Does a 1mph reduction in speed really reduce accidents by 5%?
A key phrase in the anti speed campaigner's handbook that appears again and again is
“A 1mph change in average speed causes a 5% change in accidents”
It is quoted ad nauseam is most Government anti speed publications. Now to any reasonable human being who knows anything about driving, this statement is clearly absurd, yet it alone is used to justify many of the obstructive and oppressive measures being taken against safe drivers. An ABD examination of it's origins and a demolition of the arguments used to justify it is therefore long overdue.

It originates from a study published in 1994 by the Transport Research Laboratory called "Speed, Speed Limits and Accidents" (ref S211G/RB). This document contained no new research, but simply correlates the results of previous studies around the world, which were all involved with measuring the effect of a change in the posted speed limit.

Examples are quoted of accident reductions following the introduction of reduced speed limits going back to the 30mph built up limit introduced in the UK in 1935, and on the effect on average speeds of speed limit changes.

However, the only examples of changes in accidents and mean speeds being compared at the same time are:

A simple graph follows where the change in 'before and after' mean speeds in these six examples is plotted against the percentage change in accidents from one period to the next. A line is then drawn through the points which corresponds to a 5% change in accidents per mile an hour change in mean speeds. Thus the offending sentence is born.



This argument is so full of holes it is hard to know where to begin attacking it. To be fair to the authors of the report, they make many of these points themselves, but, inevitably, these are lost when the politicians are looking for a simplistic solution to a complex and poorly researched problem. They begin by making four fundamental statistical errors which any A-level student should be familiar with.

  1. Is there a scientific causal link to back up the apparent statistical relationship?

    It is incorrect to assume causality from a statistical relationship. To illustrate this, take the assertion that wearing a baseball cap backwards is linked to a 20 point reduction in the IQ score of the wearer. A plausible statistical link. It would be obviously ridiculous to suggest that educational standards could be increased at a stroke by banning baseball caps unless it could be proven that the cap was the cause of the low intelligence rather than simply a symptom. If the ban went ahead anyway, the real causes of low achievement in school would be ignored and some rather more intelligent people would get sunstroke due being forced to venture into the midday desert sun hatless. A ludicrous scenario? It is not far from what has been done to the motorist here!

  2. Is the sample representative?

    It has been assumed that because a graph neatly fits these six examples then the same must be true of all roads. To be valid, a survey must be chosen at random from a representative section of the population. The six studies are all related to speed limit changes which had been imposed for some purpose other than the research, and so most certainly do not come into this category.

  3. Are the same things being measured?

    Since three studies refer to total accidents, two to injury accidents only and one just to fatal accidents, how they can be justifiably plotted on the same graph is beyond comprehension. Also, they relate to different road types — how can what happens in residential Hamburg in 1985 have anything to do with Finnish motorways in 1962? Add to this a myriad of different measurement techniques and other factors and you have an appalling mishmash which says nothing of value.

  4. Is the result consistent with trends in the general population?

    Any survey has to be questioned if it comes up with results that are inconsistent with trends observed in the whole population. These trends show a steady fall in accident rates and casualty rates throughout this century despite huge increases in free flowing traffic speeds. More specifically, injury accident rates fell by 30% in the UK during the 1980s whilst road speeds increased.


If the violation of general statistical rules such as these are not enough, some of the specific logical errors in this study damn it even more comprehensively.

  1. The Irrelevance of using Average Speeds

  2. The Pitfalls of using Changes in Accidents

  3. The Insanity of Combining the Two

    If this 1mph 5% law was true, the average speed on the Spanish motorways would have to be 60mph faster than that in Holland, and the West Germans with would have had to be travelling slower on their largely unlimited Autobahns than the Americans were when their freeways were limited to 55mph. Hmmm!


All of this illustrates that macro statistical techniques have been inappropriately used on subject matter where it is impossible to isolate variables and where the incidents in question are both rare and subject to complex causal factors. The conclusion of the TRL report is therefore as statistically invalid as it is rationally absurd.
Now see TRL511 to read our analysis of the report the government commissioned to try to shore up this absurdity.

"We have now sunk to a depth at which the restatement of the obvious is the first duty of intelligent men."
George Orwell