1 20mph Research – Analysis of speed outcomes in 20mph limit areas using GPS data for 12 case study areas 2nd March 2017 – RSGB Analysts Conference Jane Robinson and Richard Fernandes (Atkins) Atkins, AECOM, Professor Mike Maher 30th June 2016

2 Structure of PresentationStudy purpose and objectives, and overall methodology Existing evidence and wider trends in vehicle speeds Analysis of area-wide speed outcomes using GPS data Questions and discussion

3 Study purpose Study objectivesAtkins, Aecom and Professor Mike Maher were commissioned by the DfT in 2014 to undertake research into 20mph signed only speed limits. Purpose - To address an evidence gap regarding the effectiveness of 20mph speed limit only schemes. Study objectives To evaluate the effectiveness of 20mph speed limits, in a range of settings. To examine drivers’ and residents’ perceptions of 20mph limits. To assess the relative costs/benefits to vulnerable groups e.g. children, cyclists, the elderly. To evaluate the processes and factors which contribute to the level of effectiveness of 20mph speed limit schemes.

4 Case studies 12 case study schemes; generally implemented 2012-20158 large residential area-wide schemes 2 small-scale schemes in self-contained residential areas 2 focused on city centre areas

5 Data sources and analysisStakeholder interviews: National stakeholders (DfT, PACTS, ACPO, ABD, etc.). Local stakeholders (officers, councillors, police, public health, bus operators, interest groups) Social research (attitudes, perceptions and behaviours): Residents and drivers questionnaires (sample = ~3400) In-depth interviews with drivers (sample = ~200) 12 focus groups and online surveys with specific user groups Detailed data analysis (speed, safety and other outcomes): GPS area-wide speed data (tomtom data) Local authority spot speed data, collected using inductive loops, radar devices or similar technology STATS-19 safety data (collisions, injuries, contributory factors) Other secondary data provided by local authorities

6 Analysis of speed outcomesLet me talk you through the methodology and outcomes from our area-wide speed analysis 30th June 2016

7 Existing evidence on speed impactsSmall 20mph zones Two extensive studies undertaken by the Transport Research Laboratory (e.g. Webster & Mackie, 1996; Webster and Layfield, 2003) 20mph zones can achieve substantial reductions in average speed, of around 9-10 mph Small schemes (typically covering a few kms of road length), before speed well above 20mph (typically around 25mph), implemented to address location-specific safety issues 20mph limits Less evidence available (Portsmouth, Bristol, Edinburgh, national trial programme of advisory 20mph speed limits across Scotland) 20mph limits deliver much smaller reductions in average speed, typically around 1-2mph Large scale, lower before speeds (closer to 20mph)

8 Wider speed trends Average vehicle speeds during the weekday morning peak on locally managed 'A' roads (mph) 25.4mph (Dec 2011) -1.9mph 23.5mph (Dec 2015) % exceeding the speed limit on 30mph roads fell from 54% in to 52% in 2015 (Free flow vehicle speeds in Great Britain, DfT, 2015)

9 What is TomTom data? Anonymised GPS data Available at segment levelEach segment gives: - Sample, Speed (avg/med), Journey Time (avg/med), Distance, Every 5th Percentile Speed Sample 50 veh 80 veh 70 veh 60 veh 100 veh Speed 28mph 29mph 40mph 30mph 42mph Our analysis was based on using TomTom data. The data is from all those using TomTom products, whether this be handheld devices bought from Halfords, connected devices inbuilt into cars or the TomTom navigation app. This information is collected, anonymised and made available to query and analyse. This data is made available on road ‘segments’, which is the road network broken up into small sections of 5-200m in length. Each road segment gives information relating to every GPS probe that traversed the whole segment. It provides data such as…. (see slide). Note each segment has it’s own unique sample and statistics relating to that sample, as shown in the diagram

10 Strengths and weaknessesHistorically available Only GPS vehicles (behaviour/affluence?) Whole area, not spot locations Requires full segment traversal High area sample Low individual segment sample Over many days But aggregated days Disaggregation (to periods/segments) TomTom data was one of many ways we could have conducted this analysis, and it is worth bring upfront with the pros and cons of this approach, to help understand the findings. Strengths: Historical – means we can get the same format, and directly comparable before and after data Whole area – means we aren’t skewing to spot locations High sample – means robust sample, likely to identify the true impact Many days – so not affected by seasonality, or not biased to behaviour on a specific day Disaggregation – allows time periods and segments to be examined in turn Is mappable Weaknesses Only GPS – skewed behaviour? Do people drive differently with GPS? Requires full segment traversal – records are not kept unless they drove from end to end of segment – lose cul-de-sacs Low segment samples – compared to spot speeds maybe just 3% of sample per day Aggregated days – cannot filter down to specific days in range chosen Affluence – are GPS users a specific income band, does this affect findings?

