Review of U.S. Department of Transportation Truck Size and Weight Study - Second Report: Review of USDOT Technical Reports (2015)

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.

44 6. SAFETY The MAP-21 study charge calls for two analyses of truck safety: a comparison of accident frequency and accident risk of vehicles operating under grandfather or other exemptions from federal size and weight limits [Section 32801(a)(1)] and analysis of the safety impacts of allowing six-axle tractor-semitrailers and other alternative configurations [Section 32801(a)(5)], including the safety effect of any diversion of freight to highways that allowing operation of alternative configurations would induce [Section 32801(a)(6)(B)]. The three components of the USDOT study’s safety analysis were a comparison of crash rates of the alternative and control vehicles in states where vehicles similar to the alternative configurations now operate, comparison by simulation models of the stability of the alternative and control vehicles, and comparison of results of vehicle safety inspection conducted by enforcement officers. These are necessary components of the safety analysis and have been included in most past truck size and weight studies. The crash rate and enforcement analyses explore methods that may advance knowledge over the methods of past studies. Responsiveness to the Questions Identified by Congress The study does not respond to the legislative charge to evaluate the safety impacts of changing regulations to allow alternative configurations. Such an evaluation would require considering how the combined effects of changes in traffic volume and changes in crash rates resulting from the introduction of alternative configurations would affect the total systemwide frequency of crashes and casualties. The USDOT study explored the possibility of estimating the change in the annual number of crashes nationwide that would occur in each of its alternative configuration scenarios (Safety, 91–93). However, the study concluded that “it is not possible to draw national conclusions or present findings concerning national crash rates due to a lack of relevant crash data” (Summary, ES-6). Past truck size and

45 weight studies (e.g., USDOT 1981, II-22; TRB 1990a, 15; TRB 1990b, 129–130) included such quantitative comparisons of systemwide safety impacts of alternatives based on syntheses of information from analyses of crash data, results of research of others on crash rates, and vehicle stability simulation and testing. Methods and Data The comments below concern features of the alternative and control vehicle crash involvement rate estimates in the USDOT study that affect the credibility of the estimates, the interpretation of the results of the model of the relationship of truck crash frequency to truck traffic volume, and comparison of the vehicle stability results of the USDOT study with those of past studies. Crash Involvement Rate Analysis Systematic Uncertainty in Truck VMT Estimates The uncertainty in the estimates of VMT by truck configuration [the denominators in the crash involvement rate estimates (Safety, 26–27)] is unknown and potentially large, especially for uncommon vehicles. The distributions of VMT by vehicle configuration, road system, and state are derived from WIM data. The calculation of these distributions entailed a series of assumptions and approximations (Modal Shift, 236–241). The report does not describe any attempt to validate the estimated distributions (e.g., by comparing the estimates with observed distributions). 9 The study estimated distributions of VMT by configuration for 2 years, 2008 and 2011, but used only the 2011 VMT distribution estimates in estimating crash rates. The report (Safety, 17–18) explains that the 2008 VMT estimates were discarded 9 The committee understands that USDOT plans to publish additional documentation on data methods in a volume titled Data Acquisition and Technical Analysis Plan.

46 because “the method used to estimate configuration-specific VMT in the 2008 data differed from the 2011 data to the extent that the 2008 data were not usable as a second data point.” This observation exposes the shortcoming of the WIM-derived VMT distribution estimates: the configuration distributions are sensitive to assumptions made in processing the WIM data. As one example of difficulties in interpreting WIM data, it was pointed out to the committee 10 that one axle on some Michigan six-axle tractor-semitrailers is a lift axle that is raised from the pavement when the truck is not loaded. WIM data would record such vehicles as having five axles, while crash data would record them as six-axle vehicles. The documentation of the derivation of the vehicle class distributions does not indicate that this problem was taken into account. The tests of significance of differences in crash rates between pairs of vehicle types shown in Tables 8, 9, and 10 (Safety, 26–27) are incorrect. The test applied assumes that the values of the denominators in the ratios compared are known with certainty. In this case, the values of the denominators (VMT) are uncertain to an unknown degree. Citing consistent results in all three states for which data were obtained for the comparison of six- axle with five-axle tractor-semitrailer crash involvement rates, the USDOT report concludes: “This consistency across states lends validity to this finding” (Safety, 46). This argument would be sound if the estimates in the three states were independent. However, the crash rate estimates all depend on VMT estimates derived by the same procedure. If that procedure introduces systematic error in the VMT estimates, the consistency across states could be the result of the error. Risk Factors Not Considered The safety performance of a population of trucks depends on the characteristics of drivers, management practices of the companies that operate the vehicles, the physical condition and traffic characteristics of 10 Letter of Kirk T. Steudle, Director, Michigan Department of Transportation, to the committee, July 27, 2015. The letter states that data of the Michigan Department of Transportation suggest that the ratio of six-axle to five-axle VMT is higher than reported in the USDOT study. The committee did not examine the state data.

