believe that consideration of a total balance sheet approach that
incorporates the projection of future equity cash flows that is integrated with the interest rate
scenarios may be the best way to deal with this.
• Current standards of practice require that the actuary hold an MfAD on returns on equity
investments backing liabilities. As such, future refinements of the formula should consider
providing credit for any PfADs held for equities.
Real Estate Risk
With regards to real estate assets, again it would be helpful to have more information on the derivation
of the scenarios. We also note that our standards of practice require us to hold an MfAD on the returns
of real estate investments backing liabilities.
We have the following observations and comments about your currency risk calculations:
• It may be preferable to base the currency exposures on the statement value of assets and
liabilities as opposed to being based on the net present value of all asset and liability cash flows.
In as much as the statement values differ from the present value of cash flows, the results of
current currency formula may be dominated more by the difference in these accounting
approaches as opposed to actual currency mismatches. Even once companies have switched to
IFRS accounting, and assuming IFRS resembles the present value of cash flow approach used in
the QIS, it would certainly be simpler and more transparent to base the currency exposures on
the statement value of assets and liabilities.
• We understand that surplus held in the functional currency of a non-Canadian entity will attract
currency capital under the QIS proposal, which is in contrast to the current MCCSR guidelines.
It is good risk management to hold surplus in the same currency as the liabilities it supports and
the capital rules should not arbitrarily encourage foreign surplus to be held in Canadian dollars.
• It may be appropriate to vary the 20% factor by currency.
Liability Market Options Risk
Regrettably, time did not permit us to provide in-depth feedback on the segregated fund calculations.
Our only recommendation at this time is that there be consistency between the scenarios for market risk
for general account assets and segregated funds. While the risk profiles of the different products can be
vastly different, the underlying capital markets are the same. We hope to provide further thoughts in the
In closing, we thank you for this opportunity to comment on the QIS’s. We are willing to meet to
discuss any questions you may have and look forward to providing feedback on further iterations.
Robert C.W. Howard, B.Sc., FCIA, FSA