This paper builds on the results of a 1986 Canadian study which investigated various methods for monitoring VKT (Vehicle kilometres of travel) continuously, using a national sample of road traffic counts. The study emphasized the calculation of sub-estimates of VKT by variables of importance to accident data analysis, such as time of day, season, road type, and even occupant characteristics. The recommended solution involved a sample of counting sites stratified by road and area type, and in some circumstances by flow. It requires the implementation of some new nationally standardized methods for expanding "coverage" counts, which typically operate for only 1- 3 days in the warmer months, to VKT estimates for different levels of aggregation. Sampling and expansion error would be estimated using a new extension of replication techniques. This set of techniques would seem to offer a useful basis for developing international comparisons of VKT. It could improve comparisons between countries with a high degree of national standardization in road traffic counting, such as the UK, and countries where counting varies because it is a sub-national responsibility, such as Canada, the USA or Australia. Estimated costs are modest. With some limitations, a useful national data set appears feasible for Canada based on about 1000 short counts, approximately two-thirds of which could be "recruited" from existing counting programs. Moreover, the approach will take advantage of recently introduced improvements in counting technology, minimizing the use of the more costly observational methods.
Abstract