Originally Posted by
genec
But even in the Cross Fisher studies the information was based on what police had gathered from witnesses.. if the only witness is a motorist and dead cyclist, how accurate is the data contained in the police report... sure the motorist said the cyclist was riding the wrong way... or swerved or came flying out of the intersection... and that is what the police wrote... and that is what Cross Fisher cataloged, but if the only witness was lying in the first place... the resultant study is going to be skewed.
Velo news even reported that there was a bias of police tending to question motorists while ignoring live cyclists... again if the motorist is seeking to cover their behind, the resulting report will tend to convey the data the motorist wants the police to believe, regardless of the truth.
One can be as objective as possible while compiling the data, but if the data is based on bias and lies, then the outcome will tend to reflect that input. (Garbage in ----> Garbage out)
I am sure the review process did verify that Cross and Fisher were objective about their methods of cataloging the data... but again, if the data were flawed...
With regards to accident studies in general, I would be pretty surprised if there was wide-spread general disregard for the truth among the police. That is, I believe that their is an anti-cycling bias but that it is probably more of a subconscious bias rather than an active one. Roughly speaking, these accident studies produce results that are in the same ballpark of each other. Now there is a huge identification issue going on, but if this bias is so important small changes in police attitudes should make a big difference in the accident studies.
One can also think about the bias itself and what it means for the analysis and model fitting. My guess is that drivers probably stretch the truth more than make-up fairy tales. For example, a driver would probably say that the cyclist failed to stop at the light/intersection than completely fabricate a story where the cyclist was riding on the wrong side of the street.
Anyway, this doesn't change the fact that exposure in many of these studies is unobserved ... although I have seen a reference to one that did ... such that the probabilities/rates that we would want to know are unknown.
I still don't know why, however, one would think that U.S. accident data is so terrible such that zero inferences can be made while the data about the glorious and safe cycling in Europe is unaffected. If anything, it would seem as if a European driver has a greater incentive to lie -- since they have increased liability -- after an accident than a US/British driver.