What is the evidence that mean speed is relevant to accident rates? If accidents were due to random vehicles, then median speed would be a better measure. If the accident causing vehicles, are, say skewed towards the maximum, then mean would not be relevant. In the latter case in particular, the issue may be poor enforcement in existing 30mph zones, rather than punishing all drivers all the time.
Is it really the case that the columns ‘min’ and ‘max’ in Figure 1 are min and max speeds, or, say, 95 centile speeds?
It’s not clear what’s being correlated when you quote the correlation coefficients (R^2)

Note that the Tonbridge schemes attracted an overwhelming rejection by the locals – I don’t know about the others.

@Sam Bailey, I don’t think that you’ve got a firm grip on how to measure, mitigate and ameliorate risk. In the context of road safety, a good starting point is the micromort (one in a million chance of dying from a particular activity), which gives a good framework for comparing different risks. It’s a bit of a blunt instrument (being based on averages, you have to think of the relevant population). However, it’s an approach that’s been developed by statisticians, and I can recommend David Spiegelhalter’s videos on the risk/value trade off of eating a bacon sandwich, or going cycling (DS is an ex-president of the Royal Statistical Society, and Winton Prof of Understanding at Cambridge University). (https://bit.ly/3LJkwSA). In the context of pedestrian and driver safety, here’s a ready reckoner: https://micromorts.rip/, which shows the risk to drivers and pedestrians, which can easily be compared to ‘just living’.

Looking at the ‘ready reckoner’, car users are more at risk from traffic than pedestrians. I don’t think that the need for a crash cage stands up to much scrutiny. Not included in the list is the risk of cycling, which, from memory is ~1 micromort per 20 miles. However, overall, cycling reduces loss of QALYs from fitness degeneration to such an extent that time spent cycling is ‘free’: 15 mins cycling extends your life by 15 mins, even accounting for the risk of an accident.

To take a worked example, I live in a town of 18,000 souls. So the background death rate is around 180 per year. Between 1979 and 2021 (last year of available figures), there were 25 road deaths in 30mph zones, of which 12 were pedestrians. So in the 42 years, there were ~7,500 deaths of whom 12 were pedestrians (0.2%). It is also worth noting that the death rate (and other road accident rates) continues to fall at ~2.5% pa. Is that a big incremental risk? Is there more risk to pedestrians in accepting a lift in a 30mph zone (which is not considered high risk, I don’t think), than walking?

If there is a lot of traffic, as there is at some times of day, then it makes sense to keep the streams separate with specific crossing points whatever the speed limit (most accidents to pedestrians seem to happen where the traffic speed is well below 30mph). If there is not a lot of traffic, then it is not necessary to find a crossing – although one may wish to do that. Children can walk to school at lower risk than being at school: they are very much safer now than they used to be. It may well be that the biggest traffic risk is from the parents of other children ferrying their charges around.

A QALY (Quality Adjusted Life Year) is not a +ve or -ve thing, it’s a measure that can be used to give finer detail when looking at the frequency/impact of hazards.

I doubt that the measures represent instantaneous values during acceleration – although of course, they could – for two reasons:
1/ methodologically, it doesn’t control for the rate of acceleration, so the numbers would have very large CIs
2/ the paper includes the following description:
“4 Consumption Model Description
4.1 Average Speed Consumption Model Adjusted with Road Gradient
Average speed is a key factor in vehicles’ consumption and emissions output. Several studies have found that the relationship between the specific consumption and vehicle speed is U-shaped (Andre and Hammarstrom 2000; Smit et al. 2008). Therefore, it is possible to develop speed-based consumption factors (grams of fuel per kilometer) simply by using traffic flow records as input data for the emission model.”

I’d be interested in the research that you cite that finds 20 and 30 have similar economy levels for ICEs. I find that I’m usually in a different gear at the different speeds, so, assuming little difference in the engine efficiency, if the rpm is similar, then there ought to be a difference in gearing (m/revolution of the engine.)

I’m not sure that the reductions quoted by the different studies make sense. To quantify the discussion below, from a year ago, the annual reduction in accidents in 30mph speed limit areas is 2.9% (based on STATS19 data). This is in line with the changes quoted above, where a whole impact is described. Some of the examples constrain the impact to the area where the new limit was introduced, thus the impact of behavioural changes, which could reduce traffic levels, would also have an impact.

It’s also worth noting that the original TRL reports (esp. 421) focus on the impact of speed change on accident rates, not speed limit. It also notes that a larger impact comes from changing the variation in driving speed on a given road. In fact, the reductions in accident rates observed between 2005 and 2019 exceed those projected as possible by reducing the speeds in 30mph zones across the country in TRL421, although hardly any changes were made between the report and now.

Much of the original data comes from non-UK roads from late last century. The WHO comments also note that road accidents are mostly an issue in the developing world.

(numbers and analysis available on request)