Wednesday, July 27, 2011

"Smarter than the average..."

It is said (and I'm not sure where this started) that eighty (or seventy) percent of drivers rate themselves as "above average". This phenomenon is often ascribed to "cognitive bias"—that is, that people think they are better than they are. Psychologists explain that cognitive bias allows us to survive in the world which constantly tears down our self-esteem.

But what if that's not the case? Can 80% of drivers rate themselves above average... and be right?

Option 1: Different metrics

My interaction with other drivers shows that different drivers have different ideas of what "good driving" entails. I put myself with a group that equates good driving with courteous driving; that is, good driving doesn't frustrate (or even surprise) any fellow drivers and allows the driving system to function as efficiently as possible. After a really good drive, the driver is relaxed and confident.

A rival group of drivers believes that good driving gets you places as quickly as possible. Other drivers (especially slow ones) are obstacles to be dodged in pursuit of the quickest travel times possible. A good driver is an expert of handling and can fit in the tightest spots in freeway traffic to get there sooner. After a good drive, the driver is there quicker and is proud of his abilities.

Both groups agree that good driving results in fewer accidents. Nevertheless, they approach this goal differently. If you ask either group whether they are above average, they will compare themselves favorably to members of the other group and answer "yes". And are they wrong?

Option 2: Asymmetric distribution

Going back to the original statistic, it is claimed that 80% of drivers claim to be "above average" drivers. But this isn't necessarily even a problem. If 80% claimed to be above median drivers, we would have a problem, but this isn't the case.

Take the mean (average) national income, for instance. In 2004, it was around $60K/yr. However, the median national income was only $44K/yr. That is, 50% of the country made less than $44K, meaning far more than 50% of the country had "below-average" incomes. This makes sense; each year, some people in the United States are making millions of dollars. Since it is impossible to have a negative-million-dollar income to counter those outliers (zero is about as low as an income can go), the distribution is asymmetric, and the average is skewed.

Similarly, if we use a common driving metric of "number of accidents caused per year", some drivers are causing accidents, and may cause several. Those really bad drivers are outliers. But no matter how good a driver you are, you can't get any better than "zero accidents". With so many drivers who are at the premiere level of driving, and with such an asymmetric distribution, I don't think it's unreasonable at all for seventy or even eighty percent of drivers to be "above average".

1 comment:

  1. Peter sent this in via e-mail and asked that I post it due to Google issues:

    Two thoughts:

    1. Any distribution whose support is inifinite in only one direction (e.g. 0 to infinity) will always be skewed. In such scenarios the mean will never equal the mode, and the mean will always be towards the infinity direction. [I'd like to see a proof for or against this. Next blog post, Shayne? :)]

    2. I remember when I was in an English class in college, and the professor was asked about the grade status of the class. He said, "Well, about half of you are below average." It was funny to see how long the look of surprise and fear lasted on some of those faces.

    Technically he should have said, "about half of you are below the mode", but I'm sure he was assuming a normal distribution where mean == mode.

    ReplyDelete