The components of induced traffic
Any increase in demand which follows a reduction in price or improvement in level of service could be regarded as induced traffic but the term is usually reserved for that increase which is not explained simply by rerouting. This convention has no logical justification and is the result of the historical accident whereby analysts often had a model which was only expected to predict rerouting and so any other behioural response was "extra". The various components of induced traffic can he very different implications for scheme appraisal. It is therefore necessary, when considering how one might go about measuring induced traffic, to consider all the dimensions of response rather than that arbitrary subset which are not reflected in conventional modelling exercises.
Major problems hindering the measurement of induced traffic
1. Noise and variability in the data
Any measure Some of this is due to measurement errors and some to the inherent variability of the phenomena.The extent of measurement error is generally fairly predictable for any particular survey methodology applied in a given context and so can easily be allowed for. If appropriate equipment is used and survey procedures are correctly adhered to, this source of error can be negligible.
Traffic flows vary from hour to hour, from day to day and from month to month. Some of this variation is associated with predictable cyclical patterns and may apparently be encapsulated in typical daily flow profiles and seasonal adjustment factors (see, for example DoT 1979). In practice, however, each site will he its own cyclical patterns which will not precisely match the national erage figures which are widely used in the analysis of traffic data.
Traffic flows also vary as a function of special events and weather conditions affecting the site. These special factors, if recorded, can help to remove "noise" from the data but there will always remain an unexplained residual variation. The extent of this residual will vary from site to site.
Best practice will, of course, be to remove all the explainable variation and, by examining time series data at the site, to quantify the unexplained residual (e.g., as a coefficient of variation) and to use this to determine confidence limits for estimates of flow. Where this is not practical, coefficients of variation from comparable sites should be used.
2.Attribution of cause
Even if it is possible to deduce the amount of extra traffic associated with a policy intervention, it may not be possible to conclude whether it is caused by the intervention. We may be able, on statistical grounds to rule out the possibility of pure coincidence (or more accurately we will be able to show that the possibility of it hing happened by chance is very small) but it is still possible that the increase may be associated with the specific policy intervention without being caused by it. The real "cause" may be a wider programme of improvements which the specific policy intervention is only one element. Thus, for example, the opening of a new stretch road may be part of a wider programme of n广告ork improvements (e.g., a strategic route) or may be part of a programme of measures designed to boost the local economy -other elements of which might be the provision of development incentives, a relaxation of planning controls or a major publicity campaign. In any such case it would be extremely difficult, unless a suitable control site can be identified and monitored, to establish how much of any extra traffic is attributable to the new link.
The role of models in the measurement of induced traffic
Mathematical models can assist our estimation of induced traffic in various ways; a previous section has outlined the use of matrix estimation software to deduce O-D matrices from traffic counts while Coombe (1995) will review the use of trel demand models to make direct estimates of induc[收起]
