The NAIC’s Capital Markets Bureau monitors developments in the capital markets globally and analyzes their potential impact on the investment portfolios of US insurance companies. A list of archived Capital Markets Bureau Special Reports is available via the index
Modeling of U.S. Insurance Industry’s Holdings in Residential Mortgage-Backed Securities
With year-end 2011, the NAIC successfully completed the modeling of non-agency residential mortgage-backed securities (RMBS) held by the U.S. insurance industry. As was the case for year-end 2009 and 2010, to the extent data was available, each individual holding was modeled for expected losses using five different economic scenarios. The weighted average of those expected losses, assuming the bonds were held to maturity, were then translated into an expected recovery value. In lieu of using rating agency ratings issued by the nationally recognized statistical rating organizations (NRSROs), U.S. insurance companies were required to compare their carrying values as of year-end to the expected recovery value, to determine an NAIC designation, which was then mapped to a risk-based capital (RBC) factor. A small percentage of non-agency RMBS were not modeled for different reasons. These non-modeled securities included interest-only strips, foreign transactions and some highly complex resecuritizations. For these, U.S. insurers continued to rely on NRSRO ratings, but factored in carrying values in comparison to a fixed matrix of values.
As the NAIC Capital Markets Bureau has done in the past, the purpose of this Special Report is to consider the results of the modeling and the impact on the U.S. insurance industry’s RBC in comparison with the prior methodology that relied on NRSRO ratings, but which also assumed holdings were at par. In addition, this year, because we have the benefit of three years’ worth of results, we will compare the results over time and consider any patterns or trends over that time period. In addition to the Special Report dated April 29, 2011, “The Insurance Industry’s Investments in Residential Mortgage-backed Securities,” we also note a more recent piece published April 11, 2012, “Potential for Volatility in U.S. Insurer Holdings of Residential Mortgage-Backed Securities.”
2011 Modeling Results
As was the case for year-end 2009 and 2010, at the end of 2011, after a public exposure and comment period, the Valuation of Securities (E) Task Force formally adopted assumptions to be used in the modeling of the U.S. insurance industry’s non-agency RMBS holdings in November 2011. Those assumptions were as follows:
For year-end 2011, results for 18,459 unique CUSIPs were sent to U.S. insurers. The expected recovery values for these securities were used to determine the NAIC designation for a total exposure of $123.2 billion in book/adjusted carrying value (BACV) across the entire industry. This was divided between life and fraternal ($106.8 billion) and property, health and title ($16.4 billion). The total exposure is also compared with $150.5 billion at the end of 2009 and $127.7 billion at the end of 2010. While there was some acquisition activity during 2011, the U.S. insurance industry’s exposure to non-agency RMBS continued to decline, albeit modestly, for three primary reasons: 1) the lack of any significant new issuance in the marketplace; 2) amortizations on existing holdings; and 3) additional impairments taken during the year. Other-than-temporary impairments (OTTI) and fair value revaluations taken during the year totaled $2.9 billion in 2011. This is compared with $2.8 billion in 2010 and $15 billion in 2009.
Based on the results in expected recovery values and the industry’s year-end BACV prices, the breakdown by NAIC designation is as follows:
As was the case the prior two years, factoring in the insurer’s BACV price into the determination made a fairly substantial difference in the profile of NAIC designations. Ignoring the different carrying values and relying solely on NRSRO ratings would have yielded this breakdown:
As one example of where taking into account BACV prices is demonstrated to be a reasonable approach, compare the results for the NAIC 6 category in Table 2 vs. Table 3. In Table 2, the modeling results indicate that insurers holding those bonds should reasonably expect to recover 16.69% of par value, whereas insurers are carrying those bonds at an average BACV price of 32.79%. Based on the modeling, the probability of losing half of their carrying value is high, justifying an NAIC 6. In Table 3, the average modeling results are 58.69% and the average BACV price is 54.34% for NAIC 6. That means that the modeling suggests a very high probability of recovering full value as far as the insurer’s current exposure. If that is the case, it does not seem reasonable to be treating those holdings as an NAIC 6. The NRSRO ratings, by their nature, assume that the bond is being held at par. On the other hand, the risk profile is very different when the bond is held at a substantial discount to par. With an average BACV price of 80.07%, the industry’s current exposure is clearly at a substantial discount overall. This results in 72.3% of the industry’s holdings being assigned an NAIC designation that is higher than would have been the case if the NAIC relied solely on NRSRO ratings. It is also worth noting that 25.4% of the industry’s holdings experienced no change in the NAIC designation. In addition, a small percentage, 2.3%, of exposure, were assigned an NAIC designation that was lower than would otherwise have been indicated by the NRSRO rating.
