Actuarial group raises questions about proposed RP-2014 mortality tables
In March we posted an article reviewing the Retirement Plans Experience Committee (RPEC) of the Society of Actuaries “Exposure Draft RP-2014 Mortality Tables,” recommending new mortality tables for the valuation of defined benefit plan liabilities. The new tables significantly increase life expectancy assumptions and, if adopted by regulators and DB plan actuaries, will significantly increase DB plan liability valuations for purposes of disclosure, funding and de-risking.
On May 29, 2014, the American Academy of Actuaries sent the Society a letter raising issues with respect to the Exposure Draft concerning (1) the data used and (2) the projection methodology adopted. Given the financial significance – in terms of, e.g., earnings impact, higher contribution requirements and even (conceivably) higher PBGC premiums – the validity of the proposed RP-2014 tables is of more than just technical concern.
In this article we review the issues raised by the Academy. We realize that, in addition to being very technical, this discussion may be confusing. To keep things straight: the Society of Actuaries (SOA) is advocating the use of mortality tables that recognize significantly increased life expectancy. The Academy of Actuaries (the Academy) is raising questions about that and, by implication, questioning whether the proposed tables might not overstate life expectancy.
Summarizing, the RP-2014 tables in the SOA Exposure Draft reflect significant increases in mortality improvement relative to the current regulatory regime (RP-2000 plus Mortality Improvement Scale AA). The conclusion of the SOA is that the current rules significantly understate how long DB plan participants will live because they do not reflect changes (improvements) in life expectancy that have happened since Scale AA was published and that are expected to happen in the future.
The Exposure Draft includes a table showing the increase in liability the adoption of the new table would result in – that is, more or less, an estimate of ‘bottom line’ impact.
Percentage Change of Moving to RP-2014 (with MP-2014) from RP-2000 with generational Mortality Improvement Projection Scale AA (6% interest rate)
Thus, depending on, e.g., a plan’s current funded status, adoption of the RP-2014 tables could result in a significant increase in a plan’s liability valuation and the plan sponsor’s minimum funding obligation.
Concerns raised by the Academy
The Academy, in its May 29, 2014 letter, primarily raised issues concerning the construction of the dataset used to derive the new tables. It also raised an issue with respect to the mortality improvement rate used in the Exposure Draft.
Issues with data
What follows is a summary of the data issues raised by the Academy. (These issues are, inevitably, very technical. We are reviewing them only to give sponsors a feel for their seriousness and significance.)
The Academy’s most significant concern was the elimination of 70% of the data originally gathered. Quoting the Academy letter: “The reasons given by the RPEC for eliminating roughly 70% of the data may result in the data that was retained being biased toward lower rates of mortality and may give the impression to some readers that blocks of data were eliminated because they did not fit with anticipated experience.” [emphasis added.] Connecting the dots – ‘lower rates of mortality’ translates directly to increased life expectancy and indirectly to higher contributions.
With regard to the data the SOA used to prepare the new tables, the Academy questioned:
The exclusion of a significant amount of data as ‘outliers.’ The Academy requested more information about the data eliminated because, if most of the data eliminated for this reason reflected higher mortality, that fact might indicate an issue with the remaining data. Moreover, the Academy questioned the parameters used (“not within 10% of what was expected at a 95% confidence level”) to identify ‘outlier’ data. In this regard the Academy stated: “We are especially concerned that excluding all data from plans with ‘outlier’ ratios of actual to expected mortality may not take into account the inherent variability in mortality experience.”
The limited number of plans in the dataset. The SOA based its new tables on data from only 38 plans “with two-thirds of the retiree experience (the most important experience by far) coming from five plans. This is a small base of plans, which could skew the results towards specific industries ….”
The exclusion of public plans. The SOA eliminated “two large public employee plans on the basis that they are not representative of private employer plans because they showed higher mortality.”
The SOA’s data consolidation methodology. The SOA used “a process in which records from different years for the same participant were linked together to produce a consolidated record that could be followed through the entire exposure period, and if this linkage could not occur the data wasn’t used. We believe that a linkage is more likely to fail, with the data then excluded, in cases where a death has occurred, biasing the data retained.”
The SOA’s treatment of “inconsistent” data. The SOA excluded data “if there was ‘missing, incomplete or inconsistent’ information. One of the examples cited of ‘inconsistent’ information was when the plan used the same identifier for the participant and the resulting surviving spouse/beneficiary, and the ‘inconsistency’ was that the gender code changed. This would appear to introduce a bias toward disregarding data where deaths occurred.”
The inclusion of beneficiaries with healthy retirees. The Academy explained that a couple that chooses, e.g., a joint and survivor benefit is likely to be more healthy/live longer.
The exclusion of data from the Pension Benefit Guaranty Corporation (PBGC). “[T]he PBGC has commented that it routinely performs its own mortality studies and found the RP-2000 table, projected with scale AA and setback 1 year, to be the best fit with PBGC experience. This is a significant amount of data which could materially impact the results.”
We realize that a lot of this is hard to follow. The Academy’s point is that in a number of instances – those identified above – the SOA made choices that might bias the data towards longer life expectancies.
Mortality improvement assumption
The RP-2014 Exposure Draft included a mortality improvement rate which it indicated should be used for all plans. This rate is, in effect, a guess (or, more generously, an estimate of future experience) about how life expectancies will increase in the future. The problem with this proposal, according to the Academy, is not so much that the guess is wrong but that the SOA is taking the position that all plans must adopt it. Quoting the Academy:
Although the 1% ultimate improvement rate and the 20-year phase-in used to develop MP-2014 may be reasonable, they are speculative and other assumptions might also be reasonable. Therefore, we suggest that the language in the Exposure Drafts be softened and made more flexible (perhaps with boundaries that the RPEC deems to be reasonable) and that a tool be made available that would allow for alternative assumptions for the ultimate improvement rates over various projection periods.
Those who have problems with the SOA’s RP-2014 proposal – and that group certainly includes a number of plan sponsors – have more than one bite at this apple. The proposal is in exposure draft form. The SOA is likely to take the Academy’s challenges seriously, and it’s conceivable that the SOA may in fact change the final tables and recommendations. Even if the SOA adopts the Exposure Draft without change, however, the tables will have to be adopted/implemented by regulators – the Financial Accounting Standards Board and the Securities Exchange Commission (with respect to disclosure) and the Internal Revenue Service (with respect to minimum funding). That process may give objectors an opportunity to re-litigate issues with RP-2014, including the issues raised by the Academy in its letter.
* * *
We will continue to follow this issue.