Update, plus the Good Driver Fallacy

I just checked http://lichtenborgian.org, and the coming soon page is up, but I haven’t received any access info from Noah, nor can I see it in my FTP utility. I’ve emailed him and expect to hear from him as soon as he gets to his computer out in California. And then… then

So over on the Lacuna blog, Jeff pointed us to a book review in the Times, of Stumbling on Happiness, the main point of which in its applicability to us Lichtenbergians is that humans are an easily deluded lot, mostly by ourselves, and mostly to maintain our happiness.

The author cites the notorious statistical anomaly that “90% of drivers rate themselves above average,” which is usually used to show how self-deluded most people are. Because, clearly, 90% of drivers cannot be above the 50% mark, can they? It’s like, as is often smirkingly said, Lake Wobegon’s children. Or NCLB test scores.

But I have some questions about this much-quoted and much-derided statistic. Could it not be a fallacy itself?

For one thing, what’s the scale of “good drivership”? For the 90% figure to be wrong, there would have to be an objective rating scale of good drivership that would allow you to place all drivers along it. What counts? Accidents? Slamming on brakes? Driving with coffee? Cell phones? Five mph over the speed limit? Six?

For all I know, the original study that created this zombie factoid had such a methodology. I’m too lazy to go find it. But my next question was, were participants in the study asked to rate themselves with this scale, or were they asked to decide where they would fall on such a scale without being given any details? I suspect the latter, which is actually okay as long as then they were asked to fill out some kind of survey which then would place them accurately along the scale.

Because, and this is the important part, I think that the unexamined assumption of the 90% deal is that drivers are distributed across this scale in a normal curve, i.e., 70% of drivers cluster around the 50% mark, and the rest of us are strung out on either side, with very few awful drivers and very few perfect drivers, just like IQ.

However, I doubt that. I think it entirely possible that we are not distributed in a normal curve, but a J-curve, and in fact most of us are better drivers on an absolute scale than not. In other words, while the researchers were applying a norm reference, the real world is working with a criterion reference. Most people quite rightly examine their driving habits and say, “You know what, I’ve never had an accident, and I can’t remember the last time I even slammed on my brakes or went down a one-way street the wrong way. I did cut that guy off on the interstate last week, but he was a bad driver.” And so they rate themselves “above average,” because in fact they are.

Okay, so maybe it’s not a J-curve, which would mean that most drivers are approaching perfect, but I’d be willing to bet that it’s at least a shifted normal curve, with that main 70% clumped around the 80% mark and not the middle.

Damned statistics.

3 thoughts on “Update, plus the Good Driver Fallacy

  1. I was just thinking about this very thing in my car during my commute this week. I think the same can be said of movie theater patrons. Everyone hates the moron who can’t stop texting during the movie, or even answers their phone, and yet people do it. I think that truly the vast majority of people are pretty good drivers and good movie watchers, but the people who are bad at it are so agitating that we tend to magnify the problem in our head.

    I also would like to see some kind of driving ability plot just to compare it to a “size of vehicle” plot. My theory is that the larger car you drive, the more likely you are to be terrible at driving it. It seems like most of the people that nearly kill me drive SUVs and pickups. I was almost run off the road by 4 separate GMC Jimmys in a 2-day span this week.

  2. I’m the first to be critical of scientific thinking when it’s clearly not applicable in a situation, but in this case I think a “scientific” analysis would be informative. This was not the case here, probably; there’s no rigorous testing. Essentially, you are talking about the results of a poll, and a polling of self-characterization at that. Even if you have some “scale” for the pollee to respond to, it’s still subjective. You can lie, after all. Science could formulate some objective way to rate a driver, and those result might be interesting and probably unwelcome to many people.

Leave a Reply

Your email address will not be published.