Pensioner Longevity Data Analysis and Applications

Risk…Rewarded 1

Pensioner Longevity Data Analysis and Applications

Longevity 5 Conference | September 25, 2009

Risk…Rewarded 2

US Mortality Data and Analysis

Risk…Rewarded 3

Global Risk Services

About Hewitt Associates

Our BusinessFor almost 70 years, providing best-in-class HR Consulting and Outsourcing$3.2 billion in net revenue for fiscal 2008Located in 33 countries with approximately 23,000 associatesConsulting – Global– More than 3,000 large and mid-size companies around the world (34% revenue)Benefits Outsourcing – US-centric– Defined Benefit– Defined Contribution– Health & Welfare– More than 300 organizations worldwide (49% revenue)Multi-Process Human Resource Outsourcing – US-centric– More than 30 organizations (17% revenue)

Risk…Rewarded 4


The Implication

Focus on Data Available in US Defined Benefit Outsourcing BusinessHewitt administers pension benefits for over 100 clients– Generally large global organizations– Large over 20,000 employeesHewitt administers retirement benefits for about 15 million participants– Roughly breaks down as follows:

> 45% active employees, 40% retirees, 15% deferred vested

We are sitting on a uniquely large employee benefits database with informationParticular interest for longevity data because participants must be tracked ongoingDetailed demographic data (economic status: pay; geographic status: address)Information augmented by Social Security Death Index can create all data needed

Remainder of the discussion of our analysis will focus solely on US data

Risk…Rewarded 5


What We Have…

Clean Data on Most of the Lives in our DatabasesBetween 75-100 million life-years during 2000’s– Only 26 million life-years used in today’s analysis– Compare to previous US pensioner mortality studies

> RP-2000: 11 million life-yearsToday’s data from 31 Hewitt clients with a fairly broad industry mixFor all employees who have worked for employers while contracted with Hewitt– Indicative data: SSN, DOB, DOH, Status (active, disabled, terminated, retired)– Benefits data: pay, benefit level, total service, annuity option, etc.– Other data: city, state, ZIP, marital status, etc.

Risk…Rewarded 6


What We Have…(cont’d)

Clean Data on Most of the Lives in our DatabasesAnalysis today based on a subset looking at 4.9 million lives (subset of Hewitt data)– SSN, DOB, DOH, last-known ZIP (mailing checked to pensioners when they died)

> Allocating all life-years for each person based on last-knownResults are preliminary – may contain some noise from cashed-out or terminated non-vested

Venturing into “uncharted territory” in comparative analysisMost comparative longevity analyses have focused on Actual vs. ExpectedOurs is focused on a concurrent experience among separate subsets of data

Risk…Rewarded 7


Today’s Hypothesis

Consistent with past analyses of CDC and Census Bureau: Geography mattersNot only does geography matter, but that it matters down to ZIP code– Implicit that socio-economic status will provide high-correlation– More robust view of geographic dispersion than past analyses by state and metro-regionLooking at some of our outstanding issues: they don’t matter much– The primary problem is the accuracy of tracking participants that “leave”– We would expect any issues like this to persist consistently across ZIPIn the interest of credibility we focus on ZIP-3– A few words about the US postal system…

Risk…Rewarded 8


The US population distribution

Gradient ScaleNo Data AvailableLeast Data

Most Data

Map of ZIP-3s across the US by total population

Risk…Rewarded 9


Where Our Data is Most Robust


Most Data

Map of ZIP-3s across the US by total exposures (life-years)

Risk…Rewarded 10


How This Compares to the Actual US Population


Most Data

ZIP-3 relative Hewitt exposures vs US Pop Distribution

Risk…Rewarded 11


What Our Data Had to Say

Map of Relative Longevity in ZIP-3s across the US – white represents no deaths

Gradient ScaleNo Data AvailableShortest Lifespans

Longest Lifespans

Risk…Rewarded 12


What Our Data Had to Say (cont’d)

Top 10 ZIP-3’s (Greatest Longevity)508 – Creston, Iowa828 – Sheridan, Wyoming820 – Cheyenne, Wyoming692 – Valentine, Nebraska855 – Globe, Arizona595 – Havre, Montana999 – Ketchikan, Alaska514 – Carroll, Iowa573 – Central South Dakota328 – Orlando, Florida

Bottom 10 ZIP-3’s (Worst Longevity)481-482 – Detroit, Michigan636 – Cape Girardeau, Missouri469 – Kokomo, Indiana434-436 – Toledo, Ohio473 – Muncie, Indiana219 – Baltimore, Maryland630-631 – St. Louis, Missouri549 – Oshkosh, Wisconsin610-611 – Rockford, Illinois453-455 – Dayton, Ohio

Risk…Rewarded 13


The “Red Belt”

Quite Obviously, there is a substantive region from Texas up to Great LakesForms the collective of the Great Lakes Region, Texarkana, Deep SouthRepresents where there is generally lower average income according to US Census BureauAlso, generally a low availability of services (healthcare, infrastructure, etc.)Coasts and the Northern Plains provide meaningfully better longevity

Risk…Rewarded 14


A Brief Preview of Hewitt Data versus SOA RP-2000

More than 2.5 times the number of lives than RP-2000Less reliance on auto-maker dataGenerally consistent longevity curveMatches quite nicely, but begins to degrade towards age 60 0

0.002

0.004

0.006

0.008

30 40 50 60

Hewitt Q's RP-2000 Projected to 2005

Ages 30 to 60

Note: RP-2000 curve reflects smoothing techniques. None used on “Hewitt Q’s”.

