This report sets out to demonstrate the future potential of the CARE database as a valuable road accident research tool. Using just two of the comparable variables currently available, month and number of accidents, a greater understanding of the underlying structure of each Member State's (MS) accident process is possible. The report is based on monthly accidents between 1991 and 1995 for IRL, GB, NI, DK, NL, P and F which are currently available to IRL on a reciprocity basis. Using the CARE data each MS accident process is modelled using Box-Jenkins time series analysis, while forecasts for each MS are computed for the two years 1996-1997. The highlights of the data analysis is the regular, almost deterministic structure of GB, Nl, P and DK accident data which indicate no observable trend in accident occurrence over the time period studied. The clear downward trend of F and possibly P accidents which are the only data sets to exhibit any consistent trend, is also of interest. IRL and NI demonstrate no consistent seasonality pattern and consequently have poorer forecast precision and, in the case of NI, poorer model fit. (A)
Abstract