By Jennifer Dill, Ph.D.
Professor, Urban Studies & Planning
This week I’m at the International Travel Survey Conference in Australia. The conference happens every three years, attracting over 100 geeky people who spend time thinking about things like stated preference experiments, smartphone data collection, combining sampling frames, and respondent burden. I presented some work from our five city Green Lanes project, comparing our survey data with “objective” measures, such as videos and traffic counts. The focus was on intersections, where the protected lane is no longer separated from motor vehicles. An example of one design used in Portland, OR is shown in the adjacent figure.
Some of the comparisons are included in our report on the project. For example, we found that when asked on our intercept survey of Portland bicyclists where they should ride at the intersection shown above, only about half got the right answer (over the sharrow bike symbol). But, when we looked at the video, nearly all of the bicyclists rode where the engineers intended, at least when there were no cars in the turning lane. On the other hand, at some locations (particularly in San Francisco) drivers chose the correct location on a survey, but the cars we watched on video didn’t get it right.
We also asked bicyclists how safe they felt at the different intersections in the five cities. For the conference, I compared cyclists’ perceived level of safety with some objective measures to understand what might be influencing these perceptions. While we only had data from six intersections in San Francisco, Washington DC, and Portland, the findings were revealing and can help in future research.
Perceptions do not necessarily match “reality”
There was no meaningful correlation between the observed conflict rate and cyclists’ perception of safety. It appears that perceived safety is not correlated with observed safety, at least not using the conflict measure we developed. However, there were very few conflicts observed, and all were minor in nature. More hours of video footage over a longer period of time might be more valid as a measure. Moreover, conflicts are a surrogate for crashes. Longer term studies are needed to explore whether crash rates correlate with safety perceptions.
The designs might be working
Observed conflicts are somewhat negatively correlated with the shares of motorists and cyclists using the intersection design as intended (see table below). In other words, as more turning motor vehicles and through bicycles use the intersection treatments as intended, conflicts go down. That’s good news for the engineers’ designs.
What influences safety perceptions?
Bicyclists’ perception of safety are negatively correlated with the shares of vehicles and bicycles using the lanes correctly. In other words, as the share of road users complying with the intent of the design increases, safety perceptions go down. This seems counterintuitive. However, as the number of turning vehicles goes up, the perceptions of safety go down (see figure below). It appears that the negative effect of the absolute volume of turning vehicles is more important than the share of vehicles using the intersection correctly.
This may reveal that bicyclists see every motor vehicle as a potential threat, regardless of whether they are driving where they should be or not. It would be useful to revisit this question a year or more later; our survey took place only a few months after installation of the lane. Do bicyclists’ perceptions adjust over time as they experience the intersection more? In addition, protected bike lanes are new in the U.S. and bicyclists may assume that many motorists will not know what to do. Comparative research in cities that have extensive networks of such lanes (e.g. Copenhagen) might provide insights here.
Bicyclists’ perception of safety is also positively correlated with the width of the vehicle turning lane, indicating that more space can influence safety perceptions. As with most research, there are some limitations. We only had six intersections with both survey and video data, located in San Francisco, Portland, and Washington, DC. But, these findings provide a starting point for future research.