As recently as 2009, the Foxes found themselves relegated to England’s third-tier league, and prior to the season, bookmakers were offering 5,000-to-1 odds for anyone mad enough to pick them to win the title – the same odds quoted on Elvis Presley being found alive this year.
This may have been the biggest upset in the history of sports, but it just seemed like the new normal in 2016. In June, the Cleveland Cavaliers became the first team to ever overcome a 3-1 deficit to win the NBA Finals, breaking a 147-season championship drought for the city of Cleveland’s professional sports teams. Then in November, long-suffering Cleveland baseball fans were on the brink of celebrating the city’s second championship of the year when the Chicago Cubs came back from a 3-1 deficit of their own to win their first World Series since 1908. As a lifelong Cardinals fan, I really hope this doesn’t become a regular thing.
2016’s unexpected events were not limited to sports arenas. To the surprise of pundits and pollsters, the U.K. voted to “Brexit” the European Union, and American voters elected a reality TV host as the next president. Two rare blue lobsters were caught off the shores of Nova Scotia three days apart, and a North Carolina woman scored two major lottery prizes in a span of three months. The year even saw the first full moon on the summer solstice in almost 70 years followed by three consecutive months with “supermoons,” including the largest since 1948.
Now fear not, fans of the Foxes, Cavaliers, Cubs, Brexit, President-Elect Trump, lotteries, blue lobsters, or even supermoons – I am not here to make value judgments about these events whatsoever. Nevertheless, I am struck by the feeling that rare, unexpected, and highly improbable events are occurring with increasing frequency, in violation of the fundamental laws of probability. As an actuary, I am used to dealing with risk and uncertainty, but this level of unpredictability scares even me. I want to understand why, and more importantly, determine what insights we in the insurance industry might glean from this year’s “unexpected” events. Here is what I’ve learned:
Statistician David Hand states that “extremely improbable events are commonplace.” 2016 provided pretty persuasive plausibility for the professor’s paradoxical precept. After all, winning the lottery is unlikely enough – but twice in the same year? With an Illinois man named Gambles (natch) also scoring his second win? The genetic mutation that turns crustaceans blue occurs in only one out of every 2 million lobsters, so finding two in three days in the same waters? With a Massachusetts lobsterman also catching one... for the second time in his career?
In these situations, we see the “Law of Inevitability” and the “Law of Truly Large Numbers” at play. With respect to the lottery, someone has to win, and there are a plethora of chances to do so (yes, El Guapo, a plethora). In Illinois, for example, Mr. Gambles can play Powerball and Mega Millions twice a week, the Illinois Lotto three times a week, and each of the Pick 3, Pick 4, and Lucky Day Lottos twice a day, for a total of over 2,500 drawings per year – and over 40 other states offer similar opportunities. And while many people join in the fun when multi-state jackpots get ginormous, state lotteries are typically the dominion of a finite cadre of devoted gamblers. Reports also often stretch the definition of what it means to “win the lottery” by including smaller payouts such as those that are awarded for picking four out five numbers (i.e., the “Law of Near Enough”).
Given all of this, it is inevitable that we will occasionally see some repeat “winners.” The only surprise might be that people keep playing after they’ve already won, but human behavior often is unexpected – we’ll get to that later. Like winning the lottery, the odds of any one person finding a blue lobster are very low, but since 200 million lobsters are caught annually in the North Atlantic, at a rate of one in 2 million, we should expect around 100 of them to be blue. The real question then isn’t “Why were three blue lobsters caught in 2016?” It’s rather “What happened to the other 97?”
Extensive research shows that humans are epically bad at intuitively sensing probabilities. This can be demonstrated by our instinctively counterintuitive reactions to the “birthday paradox” (that in a room of 23 people, there is a better than 50/50 chance that two of them share the same birthday) and the “Monty Hall problem” (in a game with a prize behind one of three doors, you significantly increase your odds of winning by switching your choice from door #1 to door #2 after being shown there is no prize behind door #3). Mind. Blown.
So uncertainty is certain and humans are hard-wired to underestimate it. But there is good news: this is exactly why the insurance industry exists! We have the power to provide consumers financial security in times of great uncertainty. But with great power comes great responsibility. So while it may seem a fool’s errand to predict the likelihood and impact of black swan events like pandemics, nuclear terrorism, or sovereign default with (thankfully) little historical data to go on, it is still each insurance company’s solemn duty to ensure there is sufficient capital – or reinsurance ;) – to withstand a wide range of “unexpected” events.
