8. Final Remarks and Conclusions
This report has looked at the concept of Journey Time Reliability (JTR) and its valuation. The background is given in Section 3, and a survey of current best practice and appraisal valuations is presented in Section 4. Surveys were conducted of Scottish businesses, both of Freight Users (F) and those not involved with freight (NF). The survey results are in Section 5. In broad terms, the feeling was that JTR had got worse on the Scottish Trunk Road network in recent years, but that this had not yet become a pressing matter for the vast majority of respondents. Detailed results are presented, many disaggregated by location and activity sector.
The topics of JTR and its valuation in the literature were compared with current advice in STAG. The latter seemed well up to date with the state of the art, and no major revisions were suggested. Both in STAG and in the literature, the accepted way of encapsulating the value of JTR changes is via the measure known as the reliability ratio (RR). This is discussed in some detail in the report. It has some weaknesses, can be difficult to calculate, and can only be transported to other times and locations with great care. RR values have been calculated for both the F and NF samples.
For the Freight sample plausible RR values were obtained for all respondents, and separately for those involved in actually transporting the goods and those not. However, the useful sample size was very small and the estimated value varied greatly as assumptions were changed. The disaggregation provided extra insight, and provided estimates consistent with the literature. It was not felt that the survey value was sufficiently well based to affect the value chosen for use in appraisal. If more robust estimates are required, a much larger sample would be needed, probably involving face-to-face surveying using an Adaptive Stated Preference experiment.
For the, much larger, NF sample, a robust estimate of RR was obtained, at the bottom of the range of values suggested by the literature. This may have been because respondents were representatives of firms, rather than individual travellers. In other words, they were being asked to think of the effect of unreliability on others', rather than their own, travel. The low value is, in any event, consistent with the qualitative responses which indicated that JTR was not currently a major problem for firms. In the light of the findings, it was suggested that for appraisal purposes STAG should recommend the use of the central value found in the literature, taken to be 0.8, together with a lower value, 0.4, closer to the value found in the NF survey, 0.3.
The study also looked at the data currently available that could be used to gauge JTR on the Scottish Trunk Road Network, and the findings are reported in Section 6. The most useful source of data was found to be from Automatic Traffic data Collection sites, some of which recorded (and output) 'Vehicle–By-Vehicle' data covering the speed of passing traffic, by vehicle type. Given resources not currently available, this data could be converted to a common coding and aggregated in various ways not currently attempted. The study was able to mount a pilot investigation of this. Amongst other interesting findings, it was found possible to establish an approximate linear relationship between the standard deviation of average speeds, for days taken at monthly intervals, and the inverse of speed. It appears that around half of the variation in that standard deviation can be explained in that way. Since speed is routinely predicted, from speed/flow curves for example, this opens the way to building a forecasting model of the standard deviation of speeds incorporating additional variables such as weather conditions, road capacity, etc. Future appraisals might then be able to estimate both the mean and standard deviation of speeds on links affected by a scheme. Any change in standard deviation could be factored by RR and added to the change in travel time, then valued by using a Value of Time value as at present. There would no doubt be many other uses for such a model, but it must be stressed that it is not yet known how portable such a model would be, so that there might be a requirement to regularly collect input data which is, at present, irksome.