Research on the effect of speed on accidents has been extensively reviewed over the last decade. In this report, we draw on these reviews, looking at the issues where there is a good consensus, those where views differ, and those where views have changed in recent years. In doing so we have looked at some of the original sources on which these reviews are based, many of which are common to several, but we have also looked particularly at what the most recent research has added to this knowledge. Major reviews were published in Australia (1993), UK (1994), Canada (1997), Sweden (1997), USA (1998), New Zealand (2000) and UK (2000). The most recent major research programme was in UK, and was published in full in 2000 and 2002. The earlier reviews therefore did not have access to these data. Early parts of this research were reported in the USA review but only the New Zealand and UK reviews took substantial account of the whole programme. Over all the reviews there is a strong consensus on the evidence of the effect of speed on the consequences (i.e. the severity) of an accident, and that conclusion is well supported by the clearly understood phys ical mechanisms that occur during impact. The relation between speed and accident involvement is very much more difficult to establish. Reviewers in the latter part of the decade have become more convinced of this link, as more data has become available, particularly from the UK studies which have employed sophisticated statistical procedures to identify the underlying relationships. The basis for this greater conviction has partly been the increasing number of studies becoming available, but also the fact that different ways of looking at the link between speed and accidents produce a consistent overall picture. Each type of study has its own methodological difficulties, but their overall strength lies in the fact that all the approaches result in a similar conclusion that accident involvement, and hence the number of casualties, increases with speed, on any particular road or for any particular driver. The differences in the estimates of the size of this effect reflect the different uncertainties introduced by each method. The very simple general rule, that a 1 km/h change in speed can be expected to result in a 3% change in the number of accidents, appears to have fairly general validity as an average estimate across a wide range of studies. The effect size is fairly similar to that implied by the theoretical models developed by Andersson and Nilsson, and in both cases, the results have been confirmed by monitoring the effects of the implementation of speed reducing measures. As the size of effect has been investigated in more detail, there has been increasing awareness that the effect varies with road and traffic type. This variation may also explain some of the differences observed in the results of earlier studies. Differences have simila rly been obtained in the increase in risk for individual drivers who drive at greater than average speeds, but several reviewers have pointed out that the increase in risk from speeding is not dissimilar from that resulting from driving with alcohol levels above the legal limit. Many researchers have found it difficult to separate out the effects of average speed and of speed variation on accident occurrence. Both have been claimed by different researchers to be the primary factor in the relationship between speed and accidents. This difficulty arises because the two measures of speed are themselves related, but in different ways in different circumstances. It is generally now thought that earlier studies, which suggested increased risk resulting from drivers who adopted speeds lower than average as well as for those driving above average speeds, were mainly identifying higher risk associated with conflicts between vehicles continuing along a road at their chosen speed, and those travelling more slowly because they were turning in or out of side roads. Most of the studies of speed and accident involvement are designed to establish the link with all accidents, in order to provide robust statistical proof of the relationship. Theoretical models, based for example on stopping distances, suggest that the effect of speed on fatal and serious accidents will be even higher. The main source of empirical evidence for this comes from before and after studies where speed-reducing measures have been introduced. Where sufficient accidents can be monitored, changes in the number of fatal and serious accidents or casualties are typically 50 to 100% higher than the percentage change in all accidents. For both urban roads and rural roads, relationships have been established showing that speed has a greater effect on accidents on the poorer quality roads in each type of area, with a reducing effect as road quality improves. This effect is in addition to the more well-recognised effect that speed has a lower effect on limited access roads such as motorways than it does on non-limited access rural roads. The range of the effects is broadly similar for urban and rural areas. Australian data suggests a lower effect on rural roads than British data, probably reflecting the typical road characteristics and traffic flows in each country. The understanding that has been gained of the nature of the speed accident relationship can be used in developing speed management policies but these need also to take account of the likely changes in speed that can be achieved, and the existing accident frequency. Changes in speed following changes in posted speed limits are likely to be small. More consistent changes in speed can be obtained through physical measures or through speed camera enforcement. No evidence has been found to demonstrate the effect on safety of differential speed limits for different vehicles. The relevance of these relationships to a specific country will depend on the nature of the roads and traffic in that country, the current level of speeding and attitudes towards speed choice, the potential effectiveness of speed reducing measures, and the overall benefit that could be gained from changing speeding behaviour. A large proportion of Swedish roads are relatively long and straight with low traffic flows and few junctions. Car drivers in Sweden tend to adopt higher speeds relative to the posted speed limit than drivers in Britain and the Netherlands. So far effective measures have not been found to counter this, despite a substantial proportion of the safety benefits in the national safety plan being targeted through reduced speeding. (Author/publisher)
Abstract