Risk…Rewarded 15


A Brief Preview of Hewitt Data versus SOA RP-2000

Rates of death are 33% lower by early 70’sRP-2000 indicates mortality levels nearly twice that of Hewitt data by mid-80’s

0

0.05

0.1

0.15

60 70 80 90

Hewitt Q's RP-2000 Projected to 2005

Ages 60 to 90

Note: RP-2000 curve reflects smoothing techniques. None used on “Hewitt Q’s”.

Risk…Rewarded 16


Impact on the Aggregate S&P 500 Pension Costs

0

1000

2000

3000

US GAAP Liability

Current Hewitt Q's

(in $billions)

0

40

80

US GAAP Expense

Current Hewitt Q's

2,1672,276

55

72+5% +30%

Risk…Rewarded 17


Some Closing Thoughts

Seeking Input on Some of our Data ChallengesLooking for practitioners with experience in these large-scale mortality studiesLooking for practitioners with experience or interest in comparative longevityReactions to this data?

Risk…Rewarded 18

UK Longevity Consulting Perspective

Risk…Rewarded 19


3

3

14

31

50

51

54

0 10 20 30 40 50 60 70 80

Other investment

Currency

Bond investment

Inflation

Equity investment

Longevity

Interest rate

Longevity risk in a UK pension scheme context

Longevity as worrying as equities?

Source: Hewitt Global Risk Survey 2008 – UK Responses

Risk…Rewarded 20


■ Standard tables■ No cohort

analysis■ Immature

schemes + high net interest rates= defer thinking

■ Mortality improvements analysed by cohort

■ Fading mortality improvements still the norm

■ Mortality treated as base + improvement

■ Continued future mortality improvement taken seriously

■ Mortality rating by address (initially only bulk annuities, now individuals)

■ Per person mortality rating is standard

■ Longevity risk understood, routinely priced, even traded ?

Evolution of mortality modelling in the UK

Risk…Rewarded 21


The UK Hewitt Longevity Model

Socio-economic factors do not depend on scheme, so we canpool data between many schemesdevelop a model based on the pooled data

The UK Hewitt Longevity Model isThe use of– access to substantial pooled mortality data, plus– socio-economic information we infer from addressesto improve our estimate of individual longevity

Risk…Rewarded 22


The UK Hewitt Longevity Model Member postcodes mapped to database supplied by Experian (provider of information and analytics) to estimate member's socio-economic type A socio-economic type is a grouping of individuals having similar characteristics measured in terms of wealth and lifestyle, and therefore likely to share a similar future mortality experienceFurther grouped these socio-economic types into clusters based on the combined mortality experience of Hewitt client pension schemes so that we can model them statisticallyMap of UK illustrates how individuals with different postcodes are mapped to different mortality clustersFor each of the different socio-economic clusters, determined a mortality assumption based on the collected mortality experience of Hewitt clientsWe augment the model using Government mortality statistics for regions of UK on a per member basis.

Some insurers now provide lower payments to pensioners in more affluent areas

Risk…Rewarded 23


Longevity risk components

Idiosyncratic RiskEven if you knew the “correct” mortality rate, experience will differ, particularly in small schemes.

Basis RiskHow your scheme differs from the big population, and the difficulty of measuring this and its implications.

Systematic RiskChanges in general longevity for a big population (e.g. England and Wales, or insured lives, or SAPS)

“First person to live to 1,000 might be 60

already”

Risk…Rewarded 24


Market for longevity riskUntil recently

The main (the only?) way to reduce/remove longevity risk was via a bulk annuity type solutionCan modify future benefits but cant deal with past obligations

But NowThe Longevity swaps market existsMissing piece of risk management jigsaw?

Risk…Rewarded 25


PROVIDERPENSION

SCHEMEPENSIONER

Scheme specific longevity swaps—a simple concept

Monthly payment until pensioner

dies

Fixed leg

Floating Leg

Monthly pension payment to provider

for fixed term

Monthly payment to scheme until

pensioner dies

Risk…Rewarded 26


Structuring your Swap

Made to Measure or Off the Peg?

Index population

Aggregate,no inflation

Individual,no inflation

Individual,with inflation

Index v actual population

Aggregate v individual benefit

Assumed vactual inflation

Greater basis risk

Greater administrative complexity

Basis forswap

Basis risk

Risk…Rewarded 27


How do the longevity providers price?

OccupationPension size

Fitness Country you live inMonth of birth

Smoke

Time since retirement

Lifestyle

Current health

Sex

Age

Year of birth

Where you live

Pricing process (scheme-specific swaps):– Deduce best-estimate of mortality for the population

> Using postcode rating factors & scheme experience

– Add risk margins for the uncertainty, cost of capital and profit

http://www.clipart.com/en/close-up?o=3917228&memlevel=A&a=a&q=hour%20glass&k_mode=all&s=1&e=18&show=&c=&cid=&findincat=&g=&cc=&page=&k_exc=new%2C%20happy&pubid=

http://www.clipart.com/en/close-up?o=3965908&memlevel=A&a=a&q=healthy&k_mode=all&s=1&e=18&show=&c=&cid=&findincat=&g=&cc=&page=&k_exc=smoking&pubid=

http://www.clipart.com/en/close-up?o=4925251&memlevel=A&a=a&q=gender&k_mode=all&s=1&e=18&show=&c=&cid=&findincat=&g=&cc=&page=&k_exc=calander&pubid=

Risk…Rewarded 28


Overview of the market to dateMarket really took off in Summer 2008– Providers put specialist teams in place to target UK pension scheme trusteesBabcock and RSA deals are first swaps written directly with a UK pension scheme– Large pipeline of transactionsAlready seeing standardisation of product in the market– Have to ensure price comparisons between

providers are “like-for-like”Providers generally fall into two categories:– Investment banks

> Distribution of risk to third parties– Insurers (incl. reinsurance) companies

> Retain risk on balance sheet or reinsure

Pensioner Longevity Data Analysis and Applications

Documents