Before I delve into this essay, four observations are necessary for perspective.
My apologies to the spirit of Ray Bradbury for appropriating the title of one of his many stellar works of fiction (more than a few of which I confess to having devoured in my youth).
I put quotation marks around “wicked” to emphasize that this loaded word is being used strictly in the figurative sense. I imply no personal criticism of those who may advocate the practice about to be described. The practice itself, on the other hand, has odious implications which, if only in my view, need to be aired out and further debated lest they be embraced prematurely in risk appraisal.
I put a question mark at the end of the title because the status of the matter at issue remains uncertain. Is it already here? Should we expect its landfall soon? Or, does it remain purely theoretical at this writing?
Because of the importance of addressing this phenomenon early on, it is necessary to work off of a number of assumptions. If, as it turns out, some of these assumptions are in error, then perhaps at least a portion of the concern expressed here will be mitigated accordingly.
As I wrote at the end of the December Hot Notes , I recently received in the mail a rather “busy”8-page flyer announcing a seminar on Predictive Modeling in Underwriting, coming up in January, 2009.
The organization (World Research Group) behind this event hosted a prior gathering I am aware of – I was invited to speak and declined – along the same broad lines early this year. It would seem they are in the seminar-hosting business, with no apparent connection to our industry.
One of the program segments at this forthcoming seminar centers on a phenomenon couched as“lifestyle based analytics”…
…which sounds as innocent as it does actuarial!
It seems that several persons from the health (medical) insurance industry – none cited as having an underwriting background per se – are going to share perspectives on how such analytics might “achieve a higher degree of accuracy” in the sorting of risks.
It is unclear from what little is said here whether or not the focus is on individual, small group or large group health underwriting.
Certainly if the focus is on large group exclusively, then the concerns cited below would largely evaporate. Large group underwriting does not consider the individual person; small group underwriting does.
Further to this point, it is not entirely clear that “life style analytics” are meant to be confined to health insurance risk assessment. If one apprehends the basic concept correctly, it is no stretch to foresee how it could readily be extended to underwriting all individual and small group mortality and morbidity risk products.
Is this a matter of concern only to Americans?
I have no idea whether (a) the data intended for use in this scheme would be accessible in other countries or (b) regulations already in place elsewhere governing insurance practices would prohibit their use if, indeed, they were accessible.
In terms of the big picture implications, please keep an open mind here. Recognize that what was unthinkable yesterday has, in many fields of play, become the rule of the road the morning after.
If nothing else, forewarned is forearmed.
What Are “Life Style Analytics”?
For now, one is limited to making inferences based on the succinct description in the flyer and what has been said in a similar context in the past.
It would seem that this high-minded phrase relates to the deployment of financial transaction records – such as purchases with credit cards and possibly in other ways as well – which can be grabbed from cyberspace in what I like to think of as mankind’s “Post-Privacy Era.”
Anyone who has waved at a surveillance camera on a London street corner needs no baptism in to the notion the “personal privacy” is quickly becoming little more than a historical phenomenon.
Is it possible to fetch personal credit card use data?
Personally, I have no idea.
By inference, how could it be otherwise if this is the premise underlying the “life style analytics”gambit?
How Might These Records Be Used?
Harking back to a lecture I sat in on a few years ago at the LOMA Health Underwriting Study Group (HUSG) annual meeting, I infer that records of consumer purchases would be culled, reviewed and then enmeshed with at least some conventional risk selection assets to sort those under scrutiny into certain presumptive – with the addition of this element, “presumptuous” may be a more appropriate modifier – risk categories.
Calling to mind examples aired out at the above-mentioned HUSG presentation, if one made purchase X, an inference would be made as to how this purchase relates to health habits/lifestyle and hence to insurability.
Purchase Y might be accorded similar or vastly different attributed implications.
For example, assume one had a longstanding pattern of buying many books but no athletic equipment; then, postulate that people whose main recreational pursuit is reading are less likely, relatively speaking, to get as much exercise as peers who allocate disposable income to cross country skis and hiking paraphernalia.
This presumption, as a matter of fact, was an example given at the HUSG lecture.
A similar analogy might be divined from intercepting records of one’s grocery purchases.
- Broccoli = good
- Hot sauce = neutral
- Mars™ bars = the nefarious work of the Goddess of Adiposity!
Another example I recall from the HUSG conference went something like this:
Suppose a fellow tenders up X dollars for purchase(s) made at a gas (petrol) station.
If one knows the price per unit of the fuel and the total sum expended for same, one might infer the size of the vehicle’s gas tank…and, in so doing, pigeonhole the class of vehicle being driven. Knowing nothing about gas tank sizes, I am at the mercy of experts here!
