Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
68 This study conducted a nationwide stated-preference survey of shippers and motor carriers to estimate the value of reliability (VOR) for trucking. This value was estimated through sta- tistical models that quantified how respondents made trade-offs between costs, time, and reli- ability. A type of statistical model was used that is exceptionally flexible in being able to capture heterogeneity in the preferences of respondents and control for the unequal impact of unob- served factors on different respondents. The statistical model also reweighted the sample by the commodity and distance shares of truck shipments in the United States to improve the repre- sentativeness of the results. Reliability was measured and modeled as the 95th percentile delay, which was defined as the difference between the 95th percentile travel time and the expected travel time. This measure of reliability has various advantages over other measures, namely, that it is easy to communicate to shippers and motor carriers, has desirable modeling properties, and can be used to calculate the Travel Time Index, which is the measure of reliability most commonly used around the United States. 7.1 VOR Estimates The truck VOR recommended for general planning analyses was estimated to be $160 per shipment per hour of 95th percentile delay, or, equivalently, $9.4 per ton per hour of 95th per- centile delay. Table 4-12 shows how this value varies for different types of shipments. The main conclusion of this analysis is that the VOR depends critically on the ability of shippers and motor carriers to implement contingencies and react to delays. This conclusion leads to the following observations: â¢ Shipments of 500 miles or more have VORs that are 390 percent higher than shipments of 75 miles or less, because the risks associated with shorter shipments are typically less. â¢ Smaller companies have VORs that are 15â75 percent higher than those of larger companies, because larger companies typically have more assets, technology, and scale to mitigate the negative impacts of unreliability. â¢ Shipments to customers have VORs that are 240 percent higher than internal shipments (those within a company), because customer satisfaction and future business could be at stake. â¢ Less-than-truckload shipments have high VORs, particularly on a tonnage basis. â¢ Shipments to intermodal facilities, such as airports or rail intermodal terminals, have lower VORs than the average shipment. Statistical modeling yielded a wide range for estimates of the value of time (VOT), with esti- mates in the general model for the whole sample not being significant. Therefore, it is recom- mended that freight planners instead use the marginal cost estimate published by the American C H A P T E R 7 Conclusions
Conclusions 69 Transportation Research Institute, which for 2017 was $66.7 per hour of travel time. This value is compatible with the VOR estimates recommended by this study. 7.2 Freight Planning Applications The Reliability Valuation Framework was developed to provide guidance to analysts and planners on using the VOR estimates recommended by this study. This framework has the following components: â¢ Measurement. Guidance is provided on how to distinguish between different types of vari- ability in travel time data (idiosyncratic, systematic, and random) and isolates the variability that causes unreliability in shipment delivery schedules. Practical guidance is also provided on how to estimate route travel times from link data, so that reliability can be measured from the perspective of freight users. The approaches recommended are sensitive to the limitations of the travel time data available. â¢ Reliability modeling. Previous methodologies that have been used to model the reliability of the roadway network are described. These methodologies are critical to assessing the impact that projects might have on reliability. Ultimately, a method is recommended that relies on previously estimated statistical relationships between average travel times and 95th percentile travel times in different roadway conditions. Guidance is also provided for analysts and plan- ners estimating their own statistical relationships using local data. â¢ Valuation. Recommendations are provided for how to use VOR and VOT estimates in three common planning analyses: â Benefitâcost project evaluation: An approach is given for considering reliability alongside other important benefits from roadway projects, such as improvements in safety and air quality emissions. This approach considers the full impact of travel time uncertaintyâ which is often underestimated in typical benefitâcost analysesâon freight users. â System-level performance measurement: A methodology for quantifying the costs of trucks moving through the roadway network is provided. This methodology considers the full effect of unreliability. The metrics proposed can be used to track whether the unreliability costs of a region are decreasing or increasing. â Bottleneck identification: An approach that considers reliability is given for identifying the places in the roadway system that are causing disproportionately high costs in the move- ment of freight. 7.3 Lessons Learned in Surveying and Modeling The following lessons were learned in the estimation of the VOR: â¢ Use shorter and simpler surveys. Previous studies relied on stated-preference surveys that took too long to complete given the attention span of industry respondents. These surveys often also asked choice questions that were too complex and involved too many attributes. Most respondents are likely to have difficulties consistently comparing alternatives that involve more than three attributes, particularly if the survey takes more than 15 minutes to complete. The cognitive burden of the survey should be minimized to achieve higher-quality responses. â¢ Describe reliability appropriately in the survey. Most previous studies have described reli- ability in stated-preference surveys in ways that understate the likelihood and consequence of shipment delays. Reliability information should be presented in a format that is clear and familiar to respondents from the freight and logistics industry, such as 95th percentile delays.
70 Estimating the Value of Truck Travel Time Reliability â¢ Estimate the VOR in ways that are consistent with current planning applications. This includes defining reliability in terms of the 95th percentile delay, which is easy to calculate with existing data and agrees with other federal planning metrics that are based on 95th per- centile travel time. Other measures, such as on-time probability and the standard deviation of travel time, would be less useful in the U.S. context. â¢ Reweight the sample to improve representativeness. No previous study had reweighted the sample obtained to improve the representativeness of truck VOR estimates. Doing this, however, is critical ensuring that the estimates can be used in a wide range of planning applications. At a minimum, it is recommended that the sample be reweighted by the ship- ment distance and the commodity shares of the trucking activity being analyzed. â¢ Use statistical models that are flexible and realistic. Effort should be made to use models that do not impose unrealistic limitations on the data, such as assuming that all individuals have the same preferences or that preferences for travel time are independent of preferences for reliability. The ML models used in this study avoid these and other limitations. â¢ Clean the sample of nonserious and illogical responses. Regardless of how carefully the survey is framed, some respondents will inevitably provide useless responses because they either did not take the survey seriously or did not understand the mechanics of the choice questions. Excluding these responses was found to improve model estimates considerably. â¢ Develop survey and models that capture heterogeneity in the VOR. As seen previously, the VOR is highly sensitive to shipment distance, company size, commodity, and type of supply chain, among other factors. Care should be taken to implement surveys and estimate models that control for or capture these and other sources of heterogeneity. â¢ The reliability ratio is not a stable parameter for reliability valuation. Many previous studies have focused on calculating the reliability ratio, which is defined as the ratio between the VOR and the VOT; however, this ratio varied significantly between different models and was not stable enough for use in planning analysis. That is, having a high VOT does not necessarily imply a high VOR, and vice versa.