47 the roads the vehicles use, and the distribution of the trucks’ travel by time of day and season of the year. Therefore, in many circumstances, factors other than size and weight are likely to dominate safety comparisons among populations of vehicles. The crash rate estimates control for some of these factors by limiting the comparison to Interstate routes in individual states and by separately comparing urban and rural rates. However, the other factors influencing crash rates are not considered in the analysis. The populations of the alternative configurations and the control vehicles likely differ in some of these factors. In the states where they are in use today, the alternative configurations typically serve specialized, niche markets (e.g., an extractive industry operating entirely within the state boundaries), whereas the control vehicles are widely used in a broad variety of markets. Failure to consider differences in factors influencing crash rates may have biased the estimates of crash rate differences between the alternative and control vehicles. If the alternative configurations now used primarily in intrastate niche markets became legal for general nationwide use, the characteristics of their drivers, carriers, routes, and other features of use would resemble those of the present-day control vehicles rather than those of the present-day alternative configurations. Descriptions of Alternative Configurations in Operation The committee received information from the state of Michigan that descriptions of Michigan Interstate truck size and weight limits in the USDOT report are inaccurate. In Table 1 of the safety volume (Safety, 3), the maximum gross vehicle weight for a tractor-semitrailer in Michigan is shown as 104,000 pounds. This would be the maximum in the state for a six-axle tractor-semitrailer with the steering axle loaded to 18,000 pounds; however, tractor-semitrailers with more than six axles and weights up to 154,000 pounds operate on Michigan Interstates. Later the report states that the maximum gross vehicle weight for a six- axle tractor-semitrailer is 105,500 pounds (Safety, 13); the state informed the committee that the correct limit is 104,000 pounds. The state’s communication also notes that six-axle tractor-semitrailers in

48 Michigan typically operate with the rearmost three axles widely spread, rather than in the tridem configuration illustrated in the report and assumed in the stability analysis (Safety, ES-3, 56). 11 The description of triple-trailer use in Iowa (which does not enter into the report’s safety analysis) also requires clarification. Triples operate in Iowa, as Table 2 in the summary volume (Summary, 9) indicates, but the committee understands that they are allowed only in the Sioux City metropolitan area in the extreme northwest corner of the state. Presentation of Crash Rate Estimates in the USDOT Report The USDOT report acknowledges that the state estimates are not nationally representative, but USDOT repeatedly and prominently cites them: in the Summary volume (p. ES-6), the transmittal letter to Congress, 12 and the Safety volume summary (p. ES-6). The statements acknowledging unrepresentativeness of the state-level crash rate estimates are not accurately worded. The Summary volume (p. ES-6) states: “Crash rates for the six-axle alternative truck configuration in Washington State are significantly higher than the five-axle control truck rates. However, it is not possible to draw national conclusions or present findings concerning national crash rates due to a lack of comparable crash data in other States.” As explained above, the USDOT study does not show that the estimated rate differences are significantly different in the statistical sense. Therefore, the limitation is not simply lack of comparable data in other states, but unknown validity of the estimates in the states where data were obtained. Lack of Consideration of Crash Severity and Traffic Volume in the Safety Analysis The USDOT study reports evidence that the average severity of crashes of some alternative configurations appears to be less than that of control vehicle crashes (Safety, 94). This finding does not 11 Letter of Kirk T. Steudle, Director, Michigan Department of Transportation, to the committee, July 27, 2015. 12 Peter M. Rogoff to Hon. Bill Shuster, June 5, 2015: “FHWA was able to identify significantly higher crash rates in six-axle trucks compared to five-axle trucks in the State of Washington. . . .”

