The latest 'study of studies' supports speed camera use, but its findings are disputed by many statisticians and commentators. Dick Ahlstrom, Science Editor, reports
Yet another new study promotes the use of speed cameras as a way to reduce road crashes and deaths, but the latest research is unlikely to quell doubts about the technology.
The new report is a "study of studies", a trawl through existing international analysis of the impact of speed cameras on crashes and deaths. It was published last month in the Cochrane Database of Systematic Reviews, which publishes reports on the effects of healthcare policy.
The comprehensive review included the results of 26 separate international studies carried out in countries including Canada, Australia, Britain, New Zealand, the Netherlands, Finland, Hong Kong, Denmark, Norway and Germany.
The authors of this study acknowledge that the methodology used overall in the various studies was "generally poor", but argue in favour of the use of speed cameras given all the studies said more or less the same thing - that the use of cameras helped reduce the number of accidents.
"Despite the methodological limitations of the studies reviewed, the consistency of reported positive reductions in speed and crash outcomes across all studies suggest that are a promising intervention for reducing the number of road traffic injuries and deaths," say the authors - C Wilson, C Willis, JK Hendrikz, N Bellamy.
The confidence placed by these and other reviewers in the statistical certainties is not, however, universal. Many statisticians and commentators have attacked these studies as too weak and flawed. Either the sample size is too low or the study period too short, allowing unintentional biases to creep into the results.
Retired academic Prof John Brignell, formerly of the University of Southampton, has consistently been scathing of studies arguing in favour of speed cameras.
Brignell has made a name for himself as a debunker of bad science and dodgy statistics, but remains controversial given his dismissal of things such as human induced global warming, lung cancers caused by second-hand smoke, and the thinning of the atmospheric ozone layer.
The policy of introducing cameras is strongly backed by the UK Department of Transport and it cites studies of its own that point towards crash and injury reductions where they are used.
Brignell has claimed that departmental reports on the issue represented "a startling example of statistical chicanery from figures of road fatalities issued by the bureaucrats to justify a massive investment in these devices".
He pointed to one study presented by the UK Department of Transport that showed a dramatic decline in deaths, with 49 fewer fatalities between 1999 and 2000 where cameras had been installed. Yet in the year prior to the cameras coming on stream there was an equal fall in road fatalities of 44 "for no apparent reason at all", Brignell notes.
Brignell and other commentators have based their continuing arguments against claims being made for speed cameras on a statistical concept known as "regression to the mean" - the RTM effect.
All this really means is that big changes either up or down seen at the start of a statistical study usually iron out over time, leaving very little real change in things.
Many of the studies presented by proponents of cameras pointed to dramatic falls in accidents and deaths in the first year after installation. Often subsequent years did not show a similar improvement and in time any benefits seemed to disappear completely.
While RTM has often been ignored by groups studying the impact of speed cameras on accident statistics, ongoing and more recent studies still show improvements in accident rates when cameras are used.
The UK Department of Transport cites a four-year independent study it commissioned which showed there was still a 22 per cent reduction in collisions causing injuries after cameras were installed and 32 per cent fewer fatalities where there were cameras.
The new Cochrane review study did keep a close eye on the RTM effect however. They only selected good, quality-controlled "before and after" analyses for their report, all with at least one year's full data and many with two years' data.
The reviewers also stated they were particularly mindful of three "confounders" that might slant their study, "regression to the mean, long-term trends and changes in traffic volumes".
Even so, they still describe their findings as statistically "weak" because the studies included in their work were weak.
"Most studies did not have adequate control or discussion of potential confounders. Only a few studies controlled for RTM, long-term trends in crash rates and changes in traffic volumes," they write.
They argue "the rationale of using speed enforcement detection devices to impose safe speeds and limit maximum speeds to improve road safety is unequivocal". The authors call for better quality studies that could help decide the issue once and for all.