Suppose, further, that there is an additional sum on this charge mirroring the typical outlay for a pack of cigarettes…now, here is a worrisome hashmark for the sad-faced side of the risk ledger…
…if accurrate, that is; and assuming said cigarettes are not being purchased for someone waiting in the vehicle!
Would the holiday season purchase of a case of Jim Beam™ whiskey be a RED FLAG consistent with a too much good cheer?
Then, there are ostensible “worst case” scenarios:
- Sex “toys”
- Cuban cigars (at least in America)
- Paul Robeson CDs
I recall questioning (admittedly tongue in check) the fellow who made the HUSG presentation as follows:
If I were to purchase two tickets to see the (then, top-drawing) film Brokeback Mountain, would this be subjected to inferential dissection in a similar manner?
Let the record show that I got no answer, perhaps because of a tinge of sarcasm in my voice.
Are Health Habit/Lifestyle Choices A Valid Basis For Assessing Insurability?
Yes…and we have a wealth of clinical and epidemiological evidence to support this statement.
If someone disdains the Mediterranean diet in favor of foraging habitually on double cheeseburgers and french fries, the downstream implications have already been established in studies this underwriter has reviewed.
But it is one thing to admit to one’s dietary proclivities when questioned and quite another to presume to distinguish them based on credit card charges made at fast-food establishments!
And again, even if it can be established that less-healthy options were selected, how can we say for certain the burger and fries were consumed by the purchaser?
But these are, in the end, peripheral concerns which may or may not be manageable. The core issue here is far more insidious.
What Is The Core Issue With The “Life Style Analytics” Approach To Assessing Individuals’ Insurability?
Again, assuming that we have captured the essence of the nature and intent of this approach to underwriting, then the best way to explain the pervasive problem is by analogy…
It is said that the commandant of the famed Texas mission-cum-fortress known as the Alamo drew a line in the sand with his sabre and said something to this effect to those defending that compoundagainst an assault by the Mexican army: If you stand with me, cross this line. If not, I suggest you depart straight away as your prospects for escape are dimming as I speak.
This “line in the sand” analogy is a powerful image and I propose we draw one of our own.
Our line should be drawn between underwriting practices which satisfy appropriate prerequisites and those that fall short.
What “Appropriate Prerequisites” Do You Refer To?
The following are four criteria worthy of consideration when weighing the merits of deploying any risk appraisal asset:
- Does it confer sufficient independent protective value?
- Is it affordable?
- Can it be done within the constraints imposed by our unique milieu?
- How will it be perceived by parties to the transaction?
Now let’s apply these four criteria to what we infer about “lifestyle analytics.”
Independent protective value remains to be established…but cannot be ruled out.
Affordability is likely.
Time constraints should not apply as these data are undoubtedly accessible in or near “real time.” No other functional constraints are apparent.
Here is where the elephant enters the room!
Why Are You Opposed To “Life Style Analytics” As Described?
How would you feel about part or all of your access to premium for coverage in any line of L&H insurance being driven by presumptions based on intrusive assessment of your personal spending practices?
As an underwriter, I am taken aback.
As a consumer, I am outraged.
Allow me to diverge a moment:
In the past, I have argued that the tumor marker CEA (carcinoembryonic antigen) has fundamental drawbacks as a potential life insurance screening test.
My objection to CEA has nothing to do with its availability, ease of deployment or possible protective value. Rather, it is centered in the obvious (to me, at least) downstream consequences of using a cancer marker which is not approved for screening in a clinical context.
True, other tests we screen with are also not used in this manner clinically. Nevertheless, as we saw previously with the TAA tumor marker, push-back from applicants and their physicians, specifically where insinuations of undiagnosed – and likely metastatic – cancer are concerned, can be quite dramatic…most notably from those who are subjected to expensive work-ups and enormous stress…
…only to discover they are perfectly well; at which point experience teaches that their angst predictably turns to anger!
I have also raised eyebrows at making persistency inferences from one’s handling of credit card debt, asking rhetorically: how would you like to be told you cannot purchase life insurance because some credit reporting agency has advised your prospective insurer that you haven’t paid all of your bills on time? The arrogance of this inference trumps any statistical data regarding lapsation rates!
On anyone’s “scale of ten,” these two concerns are dwarfed by diversion of purchase records for the purpose of fueling “insinuation-based underwriting.”
Must matters really come to this in order to cost-effectively assess risk?
I think not...because as sure as the sun rises, our customers will not be forgiving of such undertakings…
…and, once consumer ire is raised, neither will those who regulate this industry.
Raising avoidable ire in this manner could open a much larger “Pandora’s box.” A recent (and unresolved) domestic brouhaha over foreign travel underwriting is a case in point.
How tragic to wind up unwittingly throwing out the baby with the bath water!
Do my arguments strike you as a tad “old-fashioned” and my position as too intransigent?
If so, perhaps it is because I turn 63 next year and have seen a great deal over the past 40 years.