Much of the field of predictive analytics is focused on modeling human behaviors. How will a block of citizens vote? How will a group of athletes perform on the field? Will consumers be positively influenced by a new marketing message? Unlike consistently predictable phenomena like planetary motion, radioactive decay, and the cuteness of puppies, humans don’t follow rigid, preordained Laws of Nature. They are often fickle, irrational, or just downright weird.
It seems indisputable that the political positions of President Obama generally lie closer to those of Hillary Clinton than those of Donald Trump. Yet in the 2016 election, many voters who supported President Obama in prior elections voted for Donald Trump. In fact, Trump won over 30% of the “blue” counties that had twice favored President Obama over his Republican rivals. We can debate countless reasons for this unexpected result, which all serve to underscore the point: people make decisions based on a wide array of factors – both logical and emotional – and thus are very complicated creatures to predict.
I would estimate that I’ve given variations of my “lemons” lecture on behavioral drivers of mortality experience well over 50 times in a dozen or so countries over the past 10 years. The reason is not that I’m too lazy to put together a different presentation. Well, that’s not the only reason. The other reason is that the truth it reveals remains timeless: human behavior, in all of its various forms, is the single most important factor impacting insurance claims experience. What personal lifestyle decisions is an applicant making? Is he anti-selectively choosing a particular company, product, or coverage amount? Is he honest on the application? Does he choose to lapse (or not lapse) his policy? Will he change his behaviors once insured? Will Tim just stop it already with these endless strings of questions? Isn’t this article long enough as it is?
Understanding a consumer’s decisions from a psychological perspective is just as important as understanding his risk factors from a medical perspective. But it goes further than that. Human-centeredinnovation in distribution, product development, or underwriting requires an understanding of the needs, attitudes, emotions, pain points, and expectations of consumers. (Unfortunately for me and my fellow actuaries, this occasionally requires talking to actual humans.)
Astonishing advances in computer science and artificial intelligence continue to blur the line between human and machine. Chess computers have long been annihilating the best human grandmasters, but in 2016 a Google algorithm named AlphaGo accomplished a much more challenging feat by besting the world champion of Go, an ancient Chinese board game with so many permutations of possible moves that it can only be won with strategic “thinking,” not brute-force computational power. This impressive analytics breakthrough was overshadowed, however, by more public failures, especially the poor predictions in advance of the Brexit vote and presidential election. Although to be fair, despite making some pretty terrible predictions throughout the election season, analytics guru Nate Silver did warn that there was “reason to think a [systematic] polling error is more likely than usual this year.”
Algorithms may eventually evolve into sentient beings that overthrow the human race, but luckily they aren’t there yet. A friend of mine wanted to test IBM Watson’s image recognition API, so he uploaded a picture of his wife. The highest probability match that Watson provided for the image was “dump truck,” even though anyone would instantly recognize the picture as a woman – and a beautiful woman at that. I guess it’s a good thing that “Human or Dump Truck?” wasn’t a category on Jeopardy.
The insurance industry is storming full speed ahead to compete in an analytics arms race, especially in the domains of digital marketing and underwriting. As we do, we should be excited about the opportunities these tools deliver for improving the consumer experience but also acknowledge their dangers. Models shouldn’t unfairly discriminate against applicants; private information should be kept private; and data sources, modeling approaches, and results should be transparent. This requires a top-down enterprise analytics strategy and a happy marriage between data scientists and business domain “translators.”
Unexpected events often seem even more unexpected because we are repeatedly told just how unexpected they are. Every Cubs fan (and Cardinals fan, hee hee) could have told you the Cubs hadn’t won since 1908, not because we all happen to be experts on the Edwardian Era, but because the sports media has drilled it into our baseball-lovin’ brains over and over again. We even make up mystical curses and superstitions to further advance the narratives: the Cubs’ Curse of the Billy Goat, the Boston Red Sox’ Curse of the Bambino, the New York Rangers’ Curse of 1940, and so on.
As discussed earlier, “rare” events are rarely rare, but that rarely stops the media from pretending like they are. Blue lobster sightings are demonstrably commonplace given the truly large number of lobsters caught each year, yet the media continues to hype their “rareness”: for proof, look no further than reports in 2013 (“rare”), 2014 (“ultra-rare”), and 2015 (“extremely rare”). But lest you think that hyping up the rareness of blue lobsters is a recent phenomenon, feel free to peruse the August 18, 1889 edition of The New York Times to learn about the “genuine blue lobster” caught in Marshfield, Massachusetts, where “nothing of the sort” had ever been seen before.
