DCCommuter

For over 30 years, researchers have been trying to show that bike lanes have safety value. It has never been done. Why?

Researchers usually cite the lack of available data, but that's only half of the picture. Let's look at how an experiment would be designed if we had perfect data available about exposure and location of accidents. According to the National Personal Transportation Survey, Americans travel about 4 billion miles by bicycle a year. Moritz in 1996 found that 5% of bicycle travel is in bike lanes. I don't think much of his survey, but I'll buy that number. About 700 cyclists a year are killed in the US.

Given those numbers, our null hypothesis would be that bike lanes are no safer than other parts of the roadway, so we would expect bike lanes to account for fatalities proportionate to miles traveled, 5%, or 35 fatalities a year. Here's the key question in experiment design: How many accidents a year would there have to be in order for us to conclude that bike lanes are statistically significantly safer? Would 34 be enough? 33? 32?

In academic research a confidence interval of 95% is required in order for a result to be considered statistically significant. This is associated with a result that is three standard deviations from the null hypothesis. In our case, the standard deviation would be expected to be 5.9 accidents/year. In order to conclude that bike lanes are safer, we would have to see 17 or fewer fatalities per year -- less than half the expected rate. In order to conclude that bike lanes are less safe, we would have to see 52 or more fatalities a year.

Bike lanes may be safer and they may be less safe, but the effect if any is small. Certainly, if bike lanes were twice as safe -- or 50% more dangerous -- we wouldn't be having this discussion. Since the effect is small, and the rate of accidents is so low, the effect on safety is not measurable.

Anyone who makes any safety claims about bike lanes is playing loose with the facts.