Synthesis of Safety for Traffic Operations - Transports Canada

Synthesis of Safety for Traffic Operations March 2003
Hauer (1997) has proposed the Empirical Bayes (EB) method of dealing with the RTTM
bias. In brief, under EB the actual crash count of the location is tempered by the mean
crash count for similar locations to produce a better long-term estimate of the safety
performance of the individual location. Application of the EB procedure requires a
relatively sizeable dataset from a number of similar locations and some statistical
expertise. In many instances practitioners may not be equipped to apply the EB methods.
These practitioners have available at least three other options for dealing with RTTM.
They are:
• Random site selection and allocation;
• Random allocation of sites with aberrant crash records to the treatment and
control groups; and
• Application of a correction factor.
These alternatives methods can greatly simplify the safety analysis. However, there are
some ethical and legal considerations associated with the first two approaches.
If a treatment is to be applied to improve safety, ethics and due diligence demand that the
treatment be applied at the locations where the treatment would do the most “good”. The
random selection of sites will result in some sites with “good” crash records being
treated, and some sites with “poor” crash records being untreated. If there is a good faith
belief that the treatment will yield a safety benefit – then this approach is inappropriate.
The above ethical/legal problem is overcome by the second method of selecting sites with
“poor” crash records and then randomly allocating them to treatment and control groups.
This eliminates the concern that sites with “good” crash records will be treated, but
maintains the concern of sites with “poor” safety records being left untreated. In this
instance, the decision to leave some sites untreated is acceptable if:
• There is a good faith belief that the treatment is ineffective (or no more effective
then what is currently in place); or
• Limited resources do not permit treating all sites.
The application of a RTTM correction factor is a method that was developed by Abbess
et al (1981), and is simpler than the EB methods. The correction factor method also
requires a relatively large dataset of locations that are similar to the treated location.
However, the mathematics are more manageable for the practitioner. The correction
factor is calculated by Equation C1.
Page 199