Donald Trump’s victory may also seem “rare” given his limited political experience, but don’t forget that American voters also elected an actor, a singer, a comedian, a bodybuilder, and a professional wrestlerto major offices in the not-too-distant past – all of which are arguably more unexpected training grounds than that of an Ivy League-educated CEO. But the media loves to drum up drama with the hopes that it will translate into ratings, and in 2016 it worked: a record 84 million Americans watched the first televised debate, and cable news channels, radio talk shows, and political blogs all recorded record levels of interest throughout the race.
And what to say of the media’s fascination with supermoons? They aren’t really rare (about 5 out of 12 full moons qualify) or unpredictable. The term was invented by an astrologer (not astronomer) and picked up by the media, perhaps because the perigee-syzygy of the Earth-Moon-Sun system wasn’t properly capturing the public’s imagination. So in a way, you could say that supermoons are the Chilean sea bass of celestial events... or not.
While the media is fun to blame, scientists and statisticians are just as guilty. Esteemed peer-reviewed journals have a strong bias toward publishing research that shows something new or unexpected rather than equally high-quality research that confirms prior findings or shows nothing particularly surprising. And an all-too-common practice referred to as p-hacking leads many researchers – usually unintentionally – to claim “statistically significant” unexpected findings that are actually neither significant nor unexpected. As economist Ronald Coase eloquently states: “Torture the data long enough and it will confess to anything.”
While relentlessly pursuing the next big innovation, insurance companies must be careful not to get swept up in the hype and hysteria of every new shiny object. View reports and data objectively with a skeptical eye. Will blockchain protocols transform the insurance industry? Well, I certainly think so – but I was also optimistic about Zoolander 2, so my predictions definitely deserve some healthy skepticism. New technologies or business ideas can deliver transformational opportunities, but are almost always dripping with uncertainty. The key then is balancing pragmatism and optimism: always expect success but be prepared to pivot or walk away if the hope turns into hype.
Distant and/or irrelevant past events often distort our perceptions of present-day probabilities. If the 1909 Chicago Cubs starting lineup was taking the field today, I will concede that they would not have much of a chance of winning the World Series. But they aren’t, and neither are any of the 1,699 other Cubs players that never won a title between 1908 and 2016. So while decades of losing may adversely affect the mental health of a team’s fans, it is largely irrelevant to the performance of a team’s current players. Sports teams are a collection of confident young athletes with no connection to their team’s history. Consider this: the average birth year for the roster of the Chicago Cubs (and, coincidentally Leicester City) is 1987, which means that these athletes have never known a world without cell phones, The Simpsons, or Donald Trump running for president (yes, he was already campaigning back then).
In addition to ushering in a disconcerting-for-Cardinals-fans era in which the Cubs are now expected to win the World Series, 2016 also marked the first time that tech companies filled the list of the world’s top five most valuable public corporations: Apple, Alphabet (Google), Microsoft, Amazon, and Facebook. Remember life before the iPad? You lived through those dark times – way back in 2010. Ubiquity comes quickly.
The winds of change are certainly blowing in insurance, and not a moment too soon. LIMRA estimates that 60 million households have a significant life insurance coverage gap, which creates a $12 trillion market need. Startup companies are seeking to disrupt the traditional ways in which insurance is distributed and underwritten. Venture capitalists invested over $2.6 billion in InsurTech startups in 2015, and although deal activity may have slowed down in 2016, I can assure you from my own personal experience that the fundamental momentum is still rapidly accelerating. Partnerships between reinsurers, carriers, startups, and data companies are becoming commonplace as everyone seeks to unlock the value of digital technologies and data analytics to better serve a massively underinsured population.
2016 proved not only that anything can happen, but that anything will happen. Predictions are often wrong, estimates of likelihood are often misunderstood, and historical paradigms are often begging to be broken. The insurance industry stands in a unique position to help the public manage risk and uncertainty, but we must remain keenly attuned to rapidly changing consumer attitudes and expectations, avoid buying into excessive hype and hyperbole, and combine computational analytics with objective expert advice and insight. While it is important to prepare for the unexpected, it is equally important to proactively shape the future. Abraham Lincoln is often credited with saying that “The best way to predict the future is to create it.” If you follow his advice, maybe 2017 won’t seem quite so unexpected.
PS: For those interested in further exploring the world of the “unexpected,” I recommend The Improbability Principle by David Hand, The Black Swan and Fooled by Randomness by Nassim Nicholas Taleb, Predictive Analytics by Eric Siegel, The Drunkard’s Walk by Leonard Mlodinow, and The Signal and the Noise by Nate Silver.
