Improving the evidence base on journey time reliability on the Trunk Road Network in Scotland
7. Implications for STAG Results from the review of current best practice and the survey of companies
7.1 The Current Position
Reliability impacts form part of the Economy criterion in the Scottish Transport Appraisal Guidance (STAG). Within the Economy Criterion, two separate analyses are undertaken, a Transport Economic Efficiency (TEE) analysis which is a cost-benefit analysis, and an Economic Activity and Location Impacts (EALI) analysis. Improvements in journey time reliability can contribute to both analyses.
A fundamental aspect of predicting the impacts of reliability is modelling the supply side - i.e. how journey time reliability will change in relation to changes in transport quality/infrastructure. STAG currently recommends the use of the ARUP INCA computer model for dual carriageways and inter-urban motorway. It also recommends the use of a model that predicts changes in the standard deviation of travel times from changes in expected travel times in urban areas. STAG contains no guidance for single carriageway rural roads. Some of the Scottish Government's stated investment priorities include upgrading such roads, including the A9 Perth to Inverness and the A96 Aberdeen to Inverness. Both of these roads suffer journey time reliability problems. INCA is also only appropriate for dual carriageways and inter-urban motorways that operate below capacity. As some dual carriageway/inter-urban motorways in Scotland already operate at or above capacity in peak times, and more capacity problems are expected in the future this presents some difficulties in appraising the impact of better reliability on dual carriageway/inter-urban motorways currently operating or expected to operate at or above capacity.
Within the TEE the value of reliability is determined using the reliability ratio (see section 4). For car journeys, this ratio relates the value of an increase in the standard deviation of journey time to the value of an equally sized increase in scheduled travel time. STAG currently only suggests a single reliability ratio of 0.8. This is for trips by car and is for all journey purposes (business, commuting, other). No recommendations are contained for freight vehicles, either LGVs or HGVs. For public transport effects, STAG defines a reliability ratio, but offers no recommended value. also Instead, it offers the analyst the possibility of calculating reductions in average lateness, with the current evidence base on the value of being late contained in the rail Passenger Demand Forecasting Handbook (ATOC, 2013).
7.2 Implications for STAG from the Review of Current Best Practice.
Current best practice was discussed in Section 4, along with the theoretical background. The results of the Expert Workshop held in 2004 in the Netherlands were known to the writers of the current version of STAG, and were appropriately incorporated. Since that time, no equivalent consensus of view that has advocated changes from those values. Some additional evidence is presented in Section 4, and this will be borne in mind when formulating recommendations for values later in this section.
It will have been noticed that the STAG definitions were there reworded, and it is recommended that those rewordings are carried forward in a revision of STAG. In particular, the following material is preferred to that currently in STAG:
The reliability ratio for car travel is defined as:
For example, one might have an estimate of the value of a travel time saving as £6/hour for some group. That means that the value of a ΔT= ‒10 minutes is estimated at £1. One now needs to know the value of reducing the standard deviation of travel times also by 10 minutes. From the literature, the consensus of opinion is that RR is often around 0.8, in which case the value of reducing the standard deviation of travel times would be about £4.80 per hour, and so the value of reducing the sd by 10 minutes would be £0.80.
For public transport however, the RR is usually defined differently. The justification for this is the existence of a timetable. Following ATOC (2013), and working in broad terms, the evidence base appears to say that one minute of average unexpected lateness is valued by passengers as being equivalent to three minutes of scheduled journey time. This value of 3 is referred to as a Lateness Factor. Since some lateness is successfully advertised to intending passengers in time for them to adjust to (egg. turning up later at a station for a train known to be running late, or catching a delayed earlier train running close to the path of the delayed intended train), the recommended average Lateness Factor is 2.5.
i.e. VOL = 2.5VOT
The reliability ratio, for public transport is defined as:
For example, one might have an estimate of the value of a travel time saving as £6/hour for some group. That means that the value an hour of lateness is £15 (using the recommended f value of 2.5). From the literature, the recommended RR value for public transport is given as 1.4. In that case the value of reducing the standard deviation of lateness would be about £21 per hour.
7.3 Implications for STAG from the Findings of the Surveys
Section 5 of the present study has found the best estimate of RR for Non-Freight, say 'cars' to be 0.3. That is right at the low end of values previously found in studies. This may be because Scotland is different, because the respondents were answering on behalf of their businesses (rather than as themselves), or because the question was misunderstood. In any event, the sample was not particularly large nor particularly targeted at estimating this value as its top priority. Accordingly, it would not be wise to do more than report the finding in STAG and suggest that it might be used in sensitivity testing alongside the currently recommended, and internationally supported, value of 0.8.
