Since 2002, road safety policy has placed great emphasis on improving drivers' control of speed and in particular on increasing the number of speed cameras in place, both fixed and mobile. It is difficult to assess the impact of mobile cameras as the effect is diffuse and not confined to one place. On the other hand, the effect of fixed cameras can be quantified more precisely by studying differences in the number of accidents at the sites where they have been installed compared with sites which are not equipped. This is the aim of this article. We propose to measure the local effect on road accidents of automatic fixed speed traps installed in France since 2003. Accidents were measured in each municipality on a monthly or quarterly basis from 1998 to 2007. We used a very flexible semi-parametric model describing the occurrence of accidents and the effect that the speed traps could have on them. The model parameters were estimated using the generalised method of nonlinear moments. We show that the installation of fixed speed traps in municipalities with fewer than 6,000 inhabitants reduced the number of injuries on the roads by 13 to 17%, and the number of deaths by 50 to 75% in the first months after they were installed. This effect decreased after six months, while still remaining significant: an 11% decrease in road accidents and a 25% to 50% decrease in deaths, compared with the situation without a fixed camera. To our knowledge, the decrease in the impact of speed traps over time has rarely been observed in the international literature, which usually looks at the short term consequences. Lastly, when the results over the period 1998-2007 in this study are extrapolated through to 2011, the installation of fixed cameras in municipalities with fewer than 6,000 inhabitants would have prevented about 740 deaths, 2,750 serious injuries and 2,230 accidents between 2003 and 2011.