Ratings

Stochastic modeling is the preferred method by which Standard & Poor's would like to reflect the living and death benefit risks in its capital model. The NAIC requires companies to rank-order their stochastic results and average the worst 10% of the results—referred to as the conditional tail expectation (CTE(90)) —as part of its approach. Standard & Poor's uses the same data to determine the CTE(90) with two key differences:

• We need various CTE results, which we correlate to rating-based target capital levels.
• We need the results with and without the benefit of hedging, as we only provide partial credit.

Standard & Poor's survey requests the total assets required (TAR)—with and without hedging—at various CTEs and the associated statutory reserves. The capital required is based on the difference of the TAR at various CTE levels minus the reserves held, and it allows 50% credit for the value of hedging. Our enhanced capital model (see "New Risk-Based Insurance Capital Model", May 23, 2007) calculates minimum capital to be held at various target ratings by stressing risks to various confidence intervals. Similarly, capital needs for variable annuities under the enhanced model are based on CTE results. As a result, CTE(90), CTE(95), CTE(97), and CTE(99) correlate with 'BBB', 'A', 'AA', and 'AAA' capital requirements, respectively. The old capital model used the CTE(90) results.

As a simple example, if a 'AA' rated company's variable annuity capital at CTE(97) was $100 million without hedging and $50 million with hedging, Standard & Poor's would require $75 million to be held in capital. (The calculated result will have a floor risk charge of zero and not allow a capital credit to be generated.) The initial credit associated with hedging programs might be adjusted as Standard & Poor's gains comfort with each insurer's specific hedge programs and broader ERM.

The CTE data should be based on the American Academy of Actuaries' prepackaged scenarios to ensure reasonably comparable results. For hedging credit, companies must use clearly defined hedging strategies. It is understood that insurers will be incorporating statistical compression techniques to reduce the tremendous burden of running 10,000 scenarios. However, insurers modeling fewer than 1,000 stochastic scenarios need to provide Standard & Poor's an explanation to support the robustness of their modeling. The number of indices used in the replication of mutual funds will also be a qualitative consideration. It is also conceivable that Standard & Poor's will allow scenarios other than the prepackaged versions, but for that to happen, a deeper, company-specific review will be required.

Another divergence from regulatory adoption is that an adjustment will be made to reflect that reserves held for variable annuity benefits—in addition to cash surrender values—are not usually available to a company in a stress situation.

As mentioned, we've found getting stochastic company results from our C-3 Phase II survey to be surprisingly challenging. Therefore, Standard & Poor's has revised its static-risk capital charges for variable annuities with living and death benefit riders. Typically the revised static risk charges are more onerous than the stochastic ones.

The static charges apply to U.S. insurance companies where stochastic results cannot reasonably be applied because of either errors or questionable standards (fewer than 1,000 scenarios, etc.). The revised static living and death benefit risk charges are now the default capital charge for living and death benefit risk. However, we will review each insurer's C-3 Phase II survey submission and will replace the new default static risk charges if we consider the stochastic survey results to be reasonable. The new static charges will apply to both the old and enhanced capital models, which we ran concurrently in 2007 for our evaluation of risk-adjusted capital for year-end 2006.

The revised static capital charges for variable annuities with living and death benefit riders are based on information available through the requirements associated with NAIC C-3 Phase II and related actuarial analysis. Deterministic scenarios have been generated to represent the various at-the-money risks associated with guaranteed minimum benefits such as with associated accumulation, income, and withdrawal riders. As with other factors in the enhanced capital model, the risk charges have been calibrated with confidence intervals associated with target rating levels ('AAA', 'AA', etc.). The factors associated with 'BBB' were applied under the old model (see table).

We understand that these new static factors will not represent every company's risk given different product designs and varying market conditions. However, Standard & Poor's believes these new estimates reflect typical risks. We will not generally make company-specific exceptions to these revised static capital charges, as we believe the best way to capture company-specific risk is through the required stochastic modeling that each insurer performs for its liabilities.

The revised static risk charges were developed following a review of a series of more than 150 stress tests. The stress tests were applied to a static model of variable annuity benefits and considered a range of specific product designs, annuitant ages, option exercise levels, mortality levels, and annual withdrawal/surrender levels. For these stress tests, we assumed business to be hedged with a program that was 75% effective, meaning that when a loss event occurred, a hedge gain was assumed to offset 75% of the loss cost (50% for the more complex guaranteed minimum income benefit) of the annuity excess benefit costs. We gave the hedging offset a haircut of 50%, which is consistent with criteria for the stochastic-based program. Companies without hedge programs will not receive the hedge offset.

The stress tests were based on the same scenarios that are the basis for the asset-related charges in the revised capital model for equities. Similar scenarios were developed for interest rate and corporate bond fund experience. Policyholder funds were assumed to be invested in a stock/bond mix with no rebalancing.

Clearly, these tests do not accurately represent any specific company's book of business. However, by using these tests, Standard & Poor's has estimated a range of relative risk levels for the various benefit types that are reflected in the revised static charges by benefit type.

In 2007, we modified the annual insurance survey we use to support our insurance capital model with input from both clients and leading actuarial consulting firms. We have requested total variable annuity account values, required reserves, TAR under C-3 Phase II, and the New York Insurance Department-based deterministic standard scenario. In addition, we request similar information by living and death benefit rider. The revised static risk charges will depend on the account values by rider. It is understood that the stochastic modeling is done on an aggregated basis and not by rider. However, Standard & Poor's is requesting estimates by rider to better understand the risk distribution. The number of scenarios modeled will also be requested and will contribute to the consideration of incorporating stochastic risk results from C-3 Phase II in lieu of the revised static variable annuity risk charges or the deterministic standard scenario.

Standard & Poor's believes the best way to reflect the risk of variable annuity living benefits in our capital model is through the NAIC's required stochastic C-3 Phase II results. However, when stochastic survey submissions are questionable, new default static charges will apply. The table shows 'BBB' risk factors under the enhanced model and by rating category under the revised capital model.

Revised Static Capital Charges For Variable Annuities With Living And Death Benefit Riders
(%)	AAA	AA	A	BBB
Return of premium death benefits	0.77	0.56	0.42	0.18
Death benefits enhanced*	3.99	3.37	2.85	1.61
Withdrawal benefits	5.52	3.46	2.73	1.37
Accumulation benefits	2.29	1.66	1.24	0.52
Income benefits	2.67	2.11	1.62	0.71
Others	3.05	2.23	1.77	0.88
*Roll-up or ratchet.