Tuesday, August 23, 2011

Correlation is not Causation

The BBC reports here on a study into cycling. I'm all for cycling and pedal my way to work as often as I can, though that's markedly less often in the winter. The article mentions the contribution that cycling makes to the economy, which I'd imagine to be absolutely negligible compared to cars. Then, when it turns to the more important health benefits, it says this:

"And [the report] says a 20% increase in cycling levels by 2015 could save millions of pounds in reduced congestion, pollution levels and NHS costs.
The report says that regular cyclists take 7.4 sick days per year, compared with 8.7 sick days for non-cyclists, saving around £128m through reduced absenteeism, with projected savings of £2bn over the next 10 years.
"

Now it's true that cycling is exercise and exercise helps keep you fit and healthy, but it doesn't follow that cyclists take fewer sick days because they cycle. It's quite likely that those who choose to cycle are already relatively young and fit, compared to those who don't, so we'd expect them to take fewer sick days. In other words, the correlation here may simply be a case of selection bias.

No comments:

Post a Comment