We have received a number of questions regarding Quantitative Impact Study #6 (QIS#6). We are providing answers to a number of the most commonly asked questions below. The questions are organized by risk.
Corrections or clarifications will be made if necessary [in brackets] to original questions and answers from earlier versions of this document. The dates of the changes or additions are indicated in the brackets. New questions and answers are appended below in each section and identified by version.
General Instructions
Version 1
G.1) The discount rate tests in QIS require a significant amount of work for products with interestsensitive cash flows (e.g. Par, UL). Some companies may be unable to complete the regular QIS discount rate tests (i.e. recalculating the results under QIS#5 discount rates)
The discount rate test measuring the volatility is now requested on a voluntary basis for the QIS#6.
Market Risk
Version 1
M.1) We appreciate the simplification to base the QIS on a pretax basis. We assume this should be consistent for all aspects of the QIS, but we note one discrepancy in the Market Risk instructions. With respect to future income tax cash flows in the interest rate risk buffer, we have found that CALM modeling and ALM practices vary across the industry, and even within companies. Future income tax cash flows may not be included in the ALM model. Therefore we recommend OSFI to allow the treatment of future income tax cash flows in QIS interest risk buffer aligning with each company’s own ALM practices.
For QIS#6, we will allow the treatment of future income tax cash flows in QIS interest risk buffer to align with each company’s own ALM practices. It is expected that for final implementation, the future income tax cash flows will be in the IRR calculation.
M.2) We note that changes have been made in QIS#6 to allow for separate interest rates for Par and NonPar business. Can OSFI provide some background on this change? How and why would the prescribed Par rates differ from the NonPar rates? Does OSFI plan to use this functionality to perform its own analysis on the QIS results? We note that changing discount rates would require reprojections of cash flows for interestsensitive business (including Par dividends).
Under IFRS, the discount rates may be different by portfolio and then for Par and Non Par blocks as well. Therefore, in anticipation of potential additional tests of the discount rates this change was made to the forms.
M.3) We would appreciate OSFI providing a full breakdown of how the base scenario spreads are derived (i.e. data source, curves used, interpolation type, spread calculation methodology, assumptions made). For the seg fund miniQIS performed in June 2014, the industry attempted to update the QIS base scenario discount rates using yearend 2013 data, based on the approach described in the Market Risk instructions. There was some uncertainty around the approach taken by OSFI to derive the Corporate A spreads. This background information would also be useful as we are asked to develop our own rates for the outofCanada discount rate test.
The corporate A rates are based on a composite of Arated curves consisting of bonds denominated in local currency. More details could be provided upon request by email.
M.4) We agree with the idea that the UL credited rates should be projected in a manner consistent with the discount rates for QIS. However, we find the additions to the QIS 6 instructions confusing. Could OSFI provide some examples of how they expect these adjustments to work?
The Corporate A returns are a proxy for IFRS discount rates which are returns net of asset credit default risk and market risk, i.e. risks unrelated to the liabilities. They replace the current CALM discount rates.
As a general approach, for all UL types, the company has to recalculate portfolio returns by replacing the company’s gross investment returns (net of C1 and Pfads) with OSFI’s Corporate A (returns) rates (i.e. replace the company's asset mix, reinvestment strategy and spreads) plus or minus the spread between the company’s gross investment returns (net of C1 and Pfads other than the C3) and the company’s CALM valuation rate, with all rates based on the CALM base scenario. Then, the credited rate would normally consider a spread equal to the difference between the CALM credited rate and the CALM valuation discount rate. The spread is calculated with these rates and then it is applied to the base and shocked QIS interest rates.
As an example, if the current rates are the following:
 Gross rate of return (net of Pfads): 5.5%
 Spread between gross rate and credited: 2%
 Credited rate: 3.5% (= 5.5%  2%)
 Valuation rate: 5%
Then if the Corporate A discount rate is 4%, the QIS gross rate of return would be (4% + 5.5%  5% =) 4.5% and the credited rate would be (4.5%  2% =) 2.5%.
In all these cases, as an approximation, the QIS discount rates can be used for the credited rates, if it is considered as not producing a material difference.
M.5) We question the following addition to the QIS#6 instructions for indexlinked RPT products: "Cash flow from investment management fees should be included in both asset and liability cash flows." Investment management fees are a source of income for the insurance company, and should be allowed to offset obligations. We do not agree that they should be treated as part of the passthrough features of the product, thus providing no benefit in the calculation of the interest buffer.
Because the related maintenance expenses are included in the liability Cfs as they should, then the fees should not be kept in the liability Cfs.
M.6) As goodwill is not specifically bifurcated to separate equity investments, we recommend that OSFI acknowledge this will be done on an approximate best efforts basis.
We acknowledge that it can be done on an approximate best effort basis.
M.7) For substantial investment Limited Partnerships, there is an inconsistency between the requirements on page 3 and the requirements within Equity Risk, where the last sentence on page 15 disallows the look through method (MCCSR Section 3.1.2) for limited partnerships. Does this statement only apply to substantial investments in limited partnerships? i.e. the intention is to replace the look through treatment with a 40% risk factor only if the limited partnership is substantial.
