Natural-born Actuaries

Friday, August 25, 2006

By Wray Herbert

You’re on hold. You want to order tickets for Superman Returns, but you have to wait your turn in the telephone box office queue. You’ve already been waiting a full minute and are getting fidgety. How long is this wait going to take? I know, forever if you’re being subjected to the Muzak version of “Horse with No Name,” but let’s assume it’s not that painful. Three minutes? Seven? Is this mindless summer movie worth wasting eleven minutes on hold?

The people at Warner Bros Pictures sure hope so. While you’re tapping your fingers, some movie executive in Hollywood is rubbing his hands together as he scans the movie’s first week revenues: $84.2 million. And the summer is young. With dollar signs in his eyes, he tries to estimate the money his company will haul in when all is said and done. He envisions his career soaring by autumn.

These two scenarios, while quite different, both involve everyday predictions. Whether it’s as trivial as a movie queue or as big as a blockbuster business deal, people are making decisions all the time based on how their minds estimate things like time and money. And often we’re forced to make these decisions without much history to guide us. How good are we at this crucial cognitive skill?

The received wisdom has long been that we’re not at all good at it. Faced with uncertainty, people tend to fall back on error-prone rules and make predictions that are off the mark. At least in the laboratory. But this conclusion struck Brown University psychologist Thomas Griffiths as questionable, simply because we see people making smart decisions every day with little to go on. He decided to test the idea in the real world.

His study was fairly simple, though it involved complex mathematical modeling. He gave people limited information about a number of different topics, and then asked them to make predictions. For example, he told some people to imagine meeting a stranger who is 18 years old. Then he asked the subjects to predict how long a life this man would have: 39 years? 83? 96?. Insurance company actuaries are paid big bucks to do this kind of estimating all the time. Others were told that an Egyptian pharaoh had been ruling for 11 years. How long would he reign? Griffiths also used the waiting time and blockbuster scenarios, among others.

There is a well established and highly regarded computer modeling system that can make optimal predictions with whatever data is available. And it is sophisticated enough to know how to distinguish between one kind of prediction and another. That is, it uses one formula for something like gross movie receipts, because it has to accommodate the fact that most movies have modest earnings, while a few blockbusters make mongo bucks. It uses a very different formula to make actuarial predictions of death, because we know that all people live on average 75 years, plus or minus.

Griffiths fed real-life historical data on all these various events into the computer, and then compared people’s actual estimates to these computer-generated predictions. On a broad range of topics—including life spans and movie revenues--people’s predictions were indistinguishable from the computer’s. It seems we're actuaries and movie moguls by nature. What’s more, he reports in the journal Psychological Science, the statistical “shape” of the real-life forecasting indicated the people have implicit beliefs about things like movies and aging, and know intuitively that they cannot be treated the same: Not to put too fine point on it, dying is far more predictable than making blockbuster movies.

But the participants weren’t perfect in all their predictions, and their mistakes shed some light on the limits of everyday cognition. For example, they consistently overestimated the reigns of Egyptian pharaohs. Griffiths was intrigued by this error, and asked the participants afterward how many years they figured a typical pharaoh ruled. Most thought around 30 years, much too high. Interestingly, they had the general model for predicting pharaohs' reigns correct; that is, they didn’t lump pharaohs’ reigns with movies, predicting blockbusters and dogs. They knew reigns were more predictable than that, because of such things as death and rules of succession and so forth. But they were basing their predictions on modern monarchs, like Queen Elizabeth II, who live much longer than pharaohs did and appear to reign forever. If they had thought to factor in disease and the primitive health care of 4000 BC, they would have been right on the mark.

It’s also possible to use these findings the other way around, Griffiths notes. That is, people’s judgments can serve as a guide to their beliefs. Another common mistake in the study is illustrative in this way. Participants were told about a person who has been married 23 years, and asked how long they thought the marriage would last. The data were useless for the purposes of the study, because 53 percent specified not a number but “forever.” This syrupy finding actually does reflect the real-life fact that about half of marriages end in divorce, but it also reflects the power of our beliefs—or hopes—to trump our generally dispassionate predictive skills.

For more insights into human nature, visit the Association for Psychological Science website at

posted by Wray Herbert @ 11:28 AM