For Freight the sample was much too small to determine STAG guidance. Currently, STAG has no recommended value for freight. The sample estimate found for RR, 0.48, is of no significance in its own right, but when disaggregated gives the interesting values of 0.21, for those firms engaged in carrying goods, and 0.71 for those firms merely shipping or receiving goods. Since all goods must be shipped and received, this suggests that a composite RR, for firms that ship, carry, and receive goods should be at least 0.71. However, that RR is only so high because not all firms in that group carried freight, and therefore had low VOT due to not directly having to pay drivers' wages and vehicle costs resulting from increases in scheduled journey times. It is therefore felt that the RR value found for all respondents (both by working with average values and by averaging over individual values) is the best guide to the composite RR for the freight movement. That value was 0.48. Were that value to be based on a large representative sample then that value should be included in STAG as the recommended value to use. In the light of the small sample size, however, it is felt right to place more reliance on international findings.
For Road Freight, Table 4.4.1 reported a wide range of estimates of RR, with some disaggregation between Shippers and Carriers. As just discussed, the Shippers will probably have had relatively low VOT values, since they do not immediately bear the costs of increases in scheduled journey times resulting from extra drivers' wages and other movement costs. Eventually, Shippers will have to cover those costs, but they will not accrue immediately to Shippers and so are unlikely to be reported. On the other hand, Shippers might also understate VOR if they do not appreciate costs of unreliability for the Receiver. However, such costs are often reflected in penalties for late arrival, and these will usually fall on the Shipper.
The lower RR values reported in the literature for Carriers and Own Account operators may well reflect higher VOT values resulting from drivers' wages. Receivers' penalty charges for late arrival will probably fall more on the Shippers than themselves. The estimated RR values for this group are therefore probably too low for a complete freight movement. The appropriate value for appraisal probably lies somewhere between the Shipper and Carrier values. Knowledge is not strong, and a range of 0.2 to 1.0 is probably a safe bet to include Road Freight RR in most circumstances. It cannot be denied that that range is centred on 0.6. If there was an absolute demand for a single value then it would be difficult to suggest a better one. More sensible would be to use 2 values 0.4 and 0.8 for sensitivity testing.
Within the EALI component of STAG there are no recommendations or guidance regarding how changes in journey time reliability may impact on the business performance (increased operating surpluses and increased production/employment). The present study has provided no quantitative relationships that might be used to extend STAG guidance in this area, but the questionnaire responses do point to the areas where unreliability has the biggest impact on company performance.
In summary, the recommendation of this report is that the following values be mentioned in STAG:
CAR JOURNEYS: Recommended value RR=0.8. If a sensitivity test value is required, take RR=0.4.
PUBLIC TRANSPORT: Not studied in this study. Note that for RAIL the ATOC (2002) range is RR=0.6 to RR=1.5. Note that the Expert Workshop of 2004 recommended RR=1.4. Note that recent is suggesting splitting by journey purpose, with RR=1 for BUSINESS and RR=0.6 for other modes. This study finds little ground for recommending different values for PT than for CARS, i.e. RR=0.8, with a sensitivity test alternative of RR=0.4.
FREIGHT: From the values available in the literature, supplemented to a limited extent by the results from the present study, the recommended best single value is RR=0.6, but with such uncertainty that a sensitivity test rand of RR=0.4 to RR=0.8 is strongly advised.
7.4 Implications for STAG of the findings from the study of ATC data.
This was investigated in Section 6. Somewhat surprisingly, given various data limitations, it was found possible to substantiate the theoretical conjecture that had been verified elsewhere, that the standard deviation of vehicle speeds was strongly linearly related to the inverse of vehicle speed. Over half of the variation in the former could be explained by the variation in the latter. Since traffic speeds are routinely forecast, the prospect of accompanying such forecasts with estimates of their spread appears feasible. Depending on context, the fit of the relationship should obviously be enhanced by the inclusion of such relevant variables as were available. For example, if forecasting for a road in December, the spread of speeds might be larger than in a neutral month. A model of standard deviation of speed could include not just the inverse of speed but also dummies for month etc. At the present time, the relevant data is not being routinely collected, and any data collection is irksome for various reasons, but this study has demonstrated that the appropriate recording is currently being done automatically at many sites. Should it be desired to calibrate models of the type described, a particular effort would need to be mounted to facilitate that, at not insignificant cost. Beyond that, real time feedback of ATC data might be used to make forecasts of traffic conditions in the following minutes, which might inform actions – including broadcasting useful advice to drivers. Exactly how much of the above is suitable for inclusion in STAG is a matter for its editors.