From what I can see, that study refers back to the “equity” study and that is where the definition of boundary roads comes from.
In more general terms, I don’t really get why boundary roads aren’t generally included. LTNs decrease traffic inside LTNs. That seems pretty self explanatory. On that basis, their impact on surrounding roads seems to be the most pertinent issue and goes to the heart of the division in the community. The reason why some people are so upset about this is that they don’t get the benefit but have to deal with the downside. I think it is this lack of equity which people find most frustrating.
So my reading of it all is this:
The equity study mapped out areas of London which had had LTNs put in between, from memory, March to Sept 2020 and were still in place in Oct 2020 (I stand to be corrected on those dates but essentially they look to find all the ones recently implemented that hadn't been removed by the time they did that study, then excluded LTNs which did not have pre-pandemic data to compare to).
They created three sets of roads:
1) Inside LTNs
2) LTN Boundary Roads
3) Roads unconnected to an LTN
The equity study sought to answer the question "who lives inside the new LTN areas?" and as such it's obvious why they wouldn't need to consider boundary roads or unconnected roads and excluded them.
The safety study took all three sets of roads and examined all of them to see how road safety has been affected.
The tweet
liquidindian has found refers to the data they used on traffic levels, which that guy says excluded certain roads. Looking at the table in the supplemental information he obviously knows where those numbers come from and that they aren't complete.
Thing is that it doesn't matter to their main claim which is about absolute numbers, not risk per trip. In their methodology they are clear about how they do this:
2. Methods
We used police injury data, which gives information on the travel mode and injury severity of road traffic injuries, plus detailed geographical coordinates for the crash location.
[2] Our primary outcome was number of injuries of any severity, both in total and by mode of travel. We present secondary analyses examining killed or seriously injured (KSI).
We used information from a range of official sources to map all new modal filters implemented from March-September 2020 in London and still in place at the end of October 2020.
[3] Based on these we manually mapped 72 LTNs and surrounding boundary roads (details in Aldred et al., n.d.: see Figures
1 and
2). We aggregated the point locations of all injuries into three mutually-exclusive groups:
- Injuries inside the LTN, defined as injuries at least 25m inside the LTN boundary.
- Injuries on LTN boundary roads, defined being located less than 25m from an LTN boundary road.
- All other injuries elsewhere in London (our comparison group).
We made pre/post comparisons of injury numbers in these LTN groups between October-December 2018 and October-December 2019 (‘pre’) versus October-December 2020 (‘post’). Using Fisher’s exact chi-squared tests we compared trends inside LTNs (group 1) and on boundary roads (group 2) to trends in injuries elsewhere in London (group 3).
So what that says to me is that to get the number of injuries, they have used all three sets of roads from the equity study. This is not the same set of roads that was used for the traffic data. There is no evidence that any LTN boundary roads are excluded from this data set, in fact the equity study says it includes all roads within 25m of an LTN implementation.
The numbers they give, from which the headline is derived, takes these roads and uses GPS data to identify which injuries happened on which roads. This is the only hard data they give.
Then they give their estimations on risk per trip, which uses the traffic data set and judges by LTN as a whole. Now the guy in the tweet knows where this traffic data came from and that it excluded certain roads and traffic is higher on those roads than this study thinks it is. I agree with
liquidindian that means risk per trip has actually decreased, since the absolute number has remained the same.
Now I agree that all boundary roads should be included in a study which includes any boundary roads (unless it's look at a very specific situation obviously) and as far as I can see, this one did when it looked at the absolute number of injuries.
(for the set of LTNs that were in place recently, still in place and had comparable data from pre-pandemic times). The traffic data they used to make an estimation of risk per trip didn't and it obviously should have, assuming they have data from pre-pandemic times to use as a baseline anyway. But that doesn't make any difference to the absolute numbers which the headline refers to.
www.forbes.com