The broad results also translate into the narrower categories within each NAIC designation based on NRSRO ratings. For bonds that would have received an NAIC 1 designation based solely on the NRSRO rating, 94.7% also received the same designation using the modeling approach, but 5.3%, or $1.6 billion, in exposure were assigned a lower designation. In the case of bonds that rated in the BBB category by the NRSROs (or an NAIC 2 equivalent), 8.9% were also assigned an NAIC 2 under the modeling approach; 85% received an NAIC 1; and 6.1% (or $520 million) were assigned a lower designation, requiring a higher RBC factor.
Focusing on those bonds rated below investment grade by the NRSROs, the ratings indicate that, assuming the bonds are held at par, these are the highest risk of not recovering their full value. As was discussed extensively in 2009 when the NAIC adopted the new procedures, the size of the loss could vary from a full impairment to a relatively nominal loss in comparison to the bond’s par value. The total exposure for these bonds was $85.2 billion in BACV. Of that total, $2.5 billion were assigned the same designation based on the modeling, 96.1% received a higher designation and 0.9% received a lower designation.
Of significance in Table 4 above is that, for those bonds deemed speculative by the NRSROs and receiving a higher NAIC designation from the modeling, the BACV price of 71.32% is significantly lower than the expected recovery value of 77.48%. That positive differential is, on average, greater for the group receiving an NAIC designation equivalent to investment grade (10.24%) and, even more so, for the group receiving an NAIC 1 (11.72%). On the other hand, for the group of bonds in this subset that received the same or lower NAIC designation based on the modeling approach, the average BACV price was higher than the expected recovery value indicating a reasonable likelihood of loss of carrying value.
Risk-Based Capital Impact
The purpose of assigning NAIC designations, whether relying solely on NRSRO-assigned ratings or through a more finely tuned modeling approach, is to map each insurer’s holding to a RBC factor. Under the current NAIC process for assigning designations for non-agency RMBS, total RBC is lower than it would have been if the NAIC had continued to rely solely on NRSRO ratings. In aggregate, the BACV price for the industry’s exposure is 80.07%, which compares favorably to the average expected recovery value of 84.21%. Without accounting for impact of the covariance component of the RBC formula, the comparison using a breakdown of the industry’s exposure based on the modeling approach is detailed in Table 5 below:
The $87.8 billion in exposure assigned an NAIC 1 designation under the modeling approach has an average BACV price of 79.78%, slightly below the overall average of 80.07%. With the substantial discount to par and the favorable comparison to the modeled expected recovery value of 87.71%, there is a substantially lower RBC requirement. The reverse is true at the lower end of Table 5, where the modeled results drive a higher RBC requirement relative to the results than would have been the case if relying solely on NRSRO ratings. This relationship is reversed if the table is based on NRSRO rating equivalents (see Table 6).
The differential in RBC has widened between the two approaches since 2009 as an increasing percentage of the industry’s non-agency RMBS holdings have been downgraded by the NRSROs, while, at the same time, the industry’s valuations have drifted downward, reflecting more conservative valuations.