The 40% risk factor applies to substantial investments in limited partnerships. For nonsubstantial investments in limited partnerships, the lookthrough method will be replaced with a 30% risk factor for this QIS, the same as nonsubstantial investments in equities. This is a change in policy that should have been mentioned in the first page of the Instructions.
M.8) The QIS 6 instructions now specify that amounts on deposit (including Par AoD) should be reflected in the interest rate risk cash flows. This is not feasible given the current valuation approaches for amounts on deposit. Amounts on deposit are generally reported as time zero nonactuarial liability, and assumptions/models for projecting the associated asset and liability cash flows may not exist. In particular, companies would require guidance on appropriate runoff assumptions for projecting the liability cash flows, and it may not be possible to split out the asset cash flows backing the amounts on deposit liabilities. Given the modeling difficulties noted in this letter, we recommend that OSFI continue to use a factorbased approach for the interest rate risk buffer on amounts on deposit, similar to MCCSR.
For the QIS, insurers can use the MCCSR factors multiplied by 1.25. A new approach will be developed for the final implementation. The spread risk line in the summary sheet of the market risk workbook should be used to input the resulting buffer for interest rate risk on amounts on deposit (for participating and nonparticipating products).
Version 2
M.9) On page 9 of Market Risk instructions (in the UL) section, fifth paragraph says: “These reinvestment assumptions and credited rates should vary appropriately with the scenario that is being tested, including the base scenario. Interest sensitive liability cash flows should use cash flows consistent with CALM but using QIS forward rates as investment return assumptions when the credited rate and product cash flow depend on the bond market rate. If they do not, use CALM credited rates. If this is not material, use CALM cash flows as an approximation. Approximations should be disclosed in the “Questions and Comments” worksheet. Asset cash flows should be consistent with the instructions above.”
In the last sentence, are you saying asset cashflows should be consistent with "Asset Cashflows" section earlier in the document (pg 6)? Or, do you mean the asset cashflows should be consistent with this paragraph, ie modeled as interest sensitive in the stress scenarios to align with the liabilities?
We mean asset cashflows should be consistent with "Asset Cashflows" section earlier in the document (p. 6), as they cannot be sensitive to interest rates.
Insurance Risk
Version 1
I.1) Calculation of Solvency Buffer: It is not practical to calculate the credits for special policyholder and reinsurance arrangements at the policy level. The new wording OSFI has added on page 7 does not address the issues outlined by the industry in our prior letter. We ask that OSFI work with the industry after QIS#6 to develop a practical approach to calculating these credits. The Final Draft introduces the sentence that solvency buffer components for each shock are to be positive. We assume this is at the geography level.
The reinsurance arrangements at the policy level means at the treaty level. The buffer components for each shock must be positive at the geography level. Please describe any negative buffer that has been put to zero in the Q&C sheets.
I.2) Mortality Risk – Level: For Life Supported business shocks (a) and (b), we believe OSFI should revert to the QIS5 terminology of mortality rates, instead of mortality cash flows. We assume that the changes in the QIS 6 Final Draft instructions were intended as clarifications only, and that there are no changes to the shocks from QIS 5.
No change was intended. In the instructions, the word “Rates” should replace the expression “Cfs”.
I.3) Longevity Risk – Trend: We ask OSFI to confirm that the additional test for business outside Canada is not required for QIS 6, and the tab “Longevity  Trend Shock (Can.)” in the insurance risk template should be left blank.
The additional test is not required.
I.4) Morbidity Risk: As 12 months is not necessarily the appropriate line in the sand as some Group contracts are longer than 12 months, we suggest the dividing line be between Individual and Group.
Insurers should, on a best efforts basis, determine the impact of not using the 12month period and make an appropriate adjustment.
I.5) Lapse Risk
(a) Designation of Lapse supported vs. lapse sensitive business
We believe the instructions should stipulate, consistent with the QIS6 excel file and with the QIS 5 approach, that there should be just one (consistent) designation applicable for all four (level, trend, vol, cat) components, not the two designations (level+trend, and vol+cat) listed in the Final Draft instructions. Furthermore, the introduction of the Cat component into the test introduces a complication since the cat shocks are additive, while the level+trend and volatility shocks are multiplicative. We would much prefer to retain the approach that is already familiar to the companies’ business units (i.e. use the level and trend and volatility shocks for the designation test).
Under this approach, consistent with QIS5, the following two tests for the designation of lapse supported vs. lapse sensitive would be conducted as follows (same direction shock for level+trend and volatility only):
 150% (20% for level+trend, 30% for vol) of best estimate lapse rates in the first year, 120% thereafter.
 50% of the best estimate lapse rates in the first year, 80% thereafter.
