SOA and Ruark: predictive analytics video and case study

I appreciate the SOA including me in their case studies of actuaries embracing predictive analytics, and I must give credit to our great team here at Ruark Consulting, our collaborators at the Goldenson Center for Actuarial Research at the University of Connecticut, and our clients.  Here is the video and article.

Those looking for some additional info can find it my article "When is Your Own Data Not Enough?" from The Actuary magazine.

 

 


Retirement Income Journal - Coverage of our Mortality Studies

https://retirementincomejournal.com/article/mortality-much-lower-in-annuities-with-living-benefits-ruark/


Ruark Consulting Releases 2018 Fixed Indexed Annuity Mortality Study

Mortality rates vary by living benefit presence & utilization

Ruark Consulting, LLC today released the results of its 2018 study of fixed indexed annuity (FIA) industry mortality. The study was based on experience from 3 million policyholders spanning the period January, 2007 through September, 2017. Fourteeen variable annuity writers participated in the study, comprising $215 billion in account value as of September, 2017.

“We have almost 50% more data than our last FIA mortality study, allowing for high credibility even when splitting results by multiple factors of influence,” said Timothy Paris, Ruark’s CEO. “We’ve also added much more detailed analysis of mortality results by living benefit presence and income behavior, contract size, tax status, interactions, and changes through time. These studies provide new and important insights into FIA mortality, particularly with the growth of living benefit experience beyond the surrender charge period.”

The company’s previous FIA mortality study was released in 2016.

Paris highlighted study enhancements in response to recent regulatory proposals. “It’s not often that life and annuity actuaries need to address new mortality and projection bases for reserves and capital,” he noted, “but now is indeed the time for that. So we’ve included analyses of industry mortality results relative to the 2012 IAM Basic Table with projection scale G2, our proprietary Ruark variable annuity mortality table, and other tabular bases to make the study results as meaningful and actionable as possible for our client companies and their actuaries.”

Study highlights include:

  • After normalizing for age, sex, and duration, Ruark observes a distinct hierarchy in mortality across living benefit presence and utilization. Highest mortality is found on contracts without guaranteed lifetime income benefits (GLIB); those with a GLIB rider that have not begun taking income have mortality at slightly below average; and GLIB contracts that are in the income phase have the lowest mortality, at 88% of average. This hierarchy is consistent with a pattern of selection on the basis of longevity benefits. Ruark also observes a difference in mortality on the basis of tax status.

  • In this study for the first time, Ruark benchmarks results against the 2012 Individual Annuity Mortality (IAM) Basic table with projection scale G2. Ruark also benchmarks results against other standard mortality tables and their proprietary 2015 Ruark Variable Annuity Mortality (RVAM) table. Standard industry mortality tables systematically overstate or understate various age-sex cohorts, even when they closely approximate aggregate mortality. In contrast, the 2015 Ruark VAM table better reflects FIA mortality both in aggregate and by age-sex cells.

  • Ruark's estimate of aggregate FIA mortality has remained stable since 2016. In contrast, VA mortality has fallen approximately 3% since 2015, driven by mortality improvement and changes in the business mix. In the case of FIA, these downward trends are offset by the effects of improved company-by-company data processing in this study, particularly with regard to spousal continuation following the death of the original policyholder.

  • FIA mortality exhibits a select-and-ultimate pattern even in the absence of individual underwriting. Mortality in the first year is 75% of average in the first contract year, jumps 20 percentage points in the second contract year, and then grows by approximately 1.6 percentage points each year thereafter. This phenomenon is consistent with the intuition that FIA buyers might be expected to be somewhat healthier in order to enter into a financial transaction that offers limited death benefits, and various forms of longevity benefits.

  • There is evidence of mortality improvement among FIA policyholders, and the extent of improvement appears to vary depending on whether or not the contract includes a GLIB rider. Contracts with a GLIB rider exhibit improvement consistent with projection scale G2; those without a rider exhibit greater annual improvement.

