Wine, Cheese, and a Lesson in Statistics

“You tell them the story.”

“No, you should tell the story.”

“But it starts with you, Amy.”

Amy Kessler, head of longevity risk transfer at Prudential, begins a tale over dinner on the third day of the 11th International Longevity Risk and Capital Markets Solutions Conference in Lyon, France. And as longevity stories go, it is a good one.

While scrutinizing UK statistics for people aged over 90, Kessler recalls, she realized the numbers didn’t match with her company’s wealth of real-world data. While some might put a call in to their actuaries, or make a request to the number crunchers at the Office for National Statistics (ONS) in London, Kessler had a better idea: Call Professor David Blake, one of the world’s foremost longevity experts, the co-creator of the longevity model Kessler was using—and occasional CIO columnist.

Blake, sitting two seats to Kessler’s right (CIO is in between—possibly the best seat in the house), picks up the story, while refilling wine glasses.

“If it was anyone else saying this data were wrong, I would have thought they were mad,” Blake says. “The ONS data are sound. We’ve always assumed that.”

It wasn’t sound. Blake and Professor Andrew Cairns (another co-creator of the same longevity model), trusting Kessler’s judgement, scratched their heads and drilled down into the numbers, going back to the 2011 and 2001 UK censuses.

In short, the ONS had inadvertently created 10,000 people aged over 90 back in 2001. These people were steadfastly refusing to die—pretty difficult to do when they didn’t exist in the first place—and were distorting the numbers.

It was funny, but perhaps you had to be there. And be interested in longevity.

Art by John CuneoOver dessert (and more local wine), the discussion moves on to the importance of data. Blake, Cairns, Kessler, and others around the table agree that we need to be more skeptical of data.

My mind flashes back to my first job, in a pension administrator’s office, manually entering fund members’ data into a computer system. Address errors, spelling errors, multiple dates of birth for the same person—such basic problems that take pension funds entering risk-transfer deals months to iron out—I saw them all.

Kessler, the perfect host, doesn’t miss a beat. “Thank you for doing that, Nick,” she says. “Those people could be our customers now.”

For Blake, of the Cass Business School in London, his battle with ONS data highlights a central theme of the conference: The combination of academic work with the “experience and intuition” that told him how to solve Kessler’s problem.

Over the past decade, Blake has been taking the conference around the world: Chicago, New York, Sydney, Frankfurt, Waterloo (Ontario), and Beijing. For longevity experts and statistics nerds alike, it is a must. Academics and pension practitioners rub shoulders and present to each other in the main auditorium. In breakout rooms, PhD students present their life expectancy models for the first time, often explaining complex concepts in a language that is not their native tongue.

It is a perfect opportunity for these students to share their theories, pick up new ideas, and—most importantly—put them into practice. Guy Coughlan, the new chief financial risk officer at the UK’s Universities Superannuation Scheme, concurs: As well as giving his own talk on the £41.6 billion ($64.1 billion) pension’s longevity statistics, he is sitting in on several academic presentations on the hunt for new ideas and perspectives.

To a mere mortal, the concepts can be mind-bending. Yijia Lin and Tianxiang Shi of the University of Nebraska present their model for a pension funding index. The presenters intend pensions to be able to purchase “buyout options” priced from this index, allowing the funds to secure a buyout at a later date if their funding level reduces. Are you with us so far?

“Address errors, spelling errors, multiple dates of birth for the same person—such basic problems that take pension funds entering risk transfer deals months to iron out—I saw them all.”The next speaker, a young Turkish PhD student, gestures toward a slide showing the most complex equations. “Of course, I don’t need to explain this to you—we are all familiar with the Lee-Carter model,” she says, nervously. The audience chuckles; the two journalists in the room stare boggle-eyed at the screen, sorry we had not made an early start on the wine.

In yet another presentation, the sizeable audience is assured that a similar page of what can only be described as algebraic gibberish can be worked out on a pocket calculator. I am certain my own device doesn’t do Bayesian probability—whatever that is.

Other presenters litter their talks with “Poisson risk,” “Gaussian kappa,” and “non-parametric beta-x.” This attendee’s brain checks out, promising to return for the biscuits and much-needed coffee.

When caffeine arrives, it becomes clear that for the students this event is also an opportunity to meet Professors Blake and Cairns, upon whose work many of their new ideas are based. The youngest attendees stand nervously behind the pair like autograph hunters, waiting for their chance to meet the masters (and perhaps ask for feedback on their own data models).

Later on in the day, and after several espressos and the obligatory glass of wine over lunch, the conference finally comes down to the level of a financial journalist. On the back of a discussion about long-term care costs, former physical education teacher Jean-Michel Ricard leads the room in a practical demonstration of the exercises he gives to retirement home residents to help improve their well-being. Logarithmic equations may be out of my reach, but deep breathing and touching my toes aren’t—at least not until they bring out the cheese board.

Outside the auditorium, the conversation comes back to data. Fresh from trying in vain to cram 60 animated slides into a 20-minute presentation, Alexander Zhavoronkov, a scientist in the field of “biogerontology,” explains to a handful of impressed attendees that Google Life Sciences is now the biggest actuarial firm in the world. With the data it has at its disposal, he states confidently, within five years its work will revolutionize longevity studies.

It’s easy to see why: A quick search for “Google Life Sciences” using a well-known internet search engine brings up a long list of technological developments in which the group is involved. Stabilizing technology for tremor sufferers, contact lenses that monitor blood sugar, and pills that can identify cancers and heart attacks. All these could be medical reality within a few years.

Chief among Google Life Sciences’ work, at least for the purposes of this conference, is a stake in DNAnexus, a “cloud-based genome informatics and data management platform.” DNAnexus is designed to collate information about the human genome to make it easier to understand, analyze, and treat genetic illnesses and aging.

Zhavoronkov talks excitedly about future advances in medical science he has come across, such as regenerative medicine that can actually make your skin younger. Take that, Laboratoire Garnier.

So as we sip good wine, ogle the cheese board, and talk in lines of letters that don’t make up words, the important point is this: All these scientific advances will be for nothing if our industry does not record and use the data it creates correctly. I hope to live long enough to see these advances make a real difference—and that the ONS doesn’t snuff me out before it happens.