(b) Volatility + Cat component
We fail to understand why the Cat risk buffer is floored at zero but the Vol risk buffer is not. The instructions pertaining to the flooring of lapse buffers is now inconsistent between volatility and catastrophe risk. We propose that both should be floored at zero, at the geography level. Flooring at the portfolio level is practically difficult, as it requires the collection of detailed portfoliolevel results from the business units and manual changes to the QIS templates as the total PV of cash flows does not reflect the portfoliolevel flooring. Furthermore, we believe that geographylevel flooring would lead to more consistent results across the industry as the interpretation of “portfolio” may vary among companies.

The lapse supported vs lapse sensitive designation only needs to be done once (this is a clarification and revision to the description in the Insurance risk instructions). This will determine the lapse designation over the lifetime of the product, which will affect how diversification is applied in the correlation matrix. For QIS#6, companies should make the designation using the following two tests to determine whether a portfolio is lapse supported or lapse sensitive (same direction of shocks assigned based on Level and Trend, and Volatility combined):
 150% (20% for level+trend, 30% for vol) of best estimate lapse rates in the first year, 120% thereafter.
 50% of the best estimate lapse rates in the first year, 80% thereafter.
The QIS#6 instructions do not request that companies use the crossover logic for the designation, but only for the application of the shock in the buffer calculation for Level and Trend.
The designation must be done by portfolio, unless there are very similar or nonmaterial portfolios. Portfolio should be defined at the most granular level that results in homogeneous portfolios (i.e., same designation for all policies in the portfolio), within practical constraints for the company. Note that the template design requires that portfolio is defined to be at least as granular as the Geography/Product Types listed in the columns of the QIS6 template.
However, as the volatility and catastrophe risk are only for one year, it is also important and necessary to make sure that the Volatility and Catastrophe shock applies in the right direction.
Under this approach, if the Volatility or Catastrophe buffers are negative at the portfolio level, then the opposite shocks should be used for Volatility and Catastrophe buffer calculation, but without changing the designation of the portfolio. If not possible to do this exact calculation, Volatility and Cat buffers must be floored to zero by portfolio prior to aggregating at the Geography level. In order to estimate the approach in the instructions on a best effort basis, the impact of using the opposite shocks for the portfolios with negative Volatility and Catastrophe buffers should be disclosed in the Comments section.
To apply the zero flooring at the Geography/Product Type level in the template, two changes are needed and must be done by the companies for this estimate:

In the Lapse Risk tab of the Insurance Risk template, flooring is currently defined at the Geographic level in rows 17 and 22 (ex. columns AQ and AR for Canada, CE and CF for USA...). These flooring formulas need to be copied and pasted into the Product Type columns (AU to CC for Canada).

Then the subtotal at Geography level has to be modified to be a sum of the Product Type inputs, similar to the aggregation of the PV cashflows in the rows above the buffer calculation. For example for Canada, copy AQ16 to AQ17, and AQ21 to AQ22 etc.
 We wanted to calculate the Catastrophe impact at a more granular level to eliminate the potential cross subsidization within the designated portfolios. Due to the extreme magnitude of the Catastrophe shocks, we are concerned that the direction of the shock over 1 year may be different from the direction of the designation defined above over the lifetime. Applying the flooring in the template at the Geography/Product level will mitigate some of this issue, but there could still be an understatement of the Catastrophe risk. When the nonfloored Catastrophe buffer is negative prior to flooring, we request that companies assess the impact of using the opposite Catastrophe shocks and disclose in the Comments section.
Version 2
I.6) Should Group Life with Waiver of Premium be subject to the mortality volatility and catastrophe risk buffers calculations in QIS? While the insurance risk instructions (page 8) state that the "mortality risk solvency buffer is not calculated for nonlife products such as WP, CI and deferred annuities", the volatility and catastrophe buffers (pages 11, 12) should be calculated for "all individual and group life insurance products that include a mortality risk."
Mortality shocks don't apply to WP because the morbidity shocks already incorporate the additional mortality risks, i.e. morbidity shocks consider both adverse morbidity and mortality experience. In other words, volatility and catastrophe WP morbidity shocks consider higher disability rates and subsequent higher mortality rates, the level WP morbidity shock considers lower termination rates and higher mortality rates (level and trend), and the trend WP morbidity shock considers no improvement of the termination rates.
Credit Risk
Version 1
C.1) For Public Bonds, we request that OSFI put back in the QIS5 reporting expedient – if it is not feasible to provide the average maturities, then companies may use the highest factor.
The company should, on a best efforts basis, approximate the required results (for example, it could use the midpoint of the maturity range).
C.2) For the Negative Reinsurance Assets, as our industry working group had different interpretations on the level of aggregation to floor assets at zero, we suggest OSFI provide clarifications.
Total reinsurance assets by reinsurer must be floored at zero, if negative, before the factor is applied to the reinsurance assets.
Summary forms
Version 2
SF.1) A few errors were found in the Summary form, in the tab TAR Comparison QIS vs. MCCSR:
 Line 21 – line 20 was not included in the formula and should have been
 Cell K17 – should refer to U222 instead of AD222
 Cell K18 – should refer to T222 instead of AC222
Please make the corrections to the forms before submitting.