In early 2013 the City of Melbourne removed a general purpose traffic lane in each direction of La Trobe Street and installed a protected bicycle lane alongside the kerb and parallel parking between the protected bicycle lane and remaining traffic lane. The bicycle lane extends, with a few exceptions, from Victoria Street to Spencer Street over a distance of around 1.8 km. Prior to the construction of the protected bicycle lane cyclist treatments varied from a conventional painted bicycle lane between kerbside parking and the general purpose traffic lane, a clearway bicycle lane against the kerb (with parking in off-peak periods) and no dedicated provision.
Figure 1: Typical provision along La Trobe St prior to treatment
Figure 1: Typical provision along La Trobe St prior to treatment
Figure 2: Typical provision along La Trobe St after treatment
Figure 2: Typical provision along La Trobe St after treatment
Along with the Albert Street bicycle lane nearby1, it has been claimed that the bicycle lane has increased cyclist injuries (for example: here). Let’s have a closer look at this claim. And further, let’s have a look at the related but unstated implication that the crash rate has increased - that is, the likelihood of being injured has increased. If true, the latter would be a profound finding given that one of the main arguments for these types of treatments is to improve safety. So it’s worth investigating.
Data
It’s well known that Police-reported crash statistics heavily underreport cyclist crashes, particularly those of lower severity. Moreover, there are problems in the definition of “serious”" injuries in particular - an injury is classified as serious if a person is sent to hospital. Once at the hospital they may be examined in the Accident & Emergency Department and sent home without being admitted, or they may be admitted and require major surgery. That’s a very broad range.
Despite these challenges the Police-reported crash statistics are about the best data we’ve got on road trauma. The data was the latest available on 12 December 2017.
Methodology
We’ve used the crash data for La Trobe Street from Victoria Street all the way to Spencer Street. We’ve then extracted all data from January 1, 2006 to December 12, 2017. We’ve further assumed the bicycle lanes were fully open on June 1, 2013 and that consutrction occurred during the first six months of 2013 such that the before period extends from 2006 to 2012 and the after period from June 2013 to 2017.2
We are dreadful at counting bicycle riders just about everywhere, and this applies especially to projects such as the Latrobe Street bicycle lanes. There are no automatic counters, and so we have no reliable indicators of exposure. As such, it is impossible for us to simply divde the number of cyclist crashes by the number/distance/time ridden along La Trobe Street before- and after the bicycle lane opened.
In the place of exposure we need a proxy. A simple, but incomplete, approach is to look at the crash trends along competing routes as well as La Trobe Street. In the case of La Trobe Street there are a few east-west alternatives, such as Victoria Street to the north and Lonsdale Street, Bourke Street and Collins Street to the south. All could conceivably be substitutes for La Trobe Street, but Bourke Street is (technically) closed to bicycle riders at the Bourke Street Mall and Victoria Street is a particularly hostile arterial road. For the purposes of this analysis we’ll look only at Lonsdale Street, which has a peak period bus lane and no dedicated cyclist provision. But it seems like a reasonably attractive alternative for confident riders, despite the presence of buses.
Results
Has the bicycle lane increased cyclist crashes on La Trobe Street?
Within the confines of La Trobe Street we can answer this one pretty definitively: yes, there is clear evidence to suggest that cyclist crashes have increased after the bicycle lane was installed. On an annualised basis I count 1.2 cyclist-involved crashes per year before treatment increasing to 8.6 per year after treatment. That’s a massive seven-fold increase.
Now, we may ask what types of crashes have increased. To do this I’ve looked at the Definition for Classifying Accidents (DCA) codes which VicRoads use to classify road user movements at the time of a crash. These are shown below, and suggest that much of the increase appears to have come from left turn side swipes and right through collisions. Here’s a couple of notable statistics about these two DCAs from this sample:
18 of the 22 left turn side swipe crashes (81%) involved a motorist turning left across the path of a bicycle rider, and
7 of the 8 right through crashes (88%) involved a motorist turning right across the path of a bicycle rider.
Figure 3: Cyclist crashes along La Trobe St by time and DCA
We can express this increase in crash types on an annualised basis as shown below. Again, it’s clear we’ve very significantly increase the rate of left turn side swipe and right through crashes. The only crash type to have decreased (indeed, been eliminated) is collisions with parked car doors. This is clearly a good thing, and is as expected given that we’ve “flipped” the position of parked cars and bicycle riders. However, overall we can only conclude that the bicycle lane has increased the rate at which riders are being struck by motorists turning across their path. Clearly, that is not a good thing.
Figure 4: Cyclist crashes by period and DCA
But before we jump to the impulsive conclusion that this can only be due to flaws within the protected bicycle lane design we need to consider another possibility: perhaps the bicycle lane has attracted riders from elsewhere, such that the overall crash rate has increased to a lesser extent, or indeed has in fact decreased. We’ll ignore the possibility that the bicycle lane has attracted all-new riding trips (i.e. diversion from other modes such as public transport or driving), as we have no empirical data on how much this may have occurred.3
This is where our comparator comes in - this this case, of Lonsdale Street. If we again extract all crashes along Lonsdale Street involving bicycle riders before and after the bicycle lane was constructed along La Trobe Street and then annualise the crashes we find the result below. This is a stark result; we’ve seen ai incrase from 1.2 to 8.6 crashes/annum along La Trobe Street but a decrease from 5.8 to 3.2 crashes/annum along Lonsdale Street. Now clearly all sorts of things have been going on along Lonsdale Street between 2006 and 2017, but the road layout has remained essentially unchanged over that period.4 It seems plausible, but certainly not proven, that the presence of the bicycle lane along La Trobe Street has attracted riders that otherwise would have ridden along Lonsdale Street, and in so doing decreased the crash counts along Lonsdale Street while increasing those along La Trobe Street.
Figure 5: Annual bicycle rider injuries before and after treatment
Now we can’t be definitive about this of course, but I’d argue that looking at the crash history along La Trobe Street in isolation, and particularly without adjusting for exposure, is misleading. It seems to me that one could fairly reasonably argue that around a third of the increase in crashes along La Trobe Street can be attributed to riders shifting from Lonsdale Street. This still leaves the issue of the other two thirds, and it seems undeniable that there is a safety issue involving motorists turning across the path of cyclists at driveways and intersections that needs addressing.
As a footnote, the average annual number of cyclist injuries across the City of Melbourne from 2006 to 2012 was 214 per annum, and since June 2013 has been 245 per annum. So there appears to have been an increase in cyclist injury burden of around 14%. Some of this can be attributed to La Trobe Street, but presumably there’s also an effect here of increasing cycling activity.
I’ll look at the evidence on this one some other time.↩
According to Nearmap, construction appeared to happen in stages between January and March 2013, so any crash that occurred prior to 2013 can reasonably be assumed to have occurred before treatment, and any crash occurring from June 2013 onwards occurred after treatment. As it turns out there are only two crashes occurring in this six month period for which we can’t be entirely certain about the bicycle lane status (out of 44 crashes overall), so the uncertainty about timelines doesn’t seem to be critical.↩
In an ideal world it would have been great to go out to the bicycle lane some period after construction (say 3 - 6 months) and ask bicycle riders what they did prior to the construction of the lane. These intercept surveys are very easy to do, and I do them as a matter of course as part of project evaluations with which I’ve been involved. Almost invariably one finds that diversion from other modes accounts for less than 20% of riders, with most having already ridden along the road (or other nearby routes).↩
The bus lane was there as far back as Streetview and Nearmap will allow me to go - to November 2007, and so I’m going to assume it was there in 2006 too.↩