49 show a causal relationship; it may reflect differences in operating environments not considered in the analysis (e.g., a larger share of the alternative configuration’s VMT may occur on congested roads, which tend to have higher crash rates but lower average crash severity than uncongested roads). Nevertheless, consideration of changes in severity along with changes in rates is necessary in assessing the safety of alternative configurations. Computing crash rates by severity (e.g., for casualty crashes and property- damage-only crashes) and by type of crash (e.g., for single-vehicle and multivehicle crashes or for rollover and run-off-road crashes), where data are sufficient, might have aided interpretation of the rates. The most meaningful measure of the safety effects of changing federal regulations is the change in the total costs of road crashes, including changes in the frequency of fatalities and nonfatal injuries, in property damage costs, and in congestion costs of crashes. The change in total costs depends on changes in crash rates, crash severities, and truck traffic volumes. Past studies have concluded that the change in truck traffic volume is likely to be the critical determinant of safety impact of the changes in regulations most commonly proposed (TRB 2002, 110). The federal study does not include an estimate of changes in crash costs. Analysis of the Relationship of Crash Involvement Rate to Traffic Volume The committee’s first report (TRB 2014, 33) commended the federal study’s plan to explore an approach to modeling truck crash risk that takes into account truck volume and total traffic volume rather than assuming that the crash involvement rate for a particular truck configuration is a constant on all roads of a particular road class. The results of the model estimation (Safety, 28) appear reasonable, indicating that truck involvement rates decline as truck volume on a road increases. However, the model has weaknesses. Segment-level annual average daily traffic (AADT) by vehicle type data were not available and had to be estimated on the basis of the strong assumption that the distribution of VMT by vehicle type is the same on all roads of the same functional class (Safety, 22). Also, the analysis did not consider factors other than

50 AADT that may affect the risk comparisons. For example, any differences between truck types in patterns of use by time of day or in frequency of use of Interstate exits would likely influence relative crash rates. Figures 2 and 3 (Safety, 29–30), illustrating the results of the crashes-versus-AADT model, are misleading. Figure 2 shows an extrapolation of the six-axle tractor-semitrailer crash frequency to traffic volumes that are 50 times greater than any observed. At the high-volume end of the curve, the confidence intervals on the extrapolated values (which the report does not indicate) are extremely large. Because the analysis did not use actual segment-level AADT by vehicle type, important risk-related factors were not considered in the analysis, and extrapolation far beyond the range of the observations is not statistically justifiable, the plots of predicted alternative configuration crashes at high traffic volume are not meaningful. The figures would be more instructive if the data points used to estimate the relationships were shown. The committee attempted to replicate points on the graphs of Figures 2 and 3 by using the estimated equation parameters and AADT data in the report but could not. Part of the difficulty is that the report does not define all the variables appearing in the model specifications. The graphs should be carefully checked before the report is put in final form. Vehicle Stability Analysis The vehicle stability analysis estimated metrics of vehicle performance during five vehicle maneuvers for each control and alternative configurations (Safety, 53), using a standard truck dynamics simulation model. The report also summarizes the results of braking tests with actual vehicles. Several of the results appear inconsistent with results of past analyses. Most notably, the rearward amplification of seven-axle and nine-axle triples in the avoidance maneuver was found to be no different from that of 28-foot and 33- foot doubles (Safety, 71). In contrast, the 2000 USDOT truck size and weight study (USDOT 2000, VIII- 12) found the rearward amplification of seven-axle A-train triples at 132,000 pounds to be substantially more severe than that of the five-axle 28-foot double at 80,000 pounds. A 1990 FHWA simulation study

51 of truck stability similarly reported much worse rearward amplification for seven-axle triples than for five-axle doubles (Fancher and Mathew 1990, 127). As the USDOT study notes (Safety, 53), greater rearward amplification is believed to be associated with greater risk of rollover during an avoidance maneuver. The study would have been strengthened if it had included examination of the sources of differences with results of past studies. In describing the performance of the four multitrailer combinations in the avoidance maneuver simulation, the USDOT report states that “all would be in danger of rolling over” and that, for the triple trailers, “the load on one end of the axle on the third trailer was completely removed for periods of less than one second.” This adverse finding does not appear in the summaries of the stability analysis in the safety report (Safety, 73–75) or in the summary volume (Summary, 50–51). The USDOT report notes that highway work zone hazards may increase in the alternative vehicle scenarios (Modal Shift, 62) but does not estimate the change in crash risk. Larger trucks may perform more poorly in construction zones than existing trucks, and introducing certain of the larger trucks would require an increase in highway construction for some period. The report’s interpretation of the results of the simulations of path deviation and high-speed off-tracking of double-trailer and triple-trailer combinations does not consider the consequences of narrower lanes in construction zones. As the USDOT report acknowledges (Safety, 58), the stability analysis does not consider the consequences of the new National Highway Traffic Safety Administration (NHTSA) regulation requiring electronic stability control on all new truck-tractors in 2017. NHTSA expects that the rule will greatly reduce the frequency of truck rollover and loss-of-control crashes [80 Federal Register 36050 (June 23, 2015)].

