Friday, September 10, 2010

Statistics Lie

It has been a problem for centuries, but as I get older it becomes more and more apparent to me how often statistics are misinterpreted and used as the basis for invalid points.

Two cases in point, both coming out of Japan.

In conversations with me, smokers for years have claimed that smoking can't be that bad for you because Japan, well-known for it's heavy-smoking population, has one of the highest life expectancies in the world. In the last few months, it has come to light that the Japanese government does not track the date of death of it's residents too well, so their life expectancy stat is skewed by counting people who have long since past away as among the living. For example, this article discusses how there are 230,000 persons listed on their rolls as being over 100 years old who cannot be located. A few that they have tried to track down are long since dead. That'll skew the numbers!

A second example: people I talk with sometimes state that Toyotas are better made than other cars and point to their lower recall rate than other cars. For years, those in automotive circles have known this is because of "secret recalls", where cars are fixed when they are brought in for service without issuing a recall. This can be thought of as a good thing or bad thing: the problem does get fixed (good), but it gets fixed without the (more) unbiased viewpoint of a third party such as the government (bad). But my point is that it skews the recall numbers. Toyota recall numbers show up as artificially low. This of course has been rectified this year when Toyota finally figured out that car problems should be publicly awknowledged and fully discussed with the US Government. And the public is finally seeing what many of us knew for years: Toyotas are no better than other makes. But the public perception was skewed by misunderstood statistics.

No comments: