Road Traffic Accidents
An experiment was carried out to investigate the effect of speed limits on road traffic accidents. The investigation took place in Sweden during the summers of 1961 and 1962 and lasted for 92 days in each summer. For part of the 92 days general speed limits of 90km/hr or 100km/hr were imposed, for the rest of the time they were not. The number of accidents resulting in personal injuries at the end of each 92-day period was reported to the police. The purpose of this report is to see if, and to what extent the number of accidents is reduced when speed limits are in place.
Several key findings emerged from the exploratory analysis. Firstly, mean analysis on the number of accidents relative to the presence of speed limits suggests that limits do reduce the number of accidents. The mean number of accidents with limits was -18.9 and 18.4 in 1961 and 1962 respectively compared with 23.9 and 22.5 without limits. Looking at these figures there is also an indication of there being fewer accidents on average in 1962 than 1961. In total, across both years, the number of accidents with limits in place was 18.4 per day compared to 23.4 without limits - this is 5 more accidents per day without limits.
With reference to the scatter plots there appears to be no link between the day number (1-92) and the number of accidents. The plots are randomly scattered and suggest no trends or correlations.
The Poisson modelling procedure provided more detailed results. The two best models were models 7 and 8 (see table 2, p.10). Model 7 considered the factors Day, Year, Limit and the interaction term Year*limit. The terms limit and year*limit were highly significant with p-values of 0.03 and 0.02 respectively. These low p-values represent the probability that the observed relationship between variables occurred by chance hence suggesting the terms are important. The model implies that speed limits reduce the log of the expected number of accidents by 0.15 and that year*limit reduces it by a further 0.25.
Model 8 considers the factors Day, Limit and the interaction term Day*Limit. The term limit has a t-value of -2.13 and a corresponding p-value of 0.02, both are highly significant and suggest limit is an important factor in determining the number of accidents. The coefficient for ‘limit’ is -0.84 suggesting that having a limit reduces the log of the expected number of accidents on each day by 0.84. Both models are suitable, but model 8 is a better model in terms of its significance and in particular its residual deviance (discussed later).
4. Exploratory Analysis
4.1 Exploratory mean analysis7
4.2 Box plots7-8
4.3 Scatter plots8-9
5. Further analysis - Regression Modelling9-12
The debate about the usefulness and effectiveness of speed limits on reducing road accidents is ongoing. A road speed limit refers to the maximum speed for vehicles allowed by law and in Sweden, where this study took place, and the UK these are very strict. National government believe there to be a strong connection between speed limits and road safety. Between 1990-1995 Sweden reduced its speed limit and at this time it was the lowest in Europe – they cited environmental rather than safety reasons for this. This limit was reversed shortly after as it was unpopular and seldom obeyed. Conversely the German motorway equivalent – the autobahns have no speed limit, only a recommended limit. However an increased liability in the case of an accident may result from exceeding the recommended limit.
A road safety...