Case study: Using a modified mortality table in place of standard tables

  • Print
  • Connect
  • Email
  • Facebook
  • Twitter
  • LinkedIn
  • Google+
By Sheila Barrett | 16 April 2015

The challenge

The Society of Actuaries’ Retirement Plans Experience Committee (RPEC) recently presented findings of a five-year mortality study. The RPEC study recommended updated mortality tables for pension plans in the form of 11 gender-specific tables, RP-2014, and a new projection scale, MP-2014. The RP-2014 tables are split as follows:

  • Employee Tables (ages 18 through 80)
    • Total (all nondisabled data), Blue Collar, White Collar, Bottom Quartile (based on salary), Top Quartile (based on salary)
  • Healthy Annuitant Tables (ages 50 through 120)
    • Total (all nondisabled data), Blue Collar, White Collar, Bottom Quartile (based on benefit amount), Top Quartile (based on benefit amount)
  • Disabled Annuitant Table (ages 18 through 120)

The new mortality and projection tables are intended to replace the existing RP-2000 tables as well as generational projection scales AA and BB, effective immediately as of October 2014.

A mid-sized client asked us to determine whether it could justify using a modified mortality table in place of the standard tables produced by the Committee. The client believed that adopting the standard RP-2014 Total Healthy tables would erroneously increase the plan’s liability because the tables appeared significantly different from the plan’s current mortality assumptions.

The Milliman solution

The client sponsors a defined benefit plan with approximately 4,000 participants, half of which are active employees. The population is composed of both white collar participants and blue collar participants. We performed an experience study by applying statistical methods to a five-year data set from January 1, 2009, through January 1, 2014. We focused on the annuitant population and considered males and females both separately and combined.

We tested multiple scenarios comparing several of the recommended RP-2014 tables as well as basic modifications to the tables, including set-forwards and setbacks. These basic modifications slide the mortality table forward or backward by a specified number of years in order to better match participant demographics. For example, some population groups are expected to have shorter lifespans due to their line of work, so setting the mortality table forward could be appropriate in order to apply higher mortality rates at younger ages when compared to the unadjusted mortality table. As an initial step in our analysis, we created Chart 1 below, which shows a summary of the observed versus expected deaths of annuitants under various proposed tables. As shown in Chart 1, only the Total Healthy Annuitant table was considered. In order to justify the use of a collared or quartile table, a plan should have an underlying population that primarily aligns with the selected collar or quartile. Our client did not have a population mix that supported the use of either of the collared tables or the quartile tables. As such, our client focused on adjustments to the Total Healthy table.

Chart 1

Healthy Annuitants Total
Exposures 5,519
Observed Deaths 206
Expected Deaths Using Current Assumption
Expected/Observed
198
96%
RP-2014 Healthy Annuitant Table Total
Expected Deaths Using Base Table
Expected/Observed
177
86%
Expected Deaths Using Table Set Forward 1 year
Expected/Observed
196
95%
Expected Deaths Using Table Set Forward 2 years
Expected/Observed
217
105%
Expected Deaths Using Table Set Forward 3 years
Expected/Observed
240
117%

Although the Total Healthy Annuitant table with a one- year set-forward appears to be a good fit, we needed to provide the client with statistical support for making mortality assumptions based on the experience of a smaller plan. Credibility theory is an actuarial analytical tool used to test and justify whether a population group is large enough to warrant creating its own mortality table based solely on plan experience (American Academy of Actuaries’ Pension Committee, 2006). We used credibility theory to determine that the plan was not large enough to create its own mortality table or even to rely solely on plan experience to warrant selecting a standard table with a set-forward. However, there was sufficient evidence to validate using a modified mortality table. We were able to assign partial credibility to the plan’s experience as opposed to the full credibility afforded large plans. Thus, we generated mortality tables by weighting the plan’s mortality experience with the standard RP-2014 Total Healthy Annuitant mortality tables.

Our modified tables applied loads to the RP-2014 Total Healthy Annuitant mortality tables based on a ratio of actual to expected deaths and partial credibility. The load modified the tables to more closely align with the plan’s experience. We presented our results to the client and the client’s auditor, and both parties assessed the observed versus expected ratios in Chart 1 and the development of modified tables based on credibility theory. Although the client’s current mortality assumption using the RP-2000 tables with Generational Projection using Scale AA appeared to be a “good fit” with an actual to expected death ratio of 96%, the client’s auditor was requiring a change to the new mortality tables recommended by RPEC. As an additional piece of information, we included Chart 2 below, which shows the impact on the plan’s accounting liability given various mortality tables.

Chart 2

Scenario Mortality Tables Expected/Observed Deaths Percent Increase in Accounting Liability (from Base) at 5.20% Discount Rate
1 RP-2000 with Generational Projection using Scale AA (Current Assumption) 96% N/A
2 RP-2014 Total Healthy Annuitant
Males loaded 6%
Females loaded 12%
93% 2.9%
3 RP-2014 Total Healthy Annuitant
Males – no set forward
Females – no set forward
86% 4.4%

Note: For each RP-2014 mortality scenario, the tables shown above were used for the healthy annuitants only. For non-annuitants, the RP-2014 Total Employee tables, male and female, with no set forwards, were used.

The outcome

As a result of our experience study, the client decided to adopt a modified mortality table supported by the limited fluctuation credibility method. The modified table applied a 6% load for males and a 12% load for females to the gender-specific RP-2014 Total Healthy Annuitant and Non-annuitant tables. The modified tables are a good fit with the plan’s experience. Partial credibility allowed the plan sponsor to rely in part on the observed deaths in the plan despite not having a population large enough for full credibility. The client also elected to adopt the MP-2014 projection scale adjusted with a 10-year convergence period and a 0.75% ultimate rate. The projection scale is applied to both the annuitant and non-annuitant mortality tables as a way of building in a margin for future mortality improvement. The modifications to the projection scale are closer to the assumptions for future mortality improvement used by the Social Security Administration in its projections, and the modifications were approved by the client’s auditor. Additional support identifying the factors behind mortality improvement assumptions can be found in the actuarial study, “Life Tables for the United States Social Security Area 1900-2100,” published in 2005 by Felicitie C. Bell and Michael L. Miller.

As a result of our experience study, we were able to statistically support the client’s use of a modified mortality table. The modified table fit more closely with the plan’s experience over the past five years, and also decreased the liability by an estimated 1.5% when compared to the plan’s liability using the unmodified RP-2014 Total Healthy Annuitant table. Although our analysis was not liability-driven, our client was pleased that we were able to provide support for a modified table that fit tightly with plan experience and avoided an unnecessary increase in plan liability. Had the client adopted an unmodified Healthy Annuitant table, this would have overstated the plan liability, which would have a negative impact on the balance sheet and plan expense for years to come. Thus, we were able to use our expertise to achieve positive outcomes for our client while complying with the auditing and actuarial standards inherent in our profession.

Reference

American Academy of Actuaries’ Pension Committee. (Nov. 2006). Substitute mortality-credibility. Retrieved from http://www.actuary.org/pdf/pension/mortality_nov06.pdf.

Bell, Felicitie C., & Miller, Michael L. (Aug. 2005). Life tables for the United States Social Security Area 1900-2100. Retrieved from http://www.ssa.gov/oact/NOTES/as120/LifeTables_Body.html.