 

Detailed study results, including company-level analytics and customized behavioral assumption models calibrated to the study data, are available for purchase by participating companies.

For further information on results, to purchase the study, or if you have any other inquiries, click here or email Timothy Paris (timothyparis@ruark.co).


Ruark Consulting Releases Variable Annuity Mortality Study Results

Mortality rates fall 3%, driven by mortality improvement and business mix

Ruark Consulting, LLC today released the results of its 2018 study of variable annuity (VA) industry mortality. The study was based on experience from 13.3 million policyholders spanning the period January, 2008 through December, 2017. Twenty-four variable annuity writers participated in the study, comprising $880 billion in account value as of December, 2017.

“We have 60% more data than our last VA mortality study, allowing for high credibility even when splitting results by multiple factors of influence,” said Timothy Paris, Ruark’s CEO. “We’ve also added much more detailed analysis of mortality results by living benefit and death benefit types, partial withdrawal and income behavior, contract size, tax status, interactions, and changes through time. These studies provide new and important insights into VA mortality, particularly with the seasoning of living benefit blocks beyond the end of surrender charge and bonus periods.”

The company’s previous VA mortality study was released in 2015.

Paris highlighted study enhancements in response to recent regulatory proposals. “It’s not often that life and annuity actuaries need to address new mortality and projection bases for reserves and capital,” he noted, “but now is indeed the time for that. So we’ve included analyses of industry mortality results relative to the 2012 IAM Basic Table with projection scale G2, our proprietary Ruark VA mortality table, and other tabular bases to make the study results as meaningful and actionable as possible for our client companies and their actuaries.”

Study highlights include:

  • Aggregate variable annuity mortality has declined 3% since 2015. The decline is largely attributable to two factors: mortality improvement and changes in the business mix. Contracts with living benefits exhibit lower mortality than average, and these contracts make up an increasing proportion of overall exposure. Whereas 44% of 2008 account value was on contracts with no living benefit, and 43% on contracts with guaranteed lifetime withdrawal benefit (GLWB) or guaranteed minimum income benefit (GMIB) riders, by 2014 the proportions were 25% and 69%, respectively. The proportion of exposure has remained roughly the same in subsequent years.

  • In this study for the first time, Ruark benchmarked VA mortality against the 2012 Individual Annuity Mortality (IAM) Basic table with projection scale G2. The 2012 IAM Basic with G2 projection is under consideration as the standard for calculating insurers' reserve and capital requirements for variable annuities. Ruark also benchmarked VA mortality against its proprietary 2015 Ruark Variable Annuity Mortality (RVAM) table, which was developed from the results of the prior study. Standard industry mortality tables systematically overstate or understate various age-sex cohorts, even when they closely approximate aggregate mortality. In contrast to the 2012 IAM table, which departs as much as 18% from experience at central ages and over 45% for younger female cohorts, the 2015 Ruark VAM table diverges by less than half those amounts.

  • Variable annuity mortality continues to evolve through time. Changes in the business mix, and mortality improvement since the prior study, have led to changes across age-sex cohorts as well as a decline in the absolute level of VA mortality.
  • A distinct hierarchy in mortality across living benefit types is apparent. Lowest mortality is found on GLWB and GMIB riders. Guaranteed minimum withdrawal benefit (GMWB) and guaranteed minimum accumulation benefit (GMAB) riders fall in a middle group, with average mortality. Those contracts without living benefits exhibit above-average mortality. This hierarchy is consistent with a pattern of policyholder selection on the basis of longevity benefits. A GLWB rider allows the policyholder to make regular withdrawals, up to a specified amount, until death – even after exhausting their account value. This benefit is most valuable to policyholders with an expectation of favorable longevity. Above-average mortality on contracts without living benefits is consistent with the intuition that less-healthy VA buyers would rationally forego a financial transaction that features such benefits -- and instead purchase a product whose main feature is a death benefit. Analysis by contract duration indicates that the selection effects wear off over time, similar to traditional life insurance products.