Comparison over Three Years
Overall, the assumptions adopted by the Valuation of Securities (E) Task Force have not varied substantially. An additional factor not depicted in Table 8 above is the timing of the expected trough. This would have the most significant impact in the Most Conservative scenarios, as a long wait before a recovery would be expected to result in a higher level of delinquencies over the time period. The troughs in each year were second quarter 2010, first quarter 2021 and third quarter 2022, respectively. The changes in assumptions had an impact on the results for the industry’s exposure from one year to the next. Focusing on the life industry’s exposure, the expected recovery value for the aggregate exposure was 83.70 in 2009, 84.03 in 2010 and 83.45 in 2011. These results were, however, likely skewed by acquisitions and dispositions during the period. It is possible to isolate the effect of the different assumptions by taking a simple average of expected recovery values for just those bonds that were modeled in each of the three years. There were 15,627 bonds that fit into that analysis. As of year-end 2009, these bonds had an average expected recovery value of 85.25. This dropped to 82.46 at year-end 2010 and to 79.20 at year-end 2011.
The decision to change the procedure for assigning NAIC designations for non-agency RMBS in 2009 was in part driven by substantial downgrade activity by the NRSROs. As of year-end 2008, 91% of the industry’s holdings carried an NAIC designation of 1 or 2, based on NRSRO ratings. That had dropped to 52% as of the second quarter of 2009. Conversely, only 2% received an NAIC 5 or NAIC 6 as of year-end 2008. This rose to 32% by mid-year 2009. This trend continued each year end from 2009 to 2011. By year-end 2011, the percentages were 31.3% for NAIC 1 and NAIC 2 as compared to 50.1% for NAIC 5 and NAIC 6.
Table 10 details the difference between expected recovery value and BACV price for each of the three years, with groupings based on NAIC designations as determined by relying solely on NRSRO ratings. Of particular note is that in 2010 and 2011, for NAIC designations 5 and 6, the industry’s holdings were held at values less than the modeled expected recovery values. Therefore, the modeling suggests that the probability of the insurers recovering their full value was high. In that case, a higher NAIC designation would seem appropriate, especially because the differentials for 2011 are comparable across all of the NAIC designation groups.
NAIC designations as determined by modeled expected recovery values in comparison with each insurer’s carrying value displays a very different profile. Investment grade issues, those assigned an NAIC 1 or NAIC 2, totaled 68.9% in 2009 and grew to 76.6% in 2011. Though still lower than the 91% that was the case at the end of 2008, this was significantly better than the profile that would exist if the NAIC continued to rely solely on NRSRO ratings for non-agency RMBS. The differentials between expected recovery value and carrying value for each successive NAIC designation also bear the expected relationship. For NAIC 1 in 2011, the expected recovery values are 7.82 points higher than the equivalent carrying values. It is a modestly negative relationship for the group of NAIC 2 bonds, and increasingly negative as one moves down the different designations.
As has been noted in previous Capital Markets Special Reports, the decision to change the process for assigning NAIC designations for non-agency RMBS resulted in a number of benefits. Most significant of these is a calibration of NAIC designations, and the RBC factors that the holdings are mapped to, to a greater level of sophistication that goes beyond simple credit risk. Continuing to rely solely on NRSRO ratings would have been particularly problematic for bonds carried at a substantial discount to par, as well as situations often cited where the potential loss of principal was small in comparison with the overall size of the holding. In addition, there were improvements in transparency and regulatory oversight of the process, as well as more accurate valuations by insurers. Assuming the NAIC continues to employ this approach, enhancements to the process may be considered. The C-1 Factor Review (E) Subgroup, which is a joint subgroup of the Valuation of Securities (E) Task Force and the Capital Adequacy (E) Task Force, is expecting to take under consideration the current framework as part of its 2012 charges.
Questions and comments are always welcome. Please contact the Capital Markets Bureau at CapitalMarkets@naic.org.
The views expressed in this publication do not necessarily represent the views of NAIC, its officers or members. NO WARRANTY IS MADE, EXPRESS OR IMPLIED, AS TO THE ACCURACY, TIMELINESS, COMPLETENESS, MERCHANTABILITY OR FITNESS FOR ANY PARTICULAR PURPOSE OF ANY OPINION OR INFORMATION GIVEN OR MADE IN THIS PUBLICATION.
© 1990 – 2018 National Association of Insurance Commissioners. All rights reserved.