fasb - yield curve analysis

[Guest VTran]

A client of mine needs to have their FASB 158 accounting completed. Their auditor wants us to do a yield curve analysis to support the choice of a discount rate. Compare the sum of discounted values using yield curve versus sum of discounted values using one single discount rate, what is the percentage difference between the two sums where we may conclude the choice of discount rate is supported by the yield curve?

[Andy the Actuary]

Some thoughts, some of which are totally off point to your question.

(1) How large a plan in terms of liabilities? Is this a publicly traded company or company subject to insurance regulation?

(2) It is the responsibility of the Company working in conjuction with the auditor to establish the economic assumptions that comprise the FASB disclosures. They are not within your juristiction.

(3) You work for the Company and not the auditor. Whereas the auditor may want this analysis, the request should come from the Company because they and not the auditor are paying your fees.

(4) Has the auditor provided their guidelines of an acceptable range of discount rates?

(5) What is the measurement date and the proposed discount rate that the auditor wants a study conducted to justify?

(6) "what is the percentage difference between the two sums where we may conclude the choice of discount rate is supported by the yield curve?" Good question with no definite answer. Which published yield curve?

[SoCalActuary]

Andy makes several excellent points.

Generally, you make a distribution projection of future payments.

You pick from the published yield curves, including the one-day snapshot, the one-month IRS average, the Citicorp table, or your other favorite choice.

Match the payments to the yield curve and solve for the equivalent rate. If your selected rate is within 0.1% of that rate, or maybe even within 0.25%, then you have done a reasonable job of setting assumptions.

Good luck.

[david rigby]

Good discussion from Andy and SoCal. Here's a few more:

- Upon solving for the "equivalent rate", do you round the result? I suggest rounding is permissible, but be careful about rounding too much, lest you lose the credibility of your technique.

- The IRS yield curve goes to 100 years. Since there aren't any such long-term bonds, the IRS has created a technique for extrapolating such rates. Some observers are skeptical (although rates beyond 50 years are usually not a significant portion of the total liability).

- The Citigroup yield curve stops at 30 years. Your plan may have a reasonable portion of its liability after that point, so think carefully about how to address that.

- The Citigroup yield curve comes with its own average rate each month, described as the equivalent rate for a "typical" plan. I have searched without success for their description of such plan. Be careful about assuming that one rate is valid for every plan.

It's not clear (to me, at least) that the original poster is an actuary. If not, I suggest engaging an actuary with appropriate FAS87/158 experience.

[Guest VTran]

By solving for one discount rate (from the yield curve), I assume the plan sponsor will not have much flexibility of choosing their discount rate anymore?

Also, a discount rate, that has already been picked, is working pretty well for one plan (liab produced by the discount rate is within .25% of the liabilities using yield curve). However, other db plans that my client has, the same discount rate would not produce liabilities which are as close to the liabilties using yield curve. So, I am wondering if there is some tolerance level out there where if the difference in liabilities is within x%, it's acceptable. Does anybody have experience with this same issue?

[david rigby]

By solving for one discount rate (from the yield curve), I assume the plan sponsor will not have much flexibility of choosing their discount rate anymore?

IMHO, auditors are usually happy to see a well-defined technique. As long as the technique is valid, then the resulting rate is valid. Use of one technique should not preclude you from fine-tuning that technique in later years.

Also, a discount rate, that has already been picked, is working pretty well for one plan (liab produced by the discount rate is within .25% of the liabilities using yield curve). However, other db plans that my client has, the same discount rate would not produce liabilities which are as close to the liabilties using yield curve. So, I am wondering if there is some tolerance level out there where if the difference in liabilities is within x%, it's acceptable. Does anybody have experience with this same issue?

A rate is not "working well" based on how close the assets and liabilities are. The discount rate is determined without regard to the assets.

[Andy the Actuary]

Final question: How come the $%&@*!( auditors continue to challenge on discount rate but I've yet to have them question the expected long term rate of return assumption -- especially when the portfolio may have been temporarily allocated to fixed investments?