Mapping Strategy

Forecasting Their Way to Durable Profits

From the front page of today's Wall Street Journal [emphasis added]:

...a little-known tool of the insurance world: Computerized catastrophe modeling. Crafted by several independent firms and used by most insurers, so-called cat models rely on complex data to estimate probable losses from hurricanes.

But regulators and other critics contend that the latest cat models -- which include assumptions about various climate changes -- are triggering higher insurance rates.

Starting in the early 1990s, cat models began to replace the industry's older tools. Previously, insurers based their rates and underwriting policies largely on historical records of past claims.

Never mind if the forecasts are accurate or not. They work. To increase profits. That's not a pejorative statement as far as the insurance companies are concerned. It's their fiduciary responsibility to maximize return for their shareholders by any legal means available. It does however, set up an interesting, albeit very long-term test of the credulity and patience of their customers. I.e., should the forecasts prove inaccurate. More from the WSJ piece [emphasis added]:

Underlying the newer cat models are scientific theories that rising sea temperatures will result in more intense, and possibly more frequent, hurricanes. The hypotheses suggest that catastrophic hurricanes like 2005's Rita, Wilma and Katrina weren't an aberration, but rather the shape of things to come.

Large reinsurance companies... were early converts to theories of global warming and cite warming of the earth's oceans when predicting massive damages from future storms.

Well of course they were! Think about it for a second.

You're the CEO of a big reinsurance company. Two scientists walk into your office. One says he's got a new computer model that predicts big future costs due to storms.

"Of course I may be wrong," says scientists number one, "After all, I'm a scientist. We deal in hypotheses. You're the executive. You need to make the call on what to do with this."

"So let me get this straight", says the CEO. "If your models are right, we get to raise our rates because our competitors, customers, regulators and the general public are all looking at the same stuff, convinced that your predictions are true? Is that what you're saying?"

Hmm, she thinks to herself (the CEO). This is pretty good. If this guy is wrong, we get to pocket the difference for decades. It would take that long for anyone to prove this guy wrong. Even better, it's not that hard a sell to the general public. We've got cover. If he's right, well... we'll keep him on retainer. Keep his firm in good shape. Drop them some plum projects to keep 'em happy... maybe even kick this over to mar-com to manage. Under that scenario, we've at least got a decent business until I retire -- no better or worse than before. All upside. No downside. Suhweet!

The CEO smiles. She turns to scientist number two. "And what do you have to say for yourself?" she asks.

"Um...", stammers scientist number two (a geologist by training), "...the fossil record over several million years, and agricultural evidence over several millennia would tend to suggest that storm activity actually decreases during warmer climatic periods. And we're just getting this new data in from deep ocean probes suggesting that 80-90% of the ocean volume is staying the same temperature or maybe getting colder. It would be a bit hasty, in my opinion, for you to raise your rates based on a set of theories and prospective models that claim precise predictability when what we're really dealing with here is massive uncertainty. If you just sign the proposal we put on your desk to study this a little further, we can refine the numbers..."

"Thanks for your advice", asks the CEO. "I'll look into it. Have a nice day."

That's a fantasy dialogue, obviously. Read the WSJ article and draw your own conclusions [emphasis added]:

Perhaps the most prominent critic to surface is Karen Clark, an economist who founded one of the first cat-modeling firms two decades ago. Today, she warns about the programs' misapplication...

Companies that rely too heavily on cat-model data "are subjecting their businesses and their customers to the volatility of computer models," says Ms. Clark, who now runs a Boston cat-model consulting business. "The models are being used as if they produce definitive answers rather than uncertain estimates." Ms. Clark says she advises clients to use them in conjunction with other factors, such as broad historical data.

To be sure, insurers themselves are facing higher rates from the reinsurance companies that backstop their claims. The reinsurers, and the financial ratings agencies that assess the health of carriers, are also using the controversial newer models.

I just love this last bit:

...some models now attempt to estimate future losses over a shorter period of time. In doing so, they may also use selective historical data. One model, for example, was reprogrammed to give greater weight to years in which ocean temperatures were particularly warm and hurricane rates were high, such as the period from 1930 to 1945 [prior to broad industrialization and CO2 increases]. That particular model resulted in higher loss estimates for the near-term.

...and therefore higher rates. The logic is circular.

Butterflies for Dummies

Check out this succinct review in last Sunday's Boston Globe of an important but oft-misunderstood concept critical to any forecasting or planning endeavor [emphasis and link added]:

The butterfly effect is a deceptively simple insight extracted from a complex modern field. As a low-profile assistant professor in MIT's department of meteorology in 1961, [the late Edward] Lorenz created an early computer program to simulate weather. One day he changed one of a dozen numbers representing atmospheric conditions, from .506127 to .506. That tiny alteration utterly transformed his long-term forecast, a point Lorenz amplified in his 1972 paper, "Predictability: Does the Flap of a Butterfly's Wings in Brazil Set Off a Tornado in Texas?"

In the paper, Lorenz claimed the large effects of tiny atmospheric events pose both a practical problem, by limiting long-term weather forecasts, and a philosophical one, by preventing us from isolating specific causes of later conditions... It is extremely hard to calculate such things with certainty... Realistically, we can't know. "It's impossible for humans to measure everything infinitely accurately," says Robert Devaney, a mathematics professor at Boston University. "And if you're off at all, the behavior of the solution could be completely off." When small imprecisions matter greatly, the world is radically unpredictable.

Moreover, Lorenz also discovered stricter limits on our knowledge, proving that even models of physical systems with a few precisely known variables, like a heated gas swirling in a box, can produce endlessly unpredictable and non-repeating effects.

"Lorenz went beyond the butterfly," says Kerry Emanuel, a professor in the department of earth, atmospheric, and planetary sciences at MIT. "To say that certain systems are not predictable, no matter how precise you make the initial conditions, is a profound statement." Instead of a vision of science in which any prediction is possible, as long as we have enough information, Lorenz's work suggested that our ability to analyze and predict the workings of the world is inherently limited.

What few articles on this subject touch upon, but I find endlessly fascinating, is how this undeniable, well-grounded scientific insight interacts with human nature -- specifically our ingrained need/desire to feel knowledgeable and in control. (I use the term 'we' in both the individual and corporate sense here.)

It has been my experience that few individuals and very few (if any) organizations ever err on the side of humility in this regard. Perceptual failures are seldom in the direction of assuming less than is actually predictable. More often, the response is something like this:

yes I know about the butterfly effect, chaos theory and all that... but it's kinda academic... and you don't know my boss... you see, I/he/she/they/we need to predict XYZ anyway...
because my/our future depends on it... so just give us a number... give us your best guess about the probability... please?

People fall back on a prediction/probability paradigm either because it's what they know (or have the planning tool-set to address) or because, even knowing that approach is flawed, they find it more comfortable... and comforting (or politically expedient in a particular organizational culture) to pretend that what is fundamentally uncertain is perhaps predictable after all. The chance of rain is 62.5%...