  • The extent to which mortality differs by living benefit type suggests that utilization of living benefits might reveal selective pressures as well, and Ruark observes that withdrawal benefit mortality differs considerably by withdrawal history. GLWB contracts that have taken no withdrawals have low mortality; those that have taken regular income withdrawals have intermediate mortality; and those that have taken excess withdrawals exhibit above-average mortality. Ruark’s conjecture is that these differences reflect policyholders’ different income and liquidity needs in response to conditions that correlate with higher mortality.

  • The effects of policy size are apparent only when results are segregated by living benefit type. Among more generous living benefit types, mortality declines with increased account value, with a difference of 7 percentage points between the smallest and largest contracts. Among older living benefit forms (GMWB and GMAB), the relationship is flat. And among contracts with no living benefits, mortality increases with account value, by 15 percentage points from the smallest to the largest contract size groups. Ruark believes that more affluent policyholders are generally more financially savvy and have greater access to expert financial advice, which translates to more savvy purchase and benefit retention behavior.

Detailed study results, including company-level analytics and customized behavioral assumption models calibrated to the study data, are available for purchase by participating companies.

For further information on results, to purchase the study, or if you have any other inquiries, click here or email Timothy Paris (timothyparis@ruark.co).


Registration is open! 2018 SOA Life and Annuity Reinsurance Seminar

If you're attending the ValAct, stay an extra day for this. If not, easy trip for anyone on the east coast.

I'm looking forward to hosting on behalf of the Reinsurance Section Council, and speaking on the topics of policyholder behavior and reinsurance.

 


When Is Your Own Data Not Enough?

Published in The Actuary magazine of the SOA, June/July 2018.

How using external data can strengthen results.

Related notes:

  • Our spring 2018 VA and FIA industry experience studies included over $1.1 trillion in current account values.  Highlights and media coverage.
  • Triennial VA and FIA mortality studies are on the way this summer.
  • We continue to be very busy with client work, developing customized predictive models that balance relevant industry-level data with company- and product-level data, with quantifiable improvement in goodness-of-fit and predictive power relative to company-only data.  Like hedging and reinsurance, this is an investment in risk management, not an expense.
  • Details to come of our plans for 2019, including expansion to other products such as traditional fixed deferred and payout annuities.
  • Contact us with feedback or to learn more about participation or purchase.

 

 


Variable annuity deferral bonuses are working

Check out the article the Retirement Income Journal published regarding our Spring 2018 variable annuity (VA) policyholder behavior study.  Go to the article.

 

All product and company names are trademarks™ or registered® trademarks of their respective holders.


Spring 2018 Variable Annuity Study Results

Insurer Incentives Delay Lifetime Benefit Commencement

SIMSBURY, CT, May 29, 2018 – Ruark Consulting, LLC today released the results of its spring 2018 studies of variable annuity (VA) policyholder behavior. The studies, which examine the factors driving surrender behavior, partial withdrawals, and annuitization, were based on experience from 13.9 million policyholders spanning the period January, 2008 through December, 2017. Twenty-five variable annuity writers participated in the study, comprising $948 billion in account value as of December, 2017.

“A variable annuity writer’s cost to provide guarantees depends on policyholder behavior, including surrender and income utilization,” said Timothy Paris, Ruark’s CEO. “Each company should ask itself the basic question – is my own data enough?”

Study highlights include:

  • As the market for guaranteed lifetime withdrawal benefit (GLWB) riders matures, it is possible to see the effects of long-dated insurer incentives to delay benefit commencement. Commencement rates are low overall, 12% in the first policy year and falling to 6-7% annually thereafter. However, usage jumps over 5 points in year 11, with the expiration of ten-year bonuses for deferring withdrawals common on many riders. In this study, we see for the first time that commencement frequency thereafter falls to an ultimate rate of about 9%. The deferral bonuses appear to have the intended effect of delaying benefit utilization. Among contracts that take a withdrawal, nearly 90% continue withdrawals in subsequent years.