52 Inspection and Violation Analysis The USDOT study compared the alternative configurations with the control vehicles with respect to rates of violation of safety rules found in roadside safety inspections conducted routinely by state enforcement officials. The analysis found that “tractor semitrailer configuration was not a significant predictor of the likelihood of a violation. That is, no significant difference was observed between the alternative tractor semitrailer configurations and the 80,000-lb. semitrailers with respect to violations, when controlling for other factors. . . .” (Summary, 53). The other factors controlled for were driver age, vehicle age, and carrier out-of-service (OOS) violation rate. This finding is important for the entire safety analysis, because it suggests that differences in operator characteristics may be the major source of safety-related differences in comparisons of groups of vehicles. As noted above, operator characteristics were not taken into account in the USDOT study’s crash rate comparisons. The report’s interpretation of the finding that operator characteristics predict violation rates is a non sequitur: “These findings have direct implications for the use of heavier combination vehicles. If carriers that enter the market using the heavier 3-S3 vehicles also have higher OOS and older equipment, this model suggests they may have higher violation rates as well” (Safety, 84). Operators with high OOS rates and old equipment that switched to the alternative configurations would have violation rates the same as those they had with their previous equipment, according to the USDOT study’s analysis. Moreover, there are no grounds for expecting that carriers with higher OOS rates or older equipment would be more likely to convert to the alternative configurations than carriers with lower rates. Any differences today between alternative and control vehicle operators in OOS rate or equipment age most probably reflect the specialized niche markets where the alternative configurations in operation today are concentrated. If the alternative configurations became the standard nationwide, their operators would have the same characteristics as present-day operators of nationally standard vehicles.

53 Recommendations The USDOT study’s effort to identify states with crash data and traffic data adequate to support estimates of crash rates for specific truck configurations was worthwhile. USDOT should continue to work with the states to develop data systems that can be used to monitor the safety performance of tractor-semitrailers operating within the federal weight limit, heavier tractor-semitrailers, and multitrailer combinations. Development and analyses of exposure data that would help to improve crash rate estimates derived by the USDOT study’s method include the following:  Validation of the WIM-derived vehicle type distributions by comparison with independent data sources (e.g., visual counts).  Sensitivity analysis to show the effect of assumptions and approximations in the derivation of the vehicle type distributions [e.g., comparison of distributions from the 2008 WIM analysis method with those using the 2011 method described in the report (Safety, 17–18)].  Comparison of crash rates derived by the study method with other credible, independent estimates.  Inclusion of more states in the estimates. Selection of the states included in the USDOT report crash rate estimates appears to have been based, in part, on availability of data required for the regression model of crash rate versus traffic volume. Whether these were the only states for which unadjusted crash rates could have been computed for various truck configurations is not clear.  Improvements in WIM data collection and analysis to allow extension of the method to non-Interstate roads and to allow computation of crash rates by time of day and traffic volume. Understanding the safety effect of changes in truck traffic volume and characteristics requires a model of the relationship of total crash frequency on a road to the traffic volume and mix of vehicle types on the road. The traditional model, in which each vehicle type is assigned a fixed crash involvement rate

54 for each road functional class, is unrealistic (TRB 1996, 69–72). USDOT should support research aimed at understanding this relationship. The models of crash frequency versus traffic volume explored in the USDOT study are a contribution to this development. References Abbreviations TRB Transportation Research Board USDOT U.S. Department of Transportation Fancher, P., and A. Mathew. 1990. Safety Implications of Various Truck Configurations—Vol. I: Technical Report. Federal Highway Administration, Jan. TRB. 1990a. Special Report 225: Truck Weight Limits: Issues and Options. National Research Council, Washington, D.C. TRB. 1990b. Special Report 227: New Trucks for Greater Productivity and Less Road Wear: An Evaluation of the Turner Proposal. National Research Council, Washington, D.C. TRB. 1996. Special Report 246: Paying Our Way: Estimating Marginal Social Costs of Freight Transportation. National Research Council, Washington, D.C. TRB. 2002. Special Report 267: Regulation of Weights, Lengths, and Widths of Commercial Motor Vehicles. National Academies, Washington, D.C. TRB. 2014. Review of U.S. Department of Transportation Truck Size and Weight Study: First Report: Review of Desk Scans. March 31. http://onlinepubs.trb.org/onlinepubs/sr/TS&WDeskScans.pdf. USDOT. 1981. An Investigation of Truck Size and Weight Limits: Final Report. Aug. USDOT. 2000. The U.S. Department of Transportation’s Comprehensive Truck Size and Weight Study: Volume III: Scenario Analysis. Aug.