http://www.tandfonline.com/doi/pdf/10.3846/16484142.2014.953205
Link works when connected to McGill network, unsure otherwise.
Seems to be open access.
A few more things:
None of this takes into account the change in traffic density. I think one thing that made a huge difference in safety, including passing distance, was to have an increase in the number of cyclists on the road. Moreover, I don’t know if drivers would tend to avoid such streets if they are not too densely distributed. Also, I think occasional cyclists feel more entitled to bike on the bike lanes, so this encourages more occasionals to bike regularly, thus increasing visibility of the cyclist population.
Anyways, I’d be curious to see how adding a traffic lane changes traffic distribution over time.
All these papers are the same: they can’t accurately model idiocy, and bad design is a major problem.