11 Methodological challengesWhat is the average speed in an area? (issues of aggregation, samples, distance) What is the right metric for measuring change in speed? (issues of slow moving vehicles) While there is lots of data, which is good, it creates a problem. How do you aggregate the data to create a picture for what is happening on 20mph roads overall? If one segment is averaging a speed of 22mph and another 25mph what is the overall picture? What if they are different lengths of segment? What if they have different sample sizes? What is the right metric for measuring change in speed? -Is it the percentage compliant with the speed limit? If so is that percentage of people or percentage of vehicle kilometres driven? -Maybe it is average speed before and after. But then which type of average? Mean will be dragged heavily by very slow moving vehicles (those looking for parking spaces, stuck in congestion, stuck behind a bus, loading/unloading etc) which don’t really tell us about how the policy is working rather tell us about the nature of the road. Median therefore might be more informative and less skewed.

12 Agreed methodology Use one year’s pre and post scheme dataSplit to peak and non-peak hours Flow and distance weight each segment’s data Use a two core metrics: Median speed Percent of vehicle distance driven compliantly Sample sizes Given the strength and weaknesses of the data and the challenges faced, we proposed the following approach: READ SLIDE In terms of samples, we are confident that this approach gave us a lot of data to analyse, considered almost 2000km of roads, and around 11M vkms of data before and 14vkms of data after Distance Before – Peak VKMs Before – Non-Peak VKMs After – Peak VKMs After – Non-Peak VKMs 1,912km 3.2 million 8.4 million 4.1 million 10.1 million

13 So what have we learnt? .

14 New 20mph limits in Residential areasCumulative speed distribution, overall (PROVISIONAL RESULTS) 85th Before = 28.2 mph 85th After = 27.0 mph Diff = -1.2 mph . Median Before = 21.1 mph Median After = 20.5 mph Diff = -0.7 mph Before 30mph compliance 91% After 20mph compliance 47% This graph shows the cumulative speed distribution across all study areas for new 20mph, signed only, roads. These are roads where the pre-scheme limit was 30mph, and is not 20mph. The distribution tells us for any speed, what percentage of vehicle kilometres driven on these roads was at or below this speed. From it we can see that: The median speed prior to the scheme was 21.1mph, far below the 30mph speed limit This has dropped just 0.7mph to 20.5mph after In terms of compliance, 91% of vehicles were compliant with the pre-scheme 30mph But only 47% are compliant with the post-scheme speed limit of 20mph. (44% were below 20mph before, so not much change) There is some change though. There is more change at higher speeds than lower speeds. The 85th percentile speed has reduced from 28.2mph to 27mph

15 New 20mph limits in City Centre areasCumulative speed distribution, overall (PROVISIONAL RESULTS) 85th Before = 25.4 mph 85th After = 23.8 mph Diff = -1.6 mph Median Before = 18.0 mph Median After = 17.1 mph Diff = -0.9 mph . Before 30mph compliance 97% After 20mph compliance 65% We also looked at the impact on city centre roads (with much lower samples) which showed a similar result. The main result is that speeds were lower anyway, and that the change in median is greater, at 0.9mph reduction.

16 Other key findings (PROVISIONAL RESULTS)There was no clear difference between peak and non peak findings (journey purpose) Roads with higher pre-scheme average speeds showed greater reduction in speed when 20mph introduced Class of road is a large factor in the speed drivers choose to travel Pre-existing traffic calming increases 20mph compliance to 67% There is no evidence of speed displacement impacts

17 Other areas of researchPerceptions and behaviour Transport outcomes – spot speeds, casualties, perceptions of the quality of the environment for walking and cycling, increase in active travel, displacement if traffic, etc. Wider impacts – health, noise and air quality, community benefits, etc. Adverse / unintended impacts

18 Questions and discussion30th June 2016