Econometrics, The Black Swan and Moneyball: Understanding the Numbers Behind the Underwriting
“Before the discovery of Australia, people in the Old World were convinced that all swans were white, an unassailable belief as it seemed completely confirmed by empirical evidence.It illustrates a severe limitation to our learning from observations or experience and the fragility of our knowledge.” -Nassim Taleb, The Black Swan
Clearly, things are not always what they seem. Economics, business, sports and gauging risk for insurance are examples of disciplines that use numbers to measure performance or the likelihood of something happening. Numbers that may explain something on the surface – a batting average, unemployment data, an income statement or a laboratory diagnostic measure can sometimes take on a different meaning after deeper analysis. Sometimes these deeper analyses help explain those ‘Black Swans’ or outliers in any discipline.
Number crunchers have turned sports into a scientific discipline that goes beyond physicists attempting to explain how and why a curveball moves. Sifting through statistics has allowed general managers to assemble sports teams that perform well for much less money than big-spending clubs by uncovering outliers and revealing patterns in the numbers and how they interact with one another.
Econometrics have emerged to the point where an easily understood number such as the unemployment rate can be picked apart to show that an economy may be better or worse than it seems by examining workweek, average wages or workforce dropouts. Finance has abundant cases where a company’s rosy earnings turned out to be a fraud leading to executives being arrested and shareholders losing money.
When Malcolm Gladwell described The Roseto Mystery in his book Outliers, we were able to see that things are not always what they seem, suggesting we need to think of things a new way. ExamOne, which helps insurers with the underwriting process, saw an opportunity to use data to look at insurance underwriting a new way and examine data differently when assessing mortality. Is it possible that certain factors of an overweight person with high cholesterol may be a good risk, but that an apparently healthy person may have silent health problems that could be uncovered?
Similar to how a baseball general manager can now more effectively calculate whether a backup outfielder is worth $1 million a season in terms of on base percentage, insurance companies can do something more equitable. Instead of pooling large groups of individuals such that healthier applicants subsidize unfairly the risks of less healthy others, carriers may be able to personalize life insurance rates. Think of paying only the premium for ‘your’ individualized risk.
Rather than slugging percentage, games played, and the number of home runs that get the scrutiny of baseball executives, insurers have traditionally evaluated high cholesterol, body mass index, and diabetes and adjusted the premiums accordingly. What was often overlooked was the connectedness of the type of health data gathered during the underwriting process. The component values of the body’s total cholesterol level are a better indicator of health rather than hitting the panic button when a person’s total cholesterol exceeds 200 mg per deciliter (mg/dL) of blood. That one score may not account for a high HDL (good cholesterol), in which 60 mg/dL or higher is ideal since it improves cardiovascular health.
As more laboratory tests are used to obtain more accurate data, the likelihood of a person’s test result deviating from the norm increases.
The relationships between the laboratory results become much more prominent because of these potential deviations. An abnormal value for a test for liver function may not be as significant as originally thought. However, combinations of laboratory data can point to heightened mortality risks, also known as ‘cryptic risk’, as well as diminished risks that reveal the ‘hidden healthy’.
Laboratory results and a variety of other physical and demographic data are used to assign a score of 0-99. Applicants with a score greater than 75 are at the highest risk, while those with a score under 75 are at the lowest risk. By looking at things differently, more than 30 percent of applicants currently excluded from preferred pools actually present lower mortality risks than the majority of preferred-qualified individuals who get the lowest insurance premiums. Conversely, more than 25 percent of preferred applicants actually exhibit higher mortality risk than most non-preferred individuals. This approach may also have some long-range implications for how health insurance is underwritten as well.
Body mass index, blood pressure and high cholesterol are the most significant criteria disqualifying insurance applicants from preferred ratings. What if we could lower premiums and still improve mortality experience?
Along with the unrecognized ‘hidden healthy’, insurance applicants can present with ‘cryptic risk.’ They may meet all of the conventional preferred criteria for insurance ratings, but the interaction of several ‘borderline normal’ data leads to higher mortality risk. There are literally millions of Black Swan insurance applicants, since they are not what they seem by traditional measures.
So whether it is baseball or insurance underwriting, it helps to question assumptions that reflect conventional wisdom. As much as batting average is important, it also pays to examine how a player hits with runners in scoring position to help piece together productive innings. Insurance underwriting demands an examination of data to reveal truths that defy conventional wisdom about health and mortality. As an insurer, this direction can give insights about overall insurability and setting the right premium for a person’s risk. An additional benefit to using personalized risk scores can determine the degree of information necessary to underwrite a policy. Some policies may be issued without an attending physician statement, a process that often delays underwriting and the opportunity for the policy to be issued and the insurer to collect premiums- paramount interest to all involved.
Baseball is not as serious as calculating the odds of someone dying prematurely, potentially leaving emotional and financial hardships for loved ones. The median household income is approximately $50,000, according to the U.S. Census Bureau (2009 data) and insurance needs are best calculated by circumstances: age of children, income, overall debts, etc. There is plenty of unmet life insurance demand that we can reach in new ways with better decision support tools. The trend line for individually owned life insurance is at historic lows. Perhaps it is time to review the same old way we have been doing things to allow insurers to assess risk more accurately.
It is as important to get it right when it comes to negotiating a big baseball contract or a routine insurance contract, since there are a millions of life insurance policies issued every year. An insurer doesn’t want to pay a life insurance claim after one or two years into a policy for health conditions that are detectable. A baseball team general manager doesn’t want to pay more than $100 million for middling performance, nor does that manager want to lavish seven-figure contracts on backup players. Smarter analysis decreases the odds of either big, avoidable claims or misguided baseball contracts.
by Troy Hartman for the March 2012 issue of Life and Health Advisor Magazine. Mr. Hartman currently serves as President of ExamOne.