@ Cody, Lies Damned Lies and Statistics
"If you look at the relation between death and speed it is roughly as follows. At 40mph 90% of struck pedestrians die. At 20mph 90% of struck pedestrians live."
That's actually the whole core of the problem. It's the reason that whenever a politician tries to improve the fatality statistics they think the answer is to reduce speed. Then they scratch their heads as to why the stats didn't improve much, so they reduce it some more, add some more penalties on etc.
If they'd looked at the two graphs that were merged to form that one graph they'd realize the problem with that graph.*
Even if speed had ZERO effect on fatalities, the way the graph is constructed means speed will still APPEAR to be the major cause of fatal accidents, and common sense tells them that a faster car has more momentum, and this tends to confirm that view.
So that graphs exaggerates the perception that speed is the cause, and it distorts the fix to be always about reducing speed.
* Basically they took 2 graphs, non fatal accidents vs speed and fatal accidents vs speed.
* The non fatal accidents are far far more common and occur mainly at slow speeds.
* The fatal accidents are far rarer, far *flatter*, and drop off at very low and very high speeds. The low speed drop off is common sense (slower car less momentum etc.), the high speed is most likely because faster roads are better separated from pedestrians so fewer chances to hit them on each km journey.
* By flatter, I mean that small changed in the speed, have small effect on the fatality rate.
* They calculated a ratio of fatal to non fatal, (they merged the 2 graphs) and it shows a sharp shift to fatal above 30mph.
* Hence it's 30mph for a reason.
* Yet if you drew a horizontal line through the fatal graph, then constructed the merged graph it would also show the same effect. The ratio would flip from non fatal to fatal above 30mph.
* Yet that fatalities graph is horizontal and speed therefore had no effect on fatalities in that scenario.
* The graph misleads them. They can see fatalities drop by a small amount, of say 10%, but the merged graph says it should be 70% so they think the speed limits must be continually broken in a clever undetected way. Hence they try smarter average cameras to catch these super sneaky speeders.
* Yet if they looked at the unmerged graph they would see it is expected to drop 10% not 70%.
* So they think that a tiny increase over the speed limit makes a big difference, yet it doesn't make a big difference.
* The same one faulty study contaminates road safety research and is quoted and requoted as the definitive study. Yet it is flawed.