  • Overall living benefit annual withdrawal frequency rates have continued to increase, and utilization has grown more efficient. Withdrawal frequency for guaranteed lifetime withdrawal benefit riders is now 25%, up nearly two points over the rate reported in Ruark Consulting’s spring 2017 study and three points over the spring 2016 value. GLWB withdrawal frequencies have increased consistently at normal retirement age and above. Most of the increase is attributable to more efficient utilization of the rider benefit, with over half of withdrawals now at or near the maximum benefit amount.

  • The effects of moneyness (account value relative to the guarantee base) on partial withdrawal behavior differ depending on circumstances. When contracts with lifetime withdrawal benefits are at- or in the money, policyholders increase the frequency of standard benefit withdrawals. This is consistent with greater benefit exercise when the benefit is more valuable. The effect is more pronounced after the expiration of deferral incentives. In contrast, when contracts move out of the money, withdrawals in excess of the maximum amount are more common. This is suggestive of policyholders taking investment gains out of the contract.

  • Overall variable annuity surrender rates in 2017 have returned upward to post-crisis levels, following a secular dip in 2016. We see three regimes in the study window: Surrenders at the shock duration (the year following the end of the surrender charge period) were nearly 30% at the onset of the 2008 economic crisis; shock rates below 10% were observed during 2016; and otherwise a post-crisis regime has prevailed, with shock rates in a range of 12-16% from 2009 through mid-2015 and 13% in 2017. The 2016 dip, first observed in Ruark’s fall 2017 study, is only partially attributable to benefits moving more in-the-money during the year; it is likely that uncertainty surrounding the DOL’s proposed Fiduciary Rule and political factors encouraged a “wait-and-see” attitude among many policyholders and advisors.

  • The presence of a living benefit rider has a notable effect on surrender rates; contracts with a lifetime benefit rider have much higher persistency than those with other types of guarantees. Also, a contract’s prior partial withdrawal history influences its persistency. Contracts with a lifetime benefit rider that have taken withdrawals in excess of the rider’s annual maximum have surrender rates three points higher overall than other contracts with those riders.  In contrast, those who have taken withdrawals no more than the rider’s maximum have the lowest surrender rates (three points lower at the shock, for example, compared to contracts who have taken no withdrawals).
  • When calculating relative value (moneyness) for lifetime withdrawal guarantees, use of a nominal measure can be deceiving. A nominal measure fails to reflect important aspects of living benefit design, and can be inflated over time as the benefit base remains unchanged even as the account value is reduced through regular withdrawals. It may be preferable in many cases to use an actuarial measure of moneyness that incorporates interest and mortality rates. Actuarial moneyness exhibits a similar dynamic effect on lapse though with notable differences in shape. As an actuarial basis pushes a contract toward out-of-the-money by discounting the guarantee, very little exposure exists for heavily ITM contracts: nominally, 77% of GLWB exposure is in-the-money, while only 10% is in-of-the-money when measured on an actuarial scale.
  • Annuitization rates on policies with guaranteed minimum income benefit (GMIB) riders continue to decline. The overall exercise rate for the riders with a 10-year waiting period is 2.1% by account value. Rates have been falling steadily since 2010, and quarterly observed rates have stayed at or below 2% since 2014. “Hybrid” rider forms that allow partial dollar-for-dollar withdrawals have much lower exercise rates than tradition forms, which reduce the benefit in a pro-rata fashion – less than 1% for hybrid, vs. 5% for pro-rata; the increasing share of exposure in the study from the hybrid type is a partial explanation for the decrease in annuitization rates over time.

Detailed study results, including company-level analytics and customized behavioral assumption models calibrated to the study data, are available for purchase by participating companies.

For further information on results, to purchase the study, or if you have any other inquiries, click here or email Timothy Paris (